Prepare the v0.5.0 release

.woodpecker/release.yml

@@ -0,0 +1,33 @@
+when:
+  - event: tag
+
+steps:
+  - name: validate-release
+    image: golang:1.25.5
+    commands:
+      - |
+        set -eu
+
+        version="$CI_COMMIT_TAG"
+        release_note="docs/releases/$version.md"
+
+        if ! printf '%s\n' "$version" | grep -E -x 'v(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)' >/dev/null; then
+          printf '%s\n' "invalid release tag: $version" >&2
+          exit 1
+        fi
+        if [ ! -s "$release_note" ]; then
+          printf '%s\n' "missing release note: $release_note" >&2
+          exit 1
+        fi
+        if ! grep -F -x "# Notarius $version" "$release_note" >/dev/null; then
+          printf '%s\n' "release note heading does not match $version" >&2
+          exit 1
+        fi
+        for heading in '## Summary' '## Compatibility' '## Upgrade' '## Changes'; do
+          if ! grep -F -x "$heading" "$release_note" >/dev/null; then
+            printf '%s\n' "release note is missing heading: $heading" >&2
+            exit 1
+          fi
+        done
+
+        ./scripts/check-release-source.sh "$version"

README.md

@@ -28,6 +28,20 @@ For the complete ordered D&D workflow, use
 [its synthetic transcript](examples/dnd-complete-transcript.json). It
 demonstrates all implemented D&D lanes and the supporting campaign references.
 
+## Install A Source Release
+
+Install a pinned source release with Go:
+
+~~~
+GOWORK=off go install \
+  gitea.maximumdirect.net/eric/notarius/cmd/notarius@<tag>
+~~~
+
+Replace `<tag>` with a stable release tag such as `vMAJOR.MINOR.PATCH`. The
+installed command's diagnostic version is described in the [CLI
+reference](docs/cli.md); maintainers preparing a release should follow [Source
+Releases](docs/release.md).
+
 ## Documentation
 
 - [CLI reference](docs/cli.md) — commands, flags, output streams, and exits.
@@ -39,6 +53,8 @@ demonstrates all implemented D&D lanes and the supporting campaign references.
   artifact formats.
 - [Subprocess consumer guide](docs/consumers/subprocess.md) — invoke Notarius
   from an orchestrator and consume a published result.
+- [Complete D&D consumer guide](docs/consumers/dnd-pipeline.md) — run the full
+  D&D pipeline as a subprocess and discover its structured artifacts.
 - [Internal overview](docs/internal/overview.md) — implemented component map
   for maintainers.
 - [Developer guide](docs/development.md) — contributor orientation and

assets/dnd/combat-turns/prompts/prompt.yaml

@@ -42,4 +42,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_combat_turns_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/enemy-events/prompts/prompt.yaml

@@ -50,4 +50,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_enemy_events_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/entity-reconciliation/schemas/dnd_entity_reconcile_llm.v1.json

@@ -1,47 +0,0 @@
-{
-  "$schema": "https://json-schema.org/draft/2020-12/schema",
-  "$id": "notarius.dnd.entity_reconcile.llm",
-  "type": "object",
-  "additionalProperties": false,
-  "required": ["duplicate_groups"],
-  "properties": {
-    "duplicate_groups": {
-      "type": "array",
-      "items": {
-        "type": "object",
-        "additionalProperties": false,
-        "required": ["members", "canonical"],
-        "properties": {
-          "members": {
-            "type": "array",
-            "items": {"$ref": "#/$defs/selector"}
-          },
-          "canonical": {"$ref": "#/$defs/selector"}
-        }
-      }
-    }
-  },
-  "$defs": {
-    "selector": {
-      "type": "object",
-      "additionalProperties": false,
-      "required": ["name", "source_refs"],
-      "properties": {
-        "name": {"type": "string", "minLength": 1},
-        "source_refs": {
-          "type": "array",
-          "minItems": 1,
-          "items": {
-            "type": "object",
-            "additionalProperties": false,
-            "required": ["start_unit_id", "end_unit_id"],
-            "properties": {
-              "start_unit_id": {"type": "integer", "minimum": 1},
-              "end_unit_id": {"type": "integer", "minimum": 1}
-            }
-          }
-        }
-      }
-    }
-  }
-}

assets/dnd/item-occurrences/prompts/prompt.yaml

@@ -42,4 +42,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_item_occurrences_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/item-occurrences/schemas/dnd_item_occurrences_llm.v1.json

@@ -22,10 +22,10 @@
             "items": {
               "type": "object",
               "additionalProperties": false,
-              "required": ["start_segment", "end_segment"],
+              "required": ["start_unit_id", "end_unit_id"],
               "properties": {
-                "start_segment": {"type": "integer"},
+                "start_unit_id": {"type": "integer"},
-                "end_segment": {"type": "integer"}
+                "end_unit_id": {"type": "integer"}
               }
             }
           }

assets/dnd/item-registry/extract/prompts/prompt.yaml

@@ -37,4 +37,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_item_registry_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/item-registry/normalize/prompts/instructions.md

@@ -1,8 +1,9 @@
-Use candidate names and cited transcript windows only to determine whether
+Determine whether candidates identify the same item type or unique designation
-candidates identify the same item type or unique designation. Do not treat
+using their contextual labels and cited transcript windows. Do not treat nearby
-nearby evidence, similar objects, or a shared owner as sufficient. Keep
+evidence, similar objects, or a shared owner as sufficient.
-currency denominations, materially different item types, and uncertain aliases
+
-separate. Do not infer an item property or uniqueness.
+Keep currency denominations and materially different item types separate. Keep
+uncertain aliases separate. Do not infer an item property or uniqueness.
 
 When selecting a canonical display name, choose one supplied candidate name
 that is the clearest established designation.

assets/dnd/item-registry/normalize/prompts/prompt.yaml

@@ -12,19 +12,19 @@ messages:
   - role: system
     content_file: ./sharedassets/common-dnd-system.md
   - role: user
-    content_file: ./instructions.md
+    content_file: ./sharedassets/protocol.md
   - role: user
-    content_file: ./sharedassets/common-dnd-entity-reconciliation.md
+    content_file: ./instructions.md
     cache_control:
       type: ephemeral
   - role: user
-    content_file: ./candidates.md
+    content_file: ./sharedassets/candidates.md
   - role: user
-    content_file: ./sharedassets/common-dnd-transcript-windows.md
+    content_file: ./sharedassets/transcript-windows.md
     cache_control:
       type: ephemeral
 output:
   format: json
   validation_mode: json_schema
-  schema_path: dnd_entity_reconcile_llm.v1.json
+  schema_path: semantic_reconciliation_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/location-occurrences/prompts/prompt.yaml

@@ -42,4 +42,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_location_occurrences_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/location-registry/extract/prompts/prompt.yaml

@@ -37,4 +37,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_location_registry_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/location-registry/normalize/prompts/candidates.md

@@ -1,2 +0,0 @@
-Location candidates:
-{{ input "candidates" }}

assets/dnd/location-registry/normalize/prompts/instructions.md

@@ -1,6 +1,8 @@
-Use candidate names and their cited transcript windows to determine whether
+Determine whether candidates identify the same physical place using their
-candidates identify the same physical place. Do not treat matching names,
+contextual labels and cited transcript windows. Do not treat matching names,
-nearby evidence, nested places, or generic labels as sufficient. Keep parent
+nearby evidence, nested places, or generic labels as sufficient.
-and child places, similarly named places, and uncertain aliases separate.
+
+Keep parent and child places separate, as well as similarly named places and
+uncertain aliases.
 
 When selecting a canonical display name, prefer the clearest established name.

assets/dnd/location-registry/normalize/prompts/prompt.yaml

@@ -12,19 +12,19 @@ messages:
   - role: system
     content_file: ./sharedassets/common-dnd-system.md
   - role: user
-    content_file: ./instructions.md
+    content_file: ./sharedassets/protocol.md
   - role: user
-    content_file: ./sharedassets/common-dnd-entity-reconciliation.md
+    content_file: ./instructions.md
     cache_control:
       type: ephemeral
   - role: user
-    content_file: ./candidates.md
+    content_file: ./sharedassets/candidates.md
   - role: user
-    content_file: ./sharedassets/common-dnd-transcript-windows.md
+    content_file: ./sharedassets/transcript-windows.md
     cache_control:
       type: ephemeral
 output:
   format: json
   validation_mode: json_schema
-  schema_path: dnd_entity_reconcile_llm.v1.json
+  schema_path: semantic_reconciliation_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/npc-occurrences/prompts/prompt.yaml

@@ -42,4 +42,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_npc_occurrences_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/npc-registry/extract/prompts/prompt.yaml

@@ -37,4 +37,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_npc_registry_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/npc-registry/normalize/prompts/candidates.md

@@ -1,3 +0,0 @@
-NPC candidates for identity comparison:
-
-{{ input "candidates" }}

assets/dnd/npc-registry/normalize/prompts/instructions.md

@@ -1,6 +1,7 @@
-Use candidate aliases and their cited transcript windows to determine whether
+Determine whether candidates refer to the same individual using their
-candidates refer to the same individual. Preserve distinct individuals even
+contextual labels and cited transcript windows. Preserve distinct individuals
-when their names are similar.
+even when their names are similar or their contextual descriptions are
+identical.
 
 When selecting a canonical display name, prefer a complete, stable proper name
 over an abbreviation. Prefer an unadorned proper name over that name plus a

assets/dnd/npc-registry/normalize/prompts/prompt.yaml

@@ -12,19 +12,19 @@ messages:
   - role: system
     content_file: ./sharedassets/common-dnd-system.md
   - role: user
-    content_file: ./instructions.md
+    content_file: ./sharedassets/protocol.md
   - role: user
-    content_file: ./sharedassets/common-dnd-entity-reconciliation.md
+    content_file: ./instructions.md
     cache_control:
       type: ephemeral
   - role: user
-    content_file: ./candidates.md
+    content_file: ./sharedassets/candidates.md
   - role: user
-    content_file: ./sharedassets/common-dnd-transcript-windows.md
+    content_file: ./sharedassets/transcript-windows.md
     cache_control:
       type: ephemeral
 output:
   format: json
   validation_mode: json_schema
-  schema_path: dnd_entity_reconcile_llm.v1.json
+  schema_path: semantic_reconciliation_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/scene-descriptions/prompts/prompt.yaml

@@ -35,4 +35,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_scene_descriptions_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/scenes/prompts/prompt.yaml

@@ -31,4 +31,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_scenes_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/shared/prompts/common-dnd-entity-reconciliation.md

@@ -1,7 +0,0 @@
-Identify only well-supported duplicate groups among the supplied candidates.
-
-Return each selected candidate's supplied contextual descriptor exactly: its
-`name` and complete ordered `source_refs`. A group must contain at least two
-supplied descriptors, and its `canonical` descriptor must be one of its
-members. Do not invent names, ranges, records, evidence, or replacement values.
-Omit any uncertain or unsafe group.

assets/dnd/shared/prompts/common-dnd-extraction-evidence.md

@@ -1,7 +1,6 @@
 Transcript units are the only evidence for extracted events and factual claims.
 Every reported factual claim must be supported by cited transcript units. Use
-integer `start_unit_id` and `end_unit_id` values from the transcript. Omit
+integer `start_unit_id` and `end_unit_id` values from the transcript.
-`source_id`; Notarius assigns the current source identity.
 
 When supporting evidence is non-contiguous, use multiple narrow ranges rather
 than a broad range that bridges unrelated conversation.

assets/dnd/shared/prompts/common-dnd-system.md

@@ -1,7 +1,5 @@
 You process Dungeons & Dragons gameplay transcripts.
 
-Rely only on the supplied inputs. They may contain transcription errors,
+As input, you will receive one or more portions of a transcript. The transcript may contain transcription errors, repeated lines, incomplete sentences, and misheard proper nouns.
-repeated lines, incomplete sentences, and misheard proper nouns.
 
-Return exactly one JSON object that conforms to the configured response schema,
+Return exactly one JSON object that conforms to the configured response schema, with no explanatory prose.
-with no explanatory prose.

assets/dnd/shared/prompts/common-dnd-transcript-chunk.md

@@ -1,5 +1,3 @@
-One extraction chunk from a Dungeons & Dragons gameplay transcript is provided
+One extraction chunk from a Dungeons & Dragons gameplay transcript is provided below. Report and infer only what is within this chunk. Its unit IDs retain their source-wide meaning.
-below. Report and infer only what is within this chunk. Its unit IDs retain
-their source-wide meaning.
 
 {{ input "transcript" }}

assets/dnd/shared/prompts/common-dnd-transcript-full.md

@@ -1,4 +1,3 @@
-The complete ordered transcript of this Dungeons & Dragons gameplay session is
+The complete ordered transcript of this Dungeons & Dragons gameplay session is provided below.
-provided below. It may contain multiple scenes.
 
 {{ input "transcript" }}

assets/dnd/shared/prompts/common-dnd-transcript-windows.md

@@ -1,6 +0,0 @@
-Selected Dungeons & Dragons gameplay transcript evidence windows are provided
-below. They may be incomplete, non-contiguous, or overlapping. Use them to
-evaluate candidate identity, but do not treat absence outside these windows as
-evidence.
-
-{{ input "transcript" }}

assets/dnd/spells/prompts/prompt.yaml

@@ -47,4 +47,4 @@ output:
   format: json
   validation_mode: json_schema
   schema_path: dnd_spells_llm.v1.json
-  repair_attempts: 0
+  repair_attempts: 1

assets/dnd/spells/prompts/spell-catalog.md

@@ -1,6 +1,6 @@
-The canonical spell-name catalog for this extraction is provided below as JSON.
+The spell catalog for this extraction is provided below as JSON. Each entry
-Return spell names using the catalog's canonical spelling exactly. Aliases and
+lists a `canonical_name` and its recognized `aliases`. If the transcript uses
-other campaign reference material are not part of this catalog input and must
+an alias, select that entry's `canonical_name`. Return spell names using the
-not be copied into the output as spell names.
+canonical spelling exactly; never return an alias as a spell name.
 
 {{ input "spell_catalog" }}

assets/generic/normalize/deduplication/prompts/candidates.md

@@ -1,2 +1,3 @@
-Item candidates:
+Candidate material:
+
 {{ input "candidates" }}

assets/generic/normalize/deduplication/prompts/instructions.md

@@ -0,0 +1,5 @@
+Identify only high-confidence duplicate entities among the supplied candidates.
+
+Preserve distinct entities even when their names are similar. Treat contextual descriptions and transcript evidence as supporting material, not as permission to merge ambiguous records.
+
+When several records are duplicates, choose as canonical the candidate with the clearest stable identity. Prefer a complete proper name over an abbreviation, and prefer an unadorned proper name over one with incidental descriptors unless the evidence establishes those descriptors as part of the name. A longer name is not inherently more canonical.

assets/generic/normalize/deduplication/prompts/prompt.yaml

@@ -0,0 +1,27 @@
+id: generic.semantic_reconciliation
+version: "v1"
+inputs:
+  - name: candidates
+    required: true
+    content_type: application/json
+  - name: transcript
+    required: true
+    content_type: application/json
+messages:
+  - role: system
+    content_file: ./system.md
+  - role: user
+    content_file: ./protocol.md
+  - role: user
+    content_file: ./instructions.md
+    cache_control:
+      type: ephemeral
+  - role: user
+    content_file: ./candidates.md
+  - role: user
+    content_file: ./transcript-windows.md
+output:
+  format: json
+  validation_mode: json_schema
+  schema_path: semantic_reconciliation_llm.v1.json
+  repair_attempts: 1

assets/generic/normalize/deduplication/prompts/protocol.md

@@ -0,0 +1,7 @@
+Use only the positive integer `candidate_id` values supplied in the candidate material.
+
+Return a duplicate group only when the evidence supports that every selected candidate describes the same underlying entity. Each group must contain at least two distinct candidate IDs, and its `canonical_candidate_id` must be one of those IDs. A candidate may appear in at most one group.
+
+Omit uncertain matches and candidates that should remain distinct. Do not invent candidates or infer an ID from list position. An empty `duplicate_groups` array is valid.
+
+The response must conform exactly to the selected JSON schema. Return IDs only: do not copy candidate names, evidence, transcript text, source identifiers, or source ranges into the response.

assets/generic/normalize/deduplication/prompts/system.md

@@ -0,0 +1,2 @@
+You reconcile structured records that may describe the same underlying entity.
+Follow the supplied protocol and return only the requested structured result.

assets/generic/normalize/deduplication/prompts/transcript-windows.md

@@ -0,0 +1,3 @@
+Transcript evidence windows:
+
+{{ input "transcript" }}

assets/generic/normalize/deduplication/schemas/semantic_reconciliation_llm.v1.json

@@ -0,0 +1,32 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "notarius.generic.semantic_reconciliation.llm",
+  "title": "notarius_semantic_reconciliation_llm_v1",
+  "type": "object",
+  "additionalProperties": false,
+  "required": ["duplicate_groups"],
+  "properties": {
+    "duplicate_groups": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "additionalProperties": false,
+        "required": ["candidate_ids", "canonical_candidate_id"],
+        "properties": {
+          "candidate_ids": {
+            "type": "array",
+            "minItems": 2,
+            "items": {
+              "type": "integer",
+              "minimum": 1
+            }
+          },
+          "canonical_candidate_id": {
+            "type": "integer",
+            "minimum": 1
+          }
+        }
+      }
+    }
+  }
+}

docs/adr/0012-resolve-opaque-entity-identifiers-deterministically.md

@@ -12,9 +12,9 @@ an opaque implementation detail. A plain name is likewise insufficient where
 multiple supplied records share that name.
 
 The LLM boundary must preserve the typed artifact and durable-schema ownership
-of [ADR-0003](0003-strongly-typed-stage-interfaces.md) and the distinction
+of [ADR-0003](0003-typed-interfaces-with-two-zone-data-model.md) and the distinction
 between disambiguating references and source evidence in
-[ADR-0009](0009-prefer-minimal-evidence-grounded-extraction-artifacts.md).
+[ADR-0009](0009-minimal-evidence-grounded-extraction-artifacts.md).
 
 ## Decision
 

docs/adr/0013-use-request-local-candidate-handles-for-semantic-reconciliation.md

@@ -0,0 +1,89 @@
+# ADR-0013: Use request-local candidate handles for semantic reconciliation
+
+**Status:** Accepted
+**Date:** 2026-08-09
+
+## Context
+
+Several typed normalize stage modules need semantic reconciliation after
+deterministic preprocessing: a model can judge whether source-backed candidates
+refer to the same underlying entity, while application code remains responsible
+for constructing the normalized artifact. Requiring the model to reproduce a
+candidate's full contextual selector makes the response larger and introduces
+avoidable formatting, ordering, and transcription failure modes.
+
+Reconciliation must preserve the exact typed artifact boundary established by
+[ADR-0003](0003-typed-interfaces-with-two-zone-data-model.md), the domain-neutral
+framework and concrete-domain dependency direction established by
+[ADR-0004](0004-package-modules-by-domain.md), and the distinction in
+[ADR-0009](0009-minimal-evidence-grounded-extraction-artifacts.md) between source
+evidence and auxiliary identity context. It also needs a concrete, narrowly
+scoped application of the request-local-label exception allowed by
+[ADR-0012](0012-resolve-opaque-entity-identifiers-deterministically.md).
+
+## Decision
+
+Semantic reconciliation will be a domain-neutral framework mechanism used by
+typed normalize stage modules. A consuming artifact family will retain
+ownership of its typed records, identity rules, consolidation policy, durable
+IDs, and domain warnings; the framework mechanism will not infer those rules
+from arbitrary data.
+
+For each reconciliation request, deterministic code will assign every eligible
+model-visible candidate a contiguous, one-based integer handle. The model may
+receive the candidate's contextual label, source references, and bounded source
+context needed to judge identity, but its structured response will identify
+candidates only by those supplied handles. A handle is local to one request,
+does not represent entity identity, and must never enter a durable artifact or
+be used to derive a durable ID.
+
+The model will propose duplicate groups and select one supplied member of each
+group as canonical. Deterministic code will resolve the handles through the
+retained request mapping, validate the complete proposal, discard unsafe
+groups, and apply only validated groups through typed domain-owned policy. The
+model will not synthesize replacement records or directly mutate an artifact.
+
+Every reconciliation prompt will combine a mandatory framework-owned protocol
+and safety policy with an explicitly selected semantic policy. The semantic
+policy may be the conservative generic policy or a domain-owned policy, but it
+cannot replace the shared response protocol or deterministic safety boundary.
+
+## Alternatives considered
+
+- Return durable application IDs. Opaque IDs do not help semantic judgment,
+  expose application identity mechanics, and make model output reproduce data
+  that deterministic code already owns.
+- Return names alone or copied contextual selectors. Names can be ambiguous,
+  while reproducing labels and source ranges adds response complexity and
+  creates mismatches without adding semantic information. Request-local
+  handles preserve exact selection without either failure mode.
+- Ask the model to return synthesized canonical replacement records. This
+  would transfer typed artifact construction, provenance consolidation, and
+  durable identity policy to a probabilistic boundary.
+- Reconcile reflection-discovered fields or arbitrary JSON. This would weaken
+  the typed artifact contract and move domain semantics into generic code.
+- Hide reconciliation inside extraction or another stage. This would obscure
+  stage ownership and create cross-stage behavior outside the fixed pipeline;
+  reconciliation remains explicit normalize-stage behavior.
+- Let each domain replace the complete prompt protocol. This would duplicate
+  safety mechanics and allow domain policy to bypass the common response and
+  validation contract.
+
+## Consequences
+
+Model responses become smaller and easier to validate, while deterministic
+application code retains authority over identity, provenance, ordering, and
+typed artifact construction. The framework requires a request-local mapping,
+bounded context preparation, a private integer response contract, proposal
+assessment, and shared prompt assets. Each consuming artifact family still
+requires a typed adapter for its irreducibly domain-specific rules.
+
+Request-local handles are deliberately unsuitable for persistence, logging as
+entity identity, checkpoint contracts, or cross-request correlation. Changes
+to shared protocol and policy assets must participate in the normal prompt,
+schema, and checkpoint fingerprint mechanisms.
+
+The shared mechanism and its initial D&D registry consumers are now
+implemented. Current behavior is documented in
+[Module Internals](../internal/modules.md#semantic-reconciliation) and
+[D&D Module Internals](../internal/dnd.md#semantic-registry-reconciliation).

docs/adr/0014-feedback-aware-validation-retries.md

@@ -0,0 +1,59 @@
+# ADR-0014: Use feedback-aware validation retries
+
+**Status:** Accepted
+**Date:** 2026-08-26
+
+## Context
+
+Validation can identify a candidate defect after a producer has returned an
+otherwise well-formed result. Retrying without the validator's deterministic,
+bounded feedback wastes the useful diagnosis, while treating validator
+execution failures as defects would ask a producer to repair conditions it
+cannot control. The mechanism must preserve typed producer ownership,
+checkpoint safety, and the repository's sensitive-data boundaries.
+
+## Decision
+
+The implementation will keep three independent budgets: the producer binding's
+outer `retries` budget, PromptKit's structured-output repair budget, and each
+validator's execution-retry budget. Validators will run sequentially in their
+configured order and aggregate both rejections and execution failures before a
+candidate disposition is selected.
+
+A correction-capable producer will provide the exact single LLM response that
+controlled its candidate using the `single_response_v1` protocol. A correction
+attempt will reconstruct the ordinary request and append exactly two fresh
+messages: that latest response as `assistant`, followed by one deterministic
+aggregate correction request as `user`. Earlier turns will not accumulate.
+
+Validator failures will not recurse into correction. Pipeline policy owns
+terminal disposition, with field-by-field producer overrides over pipeline
+defaults: structural failure and semantic rejection default to `fail_run`, and
+validator execution failure defaults to `warn_continue`. Validators can report
+facts and bounded corrective guidance, but never decide disposition.
+
+Rejected and structurally invalid candidates will not advance. A candidate
+allowed through after a validator execution failure will retain explicit
+incomplete-validation provenance and will not be checkpointed. Exact response
+and correction text remain attempt-local: they are excluded from ordinary
+errors, warnings, manifests, receipts, caches, checkpoints, and default debug
+summaries.
+
+## Alternatives considered
+
+- Retry every producer after any validation outcome. This conflates producer
+  defects with validator operational failures and wastes retry budget.
+- Let validators decide whether to continue. This would distribute pipeline
+  disposition policy across validators and undermine consistent defaults.
+- Reuse the full prior conversation. Accumulated turns introduce unbounded
+  prompt growth and make correction behavior depend on incidental history.
+- Persist raw responses to simplify diagnosis. Raw model output and correction
+  guidance may be sensitive and do not belong in durable pipeline records.
+
+## Consequences
+
+The framework gains transport-neutral correction and candidate contracts,
+producer capability checks, policy resolution, aggregated validation outcomes,
+and conservative checkpoint handling. Prompt construction remains inside the
+LLM adapter, while modules remain responsible for accurately exposing the
+single response that directly controlled a candidate.

docs/adr/0015-separate-process-warnings-from-quality-diagnostics.md

@@ -0,0 +1,67 @@
+# ADR-0015: Separate process warnings from quality diagnostics
+
+**Status:** Accepted
+**Date:** 2026-08-27
+
+## Context
+
+Notarius currently represents process degradation, incomplete validation,
+extraction-quality doubt, and routine normalization with one flat warning
+record. That makes ordinary successful runs noisy, loses the framework context
+needed to explain a finding, and gives `warning_count` no stable operational
+meaning. It also permits output encoders to add a warning after the durable
+warning file has already been written.
+
+The application needs one bounded diagnostic model that preserves exact
+occurrence counts while retaining only safe, representative samples. Fresh and
+resumed logical runs must present the same groups. The model must not alter
+validation decisions, retry budgets, rejected-output behavior, or process exit
+policy.
+
+## Decision
+
+Warnings are reserved for a completed run that advanced under an allowed
+process-level degradation or incomplete-work policy. Extraction-quality signals
+are advisories, and routine accepted transformations are observations. A
+non-degraded successful run therefore has zero actionable warnings.
+
+Modules and validators own a diagnostic's disposition, category, reason code,
+scope, and safe message. The framework adds pipeline origin, including stage,
+step, lane, module, validator, and chunk context where applicable. It then
+aggregates deterministically by disposition, category, reason code, and full
+origin. Chunk context remains on representative samples so equivalent findings
+across chunks aggregate together.
+
+Diagnostics carry exact occurrence counts, at most three distinct samples, and
+numeric omitted-sample metadata. Producers and validators are bounded to 64
+local groups. Final actionable warning groups are bounded without truncation;
+the non-warning collection may truncate represented groups while preserving an
+exact total occurrence count and explicit truncation metadata.
+
+The public contracts will be versioned: grouped actionable warnings use
+`notarius.warnings.v2`, grouped advisories and observations use
+`notarius.diagnostics.v1`, and the run receipt uses
+`notarius.run-result.v2`. Successful output encoders return logical files or
+an error; they do not add post-encoding warnings.
+
+## Alternatives considered
+
+- Keep one warning list and filter only CLI output. This would leave durable
+  consumers with the same semantically mixed, unbounded contract.
+- Map reason codes to severity in a central framework registry. This would
+  split module-owned meaning between synchronized policy tables and make new
+  diagnostic meaning implicit.
+- Preserve local omission warning records. They inflate visible group counts
+  and lose exact occurrence semantics.
+- Keep output-encoder warnings. A one-pass encoder cannot include those
+  records consistently in files it has already serialized; a two-phase encoder
+  protocol is deferred until a demonstrated need exists.
+
+## Consequences
+
+The framework gains validated diagnostic primitives, local collection,
+origin-aware aggregation, and versioned durable presentation. Existing warning
+transport remains temporarily while producers migrate. Current architecture,
+operator, integration, and internal documentation will describe the behavior
+only as each implementation step lands; this accepted decision does not claim
+that the migration is complete.

docs/cli.md

@@ -10,6 +10,7 @@ defined in [Operations](operations.md).
 
 ~~~
 notarius help
+notarius --version
 notarius run <pipeline-id> --input path/to/source.json [--json] [flags]
 notarius config validate [--config path/to/config.yml] [--pipeline pipeline-id] [--only lane-a,lane-b]
 notarius pipelines list [--config path/to/config.yml] [--json]
@@ -18,6 +19,17 @@ notarius pipelines list [--config path/to/config.yml] [--json]
 Running Notarius without arguments, or with **help**, **--help**, or **-h**,
 writes the command summary to standard output and exits with status 0.
 
+`notarius --version` is valid only as the sole root argument. It writes exactly
+`notarius <version>` followed by a newline to standard output and exits with
+status 0. A tagged `go install` build can report its main-module stable tag,
+and controlled builds can inject a stable tag at link time through
+`gitea.maximumdirect.net/eric/notarius/internal/buildinfo.Override`; an ordinary
+unversioned checkout reports `development`. Invalid injected version content is
+a runtime error with exit status 1, while extra `--version` arguments are a
+syntax error with exit status 2. This diagnostic does not replace the
+[run-result](integrations/run-result.md) or artifact contracts for downstream
+compatibility decisions.
+
 ## run
 
 ~~~
@@ -91,11 +103,14 @@ names, requiredness, and configured bindings are part of the
 Without **--json**, standard output contains the completed pipeline ID, counts
 of normalized and rejected outputs, and the output directory. A debug-enabled
 run also prints its debug-bundle path to standard output. A successful run with
-warnings reports the warning count to standard error. The published JSON bundle
+actionable process warnings reports their group and occurrence counts to
+standard error. When the selected output module publishes `warnings.json`, the
+summary also reports that durable file's path. Advisory and observation findings
+do not produce a warning line. The published JSON bundle
 is defined by the [JSON output contract](integrations/json-output.md).
 
 With **--json**, successful standard output is exactly one
-`notarius.run-result.v1` JSON document followed by a newline, with no
+`notarius.run-result.v2` JSON document followed by a newline, with no
 human-oriented status or debug-path line. Its fields and compatibility policy
 are defined by the [run-result contract](integrations/run-result.md). A caller
 must check for exit status 0 before decoding this output; a failed write can
@@ -159,7 +174,7 @@ go run ./cmd/notarius pipelines list \
 Successful commands write their primary result to standard output. Warnings and
 errors are written to standard error.
 
-For **run --json**, warnings remain on standard error and standard output is a
+For **run --json**, actionable process warnings remain on standard error and standard output is a
 machine-readable success result only. Syntax and runtime diagnostics remain on
 standard error. Parse the result only after the process exits with status 0.
 

docs/config.md

@@ -132,7 +132,11 @@ model: example-model
 Keep credentials out of the local-backend object. A PromptKit profile may name
 its credential environment variable through `api_key_env`; set that variable
 only in the run environment. PromptKit owns the
-[pinned profile-file format](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.5.0/docs/formats.md).
+[pinned profile-file format](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.9.0/docs/formats.md),
+including `base_profile` inheritance. Notarius passes profiles through without
+merging them. Filesystem profiles cannot express PromptKit's in-memory
+`APIKeyRequired` setting; an unset `api_key_env` is optional and may reach the
+provider without authorization.
 The [PromptKit upstream boundary](integrations/pkg-promptkit.md) identifies the
 supported package API, and [Operations](operations.md#operational-limits)
 describes the effective concurrency layers.
@@ -218,6 +222,8 @@ pipelines:
 | Field | Type | Default | Rules |
 | --- | --- | --- | --- |
 | **llm_profile** | string | none | Optional non-empty default PromptKit profile ID for selected LLM-backed bindings and validators. An explicitly present blank value is invalid. |
+| **structured_output_repair_attempts** | integer | prompt-owned (1 in maintained production prompts) | Optional structural-repair limit from 0 through 3 for selected LLM-backed bindings and validators. Omission leaves the prompt's declared policy in control; explicit 0 disables structural repair at that scope. |
+| **validation_policy** | object | see below | Optional terminal policy defaults for producer validation. Its fields inherit independently into chunk, extract, merge, and normalize bindings. |
 | **input** | module binding | none | Required. |
 | **chunk** | module binding | **generic** | Optional. |
 | **output** | module binding | **json** | Optional. |
@@ -237,6 +243,42 @@ run-level **--llm-profile** value first, then the binding's **llm_profile**,
 then the pipeline's **llm_profile**, and finally the PromptKit default.
 Deterministic bindings do not receive these defaults or run overrides.
 
+Structural output repair is resolved after module, validator, and `--only` lane
+selection. An object's **structured_output_repair_attempts** value takes
+precedence over the pipeline value; otherwise, an LLM-backed binding or
+validator inherits the pipeline value. If both are omitted, PromptKit uses the
+prompt's declared repair policy. The value must be an integer from 0 through 3;
+explicit `null` and non-integer values are invalid. An explicit value on a
+deterministic binding or validator is invalid, while a pipeline value simply
+does not apply to deterministic selections.
+
+`validation_policy` controls terminal disposition for one complete producer
+attempt and validator chain. It may appear on a pipeline or a **chunk**,
+**extract**, **merge**, or **normalize** module binding; input, output, and
+validator bindings reject it. Every field is optional and resolves in binding,
+pipeline, then application-default order:
+
+| Field | Values | Default |
+| --- | --- | --- |
+| **producer_structural_failure** | **fail_run**, **reject_output** | **fail_run** |
+| **semantic_rejection** | **fail_run**, **reject_output** | **fail_run** |
+| **validator_failure** | **warn_continue**, **fail_run** | **warn_continue** |
+
+The policy object and its fields must be non-null, and unknown fields are
+rejected. A deterministic producer may not explicitly set
+**producer_structural_failure** on its binding, although a pipeline-level
+default remains valid for pipelines that include LLM-backed producers.
+
+After the producer binding's retry budget is exhausted, an invalid structured
+response uses **producer_structural_failure**. One or more semantic validator
+rejections use **semantic_rejection**; rejection takes precedence over an
+exhausted validator failure or skip. With no rejection, an exhausted validator
+failure or skip uses **validator_failure**. `reject_output` records the
+terminal rejection without advancing that candidate. `warn_continue` is valid
+only for validator execution failure: it advances a structurally valid,
+otherwise unrejected result with incomplete-validation provenance and without
+making it reusable checkpoint state.
+
 A lane has these fields:
 
 | Field | Type | Default | Rules |
@@ -274,17 +316,32 @@ extract:
 | --- | --- | --- | --- |
 | **module** | string | none | Required for an object binding. Must be a registered compatible key. |
 | **llm_profile** | string | none | Optional non-empty PromptKit profile ID for an LLM-backed binding. It overrides the pipeline default unless the run supplies **--llm-profile**. |
-| **retries** | integer | 0 | Non-negative additional attempts for chunk, extract, merge, and normalize bindings. |
+| **structured_output_repair_attempts** | integer | pipeline or prompt-owned (1 in maintained production prompts) | Optional structural-repair limit from 0 through 3 for an LLM-backed binding. It overrides the pipeline value; explicit 0 disables structural repair. |
+| **validation_policy** | object | pipeline or application defaults | Optional field-by-field terminal-policy override for a chunk, extract, merge, or normalize binding. |
+| **retries** | integer | 0 | Non-negative additional complete producer attempts for chunk, extract, merge, and normalize bindings. This single budget covers operational errors, invalid structured output, module-requested normalization retry, and semantic correction. |
 | **options** | object | none | Must satisfy the selected module. |
 | **references** | map | none | Valid only on chunk, extract, merge, and normalize bindings. |
 | **validators** | list | production chain | Valid only on chunk, extract, merge, and normalize bindings. |
 
 Omitting **validators** uses the registered chain. **validators: []** selects
 an empty chain; a non-empty list replaces the chain in the listed order.
-Validator bindings accept only **module**, **llm_profile**, and **options**.
+Validator bindings accept only **module**, **llm_profile**,
-They reject **references**, **retries**, and nested **validators**. Deterministic
+**structured_output_repair_attempts**, **retries**, and **options**. Their
-validators reject an explicit **llm_profile**. Deterministic module bindings
+**retries** value is a non-negative additional validator-execution budget and
-also reject an explicit **llm_profile**.
+is valid only when the selected validator is LLM-backed. A validator retry
+rechecks the same immutable candidate; it never regenerates the producer.
+They reject
+**validation_policy**, **references**, and nested **validators**. Deterministic
+validators reject explicit **llm_profile** and
+**structured_output_repair_attempts**.
+Deterministic module bindings also reject those explicit fields.
+
+An LLM-backed chunk, extract, merge, or normalize producer with both a
+non-empty validator chain and positive **retries** must declare the supported
+single-response correction capability. Preparation rejects a configuration
+that could require semantic correction from a producer that cannot provide an
+exact prior response. A deterministic producer, or an LLM attempt that did
+not make a model call, cannot consume a semantic retry after rejection.
 
 The **json** output module accepts optional **include_chunk_map** and
 **evidence_context** settings:
@@ -319,7 +376,8 @@ Unknown outer or nested option fields are rejected, as are incompatible YAML
 types. The allowlist remains valid when a run uses lane filtering: a configured
 lane that is not active for that invocation simply contributes no evidence.
 Evidence publication is opt-in because it can persist source text and metadata.
-Its payload contract is [Published Evidence Context](integrations/evidence-context.md).
+When enabled, it publishes the selected source-unit excerpt defined by the
+[Published Evidence Context contract](integrations/evidence-context.md).
 
 ## References And Ordered Handoffs
 

docs/consumers/dnd-pipeline.md

@@ -0,0 +1,209 @@
+# Consuming The Complete D&D Pipeline
+
+Use this workflow when an orchestrator runs the maintained complete D&D
+pipeline and consumes its structured JSON artifacts. The generic
+[subprocess consumer guide](subprocess.md) owns process-level responsibilities;
+this guide connects that workflow to the complete D&D configuration, its
+Seriatim input, and its artifact inventory.
+
+The [CLI reference](../cli.md), [configuration reference](../config.md),
+[run-result receipt](../integrations/run-result.md), and
+[published JSON output contract](../integrations/json-output.md) remain the
+canonical definitions of those public interfaces.
+
+## Prepare And Validate The Deployment
+
+Start from the maintained
+[complete D&D configuration](../../examples/dnd-complete.config.yml). It uses
+the `dnd-session` pipeline and demonstrates every implemented D&D lane, ordered
+artifact handoffs, campaign references, chunk-map publication, and evidence
+context.
+
+A deployment must provide its own PromptKit profile and campaign reference
+files. Use absolute paths for service and subprocess deployments. In
+particular, observe these different resolution rules:
+
+- reference paths in YAML are resolved relative to the Notarius configuration
+  file; and
+- `promptkit.profile_file` is resolved relative to the Notarius process working
+  directory.
+
+Do not copy the repository example's relative profile path into a deployment
+without also controlling that working directory. The complete path and profile
+rules are defined in [Configuration](../config.md).
+
+Preflight the deployed configuration before processing sessions and whenever
+it changes:
+
+```sh
+notarius config validate \
+  --config /absolute/path/to/notarius.yml \
+  --pipeline dnd-session
+```
+
+Provide credentials through the environment or the documented configuration
+mechanism. Do not put credentials in command arguments, generated
+configuration, or logs.
+
+## Supply The Transcript
+
+The complete pipeline consumes a Seriatim JSON document. The
+[Seriatim input contract](../integrations/seriatim.md) defines its required
+metadata, segments, and validation rules. Preserve segment IDs: D&D artifact
+citations use those segment IDs as source-unit ranges.
+
+When the caller maintains several transcript tiers, use the final trimmed JSON
+transcript so extraction operates on the same session content presented to
+later consumers. For example, Narratio identifies this implemented artifact as
+`narratio.transcript.final_trimmed` and normally stores it at
+`transcripts/final.trimmed.json`.
+
+Notarius generates a stable prompt session from the resolved input module and
+the exact input bytes. An ordinary orchestrator should not pass `--session-id`.
+Use that override only when intentionally changing the routing relationship
+between invocations; it is not a credential or output identity.
+
+## Run Notarius
+
+Invoke the pipeline with explicit absolute paths and request its
+machine-readable receipt:
+
+```sh
+notarius run dnd-session \
+  --config /absolute/path/to/notarius.yml \
+  --input /absolute/path/to/transcripts/final.trimmed.json \
+  --output-dir /absolute/path/to/notarius-output \
+  --json
+```
+
+The caller should:
+
+- capture stdout and stderr separately;
+- propagate cancellation and impose an operator-appropriate timeout;
+- wait for process completion before interpreting stdout; and
+- retain stderr for diagnosis without copying secrets or transcript content
+  into other logs.
+
+Only exit status 0 permits decoding stdout as a receipt. Ignore stdout after a
+nonzero exit because a failed receipt write can leave partial bytes. The
+[CLI reference](../cli.md#output-streams-and-exit-statuses) defines the complete
+stream and exit-status contract.
+
+## Discover The Published Bundle
+
+Decode the successful stdout document as a supported run-result schema. For
+the current contract, `schema_version` is `notarius.run-result.v2`. Tolerate
+unknown fields allowed by that version, but reject an unsupported schema
+version.
+
+Use the receipt's absolute `output_directory` as the exact run-specific bundle
+root. Do not scan the output root for its newest directory, guess a run ID, or
+construct a bundle path. Resolve `index_file` beneath `output_directory` and
+reject an absolute logical path or any result that escapes the bundle root.
+
+The complete configuration uses the application validation defaults. A caller
+that requires fully validated D&D artifacts must also require receipt
+`validation_status: approved`; a successful `incomplete` result reflects the
+configured validator-failure continuation policy and carries its bounded
+validator provenance in `validation_summaries`.
+
+Read `index.json` and locate each requested lane in `output_files` by its exact
+`lane_id`. Do not guess a lane filename. Before decoding a payload:
+
+1. resolve its descriptor's relative `file` beneath the bundle root with the
+   same confinement check;
+2. verify the descriptor's media type and schema identity against the linked
+   artifact contract; and
+3. decode the payload according to that contract.
+
+The [published JSON output contract](../integrations/json-output.md) defines
+the index and bundle layout. Treat all paths obtained from a decoded external
+document as untrusted until confined to their documented root.
+
+## Complete Artifact Inventory
+
+When every configured lane is accepted, the complete example publishes these
+lane artifacts:
+
+| Lane ID | Purpose | Canonical contract |
+| --- | --- | --- |
+| `item-registry` | Canonical registry of encountered items and currency. | [Item registry](../integrations/dnd-item-registry-artifacts.md) |
+| `npc-registry` | Canonical registry of named NPCs. | [NPC registry](../integrations/dnd-npc-registry-artifacts.md) |
+| `location-registry` | Canonical registry of named locations. | [Location registry](../integrations/dnd-location-registry-artifacts.md) |
+| `scene-descriptions` | Classification, title, and summary for each scene. | [Scene descriptions](../integrations/dnd-scene-description-artifacts.md) |
+| `item-occurrences` | Source-grounded item discovery, acquisition, use, transfer, and loss events. | [Item occurrences](../integrations/dnd-item-occurrence-artifacts.md) |
+| `spells` | Source-grounded spell casts and casters. | [Spell casts](../integrations/dnd-spell-artifacts.md) |
+| `combat-turns` | Source-grounded combat turn participation. | [Combat turns](../integrations/dnd-combat-turn-artifacts.md) |
+| `npc-occurrences` | Source-grounded NPC interaction occurrences. | [NPC occurrences](../integrations/dnd-npc-occurrence-artifacts.md) |
+| `location-occurrences` | Source-grounded location occurrences. | [Location occurrences](../integrations/dnd-location-occurrence-artifacts.md) |
+| `enemy-events` | Source-grounded enemy combat events. | [Enemy events](../integrations/dnd-enemy-event-artifacts.md) |
+
+The JSON encoder always publishes these bundle-management files:
+
+| File | Purpose |
+| --- | --- |
+| `index.json` | Discovery document for lane and pipeline-wide artifacts. |
+| `manifest.json` | Run provenance and result summaries. |
+| `rejected.json` | Rejected pipeline outputs. |
+| `warnings.json` | Actionable process-degradation warnings. |
+| `diagnostics.json` | Advisory and observation findings for accepted artifacts. |
+
+The complete configuration also requests two pipeline-wide artifacts:
+
+- [`chunk-map.json`](../integrations/chunk-map.md), the accepted chunk plan and
+  chunk metadata; and
+- [`evidence-context.json`](../integrations/evidence-context.md), a reading
+  excerpt containing the union of selected cited source units and the
+  configured surrounding window.
+
+Discover both from their top-level `index.json` descriptors rather than
+treating them as lanes. Evidence context is convenient reading material, not
+authoritative provenance; citations in the normalized lane payloads remain the
+evidence contract.
+
+Every optional or lane file is published only when its corresponding artifact
+is available. A successful process does not guarantee that all configured
+lanes were accepted.
+
+## Decide What Counts As Consumer Success
+
+Exit status 0 means Notarius completed the pipeline and published its result
+bundle. The receipt or bundle may still report warnings, rejected outputs, or
+missing lane descriptors. A downstream consumer must define its own required
+artifact set explicitly.
+
+A caller that claims to consume the complete D&D workflow should normally
+require all ten lane IDs in the table and verify each descriptor's expected
+contract. If any required lane is missing, rejected, or incompatible, fail the
+caller's extraction step while retaining the Notarius bundle for diagnosis. A
+consumer that needs only a subset may define and document a narrower policy.
+
+Keep the successful receipt with the complete published bundle. Retain
+`manifest.json`, `rejected.json`, `warnings.json`, and captured process logs as
+required by the caller's provenance, diagnosis, and retention policies. Avoid
+selectively copying payload files without also preserving enough index and
+manifest information to identify their originating run and contracts.
+
+The transcript, lane artifacts, evidence context, manifest, debug data, and
+logs can all contain private campaign information. Apply the same access,
+publication, and retention controls used for the source transcript.
+
+## Consumer Checklist
+
+- Validate the deployed Notarius configuration and `dnd-session` pipeline.
+- Pass the final trimmed Seriatim JSON transcript with stable segment IDs.
+- Use absolute configuration, input, output-root, profile, and reference paths
+  in service deployments.
+- Capture stdout and stderr separately and enforce cancellation and timeout.
+- Parse stdout only after exit status 0.
+- Accept only supported receipt, index, and artifact schema versions while
+  tolerating permitted unknown fields.
+- Use the receipt's `output_directory`; never guess the run directory.
+- Confine `index_file` and every descriptor path to the published bundle root.
+- Discover lanes by `lane_id` and verify descriptor compatibility before
+  decoding payloads.
+- Enforce an explicit required-lane policy and inspect rejections and warnings.
+- Preserve the receipt and sufficient bundle provenance for every retained
+  artifact.
+- Protect all transcript-derived files and diagnostic streams as sensitive
+  campaign data.

docs/consumers/subprocess.md

@@ -6,6 +6,10 @@ statuses, while the [run-result receipt](../integrations/run-result.md) and
 [Published JSON Output contract](../integrations/json-output.md) own the
 durable result formats.
 
+For the maintained complete D&D workflow, including its transcript input,
+configured lane inventory, and downstream acceptance checklist, see
+[Consuming The Complete D&D Pipeline](dnd-pipeline.md).
+
 ## Run And Check The Process
 
 Optionally preflight a selected configuration and pipeline before work starts:
@@ -55,23 +59,30 @@ contract. The JSON bundle contract links to the available lane contracts.
 If `index.json` has an `evidence_context` descriptor, treat it as a
 pipeline-wide artifact rather than a lane entry. Verify its six descriptor
 fields before decoding the linked file according to the [Published Evidence
-Context contract](../integrations/evidence-context.md). Use each
+Context contract](../integrations/evidence-context.md). Decode its top-level
-`evidence_refs` entry as the citation to source material. Its surrounding
+source-unit array as a reading excerpt. Obtain authoritative citations and lane
-context range and included units explain the citation, but do not widen or
+provenance from the normalized lane artifacts; the excerpt has neither and its
-replace the cited source reference.
+nearby units do not widen a lane artifact's cited source reference.
 
 A zero exit status may still report rejected outputs, warnings, or absent
 lanes. The caller decides which lane IDs are required for its own work and
 which are optional; it should make that decision explicitly rather than infer
 failure from the receipt counts alone.
 
+When complete validation is required, also require receipt
+`validation_status: approved` and inspect `validation_summaries`. A successful
+run with `validation_status: incomplete` contains a structurally valid result
+that advanced after validator execution could not complete under the configured
+`warn_continue` policy. It is not reusable checkpoint state and should not be
+silently treated as fully reviewed by the caller.
+
 ## Preserve Provenance And Handle Data Carefully
 
 Keep the receipt with the published `manifest.json`, and retain
-`rejected.json` and `warnings.json` when review or later provenance requires
+`rejected.json`, `warnings.json`, and `diagnostics.json` when review or later provenance requires
 them. Treat the input, output bundle, cache, debug bundle, and captured process
 logs as potentially sensitive data. Apply the caller's access controls and
 retention policy, and avoid copying secrets into arguments, logs, or
 provenance records. An evidence-context artifact contains source-unit text and
-metadata, and selected lanes can cover most of an input; preserve and share it
+metadata and can cover most of an input; preserve and share it only when that
-only when that source content is authorized for the recipient.
+source content is authorized for the recipient.

docs/development.md

@@ -18,13 +18,14 @@ implemented component map.
 | Any documentation addition or revision | [Documentation Policy](policy/documentation.md) | It defines canonical homes, audiences, current-behavior rules, and maintenance requirements. |
 | Adding, changing, reviewing, or deleting tests | [Testing Policy](policy/testing.md) | It defines risk-based sufficiency, durable test boundaries, test-double guidance, and criteria for retaining tests. |
 | CLI composition or command behavior | [CLI Internals](internal/cli.md) and [CLI Reference](cli.md) | The internal guide owns composition and command flow; the reference owns public syntax. |
-| Building a subprocess caller or changing its result protocol | [Subprocess Consumer Guide](consumers/subprocess.md), [Run Result Receipt](integrations/run-result.md), and [CLI Internals](internal/cli.md) | These separate caller workflow, durable receipt contract, and CLI implementation behavior. |
+| Building a subprocess caller or changing its result protocol | [Subprocess Consumer Guide](consumers/subprocess.md), [Complete D&D Consumer Guide](consumers/dnd-pipeline.md), [Run Result Receipt](integrations/run-result.md), and [CLI Internals](internal/cli.md) | These separate generic caller workflow, the complete D&D workflow, the durable receipt contract, and CLI implementation behavior. |
 | Configuration loading, resolution, or user-visible configuration behavior | [Configuration Internals](internal/configuration.md) and [Configuration](config.md) | The internal guide owns loading and resolution mechanics; the reference owns the configuration contract. |
 | Pipeline resolution or execution | [Pipeline Internals](internal/pipeline.md) | It documents profiles, references, validation, retries, checkpoints, and runner behavior. |
 | Production modules or validators | [Module Internals](internal/modules.md), [D&D Module Internals](internal/dnd.md), and [D&D integration contracts](integrations/) | The generic guide owns extension mechanics, the D&D guide owns shared family conventions, and the contracts own durable output shapes. |
 | LLM clients, prompts, schemas, profiles, or scheduling | [LLM Runtime](internal/llm.md) | It documents the transport boundary and PromptKit integration. |
 | Output, cache, resume, or debug artifacts | [Run State Internals](internal/state.md), [Operations](operations.md), and [Configuration](config.md) | These separate implementation details, operator behavior, and configuration contracts. |
 | External input formats, artifact schemas, or durable output files | [Integration Contracts](integrations/) | Integration documents define external and durable data contracts. |
+| Release preparation, tagging, publication, or verification | [Source Releases](release.md) and [Documentation Policy](policy/documentation.md) | The release procedure owns maintainer guards and immutable-tag recovery; the policy assigns release-note ownership. |
 | Proposed or unimplemented behavior | [Roadmap](roadmap/) | Future work belongs only in roadmap documentation until implemented. |
 
 For an existing subsystem, also inspect its focused tests and the package-local

docs/integrations/dnd-spell-catalog-overlays.md

@@ -67,6 +67,12 @@ including a collision with the embedded catalog. Matching uses the catalog’s
 case, whitespace, and apostrophe normalization, so authors should avoid names
 or aliases that normalize to another spell.
 
+Spell extraction receives the effective catalog as deterministic canonical-name
+and alias pairs. An alias in the transcript selects its associated canonical
+name; the extractor is instructed to return that canonical spelling. The
+projection contains no catalog source metadata or provenance, and aliases
+remain recognition context rather than transcript evidence.
+
 The overlay is a recognition aid only. The durable spell-artifact schema and
 source-evidence rules are defined by the
 [D&D spell artifact contract](dnd-spell-artifacts.md).

docs/integrations/evidence-context.md

@@ -1,9 +1,11 @@
 # Published Evidence Context
 
 This contract defines the optional `source/evidence-context` artifact emitted
-by the production JSON output. Its configuration is owned by
+by the production JSON output. It is a selected source-unit excerpt for
-[Configuration](../config.md#module-bindings-and-validators); its logical-file
+convenient reading alongside normalized lane artifacts; it is not a second
-discovery is owned by [Published JSON Output](json-output.md).
+citation or provenance model. Its configuration is owned by
+[Configuration](../config.md#module-bindings-and-validators), and its
+logical-file discovery is owned by [Published JSON Output](json-output.md).
 
 ## Identity And Discovery
 
@@ -26,35 +28,13 @@ its absence means evidence publication was not enabled for that bundle.
 
 ## Payload
 
-The v1 payload is a JSON object with required `source_id`, `source_digest`,
+The v1 payload is a top-level JSON array of generic source units. There is no
-`window_units`, `selected_lanes`, and `contexts` fields. `selected_lanes` and
+wrapper, source-level metadata, context grouping, lane identifier, or evidence
-`contexts` are always arrays; an enabled configuration with no accepted direct
+reference in the payload. An enabled configuration with no contributing
-evidence publishes `contexts: []`.
+accepted evidence publishes `[]`.
 
 ```json
-{
+[
-  "source_id": "session-alpha",
-  "source_digest": "sha256:0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef",
-  "window_units": 1,
-  "selected_lanes": ["npc_registry", "spells"],
-  "contexts": [
-    {
-      "context_ref": {
-        "source_id": "session-alpha",
-        "start_unit_id": 10,
-        "end_unit_id": 20
-      },
-      "evidence_refs": [
-        {
-          "lane_id": "spells",
-          "source_ref": {
-            "source_id": "session-alpha",
-            "start_unit_id": 10,
-            "end_unit_id": 10
-          }
-        }
-      ],
-      "units": [
   {
     "id": 10,
     "kind": "transcript_segment",
@@ -76,41 +56,52 @@ evidence publishes `contexts: []`.
     }
   }
 ]
-    }
-  ]
-}
 ```
 
-Each context requires a `context_ref` object and `evidence_refs` and `units`
+Each source unit has required `id`, `kind`, `text`, and self `ref` fields.
-arrays. `context_ref` identifies the first and last included unit. Each
+`ref` contains `source_id`, `start_unit_id`, and `end_unit_id`, and both unit
-evidence entry contains a selected `lane_id` and an original `source_ref`. A
+endpoints identify that unit's `id`. A unit may also contain source-owned
-unit uses the existing source-unit shape: required `id`, `kind`, `text`, and
+`metadata`, an open-ended JSON object. Fixed unit and reference fields are
-self `ref`, plus optional JSON-object `metadata`. Fixed payload objects reject
+strict: consumers must reject unknown fixed fields, malformed units, invalid
-unknown fields; unit metadata may contain application-defined JSON values.
+self-references, units whose `source_id` differs from other units in the same
+excerpt, and a payload that is not the array described here.
 
-## Citations And Context
+The excerpt preserves each selected unit exactly as represented by the
+validated generic source document. It does not add evidence-context-specific
+annotations or reshape source-owned metadata.
 
-`evidence_refs` are the authoritative citations. They identify the direct
+## Selection And Citations
-references emitted by accepted normalized artifacts. `context_ref` and the
-units collection include those cited units plus nearby source units selected by
-the configured window. They are explanatory context, not widened citations.
 
-Only accepted outputs from the configured lane allowlist contribute. Rejected,
+The framework obtains direct source references only through typed evidence
-failed, absent, and lane-filtered outputs do not contribute. The artifact never
+projections of accepted normalized artifacts in the configured lane allowlist.
-contains raw input bytes, prompts, model responses, auxiliary reference
+It validates each reference against the current source document, expands its
-content, credentials, or filesystem paths.
+range by `window_units` source-unit positions on each side, clamps at document
+boundaries, and takes the union of all expanded ranges. The output contains
+each selected source unit once in source-document position order, regardless
+of numeric unit IDs. Repeated references, overlapping windows, and citations
+from multiple lanes do not duplicate a unit. Rejected, failed, absent,
+inactive, and unselected lanes contribute nothing.
 
-## Ordering And Compatibility
+Normalized lane artifacts remain authoritative for citations and for which lane
+cited a range. The excerpt has no lane attribution and must not be used to
+reconstruct it. Its included nearby units provide reading context only; they
+do not widen any citation in a lane artifact.
 
-The selected lane allowlist is lexical. Contexts and units are in source
+The excerpt contains at most every generic source unit once. It can therefore
-document position order, not numeric unit-ID order. Direct evidence entries
+equal the complete generic source document when coverage is broad or the
-are deterministically ordered by lane and source reference. Overlapping or
+window is large. No byte-, token-, or compression-size guarantee is made, and
-contiguous windows merge, and each source unit appears at most once in the
+the framework does not truncate the excerpt to meet an arbitrary size limit.
-resulting contexts.
+
+## Consumer Responsibilities And Data Handling
 
 The artifact is additive to the JSON bundle and is not a lane payload,
 normalized-output count, checkpoint, or generated reference. Consumers that
-do not need it must tolerate the absent optional descriptor. Consumers that do
+do not need it must tolerate an absent descriptor. Consumers that do use it
-use it should preserve the artifact and its schema identity with the run
+should validate the descriptor and payload before use, retain the artifact with
-provenance, and should treat its source text and metadata as sensitive durable
+its schema identity when needed for a run record, and read citations from the
-content.
+corresponding normalized lane artifacts.
+
+The excerpt contains source-unit text and source-owned metadata and is durable
+output. Treat it as sensitive source content, apply appropriate access controls
+and retention, and do not assume its selected form is materially smaller or
+less sensitive than the original input.

docs/integrations/json-output.md

@@ -9,7 +9,7 @@ Output configuration, including chunk-map and evidence-context publication, belo
 ## Bundle Layout
 
 All paths below are logical, relative, slash-separated bundle paths. The
-encoder always emits the first four JSON files below and adds lane or
+encoder always emits the first five JSON files below and adds lane or
 pipeline-wide artifact files when their corresponding artifacts are available:
 
 A subprocess caller first obtains the physical bundle root from the
@@ -21,10 +21,11 @@ root for the logical discovery described here.
 | `index.json` | Entry point that names the other published files and lane payloads. |
 | `manifest.json` | Run provenance and result summaries. |
 | `rejected.json` | Rejected pipeline outputs. |
-| `warnings.json` | Accepted-output and run warnings. |
+| `warnings.json` | Actionable process-degradation warnings. |
+| `diagnostics.json` | Accepted-artifact quality advisories and normalization observations. |
 | `lanes/<safe-lane-id>.json` | One normalized artifact payload for each lane. |
 | `chunk-map.json` | Optional accepted chunk map, when its export is enabled and available. |
-| `evidence-context.json` | Optional source-context artifact, when evidence publication is enabled. |
+| `evidence-context.json` | Optional selected source-unit excerpt, when evidence publication is enabled. |
 
 JSON files are pretty-printed with a trailing newline. Lane payloads are
 accepted only when their media type is `application/json`.
@@ -39,7 +40,8 @@ normalized lanes has this valid minimal index:
   "manifest_file": "manifest.json",
   "output_files": [],
   "rejected_file": "rejected.json",
-  "warnings_file": "warnings.json"
+  "warnings_file": "warnings.json",
+  "diagnostics_file": "diagnostics.json"
 }
 ```
 
@@ -49,6 +51,7 @@ normalized lanes has this valid minimal index:
 | `output_files` | Yes | Lane descriptors sorted by `lane_id`. |
 | `rejected_file` | Yes | Always `rejected.json`. |
 | `warnings_file` | Yes | Always `warnings.json`. |
+| `diagnostics_file` | Yes | Always `diagnostics.json`. |
 | `chunk_map` | No | Descriptor for the pipeline-wide `chunk-map.json`; never a lane descriptor. |
 | `evidence_context` | No | Descriptor for the pipeline-wide `evidence-context.json`; never a lane descriptor. |
 
@@ -92,7 +95,7 @@ group into the following externally observable summaries:
 | Run identity and result | `run_id`, `pipeline_id`, `pipeline_digest`, `schema_version`, `validation_status`, `started_at`, `completed_at` |
 | Resolved components | `input_module`, `chunker`, `extractors`, `merger`, `normalizer`, `output_encoder`, `artifact_lanes`, `validator_chains`, `module_metadata` |
 | Source and references | `source_digests`, `references` |
-| Published result summaries | `normalized_outputs`, `rejected_outputs` |
+| Published result summaries | `normalized_outputs`, `rejected_outputs`, `validation_summaries` |
 | Execution summaries | `chunk_plan`, `checkpoint_decisions`, `llm_profiles`, `metadata` |
 
 `references` records provenance such as the target, slot, origin, digest,
@@ -102,6 +105,16 @@ summarize results without embedding lane payload bytes. A chunk-plan summary is
 provenance for the plan used by this run; cache records, debug artifacts, and
 other operational state are not published as bundle files.
 
+Each `validation_summaries` entry is a bounded outcome for one producer result.
+It has required `status`, `producer_attempt_count`, and `terminal_action`;
+the stage and affected step, lane, module, or chunk identity are present when
+applicable. `status` is `complete`, `rejected`, or `incomplete`.
+`rejecting_validators`, `reason_codes`, and `incomplete_validators` preserve
+configured validator order and omit later duplicates. Entries contain no raw
+candidate response, correction guidance, validator diagnostic message, or
+artifact payload. The same shape may appear as `validation` on an affected
+rejection entry.
+
 When present, `metadata.session_id` is the effective non-secret routing
 correlation identifier used for the run. It can be visible to providers and is
 not a substitute for a cache or checkpoint identity. Its generation and
@@ -122,16 +135,57 @@ These values describe observed execution; they are not a backend-registration
 interface. Entries that differ by backend or effective reasoning remain
 distinct even when their profile, provider, and model are otherwise equal.
 
-## Rejections And Warnings
+## Rejections, Warnings, And Diagnostics
 
 `rejected.json` is always an object with a `rejected` array. Each entry has
 required `stage` and `message`; `step_id`, `lane_id`, `module_key`, `chunk_id`,
 `chunk_index`, `validator_name`, `reason_code`, `attempt_count`, and
-`diagnostic_artifact_path` are present only when applicable.
+`diagnostic_artifact_path` are present only when applicable. An entry may also
+contain the bounded `validation` summary described above; the existing singular
+validator and reason fields remain the first configured rejection for
+compatibility.
 
-`warnings.json` is always an object with a `warnings` array. Each warning has
+`warnings.json` is always the `notarius.warnings.v2` envelope:
-`reason_code` and `message`; `scope` is optional. Both arrays are empty when
+
-there is nothing to report.
+```json
+{
+  "schema_version": "notarius.warnings.v2",
+  "group_count": 0,
+  "occurrence_count": 0,
+  "groups": []
+}
+```
+
+It contains only process warnings. `group_count` is exact, and
+`occurrence_count` is the exact sum of its group occurrence counts.
+
+`diagnostics.json` is always the `notarius.diagnostics.v1` envelope:
+
+```json
+{
+  "schema_version": "notarius.diagnostics.v1",
+  "group_count": 0,
+  "occurrence_count": 0,
+  "truncated": false,
+  "unrepresented_occurrence_count": 0,
+  "groups": []
+}
+```
+
+It contains only advisory and observation groups. `group_count` counts groups
+represented in `groups`; `occurrence_count` includes both represented and
+unrepresented occurrences. When `truncated` is true,
+`unrepresented_occurrence_count` is the exact number omitted from group
+representation.
+
+Each group has `disposition`, `category`, `reason_code`, framework-owned
+`origin`, exact `occurrence_count`, bounded `samples`, and
+`omitted_sample_count`. Samples carry safe `scope` and `message`, plus a chunk
+ID and zero-based chunk index when applicable. A group retains at most three
+distinct samples. The framework fails rather than truncating actionable
+warnings beyond 128 groups; it represents at most 256 advisory/observation
+groups and records further occurrences through the diagnostic truncation
+fields above.
 
 ## Compatibility
 

docs/integrations/pkg-promptkit.md

@@ -1,11 +1,11 @@
 # PromptKit Integration
 
 Notarius pins
-[`gitea.maximumdirect.net/eric/promptkit` v0.5.0](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.5.0)
+[`gitea.maximumdirect.net/eric/promptkit` v0.9.0](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.9.0)
 as its in-process prompt engine. The upstream
-[Go package consumer guide](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.5.0/docs/consumers/pkg-promptkit.md)
+[Go package consumer guide](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.9.0/docs/consumers/pkg-promptkit.md)
 owns the public engine API, and the upstream
-[format reference](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.5.0/docs/formats.md)
+[format reference](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.9.0/docs/formats.md)
 owns prompt, profile, and schema file contracts.
 
 ## Supported Boundary
@@ -15,7 +15,8 @@ Notarius relies on the root `promptkit` package to:
 - construct an `Engine` with filesystem-backed prompt, schema, and optional
   operator and application-fallback profile sources;
 - prepare one frozen execution from a `RunRequest` with named inline artifacts,
-  variables, a direct session ID, prompt identity, and profile selection, then
+  variables, a direct session ID, prompt identity, profile selection, and
+  optional appended rendered messages, then
   record credential-redacted details and run that exact execution;
 - return rendered debug material, validated structured output, selected
   profile, backend, effective model metadata, and token usage;
@@ -26,7 +27,7 @@ Notarius relies on the root `promptkit` package to:
   admission exhaustion through `ErrCapacityExceeded`.
 
 The pinned
-[`BackendLocal`, `LocalBackend`, and `WithBackend` API](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.5.0/backends.go)
+[`BackendLocal`, `LocalBackend`, and `WithBackend` API](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.9.0/backends.go)
 owns the registration and backend-capacity contract.
 
 For one completion, the adapter calls `PrepareExecution`, takes a
@@ -52,7 +53,7 @@ Notarius sends one stable effective session through PromptKit's direct session
 field, which is authoritative for provider session behavior. It also retains
 the same value as the `session_id` prompt variable for maintained prompt
 compatibility. The generated identifier is 76 ASCII characters, within
-PromptKit v0.5.0's 256-code-point session limit. Session IDs are non-secret
+PromptKit v0.9.0's 256-code-point session limit. Session IDs are non-secret
 correlation identifiers and may be exposed to providers and provider
 observability. The CLI contract owns generation and override behavior.
 
@@ -84,7 +85,39 @@ configuration and deployment workflow are defined in
 [Configuration](../config.md#promptkit-profiles) and
 [Operations](../operations.md#promptkit-profile-deployment).
 
-Notarius supports this boundary against PromptKit v0.5.0. Its fallback source,
+PromptKit owns `base_profile` resolution under its
+[pinned format rules](https://gitea.maximumdirect.net/eric/promptkit/src/tag/v0.9.0/docs/formats.md).
+Notarius records the selected leaf identity and resolved target without parsing
+or merging inheritance. An unset filesystem `api_key_env` is optional and may
+reach the provider without authorization, which can result in a 401 or 403.
+
+PromptKit v0.9.0 accepts only the `developer`, `system`, `user`, and
+`assistant` text-chat roles after normalizing case and surrounding whitespace.
+Maintained Notarius prompt definitions use only `system` and `user`.
+
+For application-owned semantic correction, Notarius uses PromptKit v0.9.0's
+`RunRequest.AppendedMessages` after the ordinary rendered prompt. It supplies
+exactly two messages in order: the latest validated producer response with
+role `assistant`, then deterministic validation guidance with role `user`.
+It never exposes a general caller-selected role API, accumulates earlier
+correction turns, or changes the ordinary prompt prefix. Ordinary requests
+leave appended messages unset.
+
+PromptKit preserves supplied content but does not own Notarius's correction
+bounds. Notarius rejects invalid UTF-8, blank, or oversized assistant material
+(at most 1 MiB), guidance (at most 64 KiB), and combined content (at most
+1,114,112 bytes) before preparing the request. The transport-neutral
+application contract owns defensive copying and these limits. Default request
+and terminal summaries retain only safe counts, digests, identities, and usage;
+complete appended messages remain limited to the explicitly requested detailed
+debug trace.
+
+PromptKit now obtains its maintained OpenRouter and Rakestrawhome backend and
+profile catalogs from independently versioned transitive modules. Notarius
+does not import or register either catalog; PromptKit retains catalog source,
+identity, precedence, credential, and capacity ownership.
+
+Notarius supports this boundary against PromptKit v0.9.0. Its fallback source,
 prepared-execution, inspection, and typed capacity APIs are used as public
 upstream contracts; other PromptKit APIs or file-format behavior are not
 implicitly supported. A dependency upgrade requires reviewing the adapter,
@@ -97,6 +130,10 @@ pinned upstream documentation.
 module assets, maps its transport-neutral completion contract, prepares and
 executes requests, validates output, records provenance, captures debug
 material, redacts errors, and preserves timeout ownership.
+
+PromptKit provider error details do not cross the ordinary completion boundary.
+Notarius exposes a provider-neutral generation category and optional status;
+redacted provider details are retained only in requested debug material.
 [D&D Module Internals](../internal/dnd.md) owns the embedded
 `dnd-extraction` fallback profile and the maintained D&D prompt defaults.
 [Configuration](../config.md#promptkit-profiles) defines how a Notarius
@@ -105,4 +142,7 @@ the conventional local backend.
 
 PromptKit API or format changes outside this boundary are not implicitly
 supported. Updating the pinned version requires reviewing the adapter and
-profile/configuration contracts against the upstream documentation.
+profile/configuration contracts against the upstream documentation. Maintained
+production prompts use PromptKit's bounded structural-repair contract; their
+current declaration is one additional repair attempt. Notarius retains the
+transport-neutral boundary and does not expose PromptKit types to modules.

docs/integrations/run-result.md

@@ -9,36 +9,58 @@ Command syntax, streams, and exit statuses are defined in the
 
 ## Schema
 
-The current schema version is `notarius.run-result.v1`.
+The current schema version is `notarius.run-result.v2`.
 
 | Field | Required | Meaning |
 | --- | --- | --- |
-| `schema_version` | Yes | Exactly `notarius.run-result.v1`. |
+| `schema_version` | Yes | Exactly `notarius.run-result.v2`. |
 | `run_id` | Yes | The finalized Notarius run identifier. |
 | `pipeline_id` | Yes | The effective pipeline identifier. |
 | `output_directory` | Yes | Absolute path to the published, run-specific output bundle. |
 | `index_file` | For the production JSON output | Logical path `index.json`; omitted for other output modules. |
 | `normalized_output_count` | Yes | Number of final normalized outputs. |
 | `rejected_output_count` | Yes | Number of recorded rejected outputs. |
-| `warning_count` | Yes | Number of final run warnings. |
+| `warning_group_count` | Yes | Exact number of actionable warning groups. |
+| `warning_occurrence_count` | Yes | Exact occurrences represented by actionable warning groups. |
+| `diagnostic_group_count` | Yes | Number of represented advisory and observation groups. |
+| `diagnostic_occurrence_count` | Yes | Advisory and observation occurrences, including unrepresented occurrences. |
+| `diagnostics_truncated` | Yes | Whether advisory/observation group representation was truncated. |
 | `validation_status` | Yes | The final run manifest validation status. |
+| `validation_summaries` | No | Bounded per-producer validation outcomes; present when producer work ran. |
 | `debug_directory` | No | Absolute path to the run-specific debug bundle when requested debug capture completed. |
 
 For the production `json` output module, `index_file` is present only when the
 completed run returned exactly one logical output file named `index.json`.
 For another output module, its absence does not indicate a failed run.
+`validation_status` is `approved`, `rejected`, or `incomplete`; `incomplete`
+means one or more otherwise accepted results advanced under validator-failure
+`warn_continue`.
 
 ```json
 {
-  "schema_version": "notarius.run-result.v1",
+  "schema_version": "notarius.run-result.v2",
   "run_id": "run-1770000000000000000-0123456789abcdef0123456789abcdef",
   "pipeline_id": "dnd-session",
   "output_directory": "/work/results/run-1770000000000000000-0123456789abcdef0123456789abcdef",
   "index_file": "index.json",
   "normalized_output_count": 6,
   "rejected_output_count": 2,
-  "warning_count": 1,
+  "warning_group_count": 1,
-  "validation_status": "rejected"
+  "warning_occurrence_count": 2,
+  "diagnostic_group_count": 3,
+  "diagnostic_occurrence_count": 5,
+  "diagnostics_truncated": false,
+  "validation_status": "incomplete",
+  "validation_summaries": [
+    {
+      "stage": "extract",
+      "lane_id": "spells",
+      "status": "incomplete",
+      "incomplete_validators": ["dnd/spells/source_refs"],
+      "producer_attempt_count": 1,
+      "terminal_action": "warn_continue"
+    }
+  ]
 }
 ```
 
@@ -48,8 +70,11 @@ For another output module, its absence does not indicate a failed run.
 paths. They identify the paths used by Notarius and do not resolve symlinks.
 `output_directory` is the run-specific bundle, not the configured output root.
 
-The receipt is a summary and discovery document. It does not contain lane
+The receipt is a summary and discovery document. Its optional validation
-descriptors, payloads, manifest data, rejections, warnings, or file contents.
+summaries contain only stable status, identity, validator names, reason codes,
+attempt counts, and terminal actions. It does not contain lane descriptors,
+payloads, manifest payloads, rejection messages, warnings, raw model responses,
+correction guidance, or file contents.
 For the production JSON output, resolve `index_file` beneath
 `output_directory`, reject path escapes, and use the
 [Published JSON Output contract](json-output.md) to discover logical files and

docs/internal/cli.md

@@ -24,10 +24,14 @@ preparation, and runner mechanics after their inputs are supplied.
 
 ## Dispatch And Configuration Handoff
 
-The root dispatcher handles help, configuration validation, pipeline listing,
+The root dispatcher handles help, version reporting, configuration validation,
-and a pipeline run. It normalizes injectable options before dispatch so that a
+pipeline listing, and a pipeline run. Version reporting resolves build
-missing production dependency fails as a command error rather than reaching
+information through `internal/buildinfo` before production composition, so the
-execution.
+diagnostic remains available without configuration or runtime collaborators.
+The public syntax, streams, exit classes, and version semantics are defined by
+the [CLI reference](../cli.md). Other root commands normalize injectable
+options before dispatch so that a missing production dependency fails as a
+command error rather than reaching execution.
 
 Commands that need configuration use one shared loader. The CLI discovers the
 file, parses it through **internal/core/config**, starts from defaults, applies

docs/internal/configuration.md

@@ -74,10 +74,19 @@ profile-free, and no second inheritance decision occurs during execution. The
 public field definitions and precedence are owned by
 [Configuration](../config.md#pipelines).
 
+The resolver retains configured `validation_policy` overrides and derives one
+detached concrete terminal policy for the chunk producer and every lane's
+extract, merge, and normalize producers. That field-by-field inheritance is
+complete before preparation, and the effective values contribute to pipeline
+and checkpoint identity; execution does not interpret configuration defaults.
+
 The framework resolver supplies defaults, selects lanes, resolves validator
 chains, checks registered module and artifact compatibility, validates module
-options, and returns the fixed ordered pipeline shape. The resulting
+options, and returns the fixed ordered pipeline shape. Positive validator retry
-**EffectiveConfig** retains the selected ID, requested selection and reference
+budgets require an LLM-backed selected validator; deterministic validators are
+rejected during resolution. Eligible LLM-backed producer specifications also
+contribute their declared correction protocol to the resolved metadata. The
+resulting **EffectiveConfig** retains the selected ID, requested selection and reference
 changes, a clone of the input configuration, and the resolved pipeline.
 Callers may therefore retain or modify their input slices and maps without
 changing the resolved result, and later consumers cannot mutate the original
@@ -94,8 +103,8 @@ runtime error class described in the [CLI reference](../cli.md#output-streams-an
 
 The framework assigns the resolved pipeline a deterministic SHA-256 digest
 after defaults, lane selection, module bindings, reference bindings, validator
-chains, effective LLM profiles, and artifact schema identity have been
+chains, selected correction protocols, effective LLM profiles, and artifact
-resolved. The digest excludes
+schema identity have been resolved. The digest excludes
 its own stored value. It identifies resolved composition rather than raw YAML
 bytes, a debug payload, or all runtime state. The CLI records it as invocation
 provenance before execution; cache and checkpoint identity have additional

docs/internal/dnd.md

@@ -33,7 +33,9 @@ typed builder. Scene chunking, every extractor, and NPC, location, and item-regi
 normalization are registered as `llm_backed`; the remaining current D&D mergers
 and normalizers are `deterministic`. The metadata is available to catalog inspection and
 resolved-pipeline debug data and determines which selected bindings inherit the
-pipeline profile. Configuration remains the canonical owner of the exact keys,
+pipeline profile. The registry normalizers use `single_response_v1`, forwarding
+corrections to their reconciliation completion and retaining the accepted raw
+proposal only as an owned model candidate. Configuration remains the canonical owner of the exact keys,
 profile precedence, and validator order.
 
 Private structured-LLM response schemas are deliberately minimal. They reject
@@ -42,6 +44,13 @@ unknown fields, while preserving semantic candidates for deterministic
 validation. Do not promote a private response envelope into a durable schema;
 the contracts above define durable data.
 
+A D&D producer that declares `single_response_v1` forwards any supplied
+semantic correction to its structured completion and returns an owned copy of
+that completion's exact validated raw response as its model candidate. It does
+not serialize normalized artifacts to create that candidate, so deterministic
+identity, evidence, warning, and durable-schema behavior remains separate from
+the model transport material.
+
 ## Prompt Construction
 
 D&D LLM-facing content lives beneath `assets/dnd/`. Each module contributes a
@@ -84,13 +93,15 @@ replace it with a complete profile of the same ID from the configured PromptKit
 source. Deployment profile selection is documented in
 [Configuration](../config.md#promptkit-profiles).
 
-The transcript assets have distinct consumers. Scene chunking consumes the
+The D&D transcript assets have distinct consumers. Scene chunking consumes the
-complete-session `common-dnd-transcript-full.md`; extraction prompts consume
+complete-session `common-dnd-transcript-full.md`, while extraction prompts
-the current-chunk `common-dnd-transcript-chunk.md`; and NPC, location, and item
+consume the current-chunk `common-dnd-transcript-chunk.md`. NPC, location, and
-normalization consume `common-dnd-transcript-windows.md` alongside their
+item normalization instead mount the generic semantic-reconciliation
-candidate collections. Player, party, glossary, and compatible campaign
+candidate and transcript-window presentation assets. Player, party, glossary,
-references provide disambiguating context, not evidence. Reference material is
+and compatible campaign references provide disambiguating context only when
-canonically ordered before rendering so equivalent inputs remain stable.
+declared by the active prompt; they never establish evidence. Reference
+material is canonically ordered before rendering so equivalent inputs remain
+stable.
 
 Extraction prompts render the common system and identity messages first, then
 cached campaign references and the cached chunk transcript. Evidence policy and
@@ -100,10 +111,11 @@ reusable extraction prefix identical while preserving the lane-specific suffix.
 
 Scene chunking intentionally uses a different order: system, cached campaign
 references, uncached module instructions, then the final ephemeral full
-transcript. Entity normalization also has its own order: system, uncached
+transcript. Entity normalization also has its own order: D&D system, mandatory
-module instructions, ephemeral reconciliation policy, uncached candidates, and
+generic protocol, ephemeral domain semantic instructions, generic candidate
-final ephemeral transcript windows. These orders and cache controls are prompt
+presentation, and final ephemeral generic transcript windows. These orders and
-behavior; change them only through the owning manifest and prompt declaration.
+cache controls are prompt behavior; change them only through the owning
+manifest and prompt declaration.
 
 ## Evidence, Candidates, And Normalization
 
@@ -116,11 +128,19 @@ result.
 
 Default chains keep responsibilities separate: structural validators assess the
 candidate, source-reference validators resolve cited ranges against the current
-source, durable-schema validation checks an approved representation, and
+source and require extraction evidence to stay within the current chunk,
+durable-schema validation checks an approved representation, and
 relatedness validators report advisory evidence concerns. The configured order
 is documented in
 [Configuration](../config.md#production-validator-keys-and-default-chains).
 
+Every D&D rejection describes the correction in transcript-grounded domain
+terms, using contextual names, artifact fields, and source segment ranges when
+useful. The guidance must not ask the model to reproduce durable entity IDs,
+hashes, validator module keys, or reason codes. Those identifiers remain in
+ordinary validation provenance; only the actionable semantic guidance is
+eligible for the correction prompt.
+
 Enemy-event extraction additionally rejects a second `engaged` observation for
 the same comparison identity within one scene-scoped result. Normalization may
 combine results from distinct scenes, so it intentionally does not apply that
@@ -129,15 +149,49 @@ rule. Configuration owns the exact validator key and chain position.
 Normalizers are deterministic for spells, combat turns, item occurrences, NPC
 occurrences, scene descriptions, enemy events, and location occurrences. They
 canonicalize display values and evidence, use source-document order for stable
-output, and issue bounded warnings for changes or collapsed duplicates. NPC,
+output, and emit bounded normalization observations for changes or collapsed duplicates. NPC,
 item, and location registry normalizers are intentional exceptions: each first
 produces a deterministic candidate set, then may use a bounded structured-LLM
-proposal to reconcile identity groups. The proposal selects supplied
+proposal to reconcile identity groups.
-descriptors—names with their candidate source references—not durable IDs.
+
-Request-local candidate keys may support resolution internally, but are never
+## Semantic Registry Reconciliation
-included in model input or output. Colliding descriptors are ineligible, and
+
-invalid or unusable proposals retain the deterministic result with retry or
+The three registry normalizers instantiate the domain-neutral
-fallback diagnostics; the model does not directly replace durable records.
+`internal/framework/semanticreconcile` engine with default bounds. Each
+eligible candidate receives a contiguous, one-based `candidate_id` for that
+request. The model sees that handle, the candidate label and source-free
+evidence ranges, plus bounded transcript windows; it returns only duplicate
+groups of supplied handles and one supplied canonical handle per group. It
+never returns names, evidence, durable IDs, or replacement records. Identical
+labels and evidence remain independently selectable because their handles are
+distinct.
+
+The generic core owns the mandatory handle protocol, candidate and transcript
+presentation, the private response schema, source-reference validation,
+candidate and combined-material limits, structured completion, proposal
+assessment, stable group ordering, and typed plan-application mechanics. The
+D&D prompt contributes its system message and registry-specific semantic
+instructions. The generic registrar registers the shared prompt and schema;
+the D&D registrar registers each consuming prompt and the fallback profile.
+
+Fewer than two eligible candidates skips the LLM without a semantic warning.
+An exceeded bound also skips the call and preserves the deterministic
+preprocessed registry, adding the registry's bounded fallback warning. Invalid
+structured output or discarded proposal groups use the normalizer's existing
+retry contract; retry exhaustion preserves the safe deterministic or
+partially applied result and emits its bounded fallback warning. Provider,
+transport, cancellation, and context-material failures remain execution
+errors.
+
+Application remains typed and registry-owned. All three policies select the
+canonical member's normalized display name, union member evidence in source
+order, preserve ungrouped records, and derive durable identity only after
+consolidation. NPC IDs derive from the final name. Item IDs also derive from
+the final name, and a typed guard prevents currency aliases from crossing
+denominations or mixing currency with non-currency records. Location IDs
+derive from the final name and final evidence, preserving same-name,
+parent/child, and distinct physical-place identities. Registry warning scopes,
+reason codes, and postconditions remain outside the generic core.
 
 ## Generated References And Grounding
 

docs/internal/llm.md

@@ -42,6 +42,22 @@ observability. The adapter returns PromptKit’s validated raw bytes rather than
 re-encoding the decoded target. An empty optional material is represented as
 one space so its named input is retained by PromptKit.
 
+When a request includes semantic correction material, the adapter validates and
+defensively copies it before preparation, then appends exactly two messages
+after the ordinarily rendered prompt: the prior response as an assistant
+message and the correction guidance as a user message. Requests without a
+correction do not add messages or introduce caller roles. Ordinary request
+summaries record correction byte counts and digests only; complete messages are
+available solely in an explicitly requested debug trace.
+
+The adapter leaves the ordinary rendered message prefix, named inputs,
+variables, session, profile, execution overrides, prepared-execution path, and
+PromptKit repair policy unchanged for a corrected request. It never imports a
+PromptKit message type into a module or pipeline contract. PromptKit reports
+actual structural repair count and cumulative token usage per completion; the
+pipeline's safe terminal debug record projects those values without copying
+message content.
+
 Client construction may also receive a run-wide reasoning-effort override from
 the CLI factory boundary. The adapter copies the caller-owned pointer and
 creates a fresh PromptKit execution override for each request: a nil pointer
@@ -75,7 +91,7 @@ backend membership as runtime without performing generation. Fallback assets
 are mounted only when at least one source is registered. The production D&D
 registrar contributes its `dnd-extraction` fallback, and the maintained D&D
 prompts select that logical ID by default. PromptKit owns source precedence and
-profile parsing: an operator-provided matching profile takes precedence over a
+profile parsing and inheritance: an operator-provided matching profile takes precedence over a
 fallback profile without Notarius merging either document.
 When the registration is absent, a profile selecting `backend: local` fails
 inspection instead of falling back to a built-in or endpoint-only target.
@@ -102,7 +118,8 @@ because it changes scheduling rather than execution semantics.
 Production construction creates one PromptKit client and wraps it in one
 scheduled client. The scheduler has a fixed, positive permit limit, serves
 queued calls in FIFO order, and removes a queued call when its context is
-cancelled. A granted permit is released exactly once on every completion path.
+cancelled. It rechecks the caller context after admission and before dispatch.
+A granted permit is released exactly once on every completion path.
 
 The scheduled wrapper surrounds every `CompleteStructured` call, so concurrent
 lanes, pipeline retries, and LLM-backed validators share the same provider-call
@@ -141,12 +158,28 @@ arrangement and its data-only boundary are defined by
 [ADR-0011](../adr/0011-centralize-llm-assets.md), rather than by this runtime
 guide.
 
+The generic registrar is the sole production registration owner for the
+semantic-reconciliation default prompt and private response schema. The
+domain-neutral reconciliation package also exposes only its mandatory protocol
+and candidate/transcript presentation files for domain prompt manifests. D&D
+registry normalizers mount those files while retaining ownership and hashing
+of their D&D system message, semantic instructions, and complete prompt
+declaration. The response schema is therefore registered once even though
+several typed normalizers select it.
+
 Mounted prompt assets determine a module's fingerprint. The fingerprint hashes
 only the module and shared files explicitly selected by its manifest, so an
 unrelated asset does not invalidate a checkpoint. Schema loaders validate JSON,
 attach identity and digest metadata, make defensive copies, and expose
 diagnostics without raw schema bytes.
 
+Semantic-reconciliation normalizers extend this identity with the shared
+response-schema digest, framework policy version, and complete limit-policy
+digest. Their manifest metadata records the same content-free prompt, schema,
+policy, and limit identities together with domain identity and normalization
+policies. Request-local handles, source material, proposal content, and raw
+asset bytes are not checkpoint metadata.
+
 Private response schemas validate a model transport envelope. They are not the
 durable artifact schema and should not be documented as an external wire
 contract. Durable formats and compatibility rules remain in the
@@ -179,7 +212,9 @@ structured-output validation. The adapter reports an empty result, validation
 failure, empty structured body, or decode failure as
 `ErrInvalidStructuredOutput`, while retaining the returned raw bytes and debug
 material when they exist. Provider failures remain operational errors rather
-than output-validation failures.
+than output-validation failures. Apart from documented context, capacity, and
+invalid-output categories, provider error values and types do not cross the
+adapter error chain; callers receive only a credential-redacted diagnostic.
 
 When PromptKit rejects backend admission before generation, the adapter maps
 `promptkit.ErrCapacityExceeded` to
@@ -191,13 +226,29 @@ caller context takes precedence. The adapter does not retry capacity failures;
 the pipeline's existing binding attempt policy sees the operational error and
 decides whether to rerun the complete operation.
 
-Prompt-declared repair is executed within PromptKit’s structured-output flow.
+PromptKit executes structural repair within its structured-output flow. The
-The current production D&D prompt manifests set repair attempts to zero. That
+maintained production prompt manifests declare one additional repair attempt.
-setting does not replace pipeline retry behavior: a binding’s configured retry
+When a resolved binding supplies a repair value, the adapter inspects the
-count reruns its stage attempt after an error or rejection, and an exhausted
+prompt, copies its complete output contract, changes only the repair limit, and
-rejection is a recorded output rather than a provider error. The pipeline owns
+passes that complete replacement contract to PromptKit. This preserves the
-attempt lifecycle, validation chains, and retry diagnostics; see
+prompt's output format, validation mode, schema, and provider structured-output
-[Pipeline Internals](pipeline.md#validation-retries-and-output) and the
+settings.
+
+A successful repair is an ordinary successful completion, not a warning. The
+adapter reports PromptKit's actual repair count and its cumulative usage
+directly, without adding the initial and corrective counts again. Debug prompt
+material records the configured complete contract; debug response material
+records the repaired response and actual validation result. If the repair
+budget is exhausted, the adapter retains the final raw bytes and debug material
+and reports `ErrInvalidStructuredOutput`. Generation failures during an initial
+or corrective call remain provider-neutral operational errors with the same
+redaction boundary.
+
+Structural repair does not replace pipeline retry behavior: a binding's
+configured retry count reruns its complete stage attempt after an operational
+or structural error, module-requested retry, or actionable semantic rejection.
+The pipeline owns attempt lifecycle, validation chains, and retry diagnostics;
+see [Pipeline Internals](pipeline.md#validation-retries-and-output) and the
 [binding reference](../config.md#module-bindings-and-validators).
 
 ## Timeout Ownership
@@ -228,6 +279,13 @@ surfaced; when the completion already failed, its call error remains the
 result. Debug-bundle location, retention, and handling are operational concerns
 documented in [Operations](../operations.md#debug-bundles).
 
+The attempt-terminal summary is a separate safe trace record: it contains
+attempt kinds, validator outcome counts and reason codes, effective policy,
+terminal action, and repair/usage references. It excludes raw assistant
+responses and correction text. Those values can appear only in the explicitly
+requested detailed prompt and response artifacts, which require sensitive-data
+handling.
+
 Run manifests receive selected profile summaries, including optional effective
 backend and reasoning provenance, and component identities—not prompt, schema,
 source, reference, or response content. The published field semantics belong
@@ -237,6 +295,9 @@ redacted before it crosses the runtime boundary. Known-secret redaction is
 available to other runtime collaborators; it does not make prompt or response
 contents safe for general logging.
 
+Generation failures expose an application-owned category and optional HTTP
+status. Provider code, type, and message remain debug-only, after redaction.
+
 ## Failure Boundaries
 
 - Construction fails for missing asset registries, mutually exclusive profile

docs/internal/modules.md

@@ -22,6 +22,15 @@ to bindings whose declared execution class is `llm_backed` and rejects a
 binding-specific profile on a deterministic module. The user-facing precedence
 contract belongs in [Configuration](../config.md#pipelines).
 
+An eligible LLM-backed chunk, extract, merge, or normalize producer may also
+declare correction protocol `single_response_v1`. That declaration is a
+promise that the implementation accepts one attempt-local semantic correction
+and returns an owned copy of the exact one model response that directly
+controlled the candidate. It must forward correction only to its structured
+completion request; it must not manufacture prior-response material by
+serializing a normalized artifact or expose opaque application IDs. Input,
+output, validator, and deterministic specs cannot declare the protocol.
+
 Implementations that accept options must provide both an option validator and
 a builder. The validator is used while resolving configuration; the builder
 decodes the same options and constructs the implementation from the prepared
@@ -36,20 +45,38 @@ they need, register each leaf implementation, and add any family-owned assets
 or default validator chains. They return contextual errors so production
 composition fails at startup rather than at the first run.
 
+A validator that returns a completed rejection must supply two separate
+bounded values: a stable `ReasonCode` for provenance and actionable
+`CorrectionGuidance` for the producer. Guidance identifies the semantic defect
+and the constraints on one complete corrected replacement. It must not contain
+validator keys, diagnostic paths, opaque application IDs, or other internal
+identifiers. An operator-facing `Message` may explain the same event, but the
+framework never copies it into a model request. Missing or invalid guidance is
+a validator contract failure.
+
 An artifact family can register an optional typed evidence projector alongside
 its codec. The projector returns defensive copies of the artifact's direct
 generic source references and must use the codec's exact Go type. It does not
 interpret surrounding context or publish files; the pipeline validates the
 capability during preparation and the output boundary owns publication. See
 the [Published Evidence Context contract](../integrations/evidence-context.md)
-for the durable result.
+for the durable source-unit excerpt. Lane artifacts retain citation and lane
+provenance; the framework does not add either to that published excerpt.
+
+An artifact family is broader than a module: it owns the cohesive domain
+feature across its artifact type, codec, stage modules, validators, prompt
+policy, schemas, identity helpers, and reference projections. An extractor and
+normalizer in one artifact family remain independently registered modules in
+their respective pipeline stages. This ownership vocabulary does not create a
+new registry or change the fixed pipeline.
 
 ## Production Composition
 
 Production composition is intentionally split by family:
 
 - The generic registrar provides the unit chunker, generic JSON validators,
-  and JSON output encoder.
+  JSON output encoder, and shared semantic-reconciliation prompt and response
+  schema assets.
 - The Seriatim registrar provides the transcript input adapter. Its external
   input behavior is defined by the [Seriatim contract](../integrations/seriatim.md).
 - The D&D registrar provides its codecs, extractors, mergers, normalizers,
@@ -60,6 +87,42 @@ The CLI owns the composition that invokes these registrars. A module package
 may register its own family but must not assemble the CLI or make framework
 packages depend on production extensions.
 
+## Semantic Reconciliation
+
+`internal/framework/semanticreconcile` is a domain-neutral strategy used by a
+typed normalize module; it is not itself a selectable stage module. A
+source-backed artifact-family normalizer projects its deterministic records
+into contextual candidates and owned typed record envelopes, supplies its
+chosen prompt identity and resolved LLM profile, and constructs an engine with
+explicit limits. The core filters invalid evidence, assigns contiguous
+request-local integer handles, renders bounded candidate and transcript
+materials, invokes the structured-completion boundary, and assesses the
+returned duplicate groups into a stable non-overlapping plan.
+
+The normalizer then applies that plan through a typed `ApplicationPolicy`. The
+core preserves ungrouped records, contribution order, and provenance while the
+artifact family owns group guards, field and evidence consolidation, durable
+ID derivation, retry and fallback presentation, classified diagnostics, and postconditions.
+Request-local handles do not enter the typed value or durable artifact. Fewer
+than two eligible candidates skips model invocation; exceeding a candidate or
+combined-material bound preserves the deterministic result under the family's
+fallback policy. Provider, transport, cancellation, and context-construction
+failures remain execution errors.
+
+When the engine actually makes a proposal call, its typed result carries the
+owned exact proposal response under the same correction contract as other
+eligible producers. Deterministic skip, limit, and fallback outcomes carry no
+model candidate, so a later rejection applies terminal policy without spending
+an ineffective semantic retry.
+
+The core supplies a conservative generic prompt and the single private
+response schema. A domain prompt may substitute its semantic instructions but
+mounts the core-owned protocol and candidate/transcript presentation assets.
+Prompt, schema, policy, and limit identities participate in manifest metadata
+and checkpoint fingerprints. The generic registrar owns production
+registration of those shared assets; a consuming domain registrar owns only
+its domain prompt.
+
 ## Adding Or Changing A Module
 
 1. Choose the pipeline stage and the typed artifact boundary. Put external
@@ -72,9 +135,13 @@ packages depend on production extensions.
 3. Implement strict option decoding, construction, and the typed stage
    interface. Preserve caller ownership: do not retain mutable request data
    and return defensive copies where an implementation exposes stored data.
+   If declaring correction capability, forward the request correction and
+   retain only the exact validated response that controlled the result.
 4. Register the module through its typed registry helper and add it to the
    owning family registrar. Add a default validator chain only when that
    family owns the behavior; otherwise require an explicit compatible chain.
+   Every rejection path in a validator must provide actionable correction
+   guidance while retaining its stable internal reason code.
 5. Update the selectable-key and chain reference in
    [Configuration](../config.md#production-module-keys), the applicable
    integration contract, and focused tests. Keep the configuration document

docs/internal/overview.md

@@ -25,10 +25,12 @@ physical state roots.
 | Area | Implemented owners | Responsibility |
 | --- | --- | --- |
 | Executable and command boundary | **cmd/notarius**, **internal/cli** | Process entry, command dispatch, configuration discovery, production composition, runtime collaborator setup, durable file placement, and user-facing reporting. |
+| Build information | **internal/buildinfo** | Resolves a stable linked release tag or build metadata for the diagnostic root version command. |
 | Configuration | **internal/core/config** | Defaults, strict YAML parsing, environment overrides, structural validation, effective resolution, redaction, and resolved-composition summaries. |
 | Generic models | **internal/core/source**, **internal/core/artifacts**, **internal/framework/contracts** | Source documents and chunks, manifests and provenance, plus typed artifact, reference, validation, output, and structured-completion contracts. |
 | Pipeline framework | **internal/framework/pipeline** | Registries, profile and reference resolution, typed preparation, validation, retry coordination, ordered execution, handoff, and result assembly. |
 | LLM and prompt runtime | **internal/framework/llm**, **internal/framework/promptfs** | Provider-neutral structured completions, scheduling, profile recording, prompt assets, schema registration, and credential-shaped-value redaction. |
+| Semantic reconciliation | **internal/framework/semanticreconcile** | Bounded source-backed candidate preparation, request-local handle proposals, deterministic assessment, typed plan application, and reconciliation identity metadata; see [Module Internals](modules.md#semantic-reconciliation) and [D&D Module Internals](dnd.md#semantic-registry-reconciliation). |
 | Embedded LLM content | **assets** | Read-only centralized LLM-facing content, scoped by its consuming package; see [LLM Runtime](llm.md#prompt-and-schema-assets) and [D&D Module Internals](dnd.md#prompt-construction). |
 | Runtime state | **internal/core/fileio**, **internal/core/debugbundle**, **internal/framework/checkpoint**, **internal/framework/chunkplan**, **internal/framework/chunkmap**, **internal/framework/debug** | Confined atomic files, debug bundles, checkpoint and chunk-plan state, accepted chunk maps, and pipeline-facing debug recording. |
 | Production extensions | **internal/modules/generic**, **internal/modules/seriatim**, **internal/modules/dnd** | Domain-neutral extensions, Seriatim input support, and D&D extraction families registered into the production catalog. |
@@ -49,8 +51,9 @@ the CLI composition boundary.
   composition, and path safety.
 - [LLM Runtime](llm.md): structured completion, scheduling, prompt assets,
   profiles, and secret handling.
-- [Module Internals](modules.md): generic extension registration, module
+- [Module Internals](modules.md): generic extension registration, artifact
-  construction, validation, and reference mechanics.
+  families, module construction, semantic reconciliation, validation, and
+  reference mechanics.
 - [D&D Module Internals](dnd.md): shared D&D extractor conventions, generated
   reference projections, and lane-specific exceptions. Durable D&D and
   Seriatim data shapes remain in the [integration contracts](../integrations/).

docs/internal/pipeline.md

@@ -12,7 +12,7 @@ own durable output shapes. Concrete production extensions are covered by
 The pipeline framework accepts a resolved composition, registries, shared
 dependencies, input bytes, a supplied prompt session, and state/debug
 collaborators. It returns logical output files, normalized artifacts, recorded
-rejections and warnings, manifest provenance, and checkpoint decisions. The
+rejections, grouped diagnostics, manifest provenance, and checkpoint decisions. The
 CLI owns process arguments, configuration discovery, session resolution,
 physical roots, and placement of returned output files.
 
@@ -32,13 +32,26 @@ Resolution turns a configured pipeline profile into a **ResolvedPipeline**.
 It normalizes the pipeline and lane identities, applies stage defaults, selects
 requested lanes where that is supported, resolves validator chains, checks
 module capabilities and typed artifact compatibility, validates options, and
-assigns a deterministic resolved-composition digest. The resolved pipeline
+assigns a deterministic resolved-composition digest. A correction protocol is
-contains bindings and declared reference targets, not external reference bytes.
+selected from each eligible LLM-backed producer specification and becomes part
+of that resolved identity; only `single_response_v1` is currently supported.
+Preparation rejects an LLM-backed producer that combines a non-empty validator
+chain with positive producer retries unless it declares that protocol. Producers
+without validators or without retries remain valid without correction support.
+The resolved pipeline contains bindings and declared reference targets, not
+external reference bytes.
 After selection, the resolver applies command, binding, and pipeline profile
 precedence to LLM-backed bindings and validators only; prompt defaults remain
-an empty resolved binding profile. Deterministic bindings remain profile-free.
+an empty resolved binding profile. It resolves structural output repair
-These effective values are part of the digest, so execution and checkpoint
+separately: a binding's `structured_output_repair_attempts` value wins, then a
-consumers do not repeat profile inheritance.
+pipeline value applies to LLM-backed bindings and validators, and omission
+leaves the prompt-owned policy intact. An explicit repair value on a
+deterministic binding is rejected. Resolved bindings own copied repair values,
+and these effective values are part of the digest, so execution and checkpoint
+consumers do not repeat profile inheritance or configuration resolution.
+Each LLM request receives its own copy of that resolved value. PromptKit spends
+it only for structural correction inside one completion; the runner's binding
+retry policy remains the separate outer budget for complete stage attempts.
 Configuration resolution supplies the selected profile and catalog; see
 [Configuration Internals](configuration.md).
 
@@ -46,16 +59,20 @@ External reference materialization happens before preparation. The materializer
 checks that each slot is declared by the selected module, resolves a file path
 relative to the correct configuration or working-directory origin, reads
 UTF-8 text, verifies media type and size limits, and retains bounded
-provenance. A generated-artifact selector remains declared but has no bytes
+provenance. For a positive slot limit, it reads at most the limit plus one byte
-until its producing step completes.
+and rejects overflow before retaining content. A generated-artifact selector
+remains declared but has no bytes until its producing step completes.
 
 Preparation is the construction boundary. It validates the resolved shape and
 registry set, clones the resolved data, then constructs the input adapter,
-chunker, stage-local validators, every typed lane, and output encoder with
+chunker, stage-local validators, every typed lane, and output encoder. The
-cloned options, references, and shared dependencies. It also collects stable
+prepared producer metadata preserves each selected correction protocol, and
-checkpoint fingerprints. Missing registrations, incompatible typed entries,
+the resolved digest carrying that metadata participates in checkpoint identity.
-nil implementations, and constructor failures are reported before source
+Each registered builder receives its own cloned build request immediately
-parsing or any stage operation begins.
+before its module-owned code runs. Preparation also collects stable checkpoint
+fingerprints. Missing registrations, incompatible typed entries, nil
+implementations, and constructor failures are reported before source parsing
+or any stage operation begins.
 
 An output encoder can opt into source-evidence publication through its output
 policy. Preparation keeps the configured lane allowlist and active lanes
@@ -115,18 +132,52 @@ for started workers, and prevents output encoding.
 
 Every chunk, extract, merge, and normalize candidate passes its resolved
 validator chain. Validators receive immutable canonical input appropriate to
-their target: chunks, typed values, or serialized codec bytes. They may
+their target: chunks, codec-decoded typed candidates, or serialized codec
-approve, approve with warnings, reject, or fail. A rejection is an ordinary
+bytes. Each typed validator receives a newly decoded value from the one
-pipeline result; a validator error is a framework error.
+candidate serialization for that attempt, while serialized validators receive
+separately owned representation bytes and schema metadata. They may approve,
+approve with warnings, reject, fail, or be skipped when a runtime prerequisite
+is unavailable. The shared executor settles every configured validator in
+order. A failed LLM-backed validator retries only itself against the same
+immutable candidate; it does not regenerate the producer or alter the
+validator request. Rejections stop that validator, while other configured
+validators still run. The executor retains ordered results, bounded
+deduplicated correction guidance from rejections, and only the final exhausted
+failure outcome for each validator. The correction builder keeps first
+occurrence order, omits internal reason codes, validator names, and operator
+messages, and requests one complete replacement. Missing guidance or an
+oversized aggregate is a framework contract error; guidance is never inferred
+or truncated.
 
 The runner applies the binding's retry policy around a stage operation and its
-complete validation chain. It preserves warnings only from the final accepted
+complete validation chain. It preserves terminal diagnostics only from the final accepted
-or rejected attempt. Cancellation stops retries. Normalizer-specific retry
+or rejected attempt, plus one fixed validation-incomplete warning per validator whose execution
-directives consume this same budget and validate any final safe fallback through
+budget was exhausted under `warn_continue`. Cancellation stops retries.
-the normalizer chain.
+Normalizer-specific retry directives consume this same budget and validate any
+final safe fallback through the normalizer chain.
+
+The artifact-neutral producer-attempt state machine owns that shared budget,
+attempt provenance, semantic-correction material, and terminal-policy
+selection. It accepts producer and complete-validation closures, so artifact
+materialization, cache handling, checkpoints, and debug output stay at the
+operation boundary. It distinguishes operational, structural, module-requested,
+and semantic retries. A semantic retry is available only for a valid latest
+`single_response_v1` candidate; a deterministic or no-model rejection instead
+settles the semantic policy immediately. Structural-output errors alone use the
+structural policy, and validation failure without rejection settles the
+validator-failure policy without regenerating the producer.
+
+Chunk planning uses this state machine for generated plans. A rejected or
+validation-incomplete automatic cache hit is not model material and therefore
+falls through to a fresh initial generation at producer attempt one; it neither
+receives a correction, consumes retry budget, promotes cached-candidate
+diagnostics, nor overwrites the stored record. An incomplete cache validation
+under `fail_run` terminates instead. Only a newly generated, completely
+validated plan is published to the chunk-plan store. Rejected plans never
+advance, and validation-incomplete plans remain unpublishable.
 
 After terminal lane work, the runner assembles manifest provenance, normalized
-artifacts, rejections, warnings, and an optional accepted chunk map. When an
+artifacts, rejections, final grouped diagnostics, and an optional accepted chunk map. When an
 output policy selected evidence lanes, it decodes accepted serialized normalize
 outputs through their registered codecs and invokes the prepared typed
 projectors. Rejected or absent lanes contribute nothing. This reconstruction is
@@ -137,6 +188,15 @@ The CLI publishes those files only after the runner returns without a framework
 error. Logical file names and schemas are defined by the [output integration
 contracts](../integrations/).
 
+For every completed producer disposition, the runner projects one bounded
+validation summary to the manifest, the affected rejection when present, and
+the CLI result receipt. The summary records status, configured-order rejecting
+validators and reason codes, incomplete validators, producer-attempt count,
+and terminal action. It contains no operator message, correction guidance, or
+model response. `complete`, `rejected`, and `incomplete` describe the final
+candidate disposition; a run-level `incomplete` status indicates at least one
+current-run output advanced under `warn_continue`.
+
 ## Checkpoint And Debug Hooks
 
 The runner receives checkpoint and debug interfaces rather than roots. It
@@ -146,6 +206,20 @@ handoff. Generated-reference dependencies participate in checkpoint decisions.
 Selective recomputation can require a canonical accepted normalized predecessor
 before a dependent lane starts.
 
+The runner writes successful checkpoint artifacts only after complete accepted
+validation. Chunk plans follow the same rule for publication. A rejection,
+invalid structured response, or incomplete validation is never reusable state;
+the current run may still hand off an otherwise valid `warn_continue` result
+according to its terminal policy. The runner carries private reuse eligibility
+through extract, merge, normalize, and generated-reference handoff. Any stage
+derived from incomplete validation skips both checkpoint lookup and all
+checkpoint publication even when that stage's own validation completes.
+External references and fully validated generated references remain eligible.
+Attempt debug records retain safe kind,
+validator, repair-usage, policy, and terminal-decision provenance. Full
+assistant and correction content remains confined to the requested detailed
+LLM trace.
+
 Debug recording is attempt-scoped and application-owned. A failure to persist
 required debug data is a framework error. State roots, persistence, reason-code
 meanings, resume, and cleanup are intentionally owned by

docs/internal/state.md

@@ -47,6 +47,9 @@ The serialized
 they do not describe a current public state surface.
 
 Ordered-step lane checkpoints include the step identity in their storage scope.
+Accepted step and lane identities are encoded injectively before becoming
+filesystem path components, while ordinary safe identifiers retain their
+readable paths.
 When a later lane consumes a generated artifact, its dependency fingerprints
 include the producer's artifact kind, complete schema identity, media type,
 canonical content digest, and size. Ordinary resume compares those fingerprints
@@ -61,12 +64,13 @@ Ordinary resume loads extract, merge, and normalize checkpoints progressively
 and may execute later lane stages after an earlier cache miss. Selective
 recomputation instead asks the loader for the required producer's accepted
 normalize artifact. That lookup reuses the existing normalize files, requires
-workspace schema v3 plus an exact non-empty invocation identity, and deliberately
+workspace schema v4 plus an exact non-empty invocation identity, and deliberately
 does not require extract or merge checkpoint files or dependency fingerprints.
 The runner performs canonical codec and producer-provenance validation before
 cloning the artifact into normal step output. Success restores only stored
-normalize warnings and emits one normalize decision; failure retains the files,
+normalize diagnostics and emits one normalize decision; failure retains the
-records the decision, and stops without executing the producer or consumer.
+files, records the decision, and stops without executing the producer or
+consumer.
 
 The loader assigns a typed category and reason code at each validation site;
 diagnostic prose is not classified after the fact. The runner then applies
@@ -94,7 +98,7 @@ owns the operator workflow and stable reason-code meanings.
 
 `internal/core/debugbundle` allocates an explicitly requested per-run bundle
 with `summary/` and `trace/` roots. `SummaryWriter` persists redacted command,
-resolution, run, warning, and failure artifacts. `internal/framework/debug`
+resolution, run, final grouped diagnostic, and failure artifacts. `internal/framework/debug`
 implements the pipeline-facing trace recorder under the trace root.
 
 The CLI allocates a bundle before pipeline resolution and treats requested

docs/operations.md

@@ -5,6 +5,23 @@ This is the canonical guide for operating Notarius runtime state. The
 [Configuration](config.md) owns fields, defaults, and precedence. Maintainers
 who need implementation mechanics should read [Run State Internals](internal/state.md).
 
+## Source Deployment
+
+Linux is the supported deployment platform. Install a pinned source release
+with the Go version declared in `go.mod` (currently Go 1.25.5):
+
+~~~sh
+GOWORK=off go install \
+  gitea.maximumdirect.net/eric/notarius/cmd/notarius@vMAJOR.MINOR.PATCH
+~~~
+
+Pin the exact tag in deployment automation rather than following a branch.
+Use [`notarius --version`](cli.md#command-summary) as a diagnostic after
+installation; its syntax and semantics are owned by the [CLI reference](cli.md).
+The maintainer publication process, including tag guards and verification,
+belongs to [Source Releases](release.md). macOS builds are best-effort for
+development, and Windows is unsupported.
+
 ## State Surfaces
 
 Each run can use independent roots with different retention and access-control
@@ -72,6 +89,9 @@ provider call or credentials:
 notarius config validate --config /etc/notarius/config.yml --pipeline dnd-session
 ~~~
 
+An unset optional `api_key_env` reaches the provider without authorization and
+may receive a 401 or 403 response.
+
 Profile paths are currently resolved from the process working directory, not
 from the configuration file. The complete example's
 `./examples/profiles/dnd-extraction.yml` path is valid for a repository-root
@@ -85,10 +105,40 @@ and resolves configuration before module preparation and source parsing. It
 then performs any permitted cache lookup, executes the pipeline, and publishes
 logical output files only after a successful runner result.
 
-On success, the command reports the output bundle path. A warning-bearing run
+On success, the command reports the output bundle path. A run with actionable
-still succeeds and reports its warning count on standard error. Errors and
+process warnings still succeeds and reports warning-group and occurrence counts
+on standard error; advisory and observation findings do not produce a warning
+line. Errors and
 their exit classes are defined in the [CLI reference](cli.md#output-streams-and-exit-statuses).
 
+## Validation Retries And Terminal Outcomes
+
+Each producer binding has one outer **retries** budget. It covers complete
+producer attempts for operational failures, invalid structured output,
+normalizer fallback retry, and semantic correction. It is independent from
+PromptKit's structural-repair calls inside one completion and from an
+LLM-backed validator's own retry budget. A semantic correction rebuilds the
+ordinary producer request and supplies only the latest rejected model response
+plus aggregated validator guidance; it is not a conversation replay.
+
+After the applicable budgets are exhausted, the resolved
+[`validation_policy`](config.md#pipelines) determines the result. Structural
+failure and semantic rejection normally fail the run; an explicit
+`reject_output` records a rejection and allows unrelated work to finish. A
+validator execution failure normally uses `warn_continue`, which keeps an
+otherwise accepted result in the current run with `incomplete` validation
+provenance. It emits one bounded warning for every validator whose execution
+budget was exhausted. A corrected result that later passes validation does not
+retain diagnostics from abandoned attempts.
+
+Treat a successful process exit as a completed run, not as proof that every
+candidate was fully validated. Inspect the receipt's `validation_status`,
+`validation_summaries`, rejection count, and warning group and occurrence
+counts when an orchestrator requires complete validation. The durable fields
+and their meanings are owned by the
+[run-result receipt](integrations/run-result.md) and
+[published JSON output contract](integrations/json-output.md).
+
 ## Output Bundles
 
 Each successful run receives a generated safe run identifier and writes beneath:
@@ -110,9 +160,9 @@ are defined in [Accepted Chunk Map](integrations/chunk-map.md). An optional
 [evidence context](integrations/evidence-context.md) contains source-unit text
 and metadata. It is not a cache or debug artifact: retain it with the output
 bundle only for as long as consumers need it, and apply source-content access
-controls to the entire bundle. Selected lanes may collectively cite most of a
+controls to the entire bundle. Its selected source-unit excerpt may include
-transcript, so a broad allowlist can make the evidence artifact nearly as
+every source unit once when coverage is broad or its configured window is
-sensitive and large as the source itself.
+large, so do not assume a byte or token reduction or reduced sensitivity.
 
 ## Chunk-Plan Cache
 
@@ -138,7 +188,9 @@ The configured cache mode controls one invocation:
 A reused plan is still materialized and validated against the current source.
 If a prior plan no longer gives acceptable results, use a refresh run rather
 than editing cache files. Deleting a plan is recoverable but can repeat costly
-chunking work.
+chunking work. A plan accepted only under incomplete validation is not
+published, and a rejected cache hit falls through to ordinary generation rather
+than becoming a correction candidate.
 
 ## Checkpoint Recording, Resume, And Recompute
 
@@ -163,6 +215,12 @@ Reasoning-effort inheritance, replacement, and explicit clearing are distinct
 runtime identities, so checkpoints created under one state are not reused by
 either of the others.
 
+Only accepted, completely validated chunk, extract, merge, and normalize
+results are checkpointed for reuse. Rejected, structurally invalid, and
+validation-incomplete producer results remain non-reusable, even when a
+`warn_continue` result advanced during its original run. A resumed invocation
+therefore reruns that producer rather than treating degraded state as accepted.
+
 Checkpoint state is confined below an identity-specific path:
 
 ~~~
@@ -217,13 +275,16 @@ Only a [debug-enabled run](cli.md#run) creates a bundle:
 ~~~
 
 The summary contains redacted invocation and resolution information plus run,
-warning, checkpoint, chunk-plan, and terminal reporting artifacts. The trace
+final grouped diagnostic, checkpoint, chunk-plan, and terminal reporting artifacts. Attempt
-contains allowlisted application diagnostic records and can include source or
+terminal records contain bounded attempt kinds, validator outcomes, policy,
-derived application data. Neither surface is a cache input. Do not treat a
+decision, PromptKit repair count, and usage; they do not contain assistant
-debug bundle as safe to share merely because its configuration summary is
+responses or complete correction messages. The trace contains allowlisted
-redacted. Invocation metadata omits reasoning effort when it is inherited,
+application diagnostic records and can include source, model, and correction
-records the replacement value when one is supplied, and records an empty value
+content. Neither surface is a cache input. Do not treat a debug bundle as safe
-when inherited reasoning was explicitly cleared.
+to share merely because its configuration summary is redacted. Invocation
+metadata omits reasoning effort when it is inherited, records the replacement
+value when one is supplied, and records an empty value when inherited reasoning
+was explicitly cleared.
 
 Notarius never creates debug state without an explicit request and never
 automatically deletes a requested bundle. If allocation succeeds, the command
@@ -255,14 +316,29 @@ Provider execution settings and the generation timeout come from the selected
 PromptKit profile. The invocation-only **--reasoning-effort** and
 **--clear-reasoning-effort** controls may replace or clear that profile setting
 for all LLM-backed calls in one run without changing the profile. PromptKit
-v0.5.0 does not add a provider retry loop. Notarius binding retries rerun the
+structural output repair happens within one structured-completion call. Its
-complete module operation and validation chain as defined by
+effective `structured_output_repair_attempts` limit is resolved from the
-[module bindings](config.md#module-bindings-and-validators).
+selected binding, then the pipeline, then the prompt declaration; see
+[module bindings](config.md#module-bindings-and-validators). This is distinct
+from Notarius binding **retries**, which rerun the complete module operation
+and validation chain and do not consume or replenish the structural-repair
+limit. The maintained production prompts declare one repair attempt, paid only
+after a structural failure. One structured completion with repair budget **R**
+makes at most **R + 1** serial provider calls. If one stage attempt performs
+**C** structured completions, a binding with **retries: N** has a maximum of
+**(N + 1) * C * (R + 1)** provider calls; LLM-backed validators have their own
+corresponding invocation counts and budgets. This is an upper bound, not a
+promise that every call reaches a provider.
 
 Timeouts are layered. Caller cancellation is the outer authority. A positive
 effective generation timeout adds an inner request deadline, while zero
 disables only that generation deadline. The HTTP client timeout remains a
 transport-wide cap. Notarius does not add another timeout around PromptKit.
+Repairs are serial within the same caller context, so their worst-case latency
+and cost follow the provider-call bound above; provision run deadlines and
+provider budgets accordingly. Credentials remain optional unless the selected
+PromptKit profile requires one, in which case preparation fails before a
+provider call when its configured credential is unavailable.
 The pinned upstream boundary and profile-format links are in
 [PromptKit Integration](integrations/pkg-promptkit.md).
 
@@ -279,6 +355,11 @@ positive value makes the effective active local-generation bound the smaller
 of **total_llm** and that local limit, so a local limit of four permits no more
 than four active local generations.
 
+The Notarius scheduler admits one logical structured completion and holds that
+permit while PromptKit performs its serial corrective calls. PromptKit applies
+its selected-backend admission to each provider call; Notarius does not
+reacquire a permit or add another scheduler for a repair.
+
 For a positive local limit, PromptKit owns its default waiting capacity and
 admission behavior. When a PromptKit backend has admitted all active and queued
 work, a new call fails as capacity exhaustion before generation. The adapter

docs/policy/architecture.md

@@ -24,6 +24,12 @@ DAGs or a general workflow language. Every stage remains explicit; general
 chunking, merging, or normalization behavior must not be hidden inside an
 extractor.
 
+A stage module is one configured implementation of one pipeline stage. An
+artifact family is the cohesive domain feature that owns an artifact across
+the explicit stages and supporting codecs, validators, prompts, identity
+rules, and reference projections. Artifact-family ownership does not combine
+stages or alter the fixed pipeline.
+
 Input and chunking are pipeline-wide. Each selected artifact lane owns its
 extract, merge, and normalize stages, and the output stage aggregates the run's
 lane outcomes.
@@ -39,6 +45,12 @@ implementations. Domain-neutral model and framework layers provide reusable
 policy, contracts, and orchestration. Concrete input, pipeline, output, and
 validation extensions depend inward on those generic layers.
 
+Semantic reconciliation is one such domain-neutral framework mechanism. It
+prepares bounded source context, invokes a shared model-judgment protocol,
+validates proposals, and applies safe plans through typed policies supplied by
+the consuming artifact family. It does not own domain identity, durable IDs,
+warning semantics, or artifact construction rules.
+
 Generic layers must not depend on production extensions. Concrete extensions
 must not compose the application or take ownership of process behavior. The
 current packages implementing these layers are inventoried in
@@ -134,8 +146,9 @@ lanes, validators, and LLM profile: the canonical source digest selects the
 plan, while the current run still applies its configured chunk validators to
 the materialized chunks.
 
-The framework owns orchestration and handoff provenance. Modules return logical
+The framework owns orchestration, origin enrichment, aggregation, and handoff
-results and warnings; they do not own CLI reporting, physical output, cache, or
+provenance. Modules return logical results and classified diagnostics; they do
+not own CLI reporting, physical output, cache, or
 debug roots, durable file placement, or checkpoint and debug lifecycle.
 
 After pipeline-wide chunking, extraction uses bounded framework concurrency.
@@ -147,26 +160,46 @@ may overlap. The framework must not create unbounded goroutines per lane or
 chunk.
 
 Completion timing does not choose public ordering or errors. The coordinator
-orders accepted artifacts, warnings, rejections, checkpoint events, and
+orders accepted artifacts, grouped diagnostics, rejections, checkpoint events, and
 framework errors by stable pipeline scope. Rejections do not cancel unrelated
 work. A framework error cancels derived work, prevents undispatched work from
 starting, waits for started work, and prevents output encoding.
 
+Warnings are process-only signals: configuration degradation, approved fallback,
+or incomplete configured validation. Quality uncertainty and grounding findings
+are advisories; successful canonicalization and cleanup are observations.
+Modules choose that semantic classification, while the framework attaches
+origin, aggregates groups, enforces bounds, and presents final collections.
+An ordinary successful run therefore has zero warnings. See
+[ADR-0015](../adr/0015-separate-process-warnings-from-quality-diagnostics.md)
+for the decision rationale.
+
 ## Validation
 
 Validation is a framework-managed boundary around outputs from chunk, extract,
-merge, and normalize stages. Validators receive immutable stage output
+merge, and normalize stages. Validators receive immutable stage output and
-and make an explicit whole-output decision: approve, approve with warnings, or
+make an explicit whole-output decision: approve, reject, fail, or skip when a
-reject.
+runtime prerequisite is unavailable.
 
 Typed artifact validators receive the domain value directly. Chunk validators
 receive source-zone chunks, while serialized validators receive immutable
 representation bytes and declared schema metadata. A validator registered for
 one target or artifact kind cannot satisfy an incompatible selection.
 
-Rejection is a recorded pipeline outcome, not a framework execution error.
+The framework runs every applicable validator sequentially in configured order.
-Validator execution failures are framework errors. Rejected output does not
+It aggregates rejections, exhausted validator failures, and skips before the
-advance to the next stage.
+producer policy chooses a disposition. A completed rejection never advances.
+With no rejection, an exhausted validator failure may fail the run or, under
+the configured `warn_continue` policy, advance a structurally valid candidate
+with explicit incomplete-validation provenance. Validators report findings;
+they do not choose candidate disposition.
+
+A completed rejection supplies a stable reason code for internal provenance
+and bounded actionable correction guidance for the candidate producer. Reason
+codes, validator keys, and operator-facing messages remain diagnostic data;
+they are not model instructions. The framework constructs model-facing retry
+text only from the semantic guidance and fails the contract rather than
+inventing or truncating missing guidance.
 
 Default validator chains are production composition policy and are registered
 centrally by stage and module. Configuration may replace a stage-local default,
@@ -183,11 +216,29 @@ The caller of the LLM owns prompt selection, prompt inputs, response schema,
 and interpretation of structured output. Provider adapters do not own source-
 or domain-specific prompt logic.
 
+PromptKit owns bounded structural correction within one structured completion.
+Notarius owns outer stage attempts, semantic validation, and acceptance policy;
+the two budgets must remain separate.
+
+An LLM-backed producer can participate in semantic correction only when it
+declares `single_response_v1` and returns the exact one response that directly
+controlled its candidate. On an actionable rejection, the framework rebuilds
+the ordinary request and appends only the latest defective response as an
+`assistant` message plus one aggregated `user` correction message. This is a
+fresh replacement request, not a growing conversation. The retry budgets,
+terminal policy, and sensitive-data rationale are recorded in
+[ADR-0014](../adr/0014-feedback-aware-validation-retries.md).
+
 When a model selects an application entity, callers must supply a contextual
 selection and deterministically attach the opaque application identity whenever
 the selection resolves exactly. Models do not receive or reproduce opaque
-application identifiers; [ADR-0012](../adr/0012-resolve-opaque-entity-identifiers-deterministically.md)
+application identifiers. Semantic reconciliation may instead expose
-records the rationale and limited request-local-label exception.
+contiguous, one-based candidate handles that exist only for one request;
+deterministic code resolves them before typed application, and they never
+become durable identity. This is the approved request-local-label application
+of [ADR-0012](../adr/0012-resolve-opaque-entity-identifiers-deterministically.md)
+recorded by
+[ADR-0013](../adr/0013-use-request-local-candidate-handles-for-semantic-reconciliation.md).
 
 LLM calls and other external operations accept cancellation and respect
 timeouts. Concurrency control belongs in shared runtime plumbing rather than in
@@ -212,7 +263,8 @@ invalid or incompatible.
 
 Run manifests record enough resolved pipeline, module, source, reference, and
 LLM provenance to make a run auditable after configuration changes. Manifests
-record identities and summaries rather than secret or large payload content.
+record identities and bounded validation summaries rather than secret, raw
+model, correction, or large payload content.
 
 ## State, Output, And Safety
 
@@ -232,18 +284,39 @@ an invocation that explicitly requests resume. Debug is never a cache input and
 is never created without an explicit request. Pipeline modules receive
 collaborator interfaces and never physical roots.
 
+Only accepted, completely validated producer output is reusable checkpoint or
+chunk-plan state. Rejected, structurally invalid, and validation-incomplete
+results cannot become cache or checkpoint inputs, even when a
+`warn_continue` result is allowed to advance in the current run. This
+ineligibility follows derived merge and normalize results and generated
+references for the remainder of the run: current-run handoff remains allowed,
+but no dependent cache or checkpoint may be loaded or published.
+
 Writes are atomic where practical. Paths for writes, moves, overwrites, and
 deletion must be narrow and explicit. Notarius never automatically deletes
 output or requested debug bundles; cache cleanup is explicit and recoverable.
 
 Secrets must not appear in errors, logs, output, cache, debug summaries,
-traces, manifests, documentation, examples, or redacted configuration. Debug
+traces, manifests, documentation, examples, or redacted configuration. Raw
+assistant responses and complete correction messages are attempt-local and are
+excluded from ordinary durable records and summaries; the requested detailed
+debug trace is the sole diagnostic surface allowed to retain them. Debug
 collection is allowlisted to application-owned payloads and must not capture
 unrelated process environment values or filesystem content. Trace data may
 contain application data and therefore inherits its sensitivity; operators own
 access controls and retention. Physical layout and operation are defined in
 [Operations](../operations.md).
 
+## Platform And Distribution
+
+Linux is the supported deployment platform. macOS is supported only as a
+best-effort development and compilation environment, while Windows is
+unsupported. Notarius distributes source releases only: an immutable source
+tag and its checked-in release note identify a release. The project does not
+publish executable binaries, archives, installers, container images,
+checksums, signatures, or package-manager entries. Maintainer release commands
+and tag guards belong to [Source Releases](../release.md).
+
 ## Architectural Non-Goals
 
 Notarius does not aim to provide:

docs/policy/documentation.md

@@ -65,6 +65,8 @@ secret values.
 | CLI contract | `docs/cli.md` | Commands, arguments, flags, invocation semantics, and exit codes. | End-to-end operating procedures, configuration field definitions, runtime filesystem layout, module implementation details. |
 | Configuration contract | `docs/config.md` | Discovery and precedence, file schema, fields, defaults, environment overrides, validation rules, and user-selectable module or validator keys. | Complete example files, CLI syntax, runtime state lifecycle, module implementation details. |
 | Operations | `docs/operations.md` | Runtime workflows, physical filesystem and state layout, output, cache, and debug handling, resume, cleanup, permissions, recovery, and operational limits. | CLI flag syntax, configuration field definitions, logical output schemas, implementation mechanics. |
+| Source release procedure | `docs/release.md` | Maintainer release selection, candidate validation, tagging, publication guards, verification, and immutable-tag recovery. | Product installation summary, CLI version semantics, historical release summaries, CI implementation detail. |
+| Release-note history | `docs/releases/` | One checked-in historical summary for each source release made under the procedure. The note at the immutable tag is that release's record. | Current commands, behavior, contracts, and compatibility definitions. |
 | Public HTTP contract, if introduced | `docs/api.md` | Routes, authentication, media types, request and response schemas, status codes, pagination, caching, idempotency, rate limits, and HTTP retry semantics. | Client walkthroughs, upstream or downstream integration internals, implementation detail. |
 | Consumer guidance, if a public package or API is introduced | `docs/consumers/` | Task-oriented use of the public interface, minimal client examples, and consumer responsibilities. | HTTP wire semantics, external protocol contracts, internal implementation detail. |
 | External and durable integration contracts | `docs/integrations/` | External file formats and protocols, upstream and downstream contracts, logical output bundle paths and schemas, media types, and compatibility behavior. | Physical runtime placement and lifecycle, internal transformations, CLI syntax, configuration defaults. |
@@ -95,6 +97,14 @@ runtime state and how to operate or recover the application. When a workflow
 crosses these topics, choose the document that owns the task and link to the
 other contracts.
 
+### Releases
+
+`docs/release.md` owns the source-release procedure. Release notes are
+historical summaries, not current-state contract owners: the checked-in note at
+an immutable tag records that release, while current canonical documentation
+must change with the behavior it describes. Do not use a release note to defer
+or replace current documentation updates.
+
 ### Contracts And Implementation
 
 Integration and API documents define externally observable shapes and

docs/release.md

@@ -0,0 +1,164 @@
+# Source Releases
+
+This procedure is for maintainers publishing Notarius source releases. A
+release is an immutable lightweight `vMAJOR.MINOR.PATCH` tag on `main` together
+with its checked-in `docs/releases/<tag>.md` note. Tag CI validates that source
+candidate after publication; it does not publish or repair a release.
+
+Notarius publishes no binaries, archives, checksums, signatures, containers,
+package-manager entries, or Gitea release objects. Windows is not supported.
+Do not create retrospective notes for the pre-procedure `v0.1.0`, `v0.2.0`, or
+`v0.3.0` tags.
+
+## Select And Describe The Release
+
+Choose an unused stable semantic version in the form `vMAJOR.MINOR.PATCH`.
+Prereleases are not supported. Before `v1.0.0`, a minor release may change a
+documented CLI, configuration, durable artifact, integration, or operating
+contract when its note explains the impact and required operator action. A
+patch release must not intentionally break those documented contracts within
+its minor line.
+
+Create the version-matched note as part of the candidate. Every new note uses
+this structure, with concise, truthful content in each section:
+
+```markdown
+# Notarius vMAJOR.MINOR.PATCH
+
+This release ...
+
+## Summary
+
+## Compatibility
+
+## Upgrade
+
+## Changes
+```
+
+The note is a historical summary. Link to current canonical documentation for
+exact behavior, and update that documentation in the candidate rather than
+using the note as a substitute.
+
+## Prepare The Candidate
+
+Set the selected release version and disable Go workspace use for every
+candidate command:
+
+```sh
+RELEASE_VERSION=vMAJOR.MINOR.PATCH
+export RELEASE_VERSION GOWORK=off
+```
+
+Run the shared source-candidate checks from the repository. They cover module
+hygiene, tests, race tests, vet, builds, formatting, whitespace, maintained
+configuration validation, and the Linux and Darwin command-build matrix:
+
+```sh
+./scripts/check-release-source.sh "$RELEASE_VERSION"
+```
+
+Before committing, manually follow every changed local Markdown link and
+review the candidate for unintended files, generated output, credentials, or
+other unrelated changes. Commit the release note and all affected current
+documentation, then run the shared checker against that exact candidate. Push
+the candidate commit to `main` only after it succeeds. Record the exact commit
+only after that push:
+
+```sh
+RELEASE_COMMIT=$(git rev-parse 'HEAD^{commit}')
+export RELEASE_COMMIT
+```
+
+For private-module installation, configure standard `GOPRIVATE` matching this
+module and ordinary Git authentication for the hosting service before running
+the verification below. The exact authentication mechanism belongs to the
+maintainer environment; never record credentials or environment dumps in a
+release note, command history, or repository file.
+
+## Guard And Publish The Tag
+
+Fetch current remote references, then run this guard without editing the
+candidate. It requires `main`, a clean worktree and index, disabled workspace
+use, a stable release version, the recorded and pushed commit, a matching note,
+and unused local and remote tags:
+
+```sh
+git fetch origin main --tags
+
+if ! printf '%s\n' "$RELEASE_VERSION" |
+  grep -E -x 'v(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)' >/dev/null
+then
+  printf '%s\n' "invalid release version: $RELEASE_VERSION" >&2
+  exit 1
+fi
+test "$GOWORK" = off
+test "$(git branch --show-current)" = main
+test -z "$(git status --porcelain)"
+test "$RELEASE_COMMIT" = "$(git rev-parse 'HEAD^{commit}')"
+test "$RELEASE_COMMIT" = "$(git rev-parse 'origin/main^{commit}')"
+test -s "docs/releases/$RELEASE_VERSION.md"
+grep -F -x "# Notarius $RELEASE_VERSION" "docs/releases/$RELEASE_VERSION.md"
+for heading in '## Summary' '## Compatibility' '## Upgrade' '## Changes'; do
+  grep -F -x "$heading" "docs/releases/$RELEASE_VERSION.md"
+done
+if git rev-parse -q --verify "refs/tags/$RELEASE_VERSION" >/dev/null; then
+  printf '%s\n' "local tag already exists: $RELEASE_VERSION" >&2
+  exit 1
+fi
+if git ls-remote --exit-code --tags origin "refs/tags/$RELEASE_VERSION" >/dev/null 2>&1; then
+  printf '%s\n' "remote tag already exists: $RELEASE_VERSION" >&2
+  exit 1
+fi
+```
+
+Create an explicitly lightweight tag against the guarded commit, verify its
+target, and push only that tag ref:
+
+```sh
+git -c tag.gpgSign=false tag "$RELEASE_VERSION" "$RELEASE_COMMIT"
+test "$(git cat-file -t "$RELEASE_VERSION")" = commit
+test "$(git rev-parse "$RELEASE_VERSION^{commit}")" = "$RELEASE_COMMIT"
+git push origin "refs/tags/$RELEASE_VERSION:refs/tags/$RELEASE_VERSION"
+```
+
+Never use `git push --tags`, move a published tag, or delete a published tag.
+
+## Verify The Published Release
+
+Confirm that the remote tag still points at the guarded commit and that the
+note is available from the tagged tree:
+
+```sh
+REMOTE_TAG_COMMIT=$(git ls-remote origin "refs/tags/$RELEASE_VERSION" | awk '{print $1}')
+test "$REMOTE_TAG_COMMIT" = "$RELEASE_COMMIT"
+git show "$RELEASE_VERSION:docs/releases/$RELEASE_VERSION.md" >/dev/null
+```
+
+Verify a fresh source installation and its diagnostic version. The temporary
+directory confines the installed command to this check:
+
+```sh
+release_verification_dir=$(mktemp -d)
+trap 'rm -rf "$release_verification_dir"' 0 HUP INT TERM
+mkdir -p "$release_verification_dir/bin"
+GOWORK=off GOBIN="$release_verification_dir/bin" go install \
+  "gitea.maximumdirect.net/eric/notarius/cmd/notarius@$RELEASE_VERSION"
+test "$("$release_verification_dir/bin/notarius" --version)" = "notarius $RELEASE_VERSION"
+```
+
+An exact fresh checkout and `GOWORK=off go build ./cmd/notarius` is an
+equivalent source verification when local installation policy requires it.
+`notarius --version` is diagnostic only; downstream compatibility remains
+defined by the published receipt and artifact contracts.
+
+## Failure And Correction Policy
+
+If candidate validation fails before publication, fix the candidate on `main`,
+rerun the shared checker, and repeat the guards. An unpublished local tag may
+be deleted after inspection.
+
+If the remote tag or tag CI reveals a defect, leave the published tag intact.
+Fix the defect on `main`, choose a new patch version, write a new matching
+note, and repeat this procedure. Do not weaken tag immutability or add release
+assets as a workaround.

docs/releases/v0.4.0.md

@@ -0,0 +1,83 @@
+# Notarius v0.4.0
+
+This release strengthens LLM reliability and validation throughout the
+configured pipeline, upgrades the PromptKit integration, and establishes the
+source-release and downstream-consumer workflows needed for broader D&D
+pipeline integration.
+
+## Summary
+
+Notarius now distinguishes PromptKit structural-output repair from
+application-owned semantic validation retries. Producer candidates can run
+through complete deterministic validator chains, receive bounded semantic
+correction guidance, and retry under explicit stage policies. Final run
+receipts and manifests preserve bounded validation provenance, while outputs
+that advance with incomplete validation remain available to the current run
+without entering reusable checkpoint state.
+
+The release also adds a maintained complete D&D subprocess-consumer workflow,
+diagnostic build versions, and the source-only release procedure used to
+publish this version.
+
+## Compatibility
+
+- Configuration files must use schema version 4. Version 3 is not decoded or
+  rewritten; rename the top-level `scriptorium` section to `promptkit` when
+  migrating. See [Configuration](../config.md#migrating-version-3-configuration).
+- PromptKit is pinned to v0.9.0. Operator profile files use PromptKit's v0.9.0
+  format and may use its profile-inheritance support. Notarius continues to
+  resolve operator profiles before embedded fallbacks.
+- Structural-output repair and semantic stage retries are separate bounded
+  mechanisms. Maintained production prompts request one structural repair by
+  default; explicit configuration can override the supported repair count.
+- Validation policy can now fail a run, reject an output, or permit an
+  otherwise valid candidate to advance with incomplete-validation provenance.
+  The application defaults are documented in
+  [Configuration](../config.md#pipelines).
+- The `notarius.run-result.v1` receipt remains at schema version 1 and adds
+  optional validation summaries plus a required validation-status field.
+  Consumers of this pre-release contract should follow the current
+  [run-result receipt](../integrations/run-result.md).
+- Existing D&D artifact schema identities remain unchanged. Validation and
+  producer-policy changes can nevertheless cause previously accepted weak
+  candidates to retry, reject, or fail instead.
+
+## Upgrade
+
+1. Migrate every Notarius configuration to version 4 and rename `scriptorium`
+   to `promptkit`.
+2. Review deployed PromptKit profiles against the pinned v0.9.0 profile format
+   and ensure their credential environment variables are available at run
+   time.
+3. Run `notarius config validate --config <path> --pipeline <id>` before the
+   first production invocation.
+4. Review `structured_output_repair_attempts`, producer retry counts, and
+   `validation_policy` wherever the deployment needs behavior different from
+   the documented defaults.
+5. Update subprocess consumers to inspect receipt `validation_status` and to
+   tolerate the optional bounded `validation_summaries` field. A consumer that
+   requires fully validated artifacts should require `approved`.
+
+## Changes
+
+- Upgraded PromptKit from v0.5.0 through v0.9.0 and adopted profile
+  inheritance, structured-output repair, typed error classification, and the
+  correction-aware completion protocol.
+- Added pipeline and binding configuration for structural repair and terminal
+  validation policy, with strict startup validation and effective-setting
+  provenance.
+- Added feedback-aware retries for chunking, extraction, merge, normalize, and
+  semantic reconciliation producers. Retry prompts contain the exact defective
+  response and actionable semantic correction guidance without exposing
+  internal reason codes or opaque entity identifiers.
+- Added complete validator-chain execution, validator retry handling, bounded
+  warnings, terminal dispositions, and durable validation summaries.
+- Prevented validation-incomplete artifacts and all derived lineage from
+  loading or publishing reusable checkpoints while preserving same-run
+  generated-reference handoff.
+- Tightened cached chunk-plan validation so only completely validated plans are
+  reused or replace stored plans.
+- Added a machine-readable subprocess receipt workflow and complete D&D
+  consumer documentation covering all maintained artifacts.
+- Added source-release checks, immutable lightweight-tag guidance, Linux and
+  Darwin build verification, and diagnostic `notarius --version` output.

docs/releases/v0.5.0.md

@@ -0,0 +1,85 @@
+# Notarius v0.5.0
+
+This release separates actionable process warnings from extraction-quality
+advisories and routine normalization observations, giving operators a quiet
+warning channel without discarding durable diagnostic detail.
+
+## Summary
+
+Notarius now carries one validated, origin-aware diagnostic contract from
+producers and validators through retries, reusable state, output publication,
+debug summaries, run receipts, and CLI presentation. Warnings are reserved for
+process degradation or incomplete configured work. Data-quality findings are
+advisories, and successful deterministic cleanup is recorded as observations.
+An ordinary successful run therefore reports zero warnings while retaining
+bounded diagnostic provenance for later review.
+
+The framework aggregates findings deterministically by their stable identity
+and complete pipeline origin, preserves exact occurrence counts, and retains
+bounded representative samples. Warning groups fail rather than truncate;
+advisory and observation representation is bounded with explicit truncation
+metadata and exact unrepresented-occurrence counts.
+
+## Compatibility
+
+- `warnings.json` now uses the incompatible grouped
+  `notarius.warnings.v2` envelope and contains process warnings only. Consumers
+  of the former flat warning payload must migrate to the current
+  [JSON output contract](../integrations/json-output.md).
+- The new `diagnostics.json` file uses `notarius.diagnostics.v1` and contains
+  advisory and observation groups. Production `index.json` files always expose
+  both `warnings_file` and `diagnostics_file`.
+- The machine-readable run receipt is now `notarius.run-result.v2`. It replaces
+  `warning_count` with exact warning group and occurrence counts and adds
+  advisory/observation group, occurrence, and truncation fields. See the
+  current [run-result receipt](../integrations/run-result.md).
+- Custom output modules must return their complete logical file set or an
+  error. The former `OutputResult.Warnings` field has been removed; an output
+  module cannot report a warning after serializing its output.
+- Reusable state now uses `notarius.workspace.v4` and chunk-plan records use
+  `notarius.chunk-plan.v3` so they can preserve structured diagnostics. Older
+  pre-release reusable state is not reused under these contracts; start with
+  clean state when deterministic continuity with an older workspace is not
+  required.
+- Validation acceptance, semantic retry budgets, rejection policy, and D&D
+  artifact schema identities are unchanged by this release.
+
+## Upgrade
+
+1. Update subprocess consumers to require `notarius.run-result.v2` and read
+   `warning_group_count`, `warning_occurrence_count`,
+   `diagnostic_group_count`, `diagnostic_occurrence_count`, and
+   `diagnostics_truncated`.
+2. Update output-bundle consumers to decode `notarius.warnings.v2`, discover
+   `diagnostics.json` through `index.json`, and treat diagnostics as review
+   information rather than process warnings.
+3. Update any custom output module for the removal of
+   `OutputResult.Warnings`; return an error when encoding cannot complete.
+4. Clear pre-release reusable state before the first upgraded production run
+   when deterministic continuity with an older workspace is not required.
+5. Run `notarius config validate --config <path> --pipeline <id>` and perform
+   one representative run before promoting the release in an automated
+   pipeline.
+
+## Changes
+
+- Added validated diagnostic dispositions, categories, origins, stable reason
+  codes, exact occurrence counts, and bounded representative samples.
+- Added deterministic run-level aggregation with separate limits for
+  actionable warning groups and advisory/observation groups.
+- Reclassified D&D source-relatedness and unresolved-identity findings as
+  data-quality advisories and routine normalization changes as observations.
+- Preserved structured diagnostics across producer retries, validation,
+  generated-reference handoff, checkpoints, chunk-plan reuse, and debug
+  summaries while discarding superseded-attempt findings.
+- Added grouped `warnings.json`, a new grouped `diagnostics.json`, and the
+  corresponding production index entries.
+- Upgraded the machine-readable run receipt and human CLI summary to report
+  exact warning and diagnostic counts without allowing advisory volume to
+  create warning output.
+- Removed post-encoding output warnings and hardened diagnostic validation,
+  overflow handling, aggregate memory bounds, and warning-file path
+  presentation.
+- Documented diagnostic ownership, classification, operator interpretation,
+  durable contracts, and architectural invariants in ADR-0015 and the
+  canonical CLI, operations, integration, and internal documentation.

docs/roadmap/archive/warning-signal-and-presentation-audit.md

@@ -0,0 +1,648 @@
+# Warning Signal And Presentation Audit
+
+## Executive Assessment
+
+Notarius warning execution is mechanically stronger than its operator-facing
+presentation. Terminal-attempt promotion, stable ordering after concurrent
+work, checkpoint replay, validation summaries, and debug retention are all
+substantially correct. The audit found no general duplicate-append defect in
+the extract, merge, or normalize handoffs and no leakage of abandoned-attempt
+warnings into a successful result.
+
+The warning channel itself is not coherent. One flat `contracts.Warning` type
+currently represents at least four materially different concepts:
+
+- actionable degradation or incomplete validation;
+- heuristic data-quality doubt;
+- successful but potentially reviewable fallback; and
+- routine canonicalization, ordering, and deduplication observations.
+
+That conflation is the primary reason successful runs produce a count that is
+large but operationally weak. The maintained complete D&D example demonstrates
+the problem without a live provider: an approved run with no rejected outputs
+published 12 warning records, all from three advisory relatedness checks. An
+operator separately reported a successful complete D&D run with 10 outputs,
+one rejection, and 85 warnings. The production bundle for that run was not
+available in this environment, so its reason-code distribution could not be
+measured.
+
+The current implementation also has four correctness or robustness gaps:
+
+1. warning records lose stage, step, lane, module, validator, and chunk
+   provenance when promoted, which makes safe aggregation and diagnosis
+   impossible from `warnings.json` alone;
+2. there is no framework-level validation or aggregate bound, and the NPC- and
+   spell-relatedness validators bypass the D&D warning limiter entirely;
+3. warnings returned by an output encoder are added after `warnings.json` has
+   already been encoded, so the receipt, stderr, debug bundle, and published
+   warning file can disagree; and
+4. a skipped validator contributes to `incomplete` validation but does not
+   receive the warning generated for an exhausted validator failure.
+
+The recommended end state is a structured diagnostic contract with explicit
+disposition, category, origin, occurrence count, and bounded samples. Warnings
+are reserved for process-level degradation or incompleteness. LLM-judged or
+deterministically inferred extraction-quality signals are advisories, never
+warnings, and routine normalization observations remain inspectable without
+being reported as top-level warnings. An ordinary successful run in which all
+configured work completes normally should therefore report zero warnings. This
+is an architectural and durable-contract change, not merely revised CLI prose.
+
+## Evidence And Limits
+
+The audit used:
+
+- a complete static search of production `contracts.Warning` constructors,
+  reason-code constants, result fields, and promotion sites under `internal/`;
+- call-path inspection through producer attempts, validators, lane
+  coordination, chunk-plan reuse, checkpoints, output encoding, debug output,
+  CLI presentation, and run-result construction;
+- the maintained complete and minimal D&D examples with offline fake LLMs;
+- focused deterministic tests for warning bounds, semantic-reconciliation
+  fallback, `warn_continue`, semantic retries, terminal rejection, concurrency
+  ordering, and checkpoint reuse; and
+- the operator-provided observation of an 85-warning complete D&D run.
+
+No provider-backed production run was attempted because this environment has
+no API key. Consequently, the audit can establish warning mechanics, possible
+multiplicity, synthetic volume, and obvious heuristic limitations, but cannot
+estimate production frequency or the real false-positive rate of individual
+D&D advisories. Those measurements are not required to choose the recommended
+architecture; they are required before strengthening any heuristic advisory
+into a rejection or setting a numerical production acceptance target.
+
+## Complete Warning-Producer Inventory
+
+### Framework And Generic Boundaries
+
+| Producer | Reason code | Trigger and consequence | Multiplicity and bound | Current surfaces and coverage |
+| --- | --- | --- | --- | --- |
+| Reference materialization in `internal/framework/pipeline/references.go` | `empty_reference` | A bound external reference is a valid, accepted media type but contains zero bytes. The prompt may receive materially incomplete context. | One per empty bound file; finite by configuration but no shared run-level cap. | Enters `RunInput.Warnings`; reference tests protect contextual scope. |
+| Producer-attempt policy in `internal/framework/pipeline/producer_attempts.go` | `validator_execution_incomplete` | An applicable validator exhausted its execution budget and `warn_continue` accepted the otherwise valid candidate. | One per failed validator on each terminal candidate. An extract chain can multiply this by chunks and lanes. There is no global cap. | Durable warning, receipt count, stderr, debug, and validation summary. `TestWarnContinueRecordsOneWarningForEachExhaustedValidator` covers failures. |
+| Chunk, extract, merge, normalize, and output module result contracts | Module-defined | A module may return arbitrary warnings with its successful candidate. | No contract validation, message limit, per-result cap, or global cap. Current production modules are inventoried below. | Terminal-attempt filtering and concurrency ordering are well tested. |
+| Production JSON output encoder | None | The encoder copies incoming warnings into `warnings.json`; it does not currently create warnings. | Same count as its input. | JSON encoder and assembled-pipeline tests compare the incoming run warnings with the published file. |
+| Output encoder result contract | Module-defined | Any output encoder may return warnings discovered during encoding. | Unbounded by contract. No production encoder currently exercises this capability. | Appended to final `RunOutput` only after logical files were encoded; this is the cross-surface defect described in AUD-WARN-004. |
+
+Input adapters and production mergers do not currently have independent
+warning producers. Chunk-plan and checkpoint decisions are structured manifest
+or debug provenance rather than warnings. Cancellation and hard persistence,
+reference, parsing, serialization, and provider failures remain errors.
+
+### D&D Extraction Gates
+
+| Producer | Reason code | Trigger and consequence | Multiplicity and bound |
+| --- | --- | --- | --- |
+| `dnd/combat-turns` extractor | `scene_classification_unavailable` | The chunk has no exact matching scene-description classification. The extractor returns an empty accepted result and skips the LLM, so combat-turn output may be incomplete. | At most one per chunk for this lane. |
+| `dnd/enemy-events` extractor | `scene_classification_unavailable` | The same missing or mismatched scene gate causes accepted empty enemy-event output. | At most one per chunk for this lane. |
+
+An exact non-combat classification produces an intentional empty result without
+a warning. An exact combat classification proceeds normally. The two producers
+share a code and operator consequence but use different messages; their module
+origins are not retained in the final warning record.
+
+### D&D Source-Relatedness Validators
+
+All ten relatedness validators are deterministic advisories: they approve the
+candidate and warn when contextual prose or an entity name is not lexically
+present in cited text. Shape and source-reference failures are deliberately
+left to blocking validators earlier in the chain. The same relatedness
+validator is registered in both the extract and normalize default chain for
+each artifact family in `internal/modules/dnd/register/chains.go`.
+
+| Artifact family | Warning reason | Per-record trigger | Local bound | Omission reason |
+| --- | --- | --- | --- | --- |
+| Combat turns | `combat_turn_not_near_source` | Actor token sequence absent | 20 per validator invocation | `combat_turn_relatedness_warnings_omitted` |
+| Enemy events | `enemy_event_not_near_source` | Subject token sequence absent | 20 | `enemy_event_relatedness_warnings_omitted` |
+| Item occurrences | `item_occurrence_source_unrelated` | Item name token sequence absent | 20 | `item_occurrence_relatedness_warnings_omitted` |
+| Item registry | `item_not_near_source` | Item name token sequence absent | 20 | `item_relatedness_warnings_omitted` |
+| Location occurrences | `location_occurrence_not_near_source` | Location name token sequence absent | 20 | `location_occurrence_relatedness_warnings_omitted` |
+| Location registry | `location_not_near_source` | Location name token sequence absent | 20 | `location_relatedness_warnings_omitted` |
+| NPC occurrences | `npc_occurrence_not_near_source` | NPC name token sequence absent | 20 | `npc_occurrence_relatedness_warnings_omitted` |
+| NPC registry | `npc_not_near_source` | NPC name token sequence absent | **Unbounded** | None |
+| Scene descriptions | `scene_description_not_near_source` | No significant title or summary token appears; up to two findings per scene | 20 | `scene_description_relatedness_warnings_omitted` |
+| Spells | `spell_not_near_source` | Spell-name token sequence absent | **Unbounded** | None |
+
+The eight limiter-generated omission records are presentation artifacts, not
+new source-relatedness conditions. They occupy a warning slot and make list
+length differ from the actual occurrence count.
+
+### D&D Normalizers
+
+Every production D&D normalizer bounds its returned warning slice to 20 through
+`internal/modules/dnd/shared/diagnostics`, including a final omission record
+when needed. Registry semantic retries reserve space for their fallback
+warning. The following table is complete by semantically distinct condition;
+codes listed together are parallel artifact-family variants.
+
+| Condition | Reason codes | Result impact | Current classification assessment |
+| --- | --- | --- | --- |
+| Display or field whitespace/name canonicalization | `npc_fields_normalized`, `item_fields_normalized`, `location_fields_normalized`, `spell_name_canonicalized`, `combat_actor_canonicalized`, `enemy_event_name_canonicalized`, `item_occurrence_name_canonicalized`, `location_occurrence_name_canonicalized`, `scene_description_prose_normalized` | Deterministic successful mutation. The item-occurrence code can also describe `from`/`to` whitespace, not only the item name. | Routine observation. |
+| Durable ID recomputation | `npc_id_recomputed`, `item_id_recomputed`, `location_id_recomputed` | Restores the deterministic name-derived ID. | Routine observation; invalid identity is separately rejected by default chains. |
+| Source-reference sorting or deduplication | `source_references_normalized` | Sorts and removes exact duplicate references while deliberately preserving invalid references for their validators. | Routine observation. Shared code is useful but ambiguous without producer origin. |
+| Canonical record ordering | `combat_turns_reordered`, `enemy_events_reordered`, `item_occurrences_reordered`, `location_occurrences_reordered`, `npc_occurrences_reordered`, `scene_description_order_normalized` | Deterministic order changes only. | Routine observation. |
+| Exact or approved semantic duplicate consolidation | `duplicate_npc_collapsed`, `duplicate_item_collapsed`, `duplicate_location_collapsed`, `duplicate_spell_cast_collapsed`, `duplicate_combat_turn_collapsed`, `duplicate_enemy_event_collapsed`, `duplicate_item_occurrence_collapsed`, `duplicate_location_occurrence_collapsed`, `duplicate_npc_occurrence_collapsed`, `scene_description_duplicate_collapsed` | Removes duplicate records and preserves or combines canonical evidence according to the artifact policy. Registry codes cover both exact and accepted semantic consolidation. | Durable normalization observation; not normally operator-actionable. |
+| Unresolved external membership | `spell_name_unresolved`, `item_occurrence_unknown_item_id`, `location_occurrence_unknown_location_id` | The value is preserved but is not grounded in the effective catalog or registry. Default chains normally reject the same condition before normalization; it remains reachable with validator overrides or defensive direct use. | Actionable data-quality warning. |
+| Unsafe currency consolidation proposal | `item_semantic_proposal_invalid` | The proposed group is rejected and all records are preserved because denominations or currency/non-currency members are incompatible. The same code is also used internally as a retry reason. | Advisory about model proposal quality; no accepted-data loss. The control and diagnostic meanings should be separated. |
+| Semantic reconciliation unavailable or exhausted | `npc_semantic_reconciliation_exhausted`, `item_semantic_reconciliation_exhausted`, `location_semantic_reconciliation_exhausted` | The safe deterministic result is accepted, but possible semantic duplicates remain. | Actionable fallback warning. |
+| Local warning truncation | `npc_normalization_warnings_omitted`, `item_normalization_warnings_omitted`, `location_normalization_warnings_omitted`, `spell_normalization_warnings_omitted`, `combat_turn_normalization_warnings_omitted`, `enemy_event_normalization_warnings_omitted`, `item_occurrence_normalization_warnings_omitted`, `location_occurrence_normalization_warnings_omitted`, `npc_occurrence_normalization_warnings_omitted`, `scene_description_normalization_warnings_omitted` | Reports that individual records were omitted from presentation. | Group metadata, not an independent warning. |
+
+No production D&D normalization warning exposes raw model responses,
+correction guidance, or provider errors. Most dynamic names are quoted and
+truncated by the shared helper. That local discipline is not enforced by the
+generic warning contract, and the spell relatedness message does not use the
+shared truncation helper.
+
+## Warning Propagation And Surface Map
+
+```text
+external-reference warnings -----------------------------+
+                                                           |
+module candidate warnings -> validation chain warnings     |
+             |                      |                       |
+             +---- producer-attempt terminal policy -------+
+                                  |                        |
+                   accepted / terminal rejection only      |
+                                  |                        |
+              chunk or lane result in canonical order      |
+                                  |                        |
+       checkpoint record/replay and ordered step merge     |
+                                  |                        |
+                            RunOutput.Warnings <------------+
+                                  |
+                   OutputRequest -> output encoder
+                                  |             |
+                         warnings.json     OutputResult.Warnings
+                                                |
+                              appended to final RunOutput only
+                                  |
+                 receipt, stderr, final debug warning summary
+```
+
+### Attempts And Validation
+
+- `runProducerAttempts` promotes only the terminal accepted or terminal
+  rejected candidate's module and completed-validator warnings. Operational,
+  structural, semantic, and module-directed attempts that are superseded are
+  retained in attempt debug artifacts but not in the final collection.
+- A module-directed semantic-reconciliation retry adds its fallback warning
+  only when no retry remains. Earlier attempt warnings are discarded.
+- On `warn_continue`, warnings from the otherwise accepted candidate and
+  completed approved or rejected validators are retained. One fixed,
+  non-sensitive `validator_execution_incomplete` warning is added for every
+  failed validator. Skipped validators affect the validation summary and final
+  `incomplete` status but do not receive such a warning.
+- A semantic terminal rejection retains only warnings from that rejected
+  attempt. Structural rejection after producer failure cannot retain a
+  candidate warning because no valid candidate result exists.
+
+These behaviors are protected by the producer-attempt, extract-handoff,
+rejection-warning, normalize-retry, and attempt-debug tests. They are the right
+foundation for the redesign and should not be replaced with early-exit or
+all-attempt accumulation.
+
+### Concurrency And Ordering
+
+Extract jobs are dispatched chunk-first and lane-second. Results are stored by
+chunk index, finalized in ascending chunk order, and lane continuations are
+merged into a slice indexed by configured lane order. Pipeline steps run in
+configured order. The resulting public order is therefore:
+
+1. pre-run reference warnings;
+2. chunk-stage warnings;
+3. step order;
+4. configured lane order within each step;
+5. chunk order within extract;
+6. merge warnings; then
+7. normalize warnings; followed by any output-result warnings.
+
+`TestRunnerBoundsExtractJobsAndStabilizesReverseCompletion` exercises warning
+order under reversed completion. No completion-order leak was found.
+
+### Chunk Plans And Checkpoints
+
+- A reusable chunk plan stores producer warnings only. Current validators run
+  again, and their current warnings are appended. Warnings from a cached plan
+  candidate that fails current validation are discarded before regeneration.
+- Accepted extract, merge, and normalize checkpoints store the terminal
+  warnings for their stage. Reuse loads and appends those warnings once at the
+  same logical handoff. Tests compare fresh and resumed warning collections and
+  preserve their order.
+- Validation-incomplete accepted outputs are not reusable, preventing a later
+  run from silently treating incomplete validation as complete.
+- Required accepted-normalize hydration replays that normalize checkpoint's
+  warnings; checkpoint decisions separately expose that reuse occurred.
+
+The recommended redesign should keep fresh and resumed logical diagnostics
+equivalent. Whether a result was reused belongs in checkpoint provenance, not
+in the diagnostic grouping key; adding a `reused` distinction would fragment
+groups and make equivalent runs present differently.
+
+### Terminal Surfaces
+
+| Surface | Current content | Audience | Audit result |
+| --- | --- | --- | --- |
+| `RunOutput.Warnings` | Flat final slice | Framework and CLI | Canonical in-memory list, but lacks origin and bounds. |
+| Published `warnings.json` | Object containing the warnings passed into the output encoder | Durable consumers | Exact for the production JSON encoder unless the encoder itself returns warnings. |
+| `index.json` | Path to `warnings.json` | Durable consumers | Stable discovery path; no separate diagnostic-detail path. |
+| Run-result v1 | `warning_count = len(final RunOutput.Warnings)` | Subprocess callers | Count only; no group/occurrence distinction. |
+| Human stderr | `run completed with N warning(s)` | Operators | Count only and no direct detail path. Successful exit remains zero. |
+| Manifest | Validation and rejection summaries, no warning collection | Durable provenance | Correctly avoids duplicating the flat list. |
+| Debug summary `warnings.json` | Raw final warning array | Operators/developers | Includes final output-result warnings and can therefore differ from published `warnings.json`. |
+| Debug run report | Final warning count | Operators/developers | Same final slice length as receipt and stderr. |
+| Attempt/stage debug | Candidate-local warning detail and origin in path/envelope | Forensics | Sufficient to diagnose provenance, but debug capture is optional and is not a durable consumer contract. |
+
+## Empirical Measurements
+
+### Offline And Synthetic Runs
+
+| Scenario | Result | What it establishes |
+| --- | --- | --- |
+| Maintained complete D&D config and transcript with the repository's offline fake LLM | Approved, 10 normalized outputs, 0 rejected outputs, 12 warnings; receipt, stderr, and published file all reported 12 | An ordinary structurally successful workflow can be noisy without fallback or incomplete validation. |
+| Same complete run, grouped after publication | Three reason codes, seven exact `(reason, scope, message)` tuples, maximum exact-tuple repetition of three | The flat count materially overstates distinct operator conditions. Scope resets within chunks and does not identify origin. |
+| Maintained focused scene-description workflow | Approved, one normalized output, 0 warnings | The warning channel can be quiet when synthetic model text is lexically grounded. |
+| Generic warning publication contract | One warning reaches successful stderr, durable output, and debug summary | The ordinary pre-output path is consistent. |
+| NPC semantic-reconciliation candidate-limit fallback | No LLM call, all records preserved, one exhaustion warning, total warnings no greater than 20 | Fallback is bounded and materially different from routine normalization. |
+| `warn_continue` with two failed validators and one skipped validator | Validation status contains all three incomplete validators; warning slice contains two execution-incomplete records | Current warning count does not describe all incomplete validation. |
+| Retrying extract candidate | Two producer attempts; only the accepted attempt's one warning is final | Retry does not amplify abandoned warnings. |
+| Terminal semantic rejection | Only the final rejected attempt's operation and validator warnings are final | Rejection diagnostics are retained without retaining superseded warnings. |
+| Fresh versus reused extract checkpoint | Warning collections are deeply equal | Checkpoint replay does not itself amplify warnings. |
+| Spell normalizer with 21 unresolved entries | 20 records: 19 samples plus one omission record saying two additional warnings were omitted | `warning_count` is neither exact occurrence count nor distinct-condition count. |
+
+The focused audit tests passed in `internal/cli`,
+`internal/framework/pipeline`, the NPC-registry and spell normalizers, and all
+D&D packages.
+
+### Bounded Sample Review
+
+The complete offline D&D run produced:
+
+| Reason | Count | Sample | Review |
+| --- | ---: | --- | --- |
+| `location_not_near_source` | 4 | `Moon Gate` was absent from cited text | Correctly identifies deliberately unsupported fake output. Three records shared the same exact tuple because chunk and stage origin were lost. |
+| `location_occurrence_not_near_source` | 4 | A `Moon Gate` visit was absent from cited text | Correctly identifies the same unsupported registry-driven occurrence, but repeats the same operator concern across extraction and normalization. |
+| `scene_description_not_near_source` | 4 | A title or summary had no significant exact token in cited text | Mixed value. Generic `session scene` prose is ungrounded, while `Arrival` versus transcript `arrive` illustrates an expected lexical false positive. |
+
+This fake workflow is an integration fixture, not a model-quality benchmark.
+It nonetheless proves that the checks carry useful evidence while being too
+imprecise and repetitive to serve as one-warning-per-record operator alerts.
+
+### Production Evidence Still Needed
+
+The reported 85-warning run establishes that high volume occurs in practice,
+but the following remain unknown:
+
+- dominant production reason codes and stage/lane sources;
+- unique group count versus repeated occurrence count;
+- false-positive rate for each relatedness family;
+- how much volume comes from normalization observations versus advisories;
+- whether fresh and resumed production runs remain equivalent; and
+- a defensible numerical acceptance target.
+
+If further data is worthwhile, the operator can supply the v1 receipt,
+`warnings.json`, and manifest validation summaries without supplying transcript
+or lane artifacts. An initial privacy-preserving report should group by reason
+code and normalized scope family, count exact repeated tuples, and omit message
+text. Reviewing heuristic precision requires a separately approved bounded
+sample with its cited source context.
+
+## Classification Of Current Conditions
+
+| Target disposition | Current families | Result impact | Operator action | Durable placement |
+| --- | --- | --- | --- | --- |
+| **Warning** | Empty reference; validator failure or skip accepted under `warn_continue`; unavailable required scene classification; exhausted semantic reconciliation | A configured process completed under an allowed degraded or incomplete policy rather than completing normally | Correct reference/configuration, inspect provider/validator, or rerun | Actionable `warnings.json`, receipt summary, stderr summary, debug |
+| **Advisory** | Source-relatedness heuristics; unresolved spell or registry membership; guarded invalid semantic proposal; any future LLM-judged uncertainty or extraction-quality signal | Uncertain data quality or poor model proposal, but accepted data is structurally valid and deterministic guards prevented unsafe mutation | Optional model/source review; no routine action for every record | Durable diagnostic detail and debug; never a top-level warning |
+| **Observation** | Whitespace/name/ID/source-reference canonicalization; canonical ordering; exact and approved semantic duplicate consolidation | Successful intended normalization | None under normal operation | Durable bounded normalization diagnostics or debug; no stderr warning |
+| **Not a diagnostic** | Rejection, invalid structure, cancellation, persistence error, provider failure under fail-run policy | Candidate or run did not complete according to policy | Inspect rejection/error and retry or correct input/configuration | Existing rejection, validation summary, error, and debug contracts |
+
+Exact and semantic duplicate consolidation should remain distinguishable in
+category or reason metadata even though both are observations. Semantic
+reconciliation exhaustion remains a warning because a capability was not
+applied; successful approved consolidation is an observation because it is the
+normalizer's intended work.
+
+## Ranked Findings
+
+### AUD-WARN-001 — The Flat Warning Type Destroys Signal Quality
+
+- **Priority:** High operator impact; high implementation leverage.
+- **Evidence:** `contracts.Warning` has only scope, reason, and message. Routine
+  normalizer changes, heuristic doubt, fallback, and incomplete validation all
+  enter the same slice and the same CLI count. The offline complete run's 12
+  records were all advisories; the operator observed 85 records in a successful
+  real run.
+- **Impact:** Operators cannot tell whether a warning requires a rerun, a
+  configuration repair, optional review, or no action. Repeated routine output
+  trains them to ignore the channel.
+- **Recommendation:** Replace the flat semantic contract with explicit
+  `warning`, `advisory`, and `observation` dispositions plus a small category
+  vocabulary. Do not infer disposition from message text or require every
+  downstream consumer to maintain a reason-code policy table.
+
+### AUD-WARN-002 — Warning Records Lose The Origin Needed For Diagnosis And Aggregation
+
+- **Priority:** High correctness and usability impact.
+- **Evidence:** The runner knows stage, step, lane, module, validator, chunk ID,
+  and chunk index at promotion time, but `terminalWarnings` flattens module and
+  validator records into `[]contracts.Warning`. Per-chunk scopes such as
+  `locations[0]` and `occurrences[0]` then repeat without identifying their
+  chunk or producer. `source_references_normalized` is intentionally shared
+  across families and is therefore especially ambiguous.
+- **Impact:** `warnings.json` cannot answer which stage or module produced a
+  record. Message- or scope-based deduplication would merge unrelated findings
+  or retain accidental duplicates.
+- **Recommendation:** Keep module findings free of framework context, then have
+  the framework add a structured origin envelope before promotion. Validator
+  findings must retain validator identity instead of passing through
+  `validationReport.Warnings()` as a flat slice.
+
+### AUD-WARN-003 — Warning Volume Is Not End-To-End Bounded Or Validated
+
+- **Priority:** High robustness impact; medium immediate likelihood.
+- **Evidence:** D&D's `LimitWarnings` caps most individual producers at 20, but
+  NPC- and spell-relatedness return one warning per record without the helper.
+  Every extract validator is invoked per chunk, all ten relatedness checks run
+  again after normalization, and there is no run-level collector. The generic
+  contract validates neither disposition nor reason/message size, UTF-8,
+  blankness, or total records.
+- **Impact:** Warning memory and output grow with chunks, lanes, configured
+  validators, and record counts. Local omission records lose exact occurrence
+  semantics while still incrementing `warning_count`.
+- **Recommendation:** Add a generic bounded diagnostic collector that preserves
+  exact occurrence counts and bounded samples. Validate all diagnostic fields
+  at the module/framework boundary. Immediately bring NPC and spell
+  relatedness under the existing cap if the full redesign is staged.
+
+### AUD-WARN-004 — Output Encoder Warnings Make Durable Surfaces Disagree
+
+- **Priority:** Medium current impact; high contract correctness risk.
+- **Evidence:** `Runner.Run` passes existing warnings to `encoder.Encode`, then
+  the production encoder serializes `warnings.json`. Only after encoding does
+  the runner append `OutputResult.Warnings`. The receipt, stderr, debug summary,
+  and debug run report see the final slice; the already-created published file
+  cannot. No production encoder currently returns a warning, so ordinary JSON
+  runs do not trigger the defect.
+- **Impact:** A valid output-module implementation can violate the documented
+  claim that `warning_count` describes the published warning collection.
+- **Recommendation:** Remove successful output warnings from the output-module
+  contract unless a demonstrated use case requires them; encoding failures
+  should be errors and optional encoder observations should be debug data. A
+  two-phase finalize API is the viable but more complex alternative.
+
+### AUD-WARN-005 — Validation Skips Are Incomplete But Not Warned
+
+- **Priority:** Medium operator/correctness impact.
+- **Evidence:** `firstIncompleteValidation` treats failed and skipped validators
+  alike, and validation summaries include both. `incompleteValidationWarnings`
+  emits records only for `validationFailed`. The focused test demonstrates
+  three incomplete validators but two warnings.
+- **Impact:** A successful run can have `validation_status: incomplete` while
+  its warning count understates or even omits the affected validators. A caller
+  that checks only warnings receives a weaker signal than the manifest and
+  receipt status.
+- **Recommendation:** Produce one aggregated incomplete-validation warning
+  group whose occurrences cover both failure and skip, while retaining typed
+  outcome and safe reason metadata in the validation summary. Do not expose
+  provider errors or arbitrary skip prose in model or operator messages.
+
+### AUD-WARN-006 — Relatedness Checks Are Useful But Repetitive And Lexically Weak
+
+- **Priority:** Medium operator impact; low acceptance-policy urgency.
+- **Evidence:** Every family runs the advisory in both extract and normalize
+  chains. The complete fixture contains exact repeated tuples, and the checks
+  rely on exact normalized token sequences or a minimal significant-token
+  overlap. Reason naming drifts between `*_not_near_source` and
+  `*_source_unrelated`.
+- **Impact:** The checks can catch unsupported entities, but aliases, pronouns,
+  inflection, and generic scene prose create predictable false positives. Flat
+  per-record presentation magnifies them.
+- **Recommendation:** Retain the validators and their stage-local execution,
+  but classify and aggregate them as advisories. Normalize reason-code naming
+  when the diagnostic contract changes. Do not strengthen them into rejection
+  rules without a human-reviewed production evaluation.
+
+### AUD-WARN-007 — `warning_count` Has No Stable Operational Meaning
+
+- **Priority:** High downstream-contract impact.
+- **Evidence:** The receipt and CLI use `len(output.Warnings)`. One list element
+  can be an omission summary representing several hidden occurrences; repeated
+  records can represent the same condition; and skipped validators can be
+  absent. A 21-occurrence spell test produces a list length of 20.
+- **Impact:** The value is neither an exact occurrence count nor a distinct
+  warning-group count. Consumers cannot set policy or present a trustworthy
+  summary from it.
+- **Recommendation:** Introduce explicit warning-group and warning-occurrence
+  counts in a versioned receipt. Do not silently redefine the v1 field.
+
+## Recommended Target Contract And Presentation Model
+
+### Diagnostic Model
+
+Use one validated internal diagnostic model with these concepts:
+
+- **disposition:** `warning`, `advisory`, or `observation`;
+- **category:** a small enum such as `configuration`, `degradation`,
+  `validation_incomplete`, `data_quality`, `fallback`, or `normalization`;
+- **reason code:** stable semantic identity owned by the producer;
+- **origin:** framework-added phase/stage, step ID, lane ID, module key,
+  validator name, chunk ID, and chunk index when applicable;
+- **occurrence count:** exact number of matching findings;
+- **samples:** a small deterministic list of bounded scope/message pairs; and
+- **omitted sample count:** `occurrence_count - len(samples)`, represented as
+  metadata rather than another diagnostic record.
+
+Errors and rejected outputs must not become diagnostic dispositions. A warning
+means that the run completed under policy despite a process-level degradation
+or incomplete configured operation. Advisory and observation dispositions can
+describe accepted artifact quality and transformation provenance, but no
+LLM-judged extraction-quality signal may be promoted to a warning. Validation
+status remains authoritative for approval, rejection, and incomplete
+validation.
+
+### Aggregation
+
+The framework runner should own aggregation after it enriches findings with
+origin and before public output construction. Modules and validators retain
+semantic ownership of disposition, category, reason, scope, and message; they
+must not own CLI or file presentation.
+
+The default stable key should be:
+
+```text
+disposition + category + reason_code
++ phase/stage + step_id + lane_id + module_key + validator_name
+```
+
+Chunk, record scope, and message text belong in samples and must not be part of
+the group key. This groups repeated per-chunk findings without merging the same
+code across distinct producers or pipeline locations. Group order should be
+the first occurrence in the runner's existing canonical order; sample order
+should follow the same order. A final canonical sort by the complete origin key
+is also viable, but completion timing must never choose either order.
+
+Aggregation must be incremental and bounded. Producers should use a shared
+collector that counts every occurrence while retaining only bounded samples;
+the framework then merges producer groups without reconstructing counts from
+omission prose. A global maximum group count is also required, with overflow
+represented by structured aggregate metadata and with actionable groups given
+priority over lower dispositions.
+
+### Durable Files
+
+Keep one canonical home for each class:
+
+- `warnings.json` should contain versioned, grouped actionable warnings only;
+- a new `diagnostics.json` should contain versioned advisory and observation
+  groups only, avoiding duplication of warning groups;
+- `index.json` should link both files;
+- `rejected.json` and manifest validation summaries should retain their current
+  separate responsibilities; and
+- debug bundles should retain candidate-attempt detail plus the final grouped
+  projections.
+
+This is preferable to keeping all detail in `warnings.json` and filtering only
+the CLI: downstream consumers would otherwise continue to receive a semantically
+mixed warning contract, and routine observations would still dominate the
+durable file.
+
+### CLI And Receipt
+
+For a successful human run with actionable warnings, print a concise summary
+such as:
+
+```text
+notarius: run completed with 2 warning groups (7 occurrences); details=/.../warnings.json
+```
+
+Advisories and observations should not produce the warning line. Their durable
+path remains discoverable through `index.json`; a concise non-warning count can
+be added to the ordinary success line only if operator testing shows value. An
+ordinary successful run with no process degradation should write nothing to
+the warning stream even when it publishes quality advisories or normalization
+observations.
+
+Create `notarius.run-result.v2` rather than redefining v1. It should expose at
+least:
+
+- `warning_group_count`;
+- `warning_occurrence_count`; and
+- `diagnostic_group_count` for non-warning durable groups.
+
+The receipt should continue to expose validation status, validation summaries,
+and rejected-output count independently. Process exit behavior should not
+change as part of warning presentation reform.
+
+### Checkpoint Semantics
+
+Store the structured terminal diagnostic groups with accepted checkpoints and
+replay them exactly once at their logical stage. Fresh and reused runs should
+produce the same public groups and counts. Checkpoint events and debug records,
+not diagnostic identity, should disclose whether computation was reused.
+
+### Output Encoder Boundary
+
+Prefer removing `OutputResult.Warnings`. A successful output encoder should
+either return the complete logical files or fail. If future encoders genuinely
+need to produce durable post-encoding warnings, introduce an explicit
+two-phase prepare/finalize contract so those warnings can be included in the
+same published collection. Do not retain the current self-inconsistent
+one-pass capability.
+
+## Resolution Of Required Design Questions
+
+| Question | Recommendation | Viable alternative and tradeoff |
+| --- | --- | --- |
+| Explicit severity/disposition or external reason mapping? | Put validated disposition and category in the contract. | A central reason-code registry avoids payload fields but makes new modules depend on a second synchronized policy table and leaves downstream meaning implicit. |
+| Keep all detail in `warnings.json` or separate it? | Separate grouped actionable warnings from grouped advisories/observations in `diagnostics.json`. | Keep the flat durable list and aggregate only CLI output; simpler migration, but it preserves the noisy downstream contract and ambiguous count. |
+| Who owns aggregation? | Framework runner/coordinator after origin enrichment. | Output module aggregation keeps framework types smaller but duplicates policy across encoders and cannot repair missing validator origin. |
+| Stable aggregation key? | Disposition, category, reason code, and full producer origin; exclude chunk/scope/message. | Explicit producer-supplied grouping keys offer flexibility but add another identity that can drift from reason codes. Message-template grouping is brittle and unsafe. |
+| Samples and omissions? | Exact occurrence count plus deterministic bounded samples and numeric omitted-sample count. | Omission warning records preserve the current representation but inflate group counts and require prose parsing. |
+| `warning_count` semantics? | Version receipt and replace ambiguity with group and occurrence counts. | Keep v1 count as published record length and add optional fields; compatible, but two competing warning counts remain easy to misuse. |
+| Which normalization changes remain warnings? | Only exhausted process fallback. Unresolved membership is a data-quality advisory; successful canonicalization, reordering, ID repair, source-ref dedupe, and duplicate consolidation are observations. | Treat unresolved membership or semantic duplicate consolidation as warnings because they affect grounding or cardinality; more conservative, but it violates the process-only warning rule and reports accepted artifact quality as an operational failure. |
+| Source-relatedness disposition? | Grouped advisory by default; preserve current approve behavior. | Retain warning disposition or make rejection configurable. Rejection requires production precision evidence; current lexical rules are not strong enough. |
+| Checkpoint-loaded warnings? | Present the same logical groups as fresh execution and use checkpoint events for reuse provenance. | Mark groups as replayed; aids forensics but fragments aggregation and makes semantically equivalent runs differ. |
+| ADR and schema versions? | Add an ADR and version the run receipt and diagnostic files. | Treat the work as CLI-only presentation and avoid an ADR; insufficient because module contracts, output files, checkpoint payloads, and downstream fields change. |
+
+## Compatibility, Documentation, And ADR Implications
+
+The target alters public and internal contracts enough to require a new ADR.
+It should record:
+
+- the distinction among warnings, advisories, observations, rejections, and
+  errors;
+- the invariant that warnings are process-level signals, LLM-judged extraction
+  quality is never a warning, and ordinary non-degraded success has zero
+  warnings;
+- module semantic ownership versus framework origin/aggregation ownership;
+- bounded group and sample semantics;
+- fresh/checkpoint equivalence; and
+- the output-encoder decision.
+
+Implementation should introduce `notarius.run-result.v2`. The grouped warning
+and diagnostic envelopes should each carry their own schema version. Because
+the content of `warnings.json` changes incompatibly from a flat array wrapper
+to groups, release notes and the published JSON integration contract must call
+out the migration. `index.json` gains the diagnostic file path.
+
+Canonical documentation updates belong in:
+
+- `docs/cli.md` for stderr presentation only;
+- `docs/operations.md` for operator review and debug workflow;
+- `docs/integrations/json-output.md` for warning and diagnostic file schemas;
+- `docs/integrations/run-result.md` for v2 fields and compatibility;
+- `docs/consumers/subprocess.md` and `docs/consumers/dnd-pipeline.md` for
+  downstream policy checks;
+- `docs/internal/pipeline.md` for promotion, aggregation, retry, and checkpoint
+  mechanics;
+- `docs/internal/modules.md` and `docs/internal/dnd.md` for producer rules and
+  the D&D classification matrix; and
+- `docs/policy/architecture.md` for the durable ownership invariant after the
+  ADR is accepted and implemented.
+
+No configuration knob is required for the first implementation. A fixed,
+well-documented taxonomy is easier to reason about than per-reason display
+overrides. Configurable escalation or suppression can be considered only after
+production review demonstrates a concrete operator need.
+
+## Test Coverage Assessment
+
+Existing coverage worth preserving includes:
+
+- accepted-attempt and terminal-rejection warning promotion;
+- validator failure retry exhaustion and `warn_continue`;
+- module semantic retry fallback;
+- deterministic warning order under concurrent lane completion;
+- chunk-plan invalidation and discarded-cache warning behavior;
+- fresh/checkpoint warning equivalence;
+- local D&D warning caps and safe dynamic-message quoting;
+- JSON warning-file publication; and
+- CLI stderr, debug, and receipt counts.
+
+Material gaps are:
+
+- no bound test for NPC- or spell-relatedness warnings;
+- no generic warning-field or result-size validation;
+- no test for output encoder warnings versus published `warnings.json`;
+- no operator-level aggregation or bounded-sample tests;
+- no fresh/resume test for grouped counts because groups do not yet exist; and
+- no production evaluation of advisory precision.
+
+Tests should protect the semantic relationships: exact occurrence counts,
+bounded samples, deterministic group order, actionable-only warning
+presentation, and cross-surface equality. They should not assert one exact
+warning count for every complete D&D run or treat message wording as a public
+API unless the wording itself enforces a security boundary.
+
+## Audit Conclusion
+
+The application is in a good position for warning reform. Its retry,
+validation, checkpoint, and concurrency mechanics provide reliable points at
+which to attach structured diagnostics. The most valuable change is not to
+suppress individual reason codes; it is to replace the semantically flat,
+origin-free collection with bounded typed groups and to reserve the word
+“warning” for conditions that merit operator attention.
+
+Provider-backed runs would improve prioritization and help tune the D&D
+advisories, but they are not necessary to conclude that routine normalization
+and heuristic doubt should not dominate stderr or the durable warning
+contract. They should be gathered before changing heuristic acceptance policy
+or adopting a numerical production warning-volume target.

docs/roadmap/audit-plan.md

@@ -1,516 +0,0 @@
-# Codebase Audit Plan
-
-## Purpose
-
-This document defines a repository-wide audit of Notarius for correctness,
-efficiency, maintainability, and clarity. The audit should identify concrete
-improvements without treating abstraction, fewer lines, or higher test coverage
-as goals in themselves.
-
-The audit is intentionally separate from implementation. Its findings should
-be evidence-backed and sufficiently specific to support a later remediation
-roadmap, but the audit should not modify production code, tests, assets, or
-current-behavior documentation.
-
-## Governing Principles
-
-The audit must preserve the architecture and testing policies in
-`docs/policy/architecture.md` and `docs/policy/testing.md`.
-
-In particular:
-
-- Notarius remains a fixed, staged pipeline rather than a general workflow
-  engine.
-- Generic framework packages must remain domain-neutral, and production
-  modules must not acquire CLI or physical-state responsibilities.
-- Typed artifact boundaries, exact codec compatibility, deterministic ordering,
-  whole-output validation, and generated-reference provenance are correctness
-  properties, not incidental complexity to be optimized away.
-- The root `assets` package remains a content-only dependency leaf.
-- Shared helpers should protect demonstrated common semantics. Similar-looking
-  code with different ownership, error policy, identity rules, or type contracts
-  should remain separate.
-- Tests should protect durable behavior and meaningful risks. The audit should
-  not recommend tests merely to increase coverage or freeze implementation
-  details.
-- Efficiency claims must distinguish measured or structurally credible costs
-  from cosmetic line-count reductions. Optimizing local CPU work that is
-  insignificant beside an LLM call is low priority unless it also simplifies
-  correctness or applies to large inputs.
-
-## Audit Questions
-
-Every audited area should be examined through the following questions.
-
-### Correctness
-
-- Are documented architecture invariants enforced at the correct boundary?
-- Can invalid configuration, incompatible artifact types, malformed references,
-  or unavailable dependencies reach execution when they could be rejected
-  during resolution or preparation?
-- Are nil, empty, absent, rejected, failed, and canceled states distinguished
-  consistently?
-- Are stored or returned slices, maps, byte slices, options, metadata, source
-  documents, references, and artifacts defensively owned where required?
-- Are public ordering, selected errors, warnings, and checkpoint decisions
-  deterministic regardless of map or goroutine completion order?
-- Do cancellation, retry, validation, and partial-work semantics match their
-  documented ownership?
-- Do checkpoint and chunk-plan identities include every semantic dependency and
-  exclude scheduling-only or diagnostic state?
-- Can auxiliary references accidentally become source evidence, or can
-  generated references bypass codec, schema, provenance, or step-order checks?
-- Can provider-specific values, credentials, or source content escape through
-  errors, manifests, debug summaries, cache state, or logs?
-- Do schemas, codecs, candidate decoders, normalizers, and validators agree on
-  the exact durable contract without silently accepting incompatible shapes?
-
-### Duplication And Shared Mechanics
-
-- Which exact or near-duplicate implementations express the same invariant and
-  failure policy?
-- Has duplicated code already drifted in naming, nil handling, canonicalization,
-  metadata, fingerprints, validation, or diagnostics?
-- Would a helper have a natural owner and a smaller, clearer contract than the
-  duplicated callers?
-- Can an extraction preserve static typing and package ownership, or would it
-  require reflection, `any`, callbacks with many policy parameters, or a
-  domain-neutral package importing domain concepts?
-- Is repeated code required by a small interface adapter or typed registration
-  boundary and therefore clearer when left explicit?
-
-As a default heuristic, prioritize a shared helper when identical semantics
-appear in three or more production sites, or in two sites where divergence
-would create a meaningful correctness risk. Do not use that heuristic as a
-quota: one substantial duplicate may warrant extraction, while widespread
-one-line interface methods may not.
-
-### Simplicity And Idiomatic Go
-
-- Does a function combine orchestration, policy, transformation, persistence,
-  and reporting that could be separated along existing ownership boundaries?
-- Are repeated scans, sorts, conversions, clones, encodes, or decodes doing work
-  that can safely occur once?
-- Are intermediate representations necessary, or can a value be validated,
-  canonicalized, and mapped in one comprehensible pass?
-- Are maps, sets, stable sorts, generics, standard-library helpers, and error
-  wrapping used idiomatically?
-- Are abstractions earning their complexity, or are interfaces, option layers,
-  wrappers, aliases, compatibility paths, and private types left over after a
-  completed migration?
-- Are there unreachable error branches, redundant fingerprints or digests,
-  duplicated sources of truth, or accessors used only by tests?
-- Can a smaller implementation preserve exact observable behavior and safety
-  properties?
-
-### Explanatory Comments
-
-Comments should be recommended where the code is necessarily complex because
-it preserves a non-obvious invariant. Good candidates include:
-
-- concurrency coordination, cancellation, and stable error selection;
-- checkpoint identity, reuse, forced recomputation, and dependency invalidation;
-- typed erasure and restoration at framework boundaries;
-- generated-reference ordering and provenance;
-- canonicalization and identity resolution where registry evidence differs
-  from occurrence evidence;
-- prompt ordering or input identity required for backend caching; and
-- path confinement, atomic publication, redaction, or terminal error precedence.
-
-Recommend comments that explain *why* a step or ordering constraint exists and
-what would break if it changed. Do not recommend comments that narrate syntax,
-repeat a function name, duplicate current-behavior documentation, or preserve
-implementation history.
-
-### Tests
-
-- Is each consequential invariant protected at the narrowest stable boundary?
-- Are concurrency, cancellation, retries, recovery, compatibility, path safety,
-  and data-integrity behavior credibly exercised?
-- Do higher-level contract tests duplicate lower-level cases without adding
-  integration confidence?
-- Are tests coupled to private constants, helper shape, exact prose, full error
-  strings, or collaborator choreography rather than behavior?
-- Can repetitive fixtures or fakes be simplified without creating a test
-  framework more complex than the tests?
-- Would a focused fuzz test, race test, or package-level invariant test protect
-  a realistic risk better than several example tests?
-
-## Evidence And Finding Standards
-
-Static metrics and textual similarity are discovery aids, not findings. A long
-function may be a clear linear coordinator; identical methods may be useful
-typed adapters. Every reported finding must include:
-
-1. a concise title and severity;
-2. exact files and symbols;
-3. the observed behavior or structural evidence;
-4. the correctness, efficiency, maintenance, or comprehension impact;
-5. a concrete recommended direction;
-6. important invariants the remediation must preserve;
-7. focused validation that would demonstrate success; and
-8. whether the recommendation is independent or should be grouped with another
-   finding.
-
-Use these severities:
-
-- **High:** a credible risk of corrupt output, unsafe state handling, secret
-  exposure, stale reuse, deadlock, nondeterminism, or violated external
-  contract.
-- **Medium:** a plausible behavioral defect, meaningful wasted work on common
-  paths, or complexity/duplication likely to cause future correctness drift.
-- **Low:** a contained simplification, small efficiency improvement, dead code,
-  naming issue, or missing explanation with no current behavioral failure.
-
-The audit should explicitly record examined areas with no findings. This makes
-coverage visible and prevents later agents from repeatedly rediscovering the
-same safe design.
-
-## Repository Areas
-
-### 1. Architecture And Dependency Boundaries
-
-Inspect `docs/policy/architecture.md`, `docs/adr/`, `docs/internal/overview.md`,
-package imports, module registrars, and the CLI composition root.
-
-Look for:
-
-- framework or core code depending on production modules;
-- modules depending on CLI, physical roots, or provider-specific types;
-- domain knowledge placed in generic helpers;
-- duplicated registries or composition policy outside the owning registrar;
-- abstractions that turn the fixed pipeline into an implicit general graph; and
-- current code that no longer matches an accepted ADR or documented invariant.
-
-Graph-reported cross-layer calls must be traced before being classified because
-tests and interface implementations can resemble dependency inversions without
-creating a production import violation.
-
-### 2. Configuration And CLI Composition
-
-Inspect `internal/core/config`, `internal/cli`, configuration parsing and
-redaction tests, profile construction, session derivation, catalog assembly,
-reference overrides, run-result handling, terminal reporting, and maintained
-example contract tests.
-
-Pay particular attention to the currently dense paths around
-`runPipelineCommand`, configuration profile validation, selected reference
-targets, recomputation policy, and option normalization. Determine whether
-their complexity reflects necessary composition or mixed responsibilities that
-can be separated without moving policy into the framework.
-
-Verify:
-
-- file, environment, CLI, pipeline, binding, and prompt-default precedence;
-- consistent strict option and unknown-field handling;
-- session identity independence from references and pipeline-local changes;
-- effective profile and runtime fingerprint consistency;
-- redaction before errors or debug/manifest boundaries;
-- output publication only after framework success; and
-- one guarded terminalization path that preserves the primary failure.
-
-### 3. Pipeline Resolution, Preparation, And Typed Registries
-
-Inspect `internal/framework/pipeline/profile.go`, `prepare.go`, registry files,
-`options.go`, `references.go`, `handoff.go`, `construction.go`, typed contracts,
-and their focused tests.
-
-This area deserves a dedicated pass because the current graph identifies
-`ResolvePipeline`, generated-binding validation, reference-target resolution,
-and generated-reference construction as high-complexity or high-fan-in code.
-
-Verify:
-
-- static failures occur before source parsing;
-- selected and unselected lanes do not contaminate each other's requirements;
-- stage defaults and overrides have one canonical resolution path;
-- typed registration and private erasure cannot panic or accept near-matching
-  artifact types;
-- generated bindings reject cycles, forward references, ambiguity, wrong kinds,
-  and missing accepted normalized producers;
-- materialized reference bytes and options are cloned and bounded; and
-- resolved composition and prepared fingerprints include the complete semantic
-  policy exactly once.
-
-Compare input, chunker, extractor, merger, normalizer, output, validator, codec,
-evidence-projector, and validator-chain registries for shared mechanics and
-intentional differences. Repeated typed registration code is a candidate only
-if a helper can retain useful compile-time guarantees and stage-specific
-diagnostics.
-
-### 4. Pipeline Execution, Validation, Retry, And Concurrency
-
-Inspect `runner*.go`, `typed_execution.go`, `runner_typed.go`,
-`runner_concurrent.go`, validation-chain execution, normalize retry behavior,
-synchronized collaborators, and the concurrency, cancellation, retry, debug,
-and checkpoint tests.
-
-Trace complete paths rather than reviewing helper files in isolation:
-
-- source and chunk-plan selection through chunk validation;
-- deterministic chunk-first/lane-second dispatch;
-- lane extraction through merge and normalize continuations;
-- rejection versus framework-error propagation;
-- cancellation before dispatch, while queued, and while running;
-- retry attempts and warning retention;
-- stable error selection after concurrent completion;
-- checkpoint hydration back into typed execution; and
-- output suppression after a framework error.
-
-Look for goroutine leaks, unbounded work, lock-order risks, double release or
-double recording, races on shared result state, unnecessary serialization,
-and repeated canonicalization. Comments are especially valuable here when they
-explain ordering or cancellation invariants that are not apparent from local
-control flow.
-
-### 5. State, Checkpoints, Chunk Plans, Debugging, And File Safety
-
-Inspect `internal/framework/checkpoint`, `chunkplan`, `chunkmap`, `debug`,
-`evidencecontext`, `internal/core/fileio`, `debugbundle`, and their CLI
-composition.
-
-Verify:
-
-- narrow path validation and symlink-resistant confinement;
-- atomic writes and recoverable explicit cleanup;
-- separation of checkpoint recording, resume loading, chunk-plan caching, and
-  debug capture;
-- canonical encoding before content identity is trusted;
-- complete but non-secret checkpoint fingerprints;
-- correct ordinary-resume and selective-recompute behavior;
-- producer dependency invalidation across ordered steps;
-- immutable hydration and no aliasing with stored bytes;
-- debug data never influencing execution or reuse; and
-- terminal persistence failures never obscuring the primary error.
-
-Review the repeated extract/merge/normalize recorder and loader methods, path
-component validators in multiple state packages, and clone/encode/decode paths.
-Determine which repetition is a clear stage adapter and which can share a
-private primitive without weakening reason-code ownership or diagnostics.
-
-### 6. LLM Runtime, Prompt Filesystems, And Assets
-
-Inspect `internal/framework/llm`, `promptfs`, the PromptKit integration,
-scheduler, profile-source construction, prompt/schema registries, root
-`assets`, module prompt manifests, and relevant D&D shared assets.
-
-Verify:
-
-- every provider call passes through the shared scheduler and cancellation
-  removes queued calls safely;
-- PromptKit and Notarius concurrency limits compose as documented;
-- profile inspection and runtime use identical source precedence;
-- session IDs, profile-source fingerprints, prompt fingerprints, and schema
-  fingerprints reflect the intended semantic inputs;
-- secrets and provider-specific error types do not cross the boundary;
-- prompt inputs and private outputs do not expose opaque entity IDs;
-- prompt ordering, stable prefixes, and cache controls remain intentional;
-- schema loaders and filesystem adapters validate once and return defensive
-  data; and
-- the root assets package contains no business logic.
-
-Compare the LLM asset registry and prompt-filesystem adapters for duplicated
-filesystem behavior. Review repeated prompt/schema loader and metadata code in
-module packages, but reject an extraction that would centralize domain prompt
-ownership or make unrelated assets share one invalidation boundary.
-
-### 7. Generic And Seriatim Modules
-
-Inspect `internal/modules/generic` and `internal/modules/seriatim`, including
-module specs, option decoding, chunk planning, input translation, validators,
-output encoding, evidence-context publication, registration, and tests.
-
-Verify that:
-
-- external Seriatim details end at the input boundary;
-- generic chunking and output remain domain-neutral;
-- chunk plans and source units preserve source-addressed invariants;
-- output logical names are safe and deterministic;
-- output options do not bypass preparation-time compatibility checks; and
-- option decoding is strict, small, and consistent with configuration
-  validation.
-
-The graph flags generic integer option parsing and JSON output policy decoding
-as relatively complex. Examine whether that is inherent strict decoding or an
-opportunity for a smaller typed parser with equally precise diagnostics.
-
-### 8. D&D Domain Model, Codecs, And Shared Helpers
-
-Inspect `internal/modules/dnd` domain types, codecs, candidate decoders,
-identity packages, registries, shared source-reference helpers, diagnostics,
-registry resolution, entity reconciliation, mergers, registrar, and assets.
-
-Compare all ten current artifact families. Build a convention matrix covering:
-
-- module specs and execution classes;
-- constructor and option behavior;
-- manifest metadata and checkpoint fingerprints;
-- response-schema loading and private-versus-durable types;
-- source-reference conversion, canonicalization, ordering, and deduplication;
-- nil versus present-empty output;
-- codecs and strict JSON behavior;
-- registry lookup, identity derivation, immutable projections, and resolution;
-- normalizer retry/fallback behavior;
-- validators and default chains; and
-- registration, prompt assets, and documentation ownership.
-
-The graph reports many exact similarities among codec `Decode` methods,
-fingerprint/metadata methods, registry extractors, identity helpers, occurrence
-normalizers, and validators. Treat these as a prioritized review list, not an
-instruction to create one generic D&D engine. A worthwhile helper must preserve
-domain-specific identity, evidence, kind ordering, validation, diagnostics,
-and artifact typing.
-
-### 9. D&D Extraction And Normalization Flows
-
-Trace each lane end to end rather than auditing only similarly named files:
-
-- spells;
-- NPC registry and NPC occurrences;
-- combat turns and enemy events;
-- item registry and item occurrences;
-- scene descriptions; and
-- location registry and location occurrences.
-
-For registry/occurrence pairs, verify the complete semantic boundary: the model
-uses contextual evidence, deterministic code attaches opaque identity, registry
-evidence does not become occurrence evidence, and unresolved or ambiguous
-selections fail according to lane policy.
-
-Review whether any lane resolves or canonicalizes the same entity, source
-reference, or response twice; constructs unnecessary intermediate response
-forms; performs repeated sorts or scans; or retains transitional paths. Compare
-registry normalizers and occurrence normalizers for genuinely identical
-mechanics, while keeping currency, same-name location, NPC ambiguity, spell
-catalog, scene eligibility, and combat-specific policy with their owners.
-
-### 10. Test Suite And Comment Coverage
-
-Review the tests associated with every preceding area after understanding the
-production contracts. This should be a cross-cutting pass, not a request to add
-tests for every flagged function.
-
-Identify:
-
-- consequential unprotected invariants;
-- duplicated policy assertions across layers;
-- brittle tests coupled to internal constants, prompt prose, or private helper
-  shape;
-- oversized test harnesses and repeated fixtures that obscure intent;
-- race-sensitive code not exercised under `-race`;
-- parsers, canonicalizers, and path handlers where fuzzing would address a real
-  input-space risk; and
-- complex production code whose tests reveal an unclear ownership boundary.
-
-Also identify necessarily complex symbols that lack a concise invariant-level
-comment. Comment recommendations should name the exact symbol and the fact the
-comment should explain; “add more comments” is not an actionable finding.
-
-## Efficiency Evaluation
-
-The audit should consider both runtime and maintenance efficiency.
-
-For runtime efficiency, examine algorithmic behavior relative to realistic
-input dimensions: source units, chunks, lanes, references, artifacts, registry
-records, checkpoint files, and prompt assets. Prioritize repeated full-input
-passes, nested linear lookup, unnecessary JSON round trips, repeated hashing,
-large defensive copies at adjacent ownership boundaries, and serialization on
-concurrent hot paths. Preserve a defensive copy when it establishes ownership;
-removing it solely to reduce allocation is not an improvement.
-
-For maintenance efficiency, prioritize repeated policy, parallel type systems,
-duplicated error classification, scattered defaults, and migrations that left
-two ways to perform the same operation. Boilerplate is costly only when it can
-drift or obscures the semantic core. Small explicit typed adapters can be more
-maintainable than a generic abstraction.
-
-Do not recommend caching, pooling, concurrency, or a benchmark without naming
-the workload and risk it addresses. Add a benchmark only when a proposed
-optimization concerns a repeatable local path and the result would influence
-the decision.
-
-## Audit Method
-
-Each area should use the same method:
-
-1. Read its architecture/internal documentation and focused tests.
-2. Map public/package contracts and trace the main call paths.
-3. Inspect high-fan-in, high-cognitive-complexity, nested-loop, and repeated-
-   conversion symbols.
-4. Review exact and near-duplicate code side by side, including callers and
-   failure semantics.
-5. Check dependency direction, ownership, aliasing, deterministic order,
-   cancellation, and error classification.
-6. Compare tests with the risks owned at that layer.
-7. Record findings and inspected-with-no-finding areas before moving on.
-8. Run focused read-only validation when it can confirm or refute a suspected
-   problem.
-
-Prefer the repository knowledge graph for symbol discovery, call tracing, and
-similarity candidates. Use textual search for literals, diagnostics, config
-keys, asset content, and stale names. Read complete implementations and tests
-before reporting a metric-derived candidate.
-
-## Execution Sequence
-
-This document owns audit scope, questions, evidence standards, and the quality
-bar. [Staged Codebase Audit Sequence](audit-sequence.md) is the sole canonical
-owner of prompt order, stage boundaries, per-stage reading, validation commands,
-and acceptance criteria. Do not derive or maintain a second sequence here.
-
-The audit is executed as bounded prompts and writes its accumulated findings to
-`docs/roadmap/audit.md`. Later stages must build on and reconcile earlier
-evidence rather than concatenate independent reports. Implementation and
-roadmap retirement remain separate work after maintainers review the completed
-audit.
-
-## Baseline And Validation
-
-Before the first audit stage, record the commit under review and require a clean
-worktree. Refresh the code knowledge graph so renamed or deleted code does not
-produce false findings. Run the normal offline baseline:
-
-```sh
-go test ./...
-go vet ./...
-go build ./cmd/notarius
-git diff --check
-```
-
-Run `go test -race` for packages with concurrency or mutable shared state,
-especially `internal/framework/pipeline`, `internal/framework/llm`, state
-packages, and D&D registries. A repository-wide race run is appropriate for
-final verification if its cost remains reasonable.
-
-Optional diagnostic commands should be used only when relevant:
-
-- `go test -count=1` to rule out cache-masked failures;
-- `go test -shuffle=on` to detect order coupling;
-- focused fuzzing for existing or newly justified fuzz targets; and
-- focused benchmarks or profiles for a specific efficiency finding.
-
-The audit itself should not change tests to make the baseline pass. Record any
-pre-existing failure and distinguish it from an audit finding.
-
-## Deliverable Quality Bar
-
-The completed audit should:
-
-- cover every repository area listed above;
-- distinguish defects from refactoring opportunities and comment requests;
-- distinguish credible performance costs from aesthetic simplification;
-- identify intentional duplication that should remain explicit;
-- avoid recommendations that violate dependency direction or weaken typing;
-- cite exact evidence and preserve named invariants for every finding;
-- consolidate root causes rather than report many symptoms;
-- rank independent work so a later implementation plan can stage it safely;
-- recommend no code change whose expected benefit is smaller than its added
-  abstraction or test-maintenance cost; and
-- leave implementation and roadmap retirement to later work.
-
-## Open Questions
-
-None are required to begin the audit. If a later stage cannot determine whether
-behavior is intentional from code, tests, policies, ADRs, or current
-documentation, it should record the uncertainty and a recommended resolution
-rather than silently treating preference as a defect.

docs/roadmap/contextual-entity-grounding.md

@@ -1,350 +0,0 @@
-# Contextual Entity Grounding
-
-## Purpose
-
-Notarius should use an LLM for semantic interpretation of source evidence, not
-for referential-integrity work that deterministic code can perform more
-reliably. D&D prompts must therefore stop requiring models to reproduce opaque
-machine identifiers such as hash-derived entity IDs. Models should identify
-entities through human-readable, evidence-grounded context, after which
-Notarius resolves the selection and attaches the canonical durable identity.
-
-This roadmap defines the policy, affected D&D prompt families, and intended
-end state. The ordered work needed to reach that state is maintained in
-[Implementation Plan](implementation.md).
-
-## User Intent
-
-The change has two goals:
-
-- prevent otherwise useful model responses from failing because a long,
-  non-semantic string was copied incorrectly; and
-- avoid spending prompt space and model effort on exact-copy work that provides
-  no semantic value.
-
-The policy is not a ban on identifiers. Durable artifacts may continue to use
-application-owned IDs, and prompts may continue to request source-unit ranges
-that locate evidence. The policy governs which identity work is assigned to
-the model.
-
-## Policy
-
-An LLM-facing prompt input or private response schema must not require a model
-to reproduce an opaque machine identifier when Notarius can establish the same
-association deterministically.
-
-Opaque machine identifiers include cryptographic hashes, UUIDs, digests,
-database keys, durable entity IDs, and other tokens whose characters do not
-carry source-grounded meaning for the model. These values may remain in
-application state, provenance, diagnostics, checkpoints, and durable artifact
-contracts, but should be omitted from model-visible material when they do not
-help the model make a semantic decision.
-
-The intended responsibility boundary is:
-
-- the model decides which contextual entity is supported by the supplied
-  evidence and returns the bounded semantic facts requested by the module;
-- the calling module validates that the contextual selection resolves to
-  exactly one supplied candidate;
-- deterministic code supplies the canonical display value and durable entity
-  ID; and
-- existing validators continue to enforce referential integrity at later
-  artifact boundaries.
-
-Transcript `start_unit_id` and `end_unit_id` values are permitted. They are
-contextual source coordinates and form part of the evidence contract rather
-than arbitrary identity tokens. Prompt IDs, schema IDs, fingerprints, session
-IDs, and digests may also remain in runtime metadata that the model is not
-asked to reproduce.
-
-Short request-local labels are a narrowly permitted fallback only when a
-contextual selector cannot uniquely represent the available choices without
-unreasonable prompt cost. Such a label must be compact, scoped to one request,
-validated against the supplied candidate set, and never reused as a durable
-identity. Current D&D occurrence and reconciliation prompts should be designed
-without this exception; adopting it later requires a concrete demonstrated
-need and documented rationale.
-
-## Current State
-
-The initial NPC, item, and location registry extractors already follow the
-desired pattern: the model returns contextual names and evidence, and Notarius
-derives durable IDs afterward. Spells, combat turns, and enemy events use
-contextual actor names rather than requiring hash-derived NPC IDs.
-
-Two current prompt families diverge from that pattern:
-
-1. `dnd/npc-occurrences`, `dnd/item-occurrences`, and
-   `dnd/location-occurrences` place durable registry IDs in model-visible
-   projections and require the private LLM response to repeat those IDs.
-2. NPC-, item-, and location-registry normalization use the shared entity
-   reconciliation prompt, which labels candidates with opaque
-   `candidate-000001`-style keys and requires the model to copy those keys into
-   duplicate-group proposals.
-
-The durable occurrence artifacts correctly retain canonical entity IDs. The
-problem is the private model transport contract, not the published artifact
-contract.
-
-## Target Architecture
-
-### Model proposals and durable artifacts
-
-Private LLM response types must express contextual semantic proposals rather
-than reuse the durable artifact type when that type contains an opaque entity
-ID. The extractor maps a validated private response into the existing durable
-artifact only after identity resolution succeeds.
-
-No affected durable artifact kind, media type, schema ID, schema version, or
-JSON field changes as part of this work. NPC, item, and location occurrence
-artifacts continue to publish their exact canonical ID/name pair. Registry
-artifacts likewise retain their IDs and evidence.
-
-The private schemas and prompt declarations may remain at their current `v1`
-identities because Notarius is pre-release and these are not external
-contracts. Their content hashes, mapping-policy fingerprints, and affected
-prompt fingerprints must change so incompatible checkpoints are not reused.
-
-### NPC occurrence grounding
-
-The NPC occurrence prompt receives an ordered names-only projection of the
-normalized NPC registry. Its private response contains the canonical NPC name,
-occurrence kind, and current-transcript source ranges, but no `npc_id`.
-
-The extractor resolves the returned name under the existing NPC comparison
-policy. Resolution must produce exactly one registry entry. It then writes that
-entry's canonical display name and durable ID into the `dnd.NPCOccurrence`.
-An unknown or ambiguous selection invalidates the extraction operation; the
-extractor must not guess, use fuzzy matching, silently omit the record, or
-accept a partial response.
-
-### Item occurrence grounding
-
-The item occurrence prompt receives an ordered names-only projection of the
-normalized item registry. Its private response contains the canonical item
-name, occurrence kind, kind-specific fields, and current-transcript source
-ranges, but no `item_id`.
-
-The extractor resolves the returned name under the existing item comparison
-and identity policies. Resolution must produce exactly one registry entry,
-whose canonical name and durable ID are attached deterministically. Unknown or
-ambiguous selections invalidate the complete extraction operation rather than
-being guessed, repaired by similarity, or dropped.
-
-### Location occurrence grounding
-
-Location identity cannot always be resolved from a display name alone: the
-current registry intentionally permits same-name locations with distinct
-source anchors. The location occurrence prompt must therefore receive a
-contextual registry descriptor that contains the canonical display name plus
-the minimum source-grounded registry evidence needed to distinguish same-name
-records. It must not contain the durable `location:sha256:...` value.
-
-The private response uses two required selector fields: `name` and
-`registry_refs`. For a comparison-unique canonical name, `registry_refs` is an
-empty array and Notarius resolves the name under the location comparison
-policy. For a name shared by multiple registry records, `registry_refs`
-contains that record's complete canonically ordered registry ranges as
-`start_unit_id` and `end_unit_id` pairs, without `source_id`.
-
-Every model-facing registry entry uses one fixed shape with required `name`,
-`registry_refs`, and `context` fields. `context` is an array of strict objects
-containing only `unit_id` and `text`. Comparison-unique entries use empty
-`registry_refs` and `context` arrays. Same-name entries use the complete
-registry-range selector and the bounded context described below. The model
-returns only `name` and `registry_refs`; it does not reproduce `context`.
-
-For same-name groups, the projection also supplies bounded transcript units
-covered by each record's registry ranges so the model receives meaningful
-identity context rather than coordinates alone. Those ranges must resolve
-against the current source document, and the resulting contextual selectors
-must be unique. An invalid range or selector collision prevents the LLM call
-and fails the operation. Unique-name entries do not repeat registry ranges or
-context in the selector, preserving compatibility with a valid registry from
-another source when the name alone is unambiguous.
-
-The private response separately supplies current-transcript `source_refs` that
-prove the occurrence. Registry identity evidence and occurrence evidence must
-remain different fields and must never be merged. The model should omit an
-occurrence when the transcript does not support choosing among same-name
-locations. If a returned selector does not resolve to exactly one supplied
-registry record, the extractor invalidates the complete operation rather than
-guessing.
-
-### Registry reconciliation
-
-The shared entity-reconciliation input replaces opaque candidate keys with
-contextual candidate descriptors. At minimum, a descriptor contains the
-candidate's display name and its canonically ordered source-reference ranges;
-the existing transcript windows remain available for semantic judgment.
-
-Duplicate-group members and the canonical member in the private response use
-the same contextual descriptor shape. The shared reconciliation helper maps
-each descriptor back to exactly one internal candidate before assessing the
-proposal. Exact deterministic duplicates should already be removed before the
-LLM call; any remaining descriptor collision makes the affected candidate
-ineligible for model-assisted reconciliation rather than authorizing an
-arbitrary choice.
-
-Existing safety behavior remains in force: groups must contain at least two
-supplied candidates, the canonical candidate must be a member, groups must not
-overlap, and domain-specific eligibility rules remain authoritative. Invalid,
-ambiguous, or unsafe groups are discarded through the existing bounded
-fallback and diagnostic behavior. The model never directly mutates the
-durable registry.
-
-The shared private reconciliation schema and helper must remain domain-neutral
-within the D&D family. NPC-, item-, and location-specific duplicate policy
-continues to live in the owning normalizer.
-
-## Prompt And Asset Changes
-
-The following LLM-facing assets are in scope:
-
-- the prompt instructions, registry input fragments, and private response
-  schemas for NPC, item, and location occurrences;
-- the prompt manifests where input shape or selected fragments change;
-- the shared D&D entity-reconciliation fragment and private response schema;
-  and
-- the NPC-, item-, and location-registry normalization prompt inputs that use
-  the shared reconciliation contract.
-
-Affected projections must exclude durable entity IDs rather than merely stop
-mentioning them in prose. Prompt instructions should describe the contextual
-selection rule once at the narrowest owning asset and must preserve the current
-distinction between registry grounding and transcript evidence.
-
-Prompt ordering and cache controls should remain unchanged unless the new
-contextual input requires an intentional manifest change. Unrelated shared
-prompt bytes should not be edited. Prompt and schema fingerprints should
-invalidate only the operations whose selected assets or mapping semantics
-changed.
-
-## Code And Validation Changes
-
-The occurrence extractors need private response types and deterministic
-registry-resolution paths appropriate to their domain. Shared code is
-appropriate only for demonstrated mechanics that have identical semantics;
-NPC, item, and location ambiguity policies must not be forced behind a generic
-resolver merely to reduce line count.
-
-Registry projections should expose explicit model-facing methods whose names
-describe whether they are names-only or contextual identity projections. The
-existing ID/name projections may remain only for deterministic consumers that
-genuinely require them; they must no longer be wired to an LLM input.
-
-Mapping-policy and normalization-policy identifiers must be reviewed and
-advanced wherever their semantics change. Checkpoint fingerprints must cover
-the new projection content, private schema, prompt assets, and mapping policy,
-while continuing to exclude irrelevant internal implementation details.
-
-Durable occurrence normalizers and registry validators remain defense in
-depth. They continue to validate exact ID/name pairs on artifacts entering
-through checkpoints, codecs, or other boundaries even though the LLM no longer
-produces the ID directly.
-
-## Testing And Evaluation
-
-Tests should protect the behavioral boundary rather than prompt prose or
-private helper structure. The completed work should demonstrate that:
-
-- affected model-facing registry projections do not contain durable entity
-  IDs;
-- private occurrence schemas reject opaque ID fields and accept the intended
-  contextual shape;
-- valid contextual selections map to the exact canonical durable ID/name pair;
-- unknown, mismatched, and ambiguous selections fail without fuzzy matching,
-  partial acceptance, or arbitrary reassignment;
-- same-name locations remain distinguishable through contextual evidence;
-- reconciliation preserves equal-name candidates, resolves valid contextual
-  groups, and discards ambiguous or unsafe proposals;
-- registry evidence never becomes occurrence evidence;
-- durable codec, normalization, and validator behavior remains compatible; and
-- representative assembled D&D pipelines still prepare and execute with fake
-  structured-LLM responses.
-
-Do not add repository-wide prompt-prose snapshots, exact-message-count tests,
-or a change-detector test that merely scans for today's field names. Focused
-projection, schema, mapping, fallback, and integration tests are the stable
-owners of these risks. Model-quality evaluation with representative
-transcripts remains a manual development aid rather than an offline test gate.
-
-## Documentation And Architectural Record
-
-This policy is durable and applies to future modules, so it warrants
-`docs/adr/0012-resolve-opaque-entity-identifiers-deterministically.md`, which
-records:
-
-- the semantic-proposal versus referential-integrity boundary;
-- why durable opaque IDs are excluded from model response contracts;
-- why contextual evidence coordinates remain permitted;
-- the narrowly scoped request-local-label exception;
-- alternatives including durable IDs, names-only matching, and short opaque
-  handles; and
-- the consequences for private schemas, deterministic resolution, debugging,
-  and ambiguous identities.
-
-`docs/policy/architecture.md` states the general LLM boundary invariant and
-links to the ADR. `docs/internal/dnd.md` describes the concrete occurrence
-projections, contextual reconciliation selectors, resolution and failure
-behavior, and the continued separation of registry grounding from occurrence
-evidence. `docs/internal/llm.md` contains only a short clarification that
-caller-owned modules, not PromptKit or the transport adapter, resolve
-contextual model selections into application identities.
-
-The NPC, item, and location occurrence and registry integration documents must
-continue to own their durable wire contracts, while removing current claims
-that the model-facing consumer projection contains `{id,name}` or that the raw
-LLM response supplies the durable ID. They should instead explain that
-Notarius resolves contextual model output and publishes the same exact durable
-ID/name pair. No public schema examples need to remove those IDs.
-
-The generic LLM-assisted deduplication entry in `docs/roadmap/future.md` must be
-reconciled with this policy: stable IDs may exist inside deterministic state,
-but a future model-facing proposal should use contextual selectors or a
-documented request-local-label exception rather than durable IDs.
-
-## Compatibility And Operational Effects
-
-This work intentionally changes private prompt inputs, private structured
-responses, and mapping semantics. It will invalidate affected checkpoints
-through existing prompt, schema, projection, and policy fingerprints. No
-manual checkpoint migration is required.
-
-Durable D&D artifacts and generated-reference compatibility remain unchanged.
-Operators do not receive new configuration fields or CLI controls. The feature
-does not change PromptKit, provider routing, profile selection, retries,
-concurrency, or public output placement.
-
-## Non-Goals
-
-This work does not:
-
-- remove canonical IDs from durable registries or occurrence artifacts;
-- change occurrence categories, evidence rules, or registry identity policy;
-- add fuzzy, probabilistic, or embedding-based entity resolution;
-- allow registry provenance to substitute for occurrence evidence;
-- introduce a general entity graph or cross-artifact identity framework;
-- redesign unrelated D&D prompts or their schemas;
-- implement the future generic deduplication normalizer; or
-- add provider-specific prompt behavior.
-
-## Acceptance Criteria
-
-The target state is complete when:
-
-- no maintained D&D prompt requires a model to reproduce a durable opaque
-  entity ID;
-- current D&D reconciliation prompts no longer require opaque candidate keys;
-- NPC, item, and location occurrence LLM outputs are resolved
-  deterministically into their unchanged durable artifacts;
-- same-name location and reconciliation cases remain safe and unambiguous;
-- invalid contextual selections preserve the existing extraction-failure or
-  normalization-fallback semantics appropriate to their stage;
-- affected checkpoint identities change without altering public schema
-  versions;
-- focused and repository-wide tests pass offline;
-- the ADR, architecture invariant, D&D internal guide, LLM internal guide,
-  relevant integration contracts, and future roadmap accurately describe
-  their canonical portions of the implemented policy; and
-- no unrelated code, prompt behavior, or public contract changes are included.

docs/roadmap/future.md

@@ -5,13 +5,64 @@ configuration, operations, internal, and integration docs. This roadmap records
 future work only. Items are ordered roughly by current value and specificity,
 not as committed release dates.
 
+## Near-Term Validation And LLM Reliability
+
+PromptKit now owns structural output repair within one completion. Notarius
+owns stage candidates, validator chains, semantic rejection policy, bounded
+feedback-aware stage retries, validation provenance, and reusable-state
+eligibility, and the separation of actionable process warnings from quality
+diagnostics. The remaining near-term work applies those completed foundations
+to domain review and empirical evaluation.
+
+### D&D Combat Scene Semantic Validation
+
+- Add an optional production LLM-backed D&D validator that determines whether
+  proposed scene boundaries and classifications represent substantive active
+  combat correctly. Its central quality goal is that active combat is kept in
+  coherent scenes classified as `combat`, rather than split incorrectly or
+  hidden inside scenes classified as `narrative`, `recap`, or `meta`.
+- Resolve the validator's exact target before implementation. The current
+  `dnd/scenes` chunker owns only complete, gap-free source ranges, while the
+  per-chunk `dnd/scene-descriptions` extractor owns the `combat`, `narrative`,
+  `recap`, and `meta` classification. The preferred initial placement is
+  therefore an extract-stage validator for `dnd/scene-descriptions`, where it
+  can compare one proposed kind with the corresponding transcript chunk.
+- Consider a chunk-stage LLM validator only for a distinct boundary-coherence
+  question that can be answered from the complete transcript and proposed
+  range map, such as whether one continuous combat was fragmented across
+  inappropriate scene boundaries. Do not duplicate the same classification
+  judgment at both stages. Moving classification into chunk-plan annotations
+  would change the deliberately minimal, annotation-free chunk contract and
+  requires an explicit architecture review before it is selected.
+- Validate both false negatives and false positives: a non-combat kind must not
+  omit substantive active combat, and a combat kind must be supported by such
+  combat. Keep the existing deterministic downstream rule that combat-turn
+  extraction runs only for an exact `combat` scene classification; semantic
+  review improves the upstream classification but does not replace that gate.
+- Run the semantic validator through PromptKit, use a minimal required-field
+  structured response schema, and let PromptKit repair structural validator
+  output within its bounded budget. A contract-invalid final validator response
+  is a validator execution failure, not a semantic rejection and not a reason
+  to recursively validate the validator.
+- Evaluate the prompt and decision policy against a small human-reviewed set
+  containing combat setup, active turns, interruptions, multi-phase encounters,
+  brief rules discussion, aftermath, recalled combat, and false-positive
+  hostile dialogue. Measure false acceptance, false rejection, retry success,
+  added calls, latency, and token cost before placing it in the production
+  default chain.
+- An ADR is not required if classification remains owned by
+  `dnd/scene-descriptions` and the validator follows the generic validation ADR.
+  Create or supersede an ADR if the work transfers scene classification into
+  the chunker or otherwise changes stage ownership or the durable chunk-plan
+  contract.
+
 ## Near-Term D&D Pipeline
 
 ### Evaluate Spell Extraction And Normalization
 
 - Evaluate ordinary extraction retries and the completed normalization path
-  against a human-reviewed transcript set before adding repair-aware retries or
+  against a human-reviewed transcript set before and after adopting the shared
-  an LLM-backed semantic validator.
+  PromptKit repair and Notarius validation-retry policies above.
 - Maintain a small set of human-reviewed transcripts and outputs for prompt,
   validator, and normalizer development. Treat model-quality review as an
   iterative human evaluation aid, not a deterministic correctness gate.
@@ -24,28 +75,51 @@ not as committed release dates.
 
 ## Shared Normalization And Quality Work
 
-### Generic LLM-Assisted Deduplication
+The implemented source-backed core and initial D&D registry adoption are
+described by [Module Internals](../internal/modules.md#semantic-reconciliation)
+and
+[D&D Module Internals](../internal/dnd.md#semantic-registry-reconciliation).
+The sections below keep broader extensions deferred.
 
-- Add a reusable normalizer that asks an LLM to identify duplicate sets in a
+### Large-Collection Semantic Reconciliation
-  list and propose one replacement element for each set.
-- Define the minimum domain-neutral input contract, initially an ordered list
-  whose elements retain stable unique IDs as internal deterministic state.
-  Model proposals use contextual descriptors, or a specifically justified
-  request-local short label, rather than durable IDs. Artifact-kind
-  registrations or adapters may expose that structure without moving domain
-  rules into the generic package.
-- Keep mutation deterministic: parse and validate the model's duplicate groups,
-  resolve every supplied descriptor or local label exactly, reject overlapping
-  or malformed groups, prevent unrelated insertion or deletion, and apply only
-  approved replacement operations in code.
-- Preserve provenance needed for audit and downstream validation, and emit
-  warnings describing every collapsed group.
-- Evaluate batching and context-window limits before applying the normalizer to
-  large artifact collections.
 
-The model may use its own domain knowledge to judge semantic duplication; the
+- Evaluate deterministic candidate blocking only after representative registry
-generic implementation is responsible only for the common proposal contract,
+  inputs exceed the active roadmap's bounded single-request limits. Blocking
-safety checks, and deterministic application of accepted changes.
+  should use cheap, explainable signals to form plausible comparison sets while
+  preserving the possibility that a duplicate appears outside a lexical name
+  match.
+- Define correctness for candidates that appear in more than one block,
+  conflicting canonical selections, transitive identity across blocks, retry
+  isolation, and deterministic final ordering before implementation.
+- Prefer a reconciliation graph or union plan with explicit conflict checks
+  over arbitrary fixed-size slices. Never silently treat a batch boundary as
+  evidence that two candidates are distinct.
+- Record per-request bounds, block provenance, model calls, discarded
+  proposals, and final group derivation well enough to audit a collapse.
+
+### Operator-Selected Semantic Policies
+
+- Consider allowing an operator to select an approved semantic-policy prompt
+  for a typed reconciliation module without replacing the shared protocol,
+  response schema, or deterministic safety rules.
+- Define the trusted asset source, configuration syntax, compatibility checks,
+  startup validation, provenance, prompt fingerprinting, checkpoint effects,
+  and support boundary before exposing the option.
+- Prefer selection among registered, typed-policy-compatible prompt assets over
+  arbitrary filesystem prompt paths. Do not add this flexibility until an
+  operator workflow requires it; artifact-family-owned policy remains simpler
+  and safer for the initial implementation.
+
+### Broader Reconciliation Inputs And Module Selection
+
+- Revisit alternate context providers when a concrete non-source-backed entity
+  collection needs semantic reconciliation. Any extension must preserve the
+  same request-local identity, deterministic proposal validation, provenance,
+  and typed application guarantees.
+- Consider a selectable generic normalizer only if Notarius gains a real
+  domain-neutral typed artifact contract that can safely support it. Do not
+  weaken exact artifact registration or introduce reflection-based arbitrary
+  JSON mutation merely to expose a universal module key.
 
 ### Validation And Review
 
@@ -102,6 +176,18 @@ checkpoint reuse, when an older artifact may be decoded or adapted, and when a
 producer or all dependents must be recomputed. Do not add a general migration
 framework until an actual contract change requires one.
 
+### Artifact-family-oriented physical packaging
+
+[ADR-0004](../adr/0004-package-modules-by-domain.md) currently groups production
+extensions by domain and then by pipeline stage. After artifact-family
+ownership terminology is established and more families span extraction,
+normalization, validation, codecs, references, and assets, reassess whether a
+feature-first physical layout would improve navigation and reduce scattered
+changes enough to justify a repository-wide package migration. Any change must
+address Go dependency cycles, registrar ownership, stable public module keys,
+and supersession of the affected ADR-0004 decision. Conceptual artifact-family
+ownership does not by itself require this move.
+
 ## Blue-Sky Platform And Operations
 
 These ideas are intentionally less specified. Promote one into an earlier
@@ -117,7 +203,6 @@ section only after a concrete workflow, contract, and priority emerge.
 ### Distribution And Operations
 
 - Packaged release artifacts for alpha distribution.
-- A documented versioning and release process.
 - Optional generated example-output fixtures with a regeneration procedure.
 - Additional diagnostics or reporting views.
 

docs/roadmap/implementation.md

@@ -1,612 +0,0 @@
-# Contextual Entity Grounding Implementation Plan
-
-## Objective
-
-Implement [Contextual Entity Grounding](contextual-entity-grounding.md) so D&D
-LLM prompts return evidence-grounded contextual selectors while Notarius owns
-canonical entity IDs and referential integrity. Preserve every durable D&D
-artifact contract and remove opaque IDs only from model-visible inputs and
-private model responses.
-
-This plan is written for a gpt-5.6-terra coding agent. Implement the stages in
-numeric order. Each stage is intentionally scoped to one implementation prompt
-and must leave the repository buildable and its focused tests passing before
-the next stage begins.
-
-## Plan-Wide Decisions
-
-Apply these decisions throughout every stage:
-
-- Read `docs/development.md`, all files under `docs/policy/`, the feature
-  roadmap, and the focused implementation/tests named by the stage before
-  editing.
-- Preserve the fixed pipeline, typed artifact boundaries, root `assets`
-  content-only rule, module ownership, PromptKit boundary, and evidence rules.
-- Do not change the durable NPC-, item-, location-registry, or occurrence Go
-  types, JSON schemas, schema IDs, schema versions, media types, reference-slot
-  contracts, categories, or generated-reference compatibility.
-- Keep the affected prompt and private response-schema identities at `v1`.
-  They are private pre-release transport contracts; their changed content
-  hashes provide the required compatibility boundary.
-- Advance semantic policy identifiers exactly as directed in each stage. Do
-  not bump unrelated policy identifiers.
-- A contextual name match uses the entity family's existing comparison policy,
-  never fuzzy matching. A model selection must resolve to exactly one supplied
-  record before a durable ID is attached.
-- An invalid NPC, item, or location selection invalidates the complete
-  extraction operation. Do not silently drop one response record, accept a
-  partial artifact, or defer a known mapping failure to a later validator.
-- Registry provenance remains grounding only. Only the current extraction
-  chunk's `source_refs` become occurrence evidence.
-- Preserve prompt message order and cache controls unless a stage explicitly
-  directs otherwise. Edit only the selected module or shared assets; do not
-  rewrite unrelated shared prompt bytes.
-- Preserve internal opaque IDs where deterministic code needs them. The rule
-  applies to material shown to the model or requested from it, not to maps,
-  fingerprints, checkpoints, diagnostics, or durable artifacts.
-- Follow `docs/policy/testing.md`: test package-level behavior and meaningful
-  failure modes, not prompt prose, exact message counts, private helper
-  structure, or a repository-wide string-scanning change detector. All tests
-  remain deterministic, offline, and credential-free.
-- Use `apply_patch` for edits, `gofmt` changed Go files, and preserve unrelated
-  worktree changes.
-
-## Final Private Selector Contracts
-
-These shapes are implementation requirements, not public artifact schemas.
-
-### NPC occurrence response
-
-Each response record contains exactly the required fields `name`, `kind`, and
-`source_refs`. It does not contain `npc_id`. Notarius resolves `name` through
-the normalized NPC registry and writes the matched registry record's `ID` and
-canonical `Name` into the durable occurrence.
-
-### Item occurrence response
-
-Each response record contains the existing required `name`, `kind`,
-`quantity`, `from`, `to`, and `source_refs` fields. It does not contain
-`item_id`. Retain the current nullable representation and kind-specific
-semantics. Notarius resolves `name` through the normalized item registry and
-adds the matched `ID` and canonical `Name`.
-
-### Location occurrence response
-
-Each response record contains exactly the required fields `name`,
-`registry_refs`, `kind`, and `source_refs`. `registry_refs` is always an array
-of strict objects containing required integer `start_unit_id` and
-`end_unit_id`; it may be empty.
-
-- When `name` has one comparison-identity match in the supplied registry,
-  `registry_refs` must be empty and name resolution selects that record.
-- When multiple registry records share the comparison identity,
-  `registry_refs` must equal one record's complete canonically ordered source
-  ranges with `source_id` removed.
-- The model-facing registry projection uses the same `name` plus
-  `registry_refs` selector and adds a required `context` array. Unique-name
-  records project empty `registry_refs` and `context` arrays. The private
-  response does not reproduce `context`.
-- Same-name records receive bounded identity context consisting of the ordered
-  source units covered by their registry ranges. Each context element is a
-  strict object with exactly the required fields `unit_id` (integer) and
-  `text` (string); do not expose the durable location ID, source ID, digest,
-  or a replacement token.
-- Building same-name grounding validates that every registry range belongs to
-  and resolves against the current source. If two records still produce the
-  same contextual selector, grounding construction fails before the LLM call.
-- `registry_refs` never flow into the durable occurrence's `source_refs`.
-
-### Entity-reconciliation response
-
-The shared response remains an object with required `duplicate_groups`.
-Every group has required `members` and `canonical`. A member and the canonical
-selection are strict contextual objects containing:
-
-```json
-{
-  "name": "Mira Thorn",
-  "source_refs": [
-    {"start_unit_id": 12, "end_unit_id": 12}
-  ]
-}
-```
-
-The candidate prompt input uses the same descriptor and contains no `key`.
-`source_refs` is required and non-empty for every eligible candidate. The
-shared helper may retain its existing `candidate-000001`-style keys strictly
-inside Go state to preserve input-position mapping; those keys must never be
-serialized into prompt input or accepted in the private response.
-
-If two candidates produce an identical contextual descriptor, neither is
-eligible for model-assisted reconciliation because the model cannot identify
-them independently. Otherwise the helper converts returned descriptors to its
-internal candidate keys before applying all existing unknown-member,
-ineligible-member, duplicate-member, canonical-membership, overlap, retry, and
-fallback rules.
-
-## Stage 1: Convert NPC Occurrences To Name-Based Resolution
-
-### Goal
-
-Remove durable NPC IDs from the NPC-occurrence prompt and private response,
-then resolve the model's contextual name deterministically without weakening
-checkpoint identity or downstream validation.
-
-### Work
-
-1. Inspect:
-   - `assets/dnd/npc-occurrences/`;
-   - `internal/modules/dnd/extract/npcoccurrences/`;
-   - `internal/modules/dnd/npcs/registry/`;
-   - NPC-occurrence normalizer and validator checkpoint fingerprints; and
-   - their focused tests.
-2. Change `dnd_npc_occurrences_llm.v1.json` so every occurrence requires only
-   `name`, `kind`, and `source_refs`, continues to reject unknown fields, and
-   no longer declares `npc_id`.
-3. Revise the NPC-occurrence instructions to require a supplied canonical NPC
-   name and current-chunk evidence, with no instruction to copy or invent an
-   ID. Continue using the existing shared names-only NPC registry fragment and
-   preserve manifest order/cache controls.
-4. Remove `NPCID` from the private `occurrenceResponse`. After canonicalizing
-   response evidence, resolve every response name with the existing
-   `npcregistry.Registry.Lookup` comparison-key lookup. On the first unknown
-   or non-unique selection, return an extractor-scoped mapping error and no
-   value. For a match, construct the durable occurrence with the registry
-   record's exact `ID` and canonical `Name`.
-5. Change `mappingPolicy` to
-   `dnd.npc_occurrences.extract_mapping.v3`.
-6. Stop passing `IdentityPromptInput()` to the LLM; use the existing
-   names-only `PromptInput()`.
-7. Replace the misleading exported model-input API used only for identity
-   fingerprints: retain the unexported ordered `{id,name}` projection and its
-   digest, expose that value as `IdentityDigest() string`, remove
-   `IdentityPromptInput()`, and update NPC-occurrence extractor, normalizer,
-   invariant-validator, and registry-validator fingerprints to use
-   `IdentityDigest()`. The digest must still distinguish ID/name identity from
-   the names-only prompt projection.
-8. Rewrite existing focused tests around observable behavior: rendered NPC
-   registry input is names-only; the private schema rejects `npc_id`; valid
-   names acquire the registry ID; comparison-equivalent names canonicalize;
-   unknown names fail the whole extraction; registry identity fingerprints
-   remain distinct and defensive; empty registries accept only empty model
-   results. Remove tests whose only purpose was requiring the model to return
-   exact ID/name pairs.
-
-### Acceptance Criteria
-
-- No NPC-occurrence LLM input or private response contains a durable NPC ID.
-- Durable NPC occurrences still contain the exact registry ID/name pair.
-- Mapping failures remain extractor failures eligible for the configured
-  pipeline retry behavior.
-- Deterministic consumers still fingerprint the ordered registry identity,
-  while spells, combat turns, enemy events, and NPC occurrences share the
-  names-only model projection.
-
-### Validation
-
-```sh
-go fmt ./internal/modules/dnd/npcs/registry ./internal/modules/dnd/extract/npcoccurrences ./internal/modules/dnd/normalize/npcoccurrences ./internal/modules/dnd/validate/npcoccurrences/...
-go test ./internal/modules/dnd/npcs/registry ./internal/modules/dnd/extract/npcoccurrences ./internal/modules/dnd/normalize/npcoccurrences ./internal/modules/dnd/validate/npcoccurrences/...
-```
-
-This stage is suitable for one gpt-5.6-terra prompt.
-
-## Stage 2: Convert Item Occurrences To Name-Based Resolution
-
-### Goal
-
-Give item occurrences the same contextual-name/deterministic-ID boundary while
-preserving item-specific nullable fields and kind rules.
-
-### Work
-
-1. Inspect `assets/dnd/item-occurrences/`, the item occurrence extractor, the
-   item registry, the item occurrence normalizer and registry validator, and
-   their focused tests.
-2. Change `dnd_item_occurrences_llm.v1.json` to remove `item_id` from required
-   fields and properties. Preserve required `name`, `kind`, `quantity`, `from`,
-   `to`, and `source_refs`, all current enums/nullability, and strict unknown
-   field rejection.
-3. Rewrite the item registry fragment and module instructions to require the
-   supplied canonical name and current-chunk evidence without mentioning an
-   ID. Preserve prompt order and cache controls.
-4. Change the item registry's model projection from ordered `{id,name}` pairs
-   to ordered names-only objects, add a comparison-key index, and expose a
-   defensive `Lookup(name) (dnd.Item, bool)` analogous to the NPC registry.
-   Retain exact `LookupID` for durable normalizers and validators. Because item
-   IDs are derived from the item comparison identity, the names-only
-   `ProjectionDigest` remains sufficient for model input and existing
-   checkpoint consumers.
-5. Remove `ItemID` from the private response. During response canonicalization,
-   resolve every contextual name, replace it with the registry record's
-   canonical name, and attach its durable ID when constructing the final
-   `dnd.ItemOccurrence`. Unknown selections fail the complete extraction; do
-   not alter evidence or nullable-field validation ownership.
-6. Change `mappingPolicy` to
-   `dnd.item_occurrences.extract_mapping.v2`.
-7. Update focused tests to cover names-only projection, defensive comparison
-   lookup, schema rejection of `item_id`, deterministic durable mapping,
-   unknown-name failure after an otherwise valid record, empty registry/result
-   behavior, and preservation of nullable/kind-specific fields.
-
-### Acceptance Criteria
-
-- Model-visible item registry and response content contain no item hash.
-- Every accepted durable occurrence has the matched registry ID and canonical
-  name.
-- Invalid selection remains all-or-nothing, and existing normalizer/validator
-  defense in depth remains unchanged.
-
-### Validation
-
-```sh
-go fmt ./internal/modules/dnd/items/registry ./internal/modules/dnd/extract/itemoccurrences
-go test ./internal/modules/dnd/items/registry ./internal/modules/dnd/extract/itemoccurrences ./internal/modules/dnd/normalize/itemoccurrences ./internal/modules/dnd/validate/itemoccurrences/...
-```
-
-This stage is suitable for one gpt-5.6-terra prompt.
-
-## Stage 3: Add Contextual Location Grounding
-
-### Goal
-
-Replace the location registry's ID/name prompt projection with an immutable,
-source-aware grounding object that can represent same-name locations safely.
-Introduce the new path alongside the old occurrence input so this stage remains
-buildable; Stage 4 performs the atomic extractor cutover and removes the old
-path.
-
-### Work
-
-1. Inspect the location registry, location identity and source-reference
-   helpers, the generic source document index, occurrence checkpoint consumers,
-   and their focused tests.
-2. In `internal/modules/dnd/locations/registry`, define the private-model
-   types needed by both grounding and the location occurrence extractor:
-   - a returned selector with exactly `name` and `registry_refs`;
-   - a registry projection entry with exactly `name`, `registry_refs`, and
-     `context`;
-   - a source-free range with exactly `start_unit_id` and `end_unit_id`; and
-   - a context unit with exactly `unit_id` and `text`.
-   All fields are required in their private JSON shapes, and constructors and
-   accessors must make defensive copies.
-3. Add an operation-scoped immutable grounding type constructed from a resolved
-   registry and the current `*source.SourceDocument`. Its API must provide:
-   - a cloned `contracts.LLMInputMaterial` for the `location_registry` slot;
-   - deterministic resolution of a returned selector to one cloned
-     `dnd.Location`.
-   The prompt material's existing `Digest` field owns the digest of the exact
-   model projection; do not expose a second grounding-specific digest API.
-4. Construct the projection in registry order. Group entries by the existing
-   location comparison key:
-   - every projection entry has exactly the required fields `name`,
-     `registry_refs`, and `context`;
-   - comparison-unique entries use empty `registry_refs` and `context` arrays;
-   - every same-name entry uses its complete canonical source ranges stripped
-     of `source_id` and includes ordered context units covered by those ranges;
-   - each context unit contains exactly required integer `unit_id` and string
-     `text` fields, and units are deduplicated in source order; and
-   - same-name ranges must have `SourceID == doc.ID` and pass
-     `source.DocumentIndex.ValidateRef`.
-5. Fail grounding construction with a bounded, content-safe error if the
-   source is nil, a same-name range is invalid or belongs to another source,
-   a comparison key is empty, or two records produce the same selector. Do not
-   expose transcript text in the error.
-6. Resolution uses the existing comparison key. A unique-name selector is
-   accepted only with empty `registry_refs`; a same-name selector is accepted
-   only on an exact canonical range match. Reject unknown names, a non-empty
-   range list for a unique name, an empty/partial/reordered range list for an
-   ambiguous name, or any selector not present in the grounding.
-7. Separate deterministic identity fingerprinting from LLM material. Add
-   `IdentityDigest()` over the registry's ordered `{id,name}` identity
-   projection, and update the location normalizer and registry-validator
-   checkpoint consumers to use it. The new operation grounding carries the
-   model projection digest in its `LLMInputMaterial`. Retain the old ID-bearing
-   prompt accessor only as a documented transitional dependency of the
-   still-unchanged location occurrence extractor; do not add new callers.
-8. Add focused tests for unique names, same-name context and selectors,
-   canonical range order, deterministic projection/digest, defensive copies,
-   exact selector resolution, nil/foreign/invalid references, selector
-   collisions, empty registries, and identity fingerprint stability. Do not
-   assert large rendered prompt strings; decode the JSON projection and assert
-   its semantic shape.
-
-### Acceptance Criteria
-
-- The location package can build and resolve contextual selectors without
-  exposing `location_id`, `source_id`, digests, or replacement labels.
-- Same-name locations remain distinct and receive meaningful bounded context.
-- Deterministic checkpoint consumers retain an ID-sensitive fingerprint.
-- Only the existing occurrence extractor remains wired to the legacy
-  ID-bearing prompt path until Stage 4; the repository compiles and tests pass.
-
-### Validation
-
-```sh
-go fmt ./internal/modules/dnd/locations/registry ./internal/modules/dnd/normalize/locationoccurrences ./internal/modules/dnd/validate/locationoccurrences/...
-go test ./internal/modules/dnd/locations/registry ./internal/modules/dnd/normalize/locationoccurrences ./internal/modules/dnd/validate/locationoccurrences/...
-```
-
-This stage is suitable for one gpt-5.6-terra prompt.
-
-## Stage 4: Convert Location Occurrences To Contextual Resolution
-
-### Goal
-
-Wire the Stage 3 grounding object into location occurrence extraction and
-remove durable location IDs from the prompt and private response.
-
-### Work
-
-1. Inspect `assets/dnd/location-occurrences/`, the location occurrence model,
-   schema loader, extractor, canonicalization, prompt tests, and Stage 3
-   grounding tests.
-2. Change `dnd_location_occurrences_llm.v1.json` so each occurrence requires
-   exactly `name`, `registry_refs`, `kind`, and `source_refs`; remove
-   `location_id`. Keep all four occurrence kinds. Make `registry_refs` a
-   required array, including an empty array, of strict required positive
-   integer ranges. Keep occurrence `source_refs` separate and unchanged.
-3. Rewrite `location-registry.md` and module instructions to explain the two
-   selector cases, require exact supplied contextual selectors, prohibit
-   invented locations, and state that registry ranges/context are identity
-   grounding rather than occurrence evidence. Preserve manifest order and
-   cache controls.
-4. Change the private response type to `Name`, `RegistryRefs`, `Kind`, and
-   `SourceRefs`. Do not reuse `source.SourceRef` for the source-free registry
-   range type.
-5. In `Extract`, construct operation grounding from the resolved registry and
-   `req.Source` before calling the LLM, put its projection in the
-   `location_registry` input, and resolve every returned selector after
-   completion. Attach the selected registry record's exact ID and canonical
-   name to the durable occurrence while retaining only the response's
-   current-source `source_refs` as evidence.
-6. Fail the whole extraction on grounding-construction failure or the first
-   unknown, malformed, mismatched, or ambiguous selector. This replaces the
-   current behavior that can preserve unknown ID/name pairs for later
-   validators. Keep later normalizer and validator checks as defense in depth
-   for artifacts entering other boundaries.
-7. Change `mappingPolicy` to
-   `dnd.location_occurrences.extract_mapping.v2`.
-8. Remove the legacy ID-bearing registry `PromptInput` and its model-projection
-   digest once the extractor uses operation grounding. Keep the occurrence's
-   static module fingerprint based on `IdentityDigest()`, prompt/schema
-   fingerprints, and mapping policy. The operation-scoped projection is already
-   covered by source/chunk identity and configured or generated reference
-   dependencies, while its `LLMInputMaterial.Digest` identifies the exact model
-   input; do not add a second digest API or operation-aware static fingerprint.
-9. Update focused schema, prompt, extractor, canonicalization, checkpoint, and
-   generated-reference tests. Cover unique-name empty selectors, successful
-   same-name selection, failure for an unsupported ambiguous mention,
-   partial/reordered ranges, no registry-to-occurrence evidence leakage,
-   all-or-nothing failure, empty registry/result behavior, and unchanged
-   durable ordering/deduplication.
-
-### Acceptance Criteria
-
-- The location prompt and private response contain no durable location ID.
-- Unique and same-name records resolve according to the final selector
-  contract.
-- Accepted durable output is unchanged in shape and still contains an exact
-  location ID/name pair.
-- Registry context cannot become durable occurrence evidence.
-
-### Validation
-
-```sh
-go fmt ./internal/modules/dnd/extract/locationoccurrences ./internal/modules/dnd/locations/registry
-go test ./internal/modules/dnd/locations/registry ./internal/modules/dnd/extract/locationoccurrences ./internal/modules/dnd/normalize/locationoccurrences ./internal/modules/dnd/validate/locationoccurrences/...
-```
-
-This stage is suitable for one gpt-5.6-terra prompt.
-
-## Stage 5: Replace Reconciliation Keys With Contextual Descriptors
-
-### Goal
-
-Change the shared NPC/item/location registry-normalization proposal contract so
-opaque candidate keys remain internal and the model sees and returns only
-names plus evidence coordinates.
-
-### Work
-
-1. Inspect:
-   - `internal/modules/dnd/shared/entityreconcile/`;
-   - `assets/dnd/shared/prompts/common-dnd-entity-reconciliation.md`;
-   - `assets/dnd/entity-reconciliation/schemas/`;
-   - all three registry normalization manifests and prompt tests; and
-   - the NPC, item, and location registry normalizers and reconciliation tests.
-2. Introduce one exported, defensively copied contextual selector type in
-   `entityreconcile` with JSON `name` and `source_refs`, plus a strict
-   source-free range type. Use it for candidate input views and for
-   `DuplicateGroup.Members` and `.Canonical`.
-3. Keep deterministic candidate keys only inside `Materials`. During
-   `BuildContext`, validate and canonicalize candidate references as today,
-   serialize candidate views without `key`, derive a stable internal lookup
-   from canonical selector JSON to the corresponding internal candidate key,
-   and detect descriptor collisions before eligibility is established.
-   Colliding candidates must not appear in the prompt input or become
-   eligible; their records remain in deterministic normalization output.
-4. Update `Materials.Assess` to resolve every returned selector through that
-   internal lookup before running the existing group assessment. Preserve
-   existing issue categories where their meaning still applies. Treat an
-   unknown or collided descriptor as an unknown/ineligible selection, discard
-   only the affected group, and retain existing overlap handling. `SafeGroup`
-   may continue returning internal candidate keys so the three domain
-   normalizers retain their position mapping; those keys are not model-facing.
-5. Rewrite `dnd_entity_reconcile_llm.v1.json` so members and canonical are
-   strict selector objects. Require non-empty `name` structurally where the
-   current schemas do so, require `source_refs`, and make each range strict
-   with required positive integer endpoints. Preserve `duplicate_groups` and
-   the existing semantic assessment of minimum group size, membership,
-   duplicates, eligibility, and overlap rather than moving every semantic
-   failure into JSON Schema.
-6. Rewrite the shared reconciliation fragment to tell the model to return
-   supplied contextual descriptors and never invent names or ranges. Remove
-   every instruction about opaque keys. Preserve all three manifests' message
-   ordering and cache controls.
-7. Change registry normalization policy identifiers to:
-   - `dnd.npc_registry.normalize.v4`;
-   - `dnd.item_registry.normalize.v2`; and
-   - `dnd.location_registry.normalize.v2`.
-8. Update shared and domain tests to cover candidate JSON without keys,
-   contextual proposal decoding, valid selector-to-internal-key mapping,
-   equal names with different evidence, descriptor collision exclusion,
-   unknown/partial/reordered descriptors, overlapping groups, canonical
-   membership, invalid structured-output fallback, currency safety, and
-   preservation of every non-applied deterministic candidate. Update prompt
-   asset fixtures to the new selector schema; do not snapshot prompt prose.
-
-### Acceptance Criteria
-
-- No registry normalization prompt input or private response contains a
-  `candidate-*` key.
-- Internal keys remain inaccessible to the model but may still support safe
-  deterministic position mapping.
-- All existing normalizer safety, retry, fallback, warning, currency, and
-  same-name-location policies remain intact.
-- Identical contextual descriptors cannot be arbitrarily reconciled.
-
-### Validation
-
-```sh
-go fmt ./internal/modules/dnd/shared/entityreconcile ./internal/modules/dnd/normalize/npcregistry ./internal/modules/dnd/normalize/itemregistry ./internal/modules/dnd/normalize/locationregistry
-go test ./internal/modules/dnd/shared/entityreconcile ./internal/modules/dnd/normalize/npcregistry ./internal/modules/dnd/normalize/itemregistry ./internal/modules/dnd/normalize/locationregistry
-```
-
-This is the largest stage, but it is one cohesive shared-contract migration
-and is suitable for one gpt-5.6-terra prompt when implemented exactly within
-the listed packages. Do not combine it with occurrence or documentation work.
-
-## Stage 6: Record The Decision And Update Canonical Documentation
-
-### Goal
-
-Document the implemented policy in its durable architectural, internal, and
-integration homes without duplicating volatile details or presenting roadmap
-work as current behavior prematurely.
-
-### Work
-
-1. Re-read `docs/policy/documentation.md`, ADR-0003, ADR-0009, ADR-0011,
-   `docs/internal/dnd.md`, `docs/internal/llm.md`, and the six affected registry
-   and occurrence integration documents. Verify the code before describing it.
-2. Add
-   `docs/adr/0012-resolve-opaque-entity-identifiers-deterministically.md` in
-   the repository's Nygard ADR format with status `Accepted` and the actual
-   implementation date. Record the model-semantic/deterministic-identity
-   boundary, source-coordinate allowance, request-local-label exception,
-   alternatives, ambiguity behavior, and consequences. Link ADR-0003 and
-   ADR-0009 rather than repeating their complete decisions.
-3. Add a concise normative invariant under the LLM boundary in
-   `docs/policy/architecture.md`: callers use contextual model selections and
-   attach opaque application identities deterministically when possible. Link
-   ADR-0012 for rationale.
-4. Update `docs/internal/dnd.md` to replace exact model-facing `{id,name}`
-   claims with the implemented NPC/item names-only and location contextual
-   selector behavior. Document reconciliation descriptors, internal-only keys,
-   all-or-nothing occurrence mapping failures, normalization fallback, and the
-   separation between registry and occurrence evidence. Do not duplicate the
-   private JSON schemas.
-5. Add only a short ownership clarification to `docs/internal/llm.md`: the
-   calling module resolves contextual selections; PromptKit and its adapter do
-   not own entity identity.
-6. Update these durable integration contracts while preserving their public
-   ID-bearing wire examples and schema statements:
-   - `docs/integrations/dnd-npc-registry-artifacts.md`;
-   - `docs/integrations/dnd-npc-occurrence-artifacts.md`;
-   - `docs/integrations/dnd-item-registry-artifacts.md`;
-   - `docs/integrations/dnd-item-occurrence-artifacts.md`;
-   - `docs/integrations/dnd-location-registry-artifacts.md`; and
-   - `docs/integrations/dnd-location-occurrence-artifacts.md`.
-   Remove claims that LLM consumers receive `{id,name}` or that raw model
-   output supplies an ID. State that Notarius maps contextual output into the
-   unchanged exact durable pair.
-7. Revise the generic LLM-assisted deduplication entry in
-   `docs/roadmap/future.md`: stable unique IDs remain internal deterministic
-   state, while a future model proposal uses contextual descriptors or a
-   specifically justified request-local short label.
-8. Do not change README, CLI, configuration, operations, examples, or public
-   schema files; this feature has no user-selectable surface or public wire
-   change.
-
-### Acceptance Criteria
-
-- ADR-0012 owns rationale; architecture owns the normative boundary; internal
-  docs own mechanics; integration docs own unchanged durable contracts; and
-  the future roadmap no longer proposes durable IDs as the default model
-  selector.
-- No current-behavior document claims that a model copies hash-based entity
-  IDs or opaque reconciliation keys.
-- Documentation does not duplicate private schemas or implementation history.
-
-### Validation
-
-```sh
-git diff --check
-rg -n '\{id,name\}|ID/name grounding|Candidate keys are opaque|candidate-[0-9]' docs assets/dnd
-```
-
-Review every search result semantically; durable wire-contract ID/name
-requirements and internal test fixtures are not automatically errors.
-
-This stage is suitable for one gpt-5.6-terra prompt.
-
-## Stage 7: Integration Audit And Final Verification
-
-### Goal
-
-Verify the assembled D&D family, remove obsolete identity-copy paths, and
-finish with a clean, policy-compliant implementation.
-
-### Work
-
-1. Audit every maintained D&D prompt manifest, selected fragment, private
-   schema, and constructed prompt projection. Confirm that no model is asked to
-   reproduce `npc:sha256:...`, `item:sha256:...`,
-   `location:sha256:...`, `candidate-*`, a UUID, a digest, or another opaque
-   entity handle. Do not confuse runtime metadata or durable output contracts
-   with model-visible material.
-2. Trace all former APIs and fields, including `IdentityPromptInput`,
-   ID-bearing item/location prompt projections, private `NPCID`/`ItemID`/
-   `LocationID` response fields, and model-visible candidate keys. Remove dead
-   code, obsolete comments, stale test names, and unused assets. Retain
-   identity-only digests and exact durable lookup APIs used by deterministic
-   consumers.
-3. Review prompt fingerprint registration and checkpoint fingerprints. Confirm
-   that each affected prompt/schema/policy/projection change invalidates the
-   relevant operation and that unrelated D&D lanes retain their existing
-   fingerprints.
-4. Run representative production registration and multi-step pipeline tests
-   using existing fakes. Update only tests whose stable behavior changed.
-   Confirm generated NPC/item/location registry handoffs still prepare and
-   that final durable occurrences encode and validate under their existing
-   `v1` codecs.
-5. Run formatting, focused suites, full tests, vet, build, and documentation
-   whitespace checks. Fix only failures caused by this feature. Report any
-   unrelated pre-existing failure without broadening scope.
-6. Review the feature roadmap acceptance criteria one by one. Do not delete
-   `contextual-entity-grounding.md` or this implementation plan in this stage;
-   roadmap retirement is a separate maintainer action after review.
-
-### Acceptance Criteria
-
-- All feature-roadmap acceptance criteria are met.
-- The repository contains no obsolete model-facing opaque-identity path.
-- Public artifacts and generated handoffs remain compatible.
-- Tests are focused on behavior rather than prose or implementation shape.
-- The worktree contains only intentional feature and documentation changes.
-
-### Validation
-
-```sh
-go fmt ./internal/modules/dnd/...
-go test ./internal/modules/dnd/...
-go test ./internal/modules/integration/...
-go test ./...
-go vet ./...
-go build ./cmd/notarius
-git diff --check
-git status --short
-```
-
-This stage is suitable for one gpt-5.6-terra prompt.

go.mod

@@ -3,9 +3,14 @@ module gitea.maximumdirect.net/eric/notarius
 go 1.25.5
 
 require (
-	gitea.maximumdirect.net/eric/promptkit v0.5.0
+	gitea.maximumdirect.net/eric/promptkit v0.9.0
 	github.com/santhosh-tekuri/jsonschema/v6 v6.0.2
 	gopkg.in/yaml.v3 v3.0.1
 )
 
 require golang.org/x/text v0.40.0
+
+require (
+	gitea.maximumdirect.net/eric/promptkit-backend-openrouter v1.0.0 // indirect
+	gitea.maximumdirect.net/eric/promptkit-backend-rakestrawhome v1.0.0 // indirect
+)

go.sum

@@ -1,5 +1,9 @@
-gitea.maximumdirect.net/eric/promptkit v0.5.0 h1:jnpazLyyNhWrB2xzwwtUkNUfktkTdkENTwuSPnKiYrc=
+gitea.maximumdirect.net/eric/promptkit v0.9.0 h1:IpvDRC8L6xRxQ9hpuyKOmMc5b6MeLTKYyx+h1YAjy08=
-gitea.maximumdirect.net/eric/promptkit v0.5.0/go.mod h1:R95NM6fbMDGDC0/UomgnSBP6ui2ns+8SZb8bESNvrDQ=
+gitea.maximumdirect.net/eric/promptkit v0.9.0/go.mod h1:oMJ/WUJImUtwJ5e+6MAGECPYAErAkOaKel0G+3T/b4E=
+gitea.maximumdirect.net/eric/promptkit-backend-openrouter v1.0.0 h1:lc062euk2qseO//D762i3JaFyulDNML3eQQX7DkYTho=
+gitea.maximumdirect.net/eric/promptkit-backend-openrouter v1.0.0/go.mod h1:AIa7kAu2mfrRQgcspe4L+DW51WqgnALQT60lqkEywJI=
+gitea.maximumdirect.net/eric/promptkit-backend-rakestrawhome v1.0.0 h1:j9YY7wsTVjzke2kHH4YAzpU0oUpM+x+nXwl1IeS+2eg=
+gitea.maximumdirect.net/eric/promptkit-backend-rakestrawhome v1.0.0/go.mod h1:4RNS+LILDg4JbS4Ts9Lwy1C92wauXJIbeQaalps4Koo=
 github.com/dlclark/regexp2 v1.11.0 h1:G/nrcoOa7ZXlpoa/91N3X7mM3r8eIlMBBJZvsz/mxKI=
 github.com/dlclark/regexp2 v1.11.0/go.mod h1:DHkYz0B9wPfa6wondMfaivmHpzrQ3v9q8cnmRbL6yW8=
 github.com/santhosh-tekuri/jsonschema/v6 v6.0.2 h1:KRzFb2m7YtdldCEkzs6KqmJw4nqEVZGK7IN2kJkjTuQ=

internal/buildinfo/buildinfo.go

@@ -0,0 +1,36 @@
+// Package buildinfo resolves the product version embedded in a Notarius build.
+package buildinfo
+
+import (
+	"fmt"
+	"regexp"
+	"runtime/debug"
+)
+
+var stableVersion = regexp.MustCompile(`^v(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)\.(0|[1-9][0-9]*)$`)
+
+// Override is set at link time for controlled builds.
+var Override string
+
+// Version returns the release version embedded in the build, or development
+// when the build does not carry a stable release tag.
+func Version() (string, error) {
+	buildVersion := ""
+	if info, ok := debug.ReadBuildInfo(); ok {
+		buildVersion = info.Main.Version
+	}
+	return resolve(Override, buildVersion)
+}
+
+func resolve(override, buildVersion string) (string, error) {
+	if override != "" {
+		if !stableVersion.MatchString(override) {
+			return "", fmt.Errorf("build version override is not a stable release tag")
+		}
+		return override, nil
+	}
+	if stableVersion.MatchString(buildVersion) {
+		return buildVersion, nil
+	}
+	return "development", nil
+}

internal/buildinfo/buildinfo_test.go

@@ -0,0 +1,45 @@
+package buildinfo
+
+import "testing"
+
+func TestResolve(t *testing.T) {
+	tests := []struct {
+		name         string
+		override     string
+		buildVersion string
+		want         string
+		wantErr      bool
+	}{
+		{name: "stable main module version", buildVersion: "v1.2.3", want: "v1.2.3"},
+		{name: "zero version", buildVersion: "v0.0.0", want: "v0.0.0"},
+		{name: "override takes precedence", override: "v2.3.4", buildVersion: "v1.2.3", want: "v2.3.4"},
+		{name: "invalid override", override: "version", buildVersion: "v1.2.3", wantErr: true},
+		{name: "override with whitespace", override: " v1.2.3", wantErr: true},
+		{name: "leading zero major", buildVersion: "v01.2.3", want: "development"},
+		{name: "leading zero minor", buildVersion: "v1.02.3", want: "development"},
+		{name: "leading zero patch", buildVersion: "v1.2.03", want: "development"},
+		{name: "build version with whitespace", buildVersion: "v1.2.3 ", want: "development"},
+		{name: "prerelease", buildVersion: "v1.2.3-rc.1", want: "development"},
+		{name: "build suffix", buildVersion: "v1.2.3+build.1", want: "development"},
+		{name: "pseudo version", buildVersion: "v0.0.0-20260102030405-abcdef123456", want: "development"},
+		{name: "development build", buildVersion: "(devel)", want: "development"},
+		{name: "missing build information", want: "development"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got, err := resolve(tt.override, tt.buildVersion)
+			if tt.wantErr {
+				if err == nil {
+					t.Fatal("resolve() error = nil, want error")
+				}
+				return
+			}
+			if err != nil {
+				t.Fatalf("resolve() error = %v", err)
+			}
+			if got != tt.want {
+				t.Fatalf("resolve() = %q, want %q", got, tt.want)
+			}
+		})
+	}
+}

internal/cli/assembled_enemy_event_codec_contract_test.go

@@ -0,0 +1,81 @@
+package cli
+
+import (
+	"context"
+	"errors"
+	"strings"
+	"testing"
+
+	"gitea.maximumdirect.net/eric/notarius/internal/core/source"
+	"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
+	"gitea.maximumdirect.net/eric/notarius/internal/framework/pipeline"
+	"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd"
+)
+
+const invalidEnemyEventExtractorKey = "test/dnd/invalid-enemy-events"
+
+func TestAssembledEnemyEventLaneRejectsInvalidFinalArtifactDespiteValidatorOverrides(t *testing.T) {
+	components := productionTestComponents(t)
+	if err := pipeline.RegisterExtractor[dnd.EnemyEventList](components.registries.Extractors, pipeline.ModuleSpec{
+		Key:            invalidEnemyEventExtractorKey,
+		Stage:          pipeline.StageExtract,
+		ExecutionClass: contracts.ExecutionClassDeterministic,
+		Requires:       []string{"chunks", "source.transcript"},
+		Provides:       []string{"dnd.enemy_events"},
+		ArtifactKind:   dnd.EnemyEventListKind,
+	}, func() (contracts.Extractor[dnd.EnemyEventList], error) {
+		return invalidEnemyEventExtractor{}, nil
+	}); err != nil {
+		t.Fatalf("register extractor: %v", err)
+	}
+
+	accept := pipeline.ValidatorOverride{Set: true, Validators: []pipeline.ModuleBinding{pipeline.Binding("generic/always_accept")}}
+	resolved, err := pipeline.ResolvePipeline(pipeline.PipelineProfile{
+		ID:    "assembled-invalid-enemy-events",
+		Input: pipeline.Binding("seriatim"),
+		Chunk: pipeline.ModuleBinding{Module: "generic", Options: map[string]any{"max_units": 1}},
+		Artifacts: map[string]pipeline.ArtifactLaneProfile{
+			"enemy-events": {
+				Extract:   pipeline.ModuleBinding{Module: invalidEnemyEventExtractorKey, Validators: accept},
+				Normalize: pipeline.ModuleBinding{Module: pipeline.DefaultNormalizeModule, Validators: accept},
+			},
+		},
+		Output: pipeline.Binding("json"),
+	}, pipeline.ResolveOptions{}, catalogFromRegistries(components.registries))
+	if err != nil {
+		t.Fatalf("ResolvePipeline() error = %v", err)
+	}
+
+	prepared, err := pipeline.Prepare(resolved, components.registries, pipeline.ModuleDependencies{})
+	if err != nil {
+		t.Fatalf("Prepare() error = %v", err)
+	}
+	_, err = pipeline.New().Run(context.Background(), pipeline.RunInput{
+		Prepared:       prepared,
+		RawInput:       readRepositoryFile(t, "examples", "seriatim-minimal-transcript.json"),
+		ChunkCacheMode: pipeline.ChunkCacheBypass,
+	})
+	if err == nil || !strings.Contains(err.Error(), "serialize accepted extract output") || !strings.Contains(err.Error(), "must not exceed") {
+		t.Fatalf("Run() error = %v, want final durable range rejection", err)
+	}
+}
+
+type invalidEnemyEventExtractor struct{}
+
+func (invalidEnemyEventExtractor) Key() string { return invalidEnemyEventExtractorKey }
+
+func (invalidEnemyEventExtractor) ReferenceSlots() []contracts.ReferenceSlot { return nil }
+
+func (invalidEnemyEventExtractor) Extract(ctx context.Context, req contracts.TypedExtractionRequest) (contracts.TypedExtractionResult[dnd.EnemyEventList], error) {
+	if err := ctx.Err(); err != nil {
+		return contracts.TypedExtractionResult[dnd.EnemyEventList]{}, err
+	}
+	if req.Source == nil {
+		return contracts.TypedExtractionResult[dnd.EnemyEventList]{}, errors.New("assembled extractor requires source")
+	}
+	return contracts.TypedExtractionResult[dnd.EnemyEventList]{Value: dnd.EnemyEventList{Events: []dnd.EnemyEvent{{
+		Name:       "Ashfang",
+		Kind:       dnd.EnemyEventKindEngaged,
+		SourceRefs: []source.SourceRef{{SourceID: req.Source.ID, StartUnitID: 2, EndUnitID: 1}},
+	}}}}, nil
+}

internal/cli/assembled_spell_pipeline_contract_test.go

@@ -21,6 +21,10 @@ import (
 
 const assembledSpellExtractorKey = "test/dnd/spell-casts"
 
+const assembledCorrectingSpellExtractorKey = "test/dnd/correcting-spell-casts"
+
+const assembledDirectSpellValidatorKey = "test/dnd/direct-spell-correction"
+
 func TestAssembledSpellPipelineNormalizesMergedCasts(t *testing.T) {
 	registries, resolved, extractor := assembledSpellPipeline(t, assembledSpellPipelineOptions{})
 	prepared, err := pipeline.Prepare(resolved, registries, pipeline.ModuleDependencies{})
@@ -67,28 +71,39 @@ func TestAssembledSpellPipelineNormalizesMergedCasts(t *testing.T) {
 		t.Fatalf("distinct cast = %#v, want separate evidence event", distinct)
 	}
 
-	wantWarningReasons := []string{
+	wantDiagnosticReasons := []string{
 		spellnormalize.ReasonCodeSpellNameCanonicalized,
 		spellnormalize.ReasonCodeSourceReferencesNormalized,
 		spellnormalize.ReasonCodeDuplicateSpellCastCollapsed,
 		"spell_not_near_source",
 	}
-	gotWarningReasons := make([]string, len(output.Warnings))
+	gotDiagnosticReasons := make([]string, len(output.Diagnostics.Groups))
-	for index, warning := range output.Warnings {
+	for index, group := range output.Diagnostics.Groups {
-		gotWarningReasons[index] = warning.ReasonCode
+		gotDiagnosticReasons[index] = group.ReasonCode
 	}
-	if !reflect.DeepEqual(gotWarningReasons, wantWarningReasons) {
+	if !reflect.DeepEqual(gotDiagnosticReasons, wantDiagnosticReasons) {
-		t.Fatalf("warnings = %#v, want deterministic normalize and validation warnings", output.Warnings)
+		t.Fatalf("diagnostics = %#v, want deterministic normalize and validation diagnostics", output.Diagnostics)
 	}
-	if output.Warnings[2].Scope != "spell_casts[0]" || !strings.Contains(output.Warnings[2].Message, "retained input index 0") || !strings.Contains(output.Warnings[2].Message, "removed input indices [1]") {
+	if output.Diagnostics.Groups[2].Samples[0].Scope != "spell_casts[0]" || !strings.Contains(output.Diagnostics.Groups[2].Samples[0].Message, "retained input index 0") || !strings.Contains(output.Diagnostics.Groups[2].Samples[0].Message, "removed input indices [1]") {
-		t.Fatalf("duplicate warning = %#v, want retained and removed merged indices", output.Warnings[2])
+		t.Fatalf("duplicate diagnostic = %#v, want retained and removed merged indices", output.Diagnostics.Groups[2])
 	}
 
 	warningsFile := decodeAssembledOutput[struct {
-		Warnings []contracts.Warning `json:"warnings"`
+		Groups []contracts.DiagnosticGroup `json:"groups"`
 	}](t, output.OutputFiles, "warnings.json")
-	if !reflect.DeepEqual(warningsFile.Warnings, output.Warnings) {
+	if len(warningsFile.Groups) != 0 {
-		t.Fatalf("warnings file = %#v, run warnings = %#v, want manifest output path to preserve warnings", warningsFile.Warnings, output.Warnings)
+		t.Fatalf("warnings file = %#v, want no process warnings for advisory-only diagnostics", warningsFile.Groups)
+	}
+	diagnosticsFile := decodeAssembledOutput[struct {
+		SchemaVersion                string                      `json:"schema_version"`
+		GroupCount                   int                         `json:"group_count"`
+		OccurrenceCount              int                         `json:"occurrence_count"`
+		Truncated                    bool                        `json:"truncated"`
+		UnrepresentedOccurrenceCount int                         `json:"unrepresented_occurrence_count"`
+		Groups                       []contracts.DiagnosticGroup `json:"groups"`
+	}](t, output.OutputFiles, "diagnostics.json")
+	if diagnosticsFile.SchemaVersion != "notarius.diagnostics.v1" || diagnosticsFile.GroupCount != len(output.Diagnostics.Groups) || !reflect.DeepEqual(diagnosticsFile.Groups, output.Diagnostics.Groups) || diagnosticsFile.OccurrenceCount != diagnosticGroupOccurrences(output.Diagnostics.Groups)+output.Diagnostics.UnrepresentedOccurrenceCount || diagnosticsFile.Truncated != output.Diagnostics.Truncated || diagnosticsFile.UnrepresentedOccurrenceCount != output.Diagnostics.UnrepresentedOccurrenceCount {
+		t.Fatalf("diagnostics file = %#v, run diagnostics = %#v", diagnosticsFile, output.Diagnostics)
 	}
 	manifest := decodeAssembledOutput[artifacts.RunManifest](t, output.OutputFiles, "manifest.json")
 	if len(manifest.ArtifactLanes) != 1 || manifest.ArtifactLanes[0].Normalizer != spellnormalize.Key {
@@ -101,6 +116,30 @@ func TestAssembledSpellPipelineNormalizesMergedCasts(t *testing.T) {
 	}
 }
 
+func TestAssembledSpellPipelineCorrectsRejectedDirectExtraction(t *testing.T) {
+	registries, resolved, extractor := assembledCorrectingSpellPipeline(t)
+	prepared, err := pipeline.Prepare(resolved, registries, pipeline.ModuleDependencies{})
+	if err != nil {
+		t.Fatalf("Prepare() error = %v, want nil", err)
+	}
+
+	output, err := pipeline.New().Run(context.Background(), pipeline.RunInput{
+		Prepared:       prepared,
+		RawInput:       readRepositoryFile(t, "examples", "seriatim-minimal-transcript.json"),
+		ChunkCacheMode: pipeline.ChunkCacheBypass,
+	})
+	if err != nil {
+		t.Fatalf("Run() error = %v, want nil", err)
+	}
+	if len(output.Rejected) != 0 || output.Manifest.ValidationStatus != "approved" || len(output.NormalizeOutputs) != 1 {
+		t.Fatalf("run output = %#v, want corrected accepted spell output", output)
+	}
+	correction := extractor.correctionSnapshot()
+	if correction == nil || string(correction.AssistantResponse) != `{"spell":"Mysterious Burst"}` || !strings.Contains(correction.UserGuidance, "use a known spell name") || !strings.Contains(correction.UserGuidance, "complete corrected replacement") || strings.Contains(correction.UserGuidance, "unknown_spell") || strings.Contains(correction.UserGuidance, "spell is not in the catalog") {
+		t.Fatalf("extract correction = %#v, want exact rejected model response and semantic replacement guidance only", correction)
+	}
+}
+
 func TestAssembledSpellPipelineHonorsNormalizeValidatorOverride(t *testing.T) {
 	registries, resolved, _ := assembledSpellPipeline(t, assembledSpellPipelineOptions{normalizeValidatorOverride: true})
 	var normalizeChain *pipeline.ResolvedValidatorChain
@@ -127,15 +166,16 @@ func TestAssembledSpellPipelineHonorsNormalizeValidatorOverride(t *testing.T) {
 	if err != nil || output.Manifest.ValidationStatus != "approved" || len(output.Rejected) != 0 || len(output.NormalizeOutputs) != 1 {
 		t.Fatalf("Run() error = %v output = %#v, want approved override run", err, output)
 	}
-	for _, warning := range output.Warnings {
+	for _, group := range output.Diagnostics.Groups {
-		if warning.ReasonCode == "spell_not_near_source" {
+		if group.ReasonCode == "spell_not_near_source" {
-			t.Fatalf("warnings = %#v, want explicit validator override to replace default relatedness chain", output.Warnings)
+			t.Fatalf("diagnostics = %#v, want explicit validator override to replace default relatedness chain", output.Diagnostics)
 		}
 	}
 }
 
-func TestAssembledSpellPipelineRejectsUnknownSpellWithoutPromotingAttemptWarning(t *testing.T) {
+func TestAssembledSpellPipelinePromotesTerminalUnknownSpellDiagnostics(t *testing.T) {
 	registries, resolved, _ := assembledSpellPipeline(t, assembledSpellPipelineOptions{unknownSpell: true})
+	resolved.Steps[0].ArtifactLanes[0].NormalizeValidationPolicy.SemanticRejection = pipeline.SemanticRejectionRejectOutput
 	prepared, err := pipeline.Prepare(resolved, registries, pipeline.ModuleDependencies{})
 	if err != nil {
 		t.Fatalf("Prepare() error = %v, want nil", err)
@@ -161,14 +201,12 @@ func TestAssembledSpellPipelineRejectsUnknownSpellWithoutPromotingAttemptWarning
 	if !reflect.DeepEqual(rejectedFile.Rejected, output.Rejected) {
 		t.Fatalf("rejected file = %#v, run rejections = %#v, want durable rejection diagnostic", rejectedFile.Rejected, output.Rejected)
 	}
-	for _, warning := range output.Warnings {
+	if len(output.Diagnostics.Groups) != 2 || output.Diagnostics.Groups[0].ReasonCode != spellnormalize.ReasonCodeSpellNameUnresolved || output.Diagnostics.Groups[0].Samples[0].Scope != "spell_casts[0]" || output.Diagnostics.Groups[1].ReasonCode != "spell_not_near_source" {
-		if warning.ReasonCode == spellnormalize.ReasonCodeSpellNameUnresolved {
+		t.Fatalf("diagnostics = %#v, want complete terminal normalize validation diagnostics", output.Diagnostics)
-			t.Fatalf("warnings = %#v, want rejected-attempt warning to remain non-durable", output.Warnings)
-		}
 	}
 }
 
-func TestAssembledSpellPipelinePromotesUnknownSpellWarningWhenOverrideAccepts(t *testing.T) {
+func TestAssembledSpellPipelinePromotesUnknownSpellAdvisoryWhenOverrideAccepts(t *testing.T) {
 	registries, resolved, _ := assembledSpellPipeline(t, assembledSpellPipelineOptions{normalizeValidatorOverride: true, unknownSpell: true})
 	prepared, err := pipeline.Prepare(resolved, registries, pipeline.ModuleDependencies{})
 	if err != nil {
@@ -192,14 +230,20 @@ func TestAssembledSpellPipelinePromotesUnknownSpellWarningWhenOverrideAccepts(t
 	if len(normalized.SpellCasts) != 1 || normalized.SpellCasts[0].Spell != "Mysterious Burst" {
 		t.Fatalf("normalized casts = %#v, want unresolved name preserved", normalized.SpellCasts)
 	}
-	if len(output.Warnings) != 1 || output.Warnings[0].ReasonCode != spellnormalize.ReasonCodeSpellNameUnresolved || output.Warnings[0].Scope != "spell_casts[0]" {
+	if len(output.Diagnostics.Groups) != 1 || output.Diagnostics.Groups[0].ReasonCode != spellnormalize.ReasonCodeSpellNameUnresolved || output.Diagnostics.Groups[0].Samples[0].Scope != "spell_casts[0]" {
-		t.Fatalf("warnings = %#v, want promoted scoped unresolved-name warning", output.Warnings)
+		t.Fatalf("diagnostics = %#v, want promoted scoped unresolved-name diagnostic", output.Diagnostics)
 	}
 	warningsFile := decodeAssembledOutput[struct {
-		Warnings []contracts.Warning `json:"warnings"`
+		Groups []contracts.DiagnosticGroup `json:"groups"`
 	}](t, output.OutputFiles, "warnings.json")
-	if !reflect.DeepEqual(warningsFile.Warnings, output.Warnings) {
+	if len(warningsFile.Groups) != 0 {
-		t.Fatalf("warnings file = %#v, run warnings = %#v, want durable unresolved-name warning", warningsFile.Warnings, output.Warnings)
+		t.Fatalf("warnings file = %#v, want no process warnings for an advisory diagnostic", warningsFile.Groups)
+	}
+	diagnosticsFile := decodeAssembledOutput[struct {
+		Groups []contracts.DiagnosticGroup `json:"groups"`
+	}](t, output.OutputFiles, "diagnostics.json")
+	if !reflect.DeepEqual(diagnosticsFile.Groups, output.Diagnostics.Groups) {
+		t.Fatalf("diagnostics file = %#v, run diagnostics = %#v", diagnosticsFile.Groups, output.Diagnostics)
 	}
 }
 
@@ -208,6 +252,47 @@ type assembledSpellPipelineOptions struct {
 	unknownSpell               bool
 }
 
+func assembledCorrectingSpellPipeline(t *testing.T) (pipeline.Registries, pipeline.ResolvedPipeline, *assembledCorrectingSpellExtractor) {
+	t.Helper()
+	components := productionTestComponents(t)
+	extractor := &assembledCorrectingSpellExtractor{}
+	if err := pipeline.RegisterExtractor[dnd.SpellList](components.registries.Extractors, pipeline.ModuleSpec{
+		Key:                assembledCorrectingSpellExtractorKey,
+		Stage:              pipeline.StageExtract,
+		ExecutionClass:     contracts.ExecutionClassLLMBacked,
+		CorrectionProtocol: contracts.CorrectionProtocolSingleResponseV1,
+		Requires:           []string{"chunks", "source.transcript"},
+		Provides:           []string{"dnd.spell_casts"},
+		ArtifactKind:       dnd.SpellListKind,
+	}, func() (contracts.Extractor[dnd.SpellList], error) {
+		return extractor, nil
+	}); err != nil {
+		t.Fatalf("register correcting extractor: %v", err)
+	}
+	if err := pipeline.RegisterTypedValidator[dnd.SpellList](components.registries.Validators, dnd.SpellListKind, pipeline.ValidatorSpec{Key: assembledDirectSpellValidatorKey, ExecutionClass: contracts.ExecutionClassDeterministic}, func() (contracts.TypedValidator[dnd.SpellList], error) {
+		return assembledDirectSpellValidator{}, nil
+	}); err != nil {
+		t.Fatalf("register direct spell validator: %v", err)
+	}
+
+	extract := pipeline.Binding(assembledCorrectingSpellExtractorKey)
+	extract.Retries = 1
+	extract.Validators = pipeline.ValidatorOverride{Set: true, Validators: []pipeline.ModuleBinding{{Module: assembledDirectSpellValidatorKey}}}
+	resolved, err := pipeline.ResolvePipeline(pipeline.PipelineProfile{
+		ID:    "assembled-dnd-correcting-spells",
+		Input: pipeline.Binding("seriatim"),
+		Chunk: pipeline.ModuleBinding{Module: "generic", Options: map[string]any{"max_units": 1}},
+		Artifacts: map[string]pipeline.ArtifactLaneProfile{
+			"spells": {Extract: extract, Normalize: pipeline.Binding(spellnormalize.Key)},
+		},
+		Output: pipeline.Binding("json"),
+	}, pipeline.ResolveOptions{}, catalogFromRegistries(components.registries))
+	if err != nil {
+		t.Fatalf("ResolvePipeline() error = %v, want nil", err)
+	}
+	return components.registries, resolved, extractor
+}
+
 func assembledSpellPipeline(t *testing.T, options assembledSpellPipelineOptions) (pipeline.Registries, pipeline.ResolvedPipeline, *assembledSpellExtractor) {
 	t.Helper()
 	components := productionTestComponents(t)
@@ -253,6 +338,69 @@ type assembledSpellExtractor struct {
 	unknownSpell bool
 }
 
+type assembledCorrectingSpellExtractor struct {
+	mu         sync.Mutex
+	correction *contracts.SemanticCorrection
+}
+
+func (*assembledCorrectingSpellExtractor) Key() string { return assembledCorrectingSpellExtractorKey }
+
+func (*assembledCorrectingSpellExtractor) ReferenceSlots() []contracts.ReferenceSlot { return nil }
+
+func (e *assembledCorrectingSpellExtractor) Extract(_ context.Context, req contracts.TypedExtractionRequest) (contracts.TypedExtractionResult[dnd.SpellList], error) {
+	if req.Source == nil || req.Chunk == nil {
+		return contracts.TypedExtractionResult[dnd.SpellList]{}, fmt.Errorf("correcting assembled extractor requires source and chunk")
+	}
+	response := `{"spell":"accepted"}`
+	value := dnd.SpellList{SpellCasts: []dnd.SpellCast{}}
+	if req.Chunk.Index == 0 && req.Correction == nil {
+		response = `{"spell":"Mysterious Burst"}`
+		value.SpellCasts = []dnd.SpellCast{{Caster: "Aria", Spell: "Mysterious Burst", SourceRefs: []source.SourceRef{{SourceID: req.Source.ID, StartUnitID: 1, EndUnitID: 1}}}}
+	}
+	if req.Chunk.Index == 0 && req.Correction != nil {
+		correction, err := contracts.CloneSemanticCorrection(req.Correction)
+		if err != nil {
+			return contracts.TypedExtractionResult[dnd.SpellList]{}, err
+		}
+		e.mu.Lock()
+		e.correction = correction
+		e.mu.Unlock()
+		value.SpellCasts = []dnd.SpellCast{{Caster: "Aria", Spell: "Cure Wounds", SourceRefs: []source.SourceRef{{SourceID: req.Source.ID, StartUnitID: 1, EndUnitID: 1}}}}
+	}
+	candidate, err := contracts.NewModelCandidate([]byte(response), contracts.CorrectionProtocolSingleResponseV1)
+	if err != nil {
+		return contracts.TypedExtractionResult[dnd.SpellList]{}, err
+	}
+	return contracts.TypedExtractionResult[dnd.SpellList]{Value: value, ModelCandidate: candidate}, nil
+}
+
+func (e *assembledCorrectingSpellExtractor) correctionSnapshot() *contracts.SemanticCorrection {
+	e.mu.Lock()
+	defer e.mu.Unlock()
+	correction, err := contracts.CloneSemanticCorrection(e.correction)
+	if err != nil {
+		return nil
+	}
+	return correction
+}
+
+type assembledDirectSpellValidator struct{}
+
+func (assembledDirectSpellValidator) Name() string { return assembledDirectSpellValidatorKey }
+
+func (assembledDirectSpellValidator) ExecutionClass() contracts.ExecutionClass {
+	return contracts.ExecutionClassDeterministic
+}
+
+func (assembledDirectSpellValidator) Validate(_ context.Context, req contracts.TypedValidationRequest[dnd.SpellList]) (contracts.ValidationResult, error) {
+	for _, cast := range req.Value.SpellCasts {
+		if cast.Spell == "Mysterious Burst" {
+			return contracts.ValidationResult{Approved: false, ReasonCode: "unknown_spell", Message: "spell is not in the catalog", CorrectionGuidance: "use a known spell name"}, nil
+		}
+	}
+	return contracts.ValidationResult{Approved: true}, nil
+}
+
 func (e *assembledSpellExtractor) Key() string { return assembledSpellExtractorKey }
 
 func (*assembledSpellExtractor) ReferenceSlots() []contracts.ReferenceSlot { return nil }
@@ -298,6 +446,14 @@ func (e *assembledSpellExtractor) chunkIndexesSnapshot() []int {
 	return append([]int(nil), e.chunkIndexes...)
 }
 
+func diagnosticGroupOccurrences(groups []contracts.DiagnosticGroup) int {
+	count := 0
+	for _, group := range groups {
+		count += group.OccurrenceCount
+	}
+	return count
+}
+
 func decodeAssembledOutput[T any](t *testing.T, files []contracts.OutputFile, name string) T {
 	t.Helper()
 	for _, file := range files {

internal/cli/command_contract_test.go

@@ -8,6 +8,8 @@ import (
 	"path/filepath"
 	"strings"
 	"testing"
+
+	"gitea.maximumdirect.net/eric/notarius/internal/buildinfo"
 )
 
 func TestCommandHelpSpellingsWriteUsageToStdout(t *testing.T) {
@@ -38,6 +40,7 @@ func TestCommandSyntaxErrorsUseStderrAndExitTwo(t *testing.T) {
 		{name: "unknown pipelines subcommand", args: []string{"pipelines", "unknown"}, want: "unknown pipelines subcommand"},
 		{name: "malformed run flag", args: []string{"run", "demo", "--chunk_cache", "invalid"}, want: "not supported"},
 		{name: "unknown flag", args: []string{"config", "validate", "--unknown"}, want: "flag provided but not defined"},
+		{name: "version arguments", args: []string{"--version", "extra"}, want: "--version does not accept arguments"},
 	}
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
@@ -50,6 +53,33 @@ func TestCommandSyntaxErrorsUseStderrAndExitTwo(t *testing.T) {
 	}
 }
 
+func TestCommandVersionOutput(t *testing.T) {
+	previous := buildinfo.Override
+	t.Cleanup(func() { buildinfo.Override = previous })
+
+	tests := []struct {
+		name       string
+		override   string
+		wantCode   int
+		wantStdout string
+		wantStderr string
+	}{
+		{name: "development", wantStdout: "notarius development\n"},
+		{name: "release override", override: "v1.2.3", wantStdout: "notarius v1.2.3\n"},
+		{name: "invalid override", override: "release", wantCode: 1, wantStderr: "notarius: build version override is not a stable release tag\n"},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			buildinfo.Override = tt.override
+			var stdout, stderr bytes.Buffer
+			code := RunWithOptions([]string{"--version"}, &stdout, &stderr, Options{})
+			if code != tt.wantCode || stdout.String() != tt.wantStdout || stderr.String() != tt.wantStderr {
+				t.Fatalf("code=%d stdout=%q stderr=%q", code, stdout.String(), stderr.String())
+			}
+		})
+	}
+}
+
 func TestConfigDiscoveryPrefersExplicitPathThenEnvironment(t *testing.T) {
 	explicit := writeCommandConfig(t, "explicit", "alpha")
 	environment := writeCommandConfig(t, "environment", "beta")

internal/cli/dnd_enemy_events_contract_test.go

@@ -189,10 +189,18 @@ func TestMaintainedCompleteExamplePublishesRegistryBackedEntityOccurrences(t *te
 	}
 
 	evidence := readProductionJSON[evidencecontext.Document](t, filepath.Join(runRoot, "evidence-context.json"))
-	for _, laneID := range []string{"enemy-events", "npc-registry", "npc-occurrences", "item-registry", "item-occurrences", "location-registry", "location-occurrences"} {
+	if len(evidence) == 0 {
-		if !containsString(evidence.SelectedLanes, laneID) || !evidenceHasLane(evidence, laneID) {
+		t.Fatalf("evidence context = %#v, want selected source-unit evidence", evidence)
-			t.Fatalf("evidence context = %#v, want direct %s evidence", evidence, laneID)
 	}
+	seenEvidenceUnits := make(map[int]struct{}, len(evidence))
+	for _, unit := range evidence {
+		if unit.Ref.SourceID != "session-ravenfall" || unit.Ref.StartUnitID != unit.ID || unit.Ref.EndUnitID != unit.ID {
+			t.Fatalf("evidence unit = %#v, want unchanged source-unit self-reference", unit)
+		}
+		if _, exists := seenEvidenceUnits[unit.ID]; exists {
+			t.Fatalf("evidence context = %#v, want each source unit once", evidence)
+		}
+		seenEvidenceUnits[unit.ID] = struct{}{}
 	}
 
 	requests := client.requestsFor(enemyevents.PromptID)
@@ -330,7 +338,7 @@ func (client *enemyEventLLMClient) CompleteStructured(ctx context.Context, reque
 			if len(registry.Items) != 1 || registry.Items[0].Name != "Moonblade" {
 				return contracts.StructuredCompletionResponse{}, fmt.Errorf("generated item registry has %d items, want 1", len(registry.Items))
 			}
-			content = []byte(`{"occurrences":[{"name":"Moonblade","kind":"discovered","quantity":null,"from":null,"to":null,"source_refs":[{"start_segment":5,"end_segment":5}]}]}`)
+			content = []byte(`{"occurrences":[{"name":"Moonblade","kind":"discovered","quantity":null,"from":null,"to":null,"source_refs":[{"start_unit_id":5,"end_unit_id":5}]}]}`)
 		}
 	case combat.PromptID:
 		content = []byte(`{"combat_turns":[{"actor":"Kesh","turn_kind":"turn","source_refs":[{"start_unit_id":8,"end_unit_id":8}]}]}`)
@@ -388,8 +396,12 @@ func (client *enemyEventLLMClient) CompleteStructured(ctx context.Context, reque
 	if err := json.Unmarshal(content, output); err != nil {
 		return contracts.StructuredCompletionResponse{}, fmt.Errorf("populate fake structured target: %w", err)
 	}
+	snapshot, err := contracts.CloneStructuredCompletionRequest(request)
+	if err != nil {
+		return contracts.StructuredCompletionResponse{}, fmt.Errorf("clone fake request: %w", err)
+	}
 	client.mu.Lock()
-	client.requests = append(client.requests, request)
+	client.requests = append(client.requests, snapshot)
 	client.mu.Unlock()
 	return contracts.StructuredCompletionResponse{Content: content, Provider: "test", Model: "deterministic", ProfileID: request.ProfileID}, nil
 }
@@ -400,7 +412,11 @@ func (client *enemyEventLLMClient) requestsFor(promptID string) []contracts.Stru
 	var requests []contracts.StructuredCompletionRequest
 	for _, request := range client.requests {
 		if request.PromptID == promptID {
-			requests = append(requests, request)
+			snapshot, err := contracts.CloneStructuredCompletionRequest(request)
+			if err != nil {
+				panic(err)
+			}
+			requests = append(requests, snapshot)
 		}
 	}
 	return requests
@@ -415,17 +431,6 @@ func containsString(values []string, want string) bool {
 	return false
 }
 
-func evidenceHasLane(value evidencecontext.Document, laneID string) bool {
-	for _, context := range value.Contexts {
-		for _, reference := range context.EvidenceRefs {
-			if reference.LaneID == laneID {
-				return true
-			}
-		}
-	}
-	return false
-}
-
 func generatedReferenceBinding(bindings []pipeline.ReferenceBinding, slotName string) (pipeline.ReferenceBinding, bool) {
 	for _, binding := range bindings {
 		if binding.SlotName == slotName && binding.Artifact != nil {

internal/cli/dnd_scene_descriptions_contract_test.go

@@ -91,8 +91,8 @@ func TestProductionSceneDescriptionWorkflow(t *testing.T) {
 	if !reflect.DeepEqual(durable, want) {
 		t.Fatalf("durable output payload = %#v, want %#v", durable, want)
 	}
-	if len(output.Warnings) != 0 {
+	if len(output.Diagnostics.Groups) != 0 {
-		t.Fatalf("warnings = %#v, want grounded descriptions without warnings", output.Warnings)
+		t.Fatalf("diagnostics = %#v, want grounded descriptions without diagnostics", output.Diagnostics)
 	}
 }
 

internal/cli/example_contract_test.go

@@ -239,10 +239,10 @@ func TestMaintainedMinimalInvocationProducesJSONBundle(t *testing.T) {
 		t.Fatalf("rejected = %#v, want empty rejection list", rejected.Rejected)
 	}
 	warnings := readProductionJSON[struct {
-		Warnings []json.RawMessage `json:"warnings"`
+		Groups []json.RawMessage `json:"groups"`
 	}](t, filepath.Join(runRoot, "warnings.json"))
-	if len(warnings.Warnings) != 0 {
+	if len(warnings.Groups) != 0 {
-		t.Fatalf("warnings = %#v, want empty warning list", warnings.Warnings)
+		t.Fatalf("warnings = %#v, want empty warning list", warnings.Groups)
 	}
 }
 

internal/cli/production_contract_test.go

@@ -19,12 +19,15 @@ import (
 	"testing/fstest"
 	"time"
 
+	"gopkg.in/yaml.v3"
+
 	"gitea.maximumdirect.net/eric/notarius/internal/core/artifacts"
 	"gitea.maximumdirect.net/eric/notarius/internal/core/config"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/chunkmap"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/llm"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/pipeline"
+	"gitea.maximumdirect.net/eric/notarius/internal/framework/semanticreconcile"
 	"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd"
 	"gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/chunk/scenes"
 	combatcodec "gitea.maximumdirect.net/eric/notarius/internal/modules/dnd/codec/combatturns"
@@ -311,6 +314,81 @@ func TestProductionPromptAssetsPrepareWithoutProviderCredentials(t *testing.T) {
 	if _, err := pipeline.Prepare(effective.ResolvedPipeline, components.registries, pipeline.ModuleDependencies{LLM: &productionFakeLLMClient{}}); err != nil {
 		t.Fatalf("prepare production scene and spell modules: %v", err)
 	}
+
+	schemaFS, err := components.assets.SchemaFS()
+	if err != nil {
+		t.Fatalf("production schema assets: %v", err)
+	}
+	if _, err := fs.ReadFile(schemaFS, filepath.Base(semanticreconcile.SchemaAssetPath)); err != nil {
+		t.Fatalf("generic reconciliation schema asset: %v", err)
+	}
+	options, err := components.assets.PromptKitOptions()
+	if err != nil {
+		t.Fatalf("production PromptKit options: %v", err)
+	}
+	options = append(options, promptkit.WithProfiles(promptkit.OpenAICompatibleProfile(promptkit.OpenAICompatibleProfileConfig{
+		ID: "assembled-prompt-test", Endpoint: "http://127.0.0.1:1/v1", Model: "test",
+	})))
+	engine, err := promptkit.NewEngine(promptkit.Config{}, options...)
+	if err != nil {
+		t.Fatalf("production prompt engine: %v", err)
+	}
+	promptFS, err := components.assets.PromptFS()
+	if err != nil {
+		t.Fatalf("production prompt assets: %v", err)
+	}
+	type manifest struct {
+		ID      string `yaml:"id"`
+		Version string `yaml:"version"`
+		Inputs  []struct {
+			Name string `yaml:"name"`
+		} `yaml:"inputs"`
+		Messages []struct {
+			Role string `yaml:"role"`
+		} `yaml:"messages"`
+	}
+	preparedPrompts := 0
+	if err := fs.WalkDir(promptFS, ".", func(path string, entry fs.DirEntry, walkErr error) error {
+		if walkErr != nil {
+			return walkErr
+		}
+		if entry.IsDir() || filepath.Base(path) != "prompt.yaml" {
+			return nil
+		}
+		data, err := fs.ReadFile(promptFS, path)
+		if err != nil {
+			return err
+		}
+		var prompt manifest
+		if err := yaml.Unmarshal(data, &prompt); err != nil {
+			return err
+		}
+		for _, message := range prompt.Messages {
+			if message.Role != promptkit.RoleSystem && message.Role != promptkit.RoleUser {
+				return fmt.Errorf("production prompt %q uses role %q, want system or user", prompt.ID, message.Role)
+			}
+		}
+		inputs := make(map[string]promptkit.ArtifactRef, len(prompt.Inputs))
+		for _, input := range prompt.Inputs {
+			inputs[input.Name] = promptkit.Inline(`{}`)
+		}
+		prepared, err := engine.Prepare(context.Background(), promptkit.RunRequest{
+			PromptID: prompt.ID, PromptVersion: prompt.Version, ProfileID: "assembled-prompt-test", Inputs: inputs,
+		})
+		if err != nil {
+			return fmt.Errorf("prepare production prompt %q: %w", prompt.ID, err)
+		}
+		if prepared.OutputContract.RepairAttempts != 1 {
+			return fmt.Errorf("prompt %q repair attempts = %d, want 1", prompt.ID, prepared.OutputContract.RepairAttempts)
+		}
+		preparedPrompts++
+		return nil
+	}); err != nil {
+		t.Fatal(err)
+	}
+	if preparedPrompts == 0 {
+		t.Fatal("prepared no production prompts")
+	}
 }
 
 func TestProductionSpellValidatorsPrepareFromMaterializedCatalog(t *testing.T) {
@@ -774,10 +852,10 @@ func TestProductionSceneRunRecordsAnnotationFreeChunkPlanAndProvenance(t *testin
 		t.Fatalf("chunk map range annotations = %#v, want none", chunkMap.Chunks[0].Annotations)
 	}
 	warnings := readProductionJSON[struct {
-		Warnings []contracts.Warning `json:"warnings"`
+		Groups []contracts.DiagnosticGroup `json:"groups"`
 	}](t, filepath.Join(outputRoot, productionRunID, "warnings.json"))
-	if len(warnings.Warnings) != 0 {
+	if len(warnings.Groups) != 0 {
-		t.Fatalf("warnings = %#v, want none", warnings.Warnings)
+		t.Fatalf("warnings = %#v, want none", warnings.Groups)
 	}
 	if len(fake.requestsFor(scenes.PromptID)) != 1 || len(fake.requestsFor(spells.PromptID)) != 1 || len(fake.requestsFor(itemoccurrenceextract.PromptID)) != 1 {
 		t.Fatalf("fake prompt requests = %#v, want one scene, spell, and item-occurrence request", fake.requestPrompts())
@@ -1021,8 +1099,12 @@ func (client *productionFakeLLMClient) CompleteStructured(ctx context.Context, r
 	if err := json.Unmarshal(content, out); err != nil {
 		return contracts.StructuredCompletionResponse{}, fmt.Errorf("populate fake structured target: %w", err)
 	}
+	snapshot, err := contracts.CloneStructuredCompletionRequest(req)
+	if err != nil {
+		return contracts.StructuredCompletionResponse{}, fmt.Errorf("clone fake request: %w", err)
+	}
 	client.mu.Lock()
-	client.requests = append(client.requests, req)
+	client.requests = append(client.requests, snapshot)
 	client.mu.Unlock()
 	return contracts.StructuredCompletionResponse{Content: content, Provider: "test", Model: "deterministic", ProfileID: req.ProfileID}, nil
 }
@@ -1033,7 +1115,11 @@ func (client *productionFakeLLMClient) requestsFor(promptID string) []contracts.
 	var requests []contracts.StructuredCompletionRequest
 	for _, req := range client.requests {
 		if req.PromptID == promptID {
-			requests = append(requests, req)
+			snapshot, err := contracts.CloneStructuredCompletionRequest(req)
+			if err != nil {
+				panic(err)
+			}
+			requests = append(requests, snapshot)
 		}
 	}
 	return requests

internal/cli/promptkit_profiles_test.go

@@ -71,10 +71,10 @@ api_key_env: NOTARIUS_PROMPTKIT_PROFILE_INSPECTION_TEST_KEY
 		},
 		{
 			name:        "malformed profile",
-			profilePath: writeProfile(t, "malformed-profile", "id: malformed-profile\nbackend: [\n"),
+			profilePath: writeProfile(t, "malformed-profile", "id: malformed-profile\nendpoint: https://provider.example/v1?credential=forbidden\nmodel: malformed-model\n"),
 			profileID:   "malformed-profile",
 			wantErr:     []string{`PromptKit profile "malformed-profile" is invalid or unreadable`},
-			rejectErr:   []string{"malformed-profile.yaml", "backend: ["},
+			rejectErr:   []string{"malformed-profile.yaml", "credential=forbidden"},
 		},
 		{
 			name:        "invalid profile source",
@@ -157,3 +157,25 @@ func TestExplicitPromptKitProfileValidationUsesFallbackAssets(t *testing.T) {
 		t.Fatalf("validateExplicitPromptKitProfiles() error = %v, want nil", err)
 	}
 }
+
+func TestExplicitPromptKitProfileValidationRejectsInvalidInheritanceBeforeGeneration(t *testing.T) {
+	for _, profiles := range []string{
+		"id: child\nbase_profile: missing\n",
+		"id: first\nbase_profile: second\n\n---\nid: second\nbase_profile: first\n",
+	} {
+		t.Run("invalid inheritance", func(t *testing.T) {
+			profilePath := filepath.Join(t.TempDir(), "profiles.yaml")
+			if err := os.WriteFile(profilePath, []byte(profiles), 0o600); err != nil {
+				t.Fatal(err)
+			}
+			profileID := "child"
+			if strings.Contains(profiles, "id: first") {
+				profileID = "first"
+			}
+			err := validateExplicitPromptKitProfiles(context.Background(), config.Config{PromptKit: config.PromptKitConfig{ProfileFile: profilePath}}, []string{profileID}, nil)
+			if err == nil || !strings.Contains(err.Error(), "invalid or unreadable") || strings.Contains(err.Error(), profilePath) {
+				t.Fatalf("profile preflight error = %v", err)
+			}
+		})
+	}
+}

internal/cli/reference_contract_test.go

@@ -400,6 +400,9 @@ func (referenceContractCodecA) Encode(stateTestArtifact) ([]byte, error) {
 func (referenceContractCodecA) Decode([]byte) (stateTestArtifact, error) {
 	return stateTestArtifact{Value: "ok"}, nil
 }
+func (codec referenceContractCodecA) DecodeCandidate(content []byte) (stateTestArtifact, error) {
+	return codec.Decode(content)
+}
 
 func (referenceContractCodecB) Kind() contracts.ArtifactKind { return referenceContractKindBeta }
 func (referenceContractCodecB) Schema() contracts.ArtifactSchema {
@@ -415,6 +418,9 @@ func (referenceContractCodecB) Encode(stateTestArtifact) ([]byte, error) {
 func (referenceContractCodecB) Decode([]byte) (stateTestArtifact, error) {
 	return stateTestArtifact{Value: "ok"}, nil
 }
+func (codec referenceContractCodecB) DecodeCandidate(content []byte) (stateTestArtifact, error) {
+	return codec.Decode(content)
+}
 
 func referenceContractLane(t *testing.T, resolved pipeline.ResolvedPipeline, id string) pipeline.ResolvedArtifactLane {
 	t.Helper()

internal/cli/run.go

@@ -16,9 +16,11 @@ import (
 	"strings"
 	"time"
 
+	"gitea.maximumdirect.net/eric/notarius/internal/buildinfo"
 	"gitea.maximumdirect.net/eric/notarius/internal/core/artifacts"
 	"gitea.maximumdirect.net/eric/notarius/internal/core/config"
 	"gitea.maximumdirect.net/eric/notarius/internal/core/debugbundle"
+	"gitea.maximumdirect.net/eric/notarius/internal/core/fileio"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/checkpoint"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/chunkplan"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
@@ -30,6 +32,7 @@ import (
 const defaultConfigPath = "/usr/local/etc/notarius/config.yml"
 const usage = `Usage:
   notarius help
+  notarius --version
   notarius run <pipeline-id> --input path/to/source.json [--json] [flags]
   notarius config validate --config path/to/config.yml [--pipeline pipeline-id] [--only lane-a,lane-b]
   notarius pipelines list --config path/to/config.yml [--json]
@@ -61,6 +64,20 @@ func Run(args []string, stdout, stderr io.Writer) int {
 }
 
 func RunWithOptions(args []string, stdout, stderr io.Writer, opts Options) int {
+	if len(args) > 0 && args[0] == "--version" {
+		if len(args) != 1 {
+			fmt.Fprintln(stderr, "notarius: --version does not accept arguments")
+			return 2
+		}
+		version, err := buildinfo.Version()
+		if err != nil {
+			fmt.Fprintf(stderr, "notarius: %v\n", err)
+			return 1
+		}
+		fmt.Fprintf(stdout, "notarius %s\n", version)
+		return 0
+	}
+
 	var err error
 	opts, err = normalizeOptions(opts)
 	if err != nil {
@@ -147,7 +164,7 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 	machineOutput := fs.Bool("json", false, "write the successful run result as JSON")
 	debug := fs.Bool("debug", false, "write a debug bundle")
 	debugDir := fs.String("debug-dir", "", "debug bundle directory")
-	llmProfile := fs.String("llm-profile", "", "LLM profile override")
+	llmProfile := singleValueFlag{name: "--llm-profile"}
 	reasoningEffort := singleValueFlag{name: "--reasoning-effort"}
 	clearReasoningEffort := fs.Bool("clear-reasoning-effort", false, "clear the LLM profile reasoning effort")
 	resume := fs.Bool("resume", false, "reuse compatible recorded checkpoints")
@@ -157,6 +174,7 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 	referenceFlags := stringListFlag{}
 	withoutReferenceFlags := stringListFlag{}
 	fs.Var(&requestedSessionID, "session-id", "prompt session identifier")
+	fs.Var(&llmProfile, "llm-profile", "LLM profile override")
 	fs.Var(&reasoningEffort, "reasoning-effort", "reasoning effort override")
 	fs.Var(&chunkCache, "chunk_cache", "chunk plan cache mode: auto, bypass, or refresh")
 	fs.Var(&referenceFlags, "reference", "reference binding, as slot=path, chunk.slot=path, merge.slot=path, lane.slot=path, lane.extract.slot=path, lane.merge.slot=path, or lane.normalize.slot=path")
@@ -203,6 +221,10 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 		fmt.Fprintln(stderr, "notarius: --session-id must not be empty")
 		return 2
 	}
+	if llmProfile.set && strings.TrimSpace(llmProfile.value) == "" {
+		fmt.Fprintln(stderr, "notarius: --llm-profile must not be empty")
+		return 2
+	}
 	if reasoningEffort.set && *clearReasoningEffort {
 		fmt.Fprintln(stderr, "notarius: --reasoning-effort cannot be combined with --clear-reasoning-effort")
 		return 2
@@ -338,7 +360,7 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 		PipelineID:         pipelineID,
 		Only:               only,
 		Catalog:            catalog,
-		LLMProfileOverride: *llmProfile,
+		LLMProfileOverride: strings.TrimSpace(llmProfile.value),
 		ReferenceOverrides: referenceOverrides,
 		ReferenceUnbinds:   referenceUnbinds,
 	})
@@ -353,7 +375,7 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 	if err != nil {
 		return failPipelineCommand(stderr, commandState, terminalWriter, fmt.Errorf("resolve working directory: %w", err))
 	}
-	materialized, referenceWarnings, err := pipeline.MaterializeReferences(effective.ResolvedPipeline, catalog, pipeline.ReferenceMaterializationOptions{
+	materialized, referenceDiagnostics, err := pipeline.MaterializeReferences(effective.ResolvedPipeline, catalog, pipeline.ReferenceMaterializationOptions{
 		ConfigPath: loadedConfigPath,
 		WorkingDir: workingDir,
 	})
@@ -426,7 +448,7 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 	if err != nil {
 		return failPipelineCommand(stderr, commandState, terminalWriter, err)
 	}
-	checkpointRecorder, checkpointLoader, err := checkpointHandlersForRun(effective.Config.Cache.Checkpoints, opts, effective.ResolvedPipeline, prepared.CheckpointFingerprints(), llmFingerprints, rawInput, only, llmProfiles, strings.TrimSpace(*llmProfile), effectiveSessionID, runtimeOverrides, *resume)
+	checkpointRecorder, checkpointLoader, err := checkpointHandlersForRun(effective.Config.Cache.Checkpoints, opts, effective.ResolvedPipeline, prepared.CheckpointFingerprints(), llmFingerprints, rawInput, only, llmProfiles, strings.TrimSpace(llmProfile.value), effectiveSessionID, runtimeOverrides, *resume)
 	if err != nil {
 		return failPipelineCommand(stderr, commandState, terminalWriter, err)
 	}
@@ -440,7 +462,7 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 		StartedAt:        startedAt,
 		LLMProfiles:      llmProfiles,
 		Metadata:         runMetadata(effective.Config.Output.Directory, debugPath),
-		Warnings:         referenceWarnings,
+		Diagnostics:      referenceDiagnostics,
 		ChunkCacheMode:   effective.Config.Cache.ChunkPlans.Mode,
 		ChunkPlans:       chunkPlans,
 		Checkpoints:      checkpointRecorder,
@@ -449,7 +471,10 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 		Debug:            debugRecorder,
 		ExtractWorkers:   cfg.Concurrency.StageWorkers["extract"],
 	})
-	commandState.observeOutput(output)
+	diagnosticProjection, diagnosticErr := contracts.ProjectDiagnosticCollection(output.Diagnostics)
+	if diagnosticErr == nil {
+		commandState.observeOutput(output, diagnosticProjection)
+	}
 	if err != nil {
 		primaryErr := fmt.Errorf("run pipeline %q: %w", pipelineID, err)
 		if output.Manifest.PipelineID != "" {
@@ -457,15 +482,21 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 				return failPipelineCommand(stderr, commandState, terminalWriter, primaryErr, fmt.Errorf("write debug summary: %w", summaryErr))
 			}
 		}
+		if diagnosticErr != nil {
+			return failPipelineCommand(stderr, commandState, terminalWriter, primaryErr, fmt.Errorf("summarize run diagnostics: %w", diagnosticErr))
+		}
 		return failPipelineCommand(stderr, commandState, terminalWriter, primaryErr)
 	}
+	if diagnosticErr != nil {
+		return failPipelineCommand(stderr, commandState, terminalWriter, fmt.Errorf("summarize run diagnostics: %w", diagnosticErr))
+	}
 
 	if err := writePartialSummary(summary, output); err != nil {
 		return failPipelineCommand(stderr, commandState, terminalWriter, fmt.Errorf("write debug summary: %w", err))
 	}
 	var encodedResult []byte
 	if *machineOutput {
-		result, err := newRunResult(effective.ResolvedPipeline, output, runOutputDir, debugPath)
+		result, err := newRunResultWithDiagnostics(effective.ResolvedPipeline, output, runOutputDir, debugPath, diagnosticProjection)
 		if err != nil {
 			return failPipelineCommand(stderr, commandState, terminalWriter, err)
 		}
@@ -491,12 +522,25 @@ func runPipelineCommand(args []string, stdout, stderr io.Writer, opts Options) i
 			fmt.Fprintf(stdout, "debug=%s\n", debugPath)
 		}
 	}
-	if len(output.Warnings) > 0 {
+	if warningGroups := len(diagnosticProjection.Warnings); warningGroups > 0 {
-		fmt.Fprintf(stderr, "notarius: run completed with %d warning(s)\n", len(output.Warnings))
+		fmt.Fprintf(stderr, "notarius: run completed with %d warning group(s), %d occurrence(s)", warningGroups, diagnosticProjection.WarningOccurrenceCount)
+		if warningFile, ok := logicalOutputFile(output.OutputFiles, "warnings.json"); ok {
+			fmt.Fprintf(stderr, "; details=%s", filepath.Join(runOutputDir, warningFile))
+		}
+		fmt.Fprintln(stderr)
 	}
 	return 0
 }
 
+func logicalOutputFile(files []contracts.OutputFile, name string) (string, bool) {
+	for _, file := range files {
+		if file.Name == name {
+			return file.Name, true
+		}
+	}
+	return "", false
+}
+
 func writeSummary(summary *debugbundle.SummaryWriter, write func() error) error {
 	if summary == nil {
 		return nil
@@ -516,7 +560,7 @@ func writePartialSummary(summary *debugbundle.SummaryWriter, output pipeline.Run
 			return err
 		}
 	}
-	if err := summary.WriteWarnings(output.Warnings); err != nil {
+	if err := summary.WriteDiagnostics(output.Diagnostics); err != nil {
 		return err
 	}
 	return summary.WriteCheckpointEvents(output.CheckpointEvents)
@@ -752,17 +796,10 @@ func configSource(configPath string) string {
 }
 
 func writeOutputFiles(runOutputDir string, files []contracts.OutputFile) error {
-	type outputTarget struct {
-		path string
-		file contracts.OutputFile
-	}
-	targets := make([]outputTarget, 0, len(files))
 	for _, file := range files {
-		targetPath, err := outputFilePath(runOutputDir, file.Name)
+		if _, err := outputFilePath(runOutputDir, file.Name); err != nil {
-		if err != nil {
 			return err
 		}
-		targets = append(targets, outputTarget{path: targetPath, file: file})
 	}
 
 	outputParent := filepath.Dir(runOutputDir)
@@ -775,12 +812,9 @@ func writeOutputFiles(runOutputDir string, files []contracts.OutputFile) error {
 		}
 		return fmt.Errorf("create output run directory %q: %w", runOutputDir, err)
 	}
-	for _, target := range targets {
+	for _, file := range files {
-		if err := os.MkdirAll(filepath.Dir(target.path), 0o755); err != nil {
+		if err := fileio.WriteBytes(runOutputDir, file.Name, file.Bytes, 0o755, 0o644); err != nil {
-			return fmt.Errorf("create output directory %q: %w", filepath.Dir(target.path), err)
+			return fmt.Errorf("write output file %q: %w", file.Name, err)
-		}
-		if err := writeFileAtomic(target.path, target.file.Bytes, 0o644); err != nil {
-			return fmt.Errorf("write output file %q: %w", target.file.Name, err)
 		}
 	}
 	return nil
@@ -823,38 +857,6 @@ func outputFilePath(runOutputDir, logicalName string) (string, error) {
 	return target, nil
 }
 
-func writeFileAtomic(path string, data []byte, perm os.FileMode) error {
-	dir := filepath.Dir(path)
-	temp, err := os.CreateTemp(dir, "."+filepath.Base(path)+".tmp-*")
-	if err != nil {
-		return err
-	}
-	tempPath := temp.Name()
-	removeTemp := true
-	defer func() {
-		if removeTemp {
-			_ = os.Remove(tempPath)
-		}
-	}()
-
-	if _, err := temp.Write(data); err != nil {
-		_ = temp.Close()
-		return err
-	}
-	if err := temp.Chmod(perm); err != nil {
-		_ = temp.Close()
-		return err
-	}
-	if err := temp.Close(); err != nil {
-		return err
-	}
-	if err := os.Rename(tempPath, path); err != nil {
-		return err
-	}
-	removeTemp = false
-	return nil
-}
-
 func reorderRunArgs(args []string) []string {
 	var flags []string
 	var positionals []string

internal/cli/run_contract_test.go

@@ -43,6 +43,12 @@ func TestRunControlsRejectSyntaxWithoutAllocatingState(t *testing.T) {
 		{name: "blank session ID", args: func(roots stateTestRoots) []string {
 			return []string{"run", "sample", "--config", roots.config, "--input", roots.input, "--session-id", ""}
 		}},
+		{name: "blank LLM profile", args: func(roots stateTestRoots) []string {
+			return []string{"run", "sample", "--config", roots.config, "--input", roots.input, "--llm-profile", ""}
+		}},
+		{name: "whitespace LLM profile", args: func(roots stateTestRoots) []string {
+			return []string{"run", "sample", "--config", roots.config, "--input", roots.input, "--llm-profile", " \t "}
+		}},
 		{name: "multiple pipeline IDs", args: func(roots stateTestRoots) []string {
 			return []string{"run", "sample", "extra", "--config", roots.config, "--input", roots.input}
 		}},
@@ -584,10 +590,11 @@ func TestRunWarningsRemainSuccessfulAndReachDurableSurfaces(t *testing.T) {
 	roots := newStateTestRoots(t)
 	harness := newStateTestHarness()
 	harness.includeWarnings = true
-	harness.chunkWarnings = []contracts.Warning{{Scope: "chunk", ReasonCode: "contract-warning", Message: "warning retained"}}
+	harness.includeWarningFile = true
+	harness.chunkDiagnostics = []contracts.ProducerDiagnostic{stateTestDiagnostic("chunk", "contract-warning", "warning retained")}
 	var stdout, stderr bytes.Buffer
 	code := RunWithOptions([]string{"run", "sample", "--config", roots.config, "--input", roots.input, "--chunk_cache", "bypass", "--debug"}, &stdout, &stderr, harness.options())
-	if code != 0 || !strings.Contains(stdout.String(), "outputs=1") || !strings.Contains(stderr.String(), "1 warning(s)") {
+	if code != 0 || !strings.Contains(stdout.String(), "outputs=1") || !strings.Contains(stderr.String(), "1 warning group(s), 1 occurrence(s)") || !strings.Contains(stderr.String(), "warnings.json") {
 		t.Fatalf("code=%d stdout=%q stderr=%q", code, stdout.String(), stderr.String())
 	}
 	outputPath := filepath.Join(onlyChildDir(t, roots.output), "result.json")
@@ -595,11 +602,12 @@ func TestRunWarningsRemainSuccessfulAndReachDurableSurfaces(t *testing.T) {
 	if err != nil || !strings.Contains(string(output), "contract-warning") {
 		t.Fatalf("durable output = %q, %v", output, err)
 	}
+	assertFile(t, filepath.Join(filepath.Dir(outputPath), "warnings.json"))
 	bundle := onlyChildDir(t, roots.debug)
-	var warnings []contracts.Warning
+	var diagnostics contracts.DiagnosticCollection
-	readStateTestSummaryJSON(t, bundle, "warnings.json", &warnings)
+	readStateTestSummaryJSON(t, bundle, "final-diagnostics.json", &diagnostics)
-	if len(warnings) != 1 || warnings[0].ReasonCode != "contract-warning" {
+	if len(diagnostics.Groups) != 1 || diagnostics.Groups[0].ReasonCode != "contract-warning" {
-		t.Fatalf("debug warnings = %#v", warnings)
+		t.Fatalf("debug diagnostics = %#v", diagnostics)
 	}
 }
 

internal/cli/run_id_test.go

@@ -38,10 +38,36 @@ func TestWriteOutputFilesSupportsNestedLogicalPaths(t *testing.T) {
 	if err := writeOutputFiles(runPath, []contracts.OutputFile{{Name: "nested/result.json", Bytes: []byte("result")}}); err != nil {
 		t.Fatal(err)
 	}
-	data, err := os.ReadFile(filepath.Join(runPath, "nested", "result.json"))
+	resultPath := filepath.Join(runPath, "nested", "result.json")
+	data, err := os.ReadFile(resultPath)
 	if err != nil || string(data) != "result" {
 		t.Fatalf("nested output = %q, %v", data, err)
 	}
+	for path, want := range map[string]os.FileMode{runPath: 0o755, filepath.Join(runPath, "nested"): 0o755, resultPath: 0o644} {
+		info, err := os.Stat(path)
+		if err != nil {
+			t.Fatal(err)
+		}
+		mode := info.Mode().Perm()
+		if mode&^want != 0 {
+			t.Fatalf("%s mode = %#o, must not be broader than %#o", path, mode, want)
+		}
+		if info.IsDir() && mode&0o700 != 0o700 {
+			t.Fatalf("%s mode = %#o, want owner access", path, mode)
+		}
+		if !info.IsDir() && mode != want {
+			t.Fatalf("%s mode = %#o, want %#o", path, mode, want)
+		}
+	}
+	entries, err := os.ReadDir(filepath.Join(runPath, "nested"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, entry := range entries {
+		if strings.Contains(entry.Name(), ".tmp-") {
+			t.Fatalf("temporary file remains: %s", entry.Name())
+		}
+	}
 }
 
 func TestWriteOutputFilesRejectsUnsafeNamesBeforeAllocatingRunDirectory(t *testing.T) {
@@ -75,7 +101,7 @@ func TestWriteOutputFilesRetainsNewPartialDirectoryAndPreservesSibling(t *testin
 		{Name: "blocked", Bytes: []byte("partial output")},
 		{Name: "blocked/nested.json", Bytes: []byte("unreachable")},
 	})
-	if err == nil || !strings.Contains(err.Error(), "create output directory") {
+	if err == nil || !strings.Contains(err.Error(), `write output file "blocked/nested.json"`) {
 		t.Fatalf("writeOutputFiles() error = %v, want later directory failure", err)
 	}
 	if got, err := os.ReadFile(filepath.Join(runPath, "blocked")); err != nil || string(got) != "partial output" {

internal/cli/run_result.go

@@ -7,10 +7,12 @@ import (
 	"path/filepath"
 	"strings"
 
+	"gitea.maximumdirect.net/eric/notarius/internal/core/artifacts"
+	"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/pipeline"
 )
 
-const runResultSchemaVersion = "notarius.run-result.v1"
+const runResultSchemaVersion = "notarius.run-result.v2"
 
 type runResult struct {
 	SchemaVersion             string                        `json:"schema_version"`
@@ -20,12 +22,25 @@ type runResult struct {
 	IndexFile                 string                        `json:"index_file,omitempty"`
 	NormalizedOutputCount     int                           `json:"normalized_output_count"`
 	RejectedOutputCount       int                           `json:"rejected_output_count"`
-	WarningCount          int    `json:"warning_count"`
+	WarningGroupCount         int                           `json:"warning_group_count"`
+	WarningOccurrenceCount    int                           `json:"warning_occurrence_count"`
+	DiagnosticGroupCount      int                           `json:"diagnostic_group_count"`
+	DiagnosticOccurrenceCount int                           `json:"diagnostic_occurrence_count"`
+	DiagnosticsTruncated      bool                          `json:"diagnostics_truncated"`
 	ValidationStatus          string                        `json:"validation_status"`
+	ValidationSummaries       []artifacts.ValidationSummary `json:"validation_summaries,omitempty"`
 	DebugDirectory            string                        `json:"debug_directory,omitempty"`
 }
 
 func newRunResult(resolved pipeline.ResolvedPipeline, output pipeline.RunOutput, outputDirectory, debugDirectory string) (runResult, error) {
+	diagnosticProjection, err := contracts.ProjectDiagnosticCollection(output.Diagnostics)
+	if err != nil {
+		return runResult{}, fmt.Errorf("summarize run diagnostics: %w", err)
+	}
+	return newRunResultWithDiagnostics(resolved, output, outputDirectory, debugDirectory, diagnosticProjection)
+}
+
+func newRunResultWithDiagnostics(resolved pipeline.ResolvedPipeline, output pipeline.RunOutput, outputDirectory, debugDirectory string, diagnosticProjection contracts.DiagnosticProjection) (runResult, error) {
 	if strings.TrimSpace(output.Manifest.RunID) == "" {
 		return runResult{}, fmt.Errorf("run result requires a run ID")
 	}
@@ -57,8 +72,13 @@ func newRunResult(resolved pipeline.ResolvedPipeline, output pipeline.RunOutput,
 		OutputDirectory:           absOutputDirectory,
 		NormalizedOutputCount:     len(output.NormalizeOutputs),
 		RejectedOutputCount:       len(output.Rejected),
-		WarningCount:          len(output.Warnings),
+		WarningGroupCount:         len(diagnosticProjection.Warnings),
+		WarningOccurrenceCount:    diagnosticProjection.WarningOccurrenceCount,
+		DiagnosticGroupCount:      len(diagnosticProjection.Diagnostics),
+		DiagnosticOccurrenceCount: diagnosticProjection.DiagnosticOccurrenceCount,
+		DiagnosticsTruncated:      output.Diagnostics.Truncated,
 		ValidationStatus:          output.Manifest.ValidationStatus,
+		ValidationSummaries:       cloneValidationSummaries(output.Manifest.ValidationSummaries),
 	}
 
 	if strings.TrimSpace(debugDirectory) != "" {
@@ -85,6 +105,17 @@ func newRunResult(resolved pipeline.ResolvedPipeline, output pipeline.RunOutput,
 	return result, nil
 }
 
+func cloneValidationSummaries(summaries []artifacts.ValidationSummary) []artifacts.ValidationSummary {
+	if len(summaries) == 0 {
+		return nil
+	}
+	cloned := make([]artifacts.ValidationSummary, len(summaries))
+	for index, summary := range summaries {
+		cloned[index] = artifacts.CloneValidationSummary(summary)
+	}
+	return cloned
+}
+
 func encodeRunResult(result runResult) ([]byte, error) {
 	encoded, err := json.Marshal(result)
 	if err != nil {

internal/cli/run_result_command_test.go

@@ -27,7 +27,7 @@ func TestMaintainedMinimalInvocationEmitsRunResult(t *testing.T) {
 	}
 
 	receipt := decodeRunResultDocument(t, stdout.String())
-	if got := receipt["schema_version"]; got != "notarius.run-result.v1" {
+	if got := receipt["schema_version"]; got != "notarius.run-result.v2" {
 		t.Fatalf("schema_version = %q", got)
 	}
 	if got := receipt["run_id"]; got != productionRunID {
@@ -45,8 +45,8 @@ func TestMaintainedMinimalInvocationEmitsRunResult(t *testing.T) {
 	if got := receipt["rejected_output_count"]; got != float64(0) {
 		t.Fatalf("rejected_output_count = %v", got)
 	}
-	if got := receipt["warning_count"]; got != float64(0) {
+	if got := receipt["warning_group_count"]; got != float64(0) || receipt["warning_occurrence_count"] != float64(0) || receipt["diagnostic_group_count"] != float64(0) || receipt["diagnostic_occurrence_count"] != float64(0) || receipt["diagnostics_truncated"] != false {
-		t.Fatalf("warning_count = %v", got)
+		t.Fatalf("diagnostic counts = %#v", receipt)
 	}
 	if got := receipt["validation_status"]; got != "approved" {
 		t.Fatalf("validation_status = %q", got)
@@ -63,19 +63,19 @@ func TestMaintainedMinimalInvocationEmitsRunResult(t *testing.T) {
 func TestRunResultReportsWarningsAndDebugBundle(t *testing.T) {
 	roots := newStateTestRoots(t)
 	harness := newStateTestHarness()
-	harness.chunkWarnings = []contracts.Warning{{Scope: "chunk", ReasonCode: "contract-warning", Message: "warning retained"}}
+	harness.chunkDiagnostics = []contracts.ProducerDiagnostic{stateTestDiagnostic("chunk", "contract-warning", "warning retained")}
 	var stdout, stderr bytes.Buffer
 	code := RunWithOptions([]string{
 		"run", "sample", "--config", roots.config, "--input", roots.input,
 		"--chunk_cache", "bypass", "--debug", "--json",
 	}, &stdout, &stderr, harness.options())
-	if code != 0 || !strings.Contains(stderr.String(), "1 warning(s)") {
+	if code != 0 || !strings.Contains(stderr.String(), "1 warning group(s), 1 occurrence(s)") || strings.Contains(stderr.String(), "warnings.json") {
 		t.Fatalf("code=%d stdout=%q stderr=%q", code, stdout.String(), stderr.String())
 	}
 
 	receipt := decodeRunResultDocument(t, stdout.String())
-	if got := receipt["warning_count"]; got != float64(1) {
+	if got := receipt["warning_group_count"]; got != float64(1) || receipt["warning_occurrence_count"] != float64(1) || receipt["diagnostic_group_count"] != float64(0) || receipt["diagnostic_occurrence_count"] != float64(0) || receipt["diagnostics_truncated"] != false {
-		t.Fatalf("warning_count = %v", got)
+		t.Fatalf("diagnostic counts = %#v", receipt)
 	}
 	debugDirectory, ok := receipt["debug_directory"].(string)
 	if !ok || !filepath.IsAbs(debugDirectory) || debugDirectory != onlyChildDir(t, roots.debug) {
@@ -92,7 +92,7 @@ func TestRunResultReportsSuccessfulRejection(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	configBytes = []byte(replaceRequiredOnce(t, string(configBytes), "        normalize: test/normalize\n", "        normalize:\n          module: test/normalize\n          validators:\n            - generic/always_reject\n"))
+	configBytes = []byte(replaceRequiredOnce(t, string(configBytes), "        normalize: test/normalize\n", "        normalize:\n          module: test/normalize\n          validators:\n            - generic/always_reject\n          validation_policy:\n            semantic_rejection: reject_output\n"))
 	if err := os.WriteFile(roots.config, configBytes, 0o600); err != nil {
 		t.Fatal(err)
 	}

internal/cli/run_result_test.go

@@ -52,8 +52,11 @@ func TestRunResultEncodesRequiredFieldsAndCounts(t *testing.T) {
 	if got := decoded["rejected_output_count"]; got != float64(1) {
 		t.Fatalf("rejected_output_count = %v", got)
 	}
-	if got := decoded["warning_count"]; got != float64(1) {
+	if got := decoded["warning_group_count"]; got != float64(1) || decoded["warning_occurrence_count"] != float64(1) || decoded["diagnostic_group_count"] != float64(0) || decoded["diagnostic_occurrence_count"] != float64(0) || decoded["diagnostics_truncated"] != false {
-		t.Fatalf("warning_count = %v", got)
+		t.Fatalf("diagnostic counts = %#v", decoded)
+	}
+	if got := decoded["validation_summaries"]; got != nil {
+		t.Fatalf("validation_summaries = %#v, want omitted when empty", got)
 	}
 	if got := decoded["output_directory"]; got != filepath.Join(mustWorkingDirectory(t), "relative-output") {
 		t.Fatalf("output_directory = %q", got)
@@ -112,6 +115,26 @@ func TestRunResultOmitsIndexFileForOtherOutputModules(t *testing.T) {
 	}
 }
 
+func TestRunResultProjectsOwnedValidationSummaries(t *testing.T) {
+	output := testRunOutput()
+	output.Manifest.ValidationSummaries = []artifacts.ValidationSummary{{Status: "incomplete", IncompleteValidators: []string{"validator"}, ProducerAttemptCount: 1, TerminalAction: "warn_continue"}}
+	result, err := newRunResult(testResolvedPipeline(pipeline.DefaultOutputModule), output, "output", "")
+	if err != nil {
+		t.Fatal(err)
+	}
+	output.Manifest.ValidationSummaries[0].IncompleteValidators[0] = "caller mutation"
+	if got := result.ValidationSummaries[0].IncompleteValidators; len(got) != 1 || got[0] != "validator" {
+		t.Fatalf("result validation summaries = %#v", result.ValidationSummaries)
+	}
+	encoded, err := encodeRunResult(result)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !bytes.Contains(encoded, []byte(`"validation_summaries":[{"status":"incomplete","incomplete_validators":["validator"],"producer_attempt_count":1,"terminal_action":"warn_continue"}]`)) {
+		t.Fatalf("encoded result = %s", encoded)
+	}
+}
+
 func TestRunResultRequiresOneProductionIndexFile(t *testing.T) {
 	tests := []struct {
 		name  string
@@ -159,7 +182,14 @@ func testRunOutput() pipeline.RunOutput {
 		Manifest:         artifacts.RunManifest{RunID: "run-123", PipelineID: "sample", ValidationStatus: "rejected"},
 		NormalizeOutputs: []contracts.SerializedOutput{{}, {}},
 		Rejected:         []contracts.RejectedOutput{{}},
-		Warnings:         []contracts.Warning{{}},
+		Diagnostics: contracts.DiagnosticCollection{Groups: []contracts.DiagnosticGroup{{
+			Disposition:     contracts.DiagnosticDispositionWarning,
+			Category:        contracts.DiagnosticCategoryFallback,
+			ReasonCode:      "fallback",
+			Origin:          contracts.DiagnosticOrigin{Stage: contracts.DiagnosticOriginStageNormalize, StepID: "step", LaneID: "lane", ModuleKey: "module"},
+			OccurrenceCount: 1,
+			Samples:         []contracts.DiagnosticSample{{Scope: "scope", Message: "message"}},
+		}}},
 		OutputFiles: []contracts.OutputFile{{Name: "index.json"}},
 	}
 }

internal/cli/run_terminal.go

@@ -6,6 +6,7 @@ import (
 	"io"
 
 	"gitea.maximumdirect.net/eric/notarius/internal/core/debugbundle"
+	"gitea.maximumdirect.net/eric/notarius/internal/framework/contracts"
 	"gitea.maximumdirect.net/eric/notarius/internal/framework/pipeline"
 )
 
@@ -33,13 +34,17 @@ func (s *pipelineCommandState) setDebugPath(debugPath string) {
 	}
 }
 
-func (s *pipelineCommandState) observeOutput(output pipeline.RunOutput) {
+func (s *pipelineCommandState) observeOutput(output pipeline.RunOutput, diagnostics contracts.DiagnosticProjection) {
 	if s == nil {
 		return
 	}
 	s.report.OutputCount = len(output.NormalizeOutputs)
 	s.report.RejectedCount = len(output.Rejected)
-	s.report.WarningCount = len(output.Warnings)
+	s.report.WarningGroupCount = len(diagnostics.Warnings)
+	s.report.WarningOccurrenceCount = diagnostics.WarningOccurrenceCount
+	s.report.DiagnosticGroupCount = len(diagnostics.Diagnostics)
+	s.report.DiagnosticOccurrenceCount = diagnostics.DiagnosticOccurrenceCount
+	s.report.DiagnosticsTruncated = output.Diagnostics.Truncated
 	s.report.ValidationStatus = output.Manifest.ValidationStatus
 }
 

internal/cli/spell_catalog_identity_contract_test.go

@@ -238,7 +238,7 @@ func TestChangedSemanticSpellCatalogFingerprintCannotResumeRecordedCheckpoint(t
 			Kind: dnd.SpellListKind, Schema: normalizeSchema, MediaType: "application/json", Content: []byte(`{"spell_casts":[]}`),
 		},
 	}
-	if err := recorder.NormalizeSucceeded("spells", spellnormalize.Key, normalizeDependencies, normalizeArtifact, nil); err != nil {
+	if err := recorder.NormalizeSucceeded("spells", spellnormalize.Key, normalizeDependencies, normalizeArtifact); err != nil {
 		t.Fatal(err)
 	}
 
@@ -410,8 +410,8 @@ func TestMaintainedProductionOverlayRunAlignsGroundingValidationAndProvenance(t
 		t.Fatalf("spell requests = %d, want one", len(requests))
 	}
 	catalogInput, ok := requests[0].Inputs[spellcatalog.SpellCatalogReferenceSlot]
-	if !ok || !strings.Contains(string(catalogInput.Content), "Aegis of Emberfall") || strings.Contains(string(catalogInput.Content), "Emberfall Aegis") {
+	if !ok || !strings.Contains(string(catalogInput.Content), `"canonical_name":"Aegis of Emberfall"`) || !strings.Contains(string(catalogInput.Content), `"aliases":["Emberfall Aegis"]`) {
-		t.Fatalf("spell catalog prompt input = %#v, want canonical overlay name without alias", catalogInput)
+		t.Fatalf("spell catalog prompt input = %#v, want canonical overlay name and recognition alias", catalogInput)
 	}
 	artifact := readProductionJSON[dnd.SpellList](t, filepath.Join(runRoot, "lanes", "spells.json"))
 	if len(artifact.SpellCasts) != 1 || artifact.SpellCasts[0].Spell != "Aegis of Emberfall" {

internal/cli/spell_catalog_retry_contract_test.go

@@ -24,7 +24,7 @@ func TestProductionSpellCatalogValidationRetries(t *testing.T) {
 		wantCalls        int
 		wantRejected     bool
 		wantSpell        string
-		wantWarningCode string
+		wantAdvisoryCode string
 	}{
 		{
 			name: "unknown spell remains rejected after exhaustion",
@@ -44,7 +44,7 @@ func TestProductionSpellCatalogValidationRetries(t *testing.T) {
 			},
 			wantCalls:        2,
 			wantSpell:        "Aegis of Emberfall",
-			wantWarningCode: "spell_not_near_source",
+			wantAdvisoryCode: "spell_not_near_source",
 		},
 	}
 
@@ -65,6 +65,7 @@ func TestProductionSpellCatalogValidationRetries(t *testing.T) {
 				t.Fatalf("materialize production references: %v", err)
 			}
 			materialized.Steps[0].ArtifactLanes[0].Extract.Retries = retries
+			materialized.Steps[0].ArtifactLanes[0].ExtractValidationPolicy.SemanticRejection = pipeline.SemanticRejectionRejectOutput
 
 			llmClient := &catalogRetryLLMClient{responses: tt.responses}
 			prepared, err := pipeline.Prepare(materialized, components.registries, pipeline.ModuleDependencies{LLM: llmClient})
@@ -91,8 +92,8 @@ func TestProductionSpellCatalogValidationRetries(t *testing.T) {
 				if rejection.ReasonCode != "unknown_spell" || rejection.AttemptCount != retries+1 {
 					t.Fatalf("rejection = %#v, want exhausted unknown-spell rejection", rejection)
 				}
-				if len(output.Warnings) != 0 {
+				if len(output.Diagnostics.Groups) != 1 || output.Diagnostics.Groups[0].ReasonCode != "spell_not_near_source" {
-					t.Fatalf("warnings = %#v, want no warnings from rejected attempts", output.Warnings)
+					t.Fatalf("diagnostics = %#v, want complete terminal validation diagnostic", output.Diagnostics)
 				}
 				return
 			}
@@ -107,8 +108,16 @@ func TestProductionSpellCatalogValidationRetries(t *testing.T) {
 			if len(value.SpellCasts) != 1 || value.SpellCasts[0].Spell != tt.wantSpell {
 				t.Fatalf("normalized spell list = %#v, want accepted overlay spell", value)
 			}
-			if len(output.Warnings) != 2 || output.Warnings[0].ReasonCode != tt.wantWarningCode || output.Warnings[1].ReasonCode != tt.wantWarningCode {
+			if len(output.Diagnostics.Groups) != 2 || output.Diagnostics.Groups[0].ReasonCode != tt.wantAdvisoryCode {
-				t.Fatalf("warnings = %#v, want accepted-attempt warnings from extract and normalize validation", output.Warnings)
+				t.Fatalf("diagnostics = %#v, want terminal extract and normalize diagnostics", output.Diagnostics)
+			}
+			if len(output.Diagnostics.Groups) != 2 {
+				t.Fatalf("diagnostics = %#v, want extract and normalize data-quality advisories", output.Diagnostics)
+			}
+			for _, diagnostic := range output.Diagnostics.Groups {
+				if diagnostic.Disposition != contracts.DiagnosticDispositionAdvisory || diagnostic.ReasonCode != tt.wantAdvisoryCode {
+					t.Fatalf("diagnostics = %#v, want only data-quality advisories", output.Diagnostics)
+				}
 			}
 		})
 	}

internal/cli/state_hardening_test.go

@@ -349,7 +349,7 @@ func TestRunUsesOneInjectedIdentityForDebugOutputAndManifest(t *testing.T) {
 		t.Fatalf("debug invocation session = %q, want %q", invocation.SessionID, wantSessionID)
 	}
 	report := readStateTestRunReport(t, debugPath)
-	if !report.Succeeded || report.RunID != runID || report.PipelineID != "sample" || report.OutputPath != outputPath || report.DebugPath != debugPath || report.OutputCount != 1 || report.RejectedCount != 0 || report.WarningCount != 0 || report.ValidationStatus != "approved" {
+	if !report.Succeeded || report.RunID != runID || report.PipelineID != "sample" || report.OutputPath != outputPath || report.DebugPath != debugPath || report.OutputCount != 1 || report.RejectedCount != 0 || report.WarningGroupCount != 0 || report.WarningOccurrenceCount != 0 || report.DiagnosticGroupCount != 0 || report.DiagnosticOccurrenceCount != 0 || report.DiagnosticsTruncated || report.ValidationStatus != "approved" {
 		t.Fatalf("success report = %#v", report)
 	}
 	if !strings.Contains(result.stdout, "outputs=1 rejected=0") {
@@ -431,7 +431,7 @@ func TestRunWritesTerminalArtifactsForResolutionPipelineAndOutputFailures(t *tes
 			bundlePath := onlyChildDir(t, roots.debug)
 			runID := filepath.Base(bundlePath)
 			report := readStateTestRunReport(t, bundlePath)
-			if report.Succeeded || report.RunID != runID || report.PipelineID != tc.pipelineID || report.OutputPath != filepath.Join(roots.output, runID) || report.DebugPath != bundlePath || report.OutputCount != tc.wantOutputs || report.RejectedCount != 0 || report.WarningCount != 0 || report.ValidationStatus != tc.wantValidation {
+			if report.Succeeded || report.RunID != runID || report.PipelineID != tc.pipelineID || report.OutputPath != filepath.Join(roots.output, runID) || report.DebugPath != bundlePath || report.OutputCount != tc.wantOutputs || report.RejectedCount != 0 || report.WarningGroupCount != 0 || report.WarningOccurrenceCount != 0 || report.DiagnosticGroupCount != 0 || report.DiagnosticOccurrenceCount != 0 || report.DiagnosticsTruncated || report.ValidationStatus != tc.wantValidation {
 				t.Fatalf("failure report = %#v", report)
 			}
 			errorLog, err := os.ReadFile(filepath.Join(bundlePath, "summary", "error.log"))
@@ -445,7 +445,7 @@ func TestRunWritesTerminalArtifactsForResolutionPipelineAndOutputFailures(t *tes
 func TestRunRetainsPartialPipelineOutcomeInFailureSummary(t *testing.T) {
 	roots := newStateTestRoots(t)
 	harness := newStateTestHarness()
-	harness.chunkWarnings = []contracts.Warning{{Scope: "chunk", ReasonCode: "partial-warning", Message: "warning retained before failure"}}
+	harness.chunkDiagnostics = []contracts.ProducerDiagnostic{stateTestDiagnostic("chunk", "partial-warning", "warning retained before failure")}
 	harness.extractErr = errors.New("synthetic partial pipeline failure")
 
 	result := runStateTest(t, roots, harness.options(), true, true, "bypass")
@@ -454,7 +454,7 @@ func TestRunRetainsPartialPipelineOutcomeInFailureSummary(t *testing.T) {
 	}
 	bundlePath := onlyChildDir(t, roots.debug)
 	report := readStateTestRunReport(t, bundlePath)
-	if report.Succeeded || report.OutputCount != 0 || report.RejectedCount != 0 || report.WarningCount != 1 || report.ValidationStatus != "failed" {
+	if report.Succeeded || report.OutputCount != 0 || report.RejectedCount != 0 || report.WarningGroupCount != 1 || report.WarningOccurrenceCount != 1 || report.DiagnosticGroupCount != 0 || report.DiagnosticOccurrenceCount != 0 || report.DiagnosticsTruncated || report.ValidationStatus != "failed" {
 		t.Fatalf("partial failure report = %#v", report)
 	}
 
@@ -463,10 +463,10 @@ func TestRunRetainsPartialPipelineOutcomeInFailureSummary(t *testing.T) {
 	if manifest.RunID != report.RunID || manifest.PipelineID != "sample" || manifest.ValidationStatus != "failed" {
 		t.Fatalf("partial manifest = %#v", manifest)
 	}
-	var warnings []contracts.Warning
+	var diagnostics contracts.DiagnosticCollection
-	readStateTestSummaryJSON(t, bundlePath, "warnings.json", &warnings)
+	readStateTestSummaryJSON(t, bundlePath, "final-diagnostics.json", &diagnostics)
-	if len(warnings) != 1 || warnings[0].ReasonCode != "partial-warning" {
+	if len(diagnostics.Groups) != 1 || diagnostics.Groups[0].ReasonCode != "partial-warning" {
-		t.Fatalf("partial warnings = %#v", warnings)
+		t.Fatalf("partial diagnostics = %#v", diagnostics)
 	}
 	var events []pipeline.CheckpointEvent
 	readStateTestSummaryJSON(t, bundlePath, "checkpoint-events.json", &events)
@@ -863,11 +863,12 @@ type stateTestHarness struct {
 	chunkCalls, extractCalls int
 	runIDCalls               uint64
 	extractErr               error
-	chunkWarnings            []contracts.Warning
+	chunkDiagnostics         []contracts.ProducerDiagnostic
 	moduleProfiles           []string
 	sessionIDs               []string
-	outputWarnings           []contracts.Warning
+	outputDiagnostics        contracts.DiagnosticCollection
 	includeWarnings          bool
+	includeWarningFile       bool
 }
 
 func newStateTestHarness() *stateTestHarness { return &stateTestHarness{} }
@@ -892,7 +893,7 @@ func (h *stateTestHarness) options() Options {
 		panic(err)
 	}
 	if err := registries.Outputs.RegisterWithSpec(pipeline.ModuleSpec{Key: "test/output", Stage: pipeline.StageOutput, ExecutionClass: contracts.ExecutionClassDeterministic, Requires: []string{"normalized"}, Provides: []string{"output"}}, func() (contracts.OutputEncoder, error) {
-		return stateTestOutput{harness: h, includeWarnings: h.includeWarnings}, nil
+		return stateTestOutput{harness: h, includeDiagnostics: h.includeWarnings}, nil
 	}); err != nil {
 		panic(err)
 	}
@@ -931,7 +932,7 @@ func (c stateTestChunker) Plan(_ context.Context, req contracts.ChunkRequest) (c
 	c.harness.mu.Lock()
 	c.harness.chunkCalls++
 	c.harness.mu.Unlock()
-	return contracts.ChunkPlanResult{Plan: source.ChunkPlan{SourceDigest: req.Source.Digest, Ranges: []source.ChunkRange{{StartUnitID: 1, EndUnitID: 1}}}, Warnings: append([]contracts.Warning(nil), c.harness.chunkWarnings...)}, nil
+	return contracts.ChunkPlanResult{Plan: source.ChunkPlan{SourceDigest: req.Source.Digest, Ranges: []source.ChunkRange{{StartUnitID: 1, EndUnitID: 1}}}, Diagnostics: contracts.CloneProducerDiagnostics(c.harness.chunkDiagnostics)}, nil
 }
 
 const stateTestArtifactKind contracts.ArtifactKind = "test/artifact"
@@ -955,6 +956,9 @@ func (stateTestCodec) Encode(v stateTestArtifact) ([]byte, error) {
 func (stateTestCodec) Decode([]byte) (stateTestArtifact, error) {
 	return stateTestArtifact{Value: "ok"}, nil
 }
+func (codec stateTestCodec) DecodeCandidate(content []byte) (stateTestArtifact, error) {
+	return codec.Decode(content)
+}
 
 type stateTestExtractor struct{ harness *stateTestHarness }
 
@@ -997,19 +1001,27 @@ func (n stateTestNormalizer) Normalize(_ context.Context, req contracts.TypedNor
 
 type stateTestOutput struct {
 	harness            *stateTestHarness
-	includeWarnings bool
+	includeDiagnostics bool
 }
 
 func (o stateTestOutput) Key() string { return "test/output" }
 func (o stateTestOutput) Encode(_ context.Context, req contracts.OutputRequest) (contracts.OutputResult, error) {
 	o.harness.mu.Lock()
-	o.harness.outputWarnings = append([]contracts.Warning(nil), req.Warnings...)
+	o.harness.outputDiagnostics = contracts.CloneDiagnosticCollection(req.Diagnostics)
 	o.harness.mu.Unlock()
 	data := []byte("{\"ok\":true}\n")
-	if o.includeWarnings && len(req.Warnings) > 0 {
+	if o.includeDiagnostics && len(req.Diagnostics.Groups) > 0 {
-		data = []byte(fmt.Sprintf("{\"ok\":true,\"warnings\":%q}\n", req.Warnings[0].ReasonCode))
+		data = []byte(fmt.Sprintf("{\"ok\":true,\"diagnostics\":%q}\n", req.Diagnostics.Groups[0].ReasonCode))
 	}
-	return contracts.OutputResult{Files: []contracts.OutputFile{{Name: "result.json", Bytes: data}}}, nil
+	files := []contracts.OutputFile{{Name: "result.json", Bytes: data}}
+	if o.harness.includeWarningFile {
+		files = append(files, contracts.OutputFile{Name: "warnings.json", Bytes: []byte("{\"warnings\":true}\n")})
+	}
+	return contracts.OutputResult{Files: files}, nil
+}
+
+func stateTestDiagnostic(scope, reasonCode, message string) contracts.ProducerDiagnostic {
+	return contracts.ProducerDiagnostic{Disposition: contracts.DiagnosticDispositionWarning, Category: contracts.DiagnosticCategoryDegradation, ReasonCode: reasonCode, OccurrenceCount: 1, Samples: []contracts.DiagnosticSample{{Scope: scope, Message: message}}}
 }
 
 type failingDebugRecorder struct{}

internal/core/artifacts/artifacts.go

@@ -104,6 +104,33 @@ type RejectedOutputManifest struct {
 	Message                string             `json:"message,omitempty"`
 	AttemptCount           int                `json:"attempt_count,omitempty"`
 	DiagnosticArtifactPath string             `json:"diagnostic_artifact_path,omitempty"`
+	Validation             *ValidationSummary `json:"validation,omitempty"`
+}
+
+// ValidationSummary is the bounded, durable outcome of validating one
+// producer result. It deliberately contains identities and stable codes, not
+// model responses, corrective guidance, validator diagnostics, or payloads.
+type ValidationSummary struct {
+	Stage                string   `json:"stage,omitempty"`
+	StepID               string   `json:"step_id,omitempty"`
+	LaneID               string   `json:"lane_id,omitempty"`
+	ModuleKey            string   `json:"module_key,omitempty"`
+	ChunkID              string   `json:"chunk_id,omitempty"`
+	ChunkIndex           int      `json:"chunk_index,omitempty"`
+	Status               string   `json:"status"`
+	RejectingValidators  []string `json:"rejecting_validators,omitempty"`
+	ReasonCodes          []string `json:"reason_codes,omitempty"`
+	IncompleteValidators []string `json:"incomplete_validators,omitempty"`
+	ProducerAttemptCount int      `json:"producer_attempt_count"`
+	TerminalAction       string   `json:"terminal_action"`
+}
+
+// CloneValidationSummary returns an independently owned durable summary.
+func CloneValidationSummary(summary ValidationSummary) ValidationSummary {
+	summary.RejectingValidators = append([]string(nil), summary.RejectingValidators...)
+	summary.ReasonCodes = append([]string(nil), summary.ReasonCodes...)
+	summary.IncompleteValidators = append([]string(nil), summary.IncompleteValidators...)
+	return summary
 }
 
 type CheckpointDecisionManifest struct {
@@ -163,6 +190,7 @@ type RunManifest struct {
 	References          []ReferenceProvenance        `json:"references,omitempty"`
 	NormalizedOutputs   []NormalizedOutputManifest   `json:"normalized_outputs,omitempty"`
 	RejectedOutputs     []RejectedOutputManifest     `json:"rejected_outputs,omitempty"`
+	ValidationSummaries []ValidationSummary          `json:"validation_summaries,omitempty"`
 	CheckpointDecisions []CheckpointDecisionManifest `json:"checkpoint_decisions,omitempty"`
 	LLMProfiles         []LLMProfileManifest         `json:"llm_profiles,omitempty"`
 	Metadata            map[string]any               `json:"metadata,omitempty"`

internal/core/config/config.go

@@ -116,6 +116,11 @@ func (c *ConcurrencyConfig) recomputeStageWorkerDefaults() {
 
 func clonePipelineProfile(in pipeline.PipelineProfile) pipeline.PipelineProfile {
 	out := in
+	out.ValidationPolicy = cloneValidationPolicyOverride(in.ValidationPolicy)
+	if in.StructuredOutputRepairAttempts != nil {
+		value := *in.StructuredOutputRepairAttempts
+		out.StructuredOutputRepairAttempts = &value
+	}
 	out.Input = cloneModuleBinding(in.Input)
 	out.Chunk = cloneModuleBinding(in.Chunk)
 	out.Output = cloneModuleBinding(in.Output)
@@ -196,6 +201,11 @@ func cloneReferenceSource(in pipeline.ReferenceSource) pipeline.ReferenceSource
 
 func cloneModuleBinding(in pipeline.ModuleBinding) pipeline.ModuleBinding {
 	out := in
+	out.ValidationPolicy = cloneValidationPolicyOverride(in.ValidationPolicy)
+	if in.StructuredOutputRepairAttempts != nil {
+		value := *in.StructuredOutputRepairAttempts
+		out.StructuredOutputRepairAttempts = &value
+	}
 	if len(in.Options) > 0 {
 		out.Options = cloneOptions(in.Options)
 	}
@@ -204,6 +214,26 @@ func cloneModuleBinding(in pipeline.ModuleBinding) pipeline.ModuleBinding {
 	return out
 }
 
+func cloneValidationPolicyOverride(in *pipeline.ValidationPolicyOverride) *pipeline.ValidationPolicyOverride {
+	if in == nil {
+		return nil
+	}
+	out := *in
+	if in.ProducerStructuralFailure != nil {
+		value := *in.ProducerStructuralFailure
+		out.ProducerStructuralFailure = &value
+	}
+	if in.SemanticRejection != nil {
+		value := *in.SemanticRejection
+		out.SemanticRejection = &value
+	}
+	if in.ValidatorFailure != nil {
+		value := *in.ValidatorFailure
+		out.ValidatorFailure = &value
+	}
+	return &out
+}
+
 func cloneValidatorOverride(in pipeline.ValidatorOverride) pipeline.ValidatorOverride {
 	out := pipeline.ValidatorOverride{Set: in.Set}
 	if len(in.Validators) > 0 {

internal/core/config/effective_config_contract_test.go

@@ -427,6 +427,10 @@ func (effectiveCodec) Decode(content []byte) (effectiveArtifact, error) {
 	return value, err
 }
 
+func (codec effectiveCodec) DecodeCandidate(content []byte) (effectiveArtifact, error) {
+	return codec.Decode(content)
+}
+
 type effectiveInput struct{ key string }
 
 func (m effectiveInput) Key() string { return m.key }

internal/core/config/file_config.go

@@ -3,6 +3,7 @@ package config
 import (
 	"bytes"
 	"fmt"
+	"io"
 	"os"
 	"path/filepath"
 	"sort"
@@ -35,6 +36,8 @@ type FilePromptKitLocalBackendConfig struct {
 
 type FilePipelineProfile struct {
 	LLMProfile                     *string                            `yaml:"llm_profile,omitempty"`
+	StructuredOutputRepairAttempts *int                               `yaml:"structured_output_repair_attempts,omitempty"`
+	ValidationPolicy               *pipeline.ValidationPolicyOverride `yaml:"validation_policy,omitempty"`
 	Input                          fileModuleBinding                  `yaml:"input"`
 	Chunk                          *fileModuleBinding                 `yaml:"chunk,omitempty"`
 	Artifacts                      map[string]FileArtifactLaneProfile `yaml:"artifacts,omitempty"`
@@ -47,15 +50,25 @@ type FilePipelineProfile struct {
 }
 
 func (p *FilePipelineProfile) UnmarshalYAML(node *yaml.Node) error {
+	if err := validateStructuredOutputRepairAttemptsNode(node, "pipeline profile"); err != nil {
+		return err
+	}
 	type plainFilePipelineProfile FilePipelineProfile
 	var decoded plainFilePipelineProfile
 	seen, err := decodeKnownMapping(node, &decoded, map[string]struct{}{
-		"llm_profile": {}, "input": {}, "chunk": {}, "artifacts": {}, "steps": {}, "output": {}, "references": {},
+		"llm_profile": {}, "structured_output_repair_attempts": {}, "validation_policy": {}, "input": {}, "chunk": {}, "artifacts": {}, "steps": {}, "output": {}, "references": {},
 	}, "pipeline profile")
 	if err != nil {
 		return err
 	}
 	*p = FilePipelineProfile(decoded)
+	if validationPolicyNode, ok := mappingValue(node, "validation_policy"); ok {
+		policy, err := parseValidationPolicy(validationPolicyNode, "pipeline profile")
+		if err != nil {
+			return err
+		}
+		p.ValidationPolicy = policy
+	}
 	_, p.artifactsSet = seen["artifacts"]
 	_, p.stepsSet = seen["steps"]
 	_, p.llmProfileSet = seen["llm_profile"]
@@ -148,6 +161,8 @@ type FileDebugConfig struct {
 type fileModuleBinding struct {
 	Module                         string
 	LLMProfile                     string
+	StructuredOutputRepairAttempts *int
+	ValidationPolicy               *pipeline.ValidationPolicyOverride
 	Retries                        int
 	Options                        map[string]any
 	References                     map[string]fileReferenceSource
@@ -244,9 +259,14 @@ func (b *fileModuleBinding) UnmarshalYAML(node *yaml.Node) error {
 		b.Module = strings.TrimSpace(module)
 		return nil
 	case yaml.MappingNode:
+		seen := make(map[string]struct{}, len(node.Content)/2)
 		for i := 0; i < len(node.Content); i += 2 {
 			keyNode := node.Content[i]
 			valueNode := node.Content[i+1]
+			if _, exists := seen[keyNode.Value]; exists {
+				return fmt.Errorf("module binding field %q is duplicated", keyNode.Value)
+			}
+			seen[keyNode.Value] = struct{}{}
 			switch keyNode.Value {
 			case "module":
 				var module string
@@ -263,6 +283,18 @@ func (b *fileModuleBinding) UnmarshalYAML(node *yaml.Node) error {
 				if b.LLMProfile == "" {
 					return fmt.Errorf("llm_profile must not be empty when set")
 				}
+			case "structured_output_repair_attempts":
+				attempts, err := parseStructuredOutputRepairAttempts(valueNode, "module binding")
+				if err != nil {
+					return err
+				}
+				b.StructuredOutputRepairAttempts = attempts
+			case "validation_policy":
+				policy, err := parseValidationPolicy(valueNode, "module binding")
+				if err != nil {
+					return err
+				}
+				b.ValidationPolicy = policy
 			case "retries":
 				var retries int
 				if err := valueNode.Decode(&retries); err != nil {
@@ -305,13 +337,100 @@ func (b fileModuleBinding) toPipelineBinding() pipeline.ModuleBinding {
 	return pipeline.ModuleBinding{
 		Module:                         strings.TrimSpace(b.Module),
 		LLMProfile:                     strings.TrimSpace(b.LLMProfile),
+		StructuredOutputRepairAttempts: cloneStructuredOutputRepairAttempts(b.StructuredOutputRepairAttempts),
+		ValidationPolicy:               cloneValidationPolicyOverride(b.ValidationPolicy),
 		Retries:                        b.Retries,
 		Options:                        cloneOptions(b.Options),
 		References:                     fileReferenceSourcesToPipeline(b.References),
-		Validators: b.Validators,
+		Validators:                     cloneValidatorOverride(b.Validators),
 	}
 }
 
+func mappingValue(node *yaml.Node, key string) (*yaml.Node, bool) {
+	for i := 0; i < len(node.Content); i += 2 {
+		if node.Content[i].Value == key {
+			return node.Content[i+1], true
+		}
+	}
+	return nil, false
+}
+
+func parseValidationPolicy(node *yaml.Node, context string) (*pipeline.ValidationPolicyOverride, error) {
+	if node == nil || node.Tag == "!!null" || node.Kind != yaml.MappingNode {
+		return nil, fmt.Errorf("%s validation_policy must be an object", context)
+	}
+	policy := &pipeline.ValidationPolicyOverride{}
+	seen := make(map[string]struct{}, len(node.Content)/2)
+	for i := 0; i < len(node.Content); i += 2 {
+		key := node.Content[i].Value
+		value := node.Content[i+1]
+		if _, exists := seen[key]; exists {
+			return nil, fmt.Errorf("%s validation_policy field %q is duplicated", context, key)
+		}
+		seen[key] = struct{}{}
+		if value.Tag == "!!null" || value.Kind != yaml.ScalarNode || value.Tag != "!!str" {
+			return nil, fmt.Errorf("%s validation_policy.%s must be a string", context, key)
+		}
+		switch key {
+		case "producer_structural_failure":
+			value := pipeline.ProducerStructuralFailureAction(value.Value)
+			policy.ProducerStructuralFailure = &value
+		case "semantic_rejection":
+			value := pipeline.SemanticRejectionAction(value.Value)
+			policy.SemanticRejection = &value
+		case "validator_failure":
+			value := pipeline.ValidatorFailureAction(value.Value)
+			policy.ValidatorFailure = &value
+		default:
+			return nil, fmt.Errorf("field %s not found in %s validation_policy", key, context)
+		}
+	}
+	if err := policy.Validate(); err != nil {
+		return nil, fmt.Errorf("%s validation_policy: %w", context, err)
+	}
+	return policy, nil
+}
+
+func validateStructuredOutputRepairAttemptsNode(node *yaml.Node, context string) error {
+	if node.Kind != yaml.MappingNode {
+		return fmt.Errorf("%s must be an object", context)
+	}
+	for i := 0; i < len(node.Content); i += 2 {
+		if node.Content[i].Value != "structured_output_repair_attempts" {
+			continue
+		}
+		if _, err := parseStructuredOutputRepairAttempts(node.Content[i+1], context); err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func parseStructuredOutputRepairAttempts(node *yaml.Node, context string) (*int, error) {
+	if node.Tag == "!!null" {
+		return nil, fmt.Errorf("%s structured_output_repair_attempts must not be null", context)
+	}
+	if node.Kind != yaml.ScalarNode || node.Tag != "!!int" {
+		return nil, fmt.Errorf("%s structured_output_repair_attempts must be an integer", context)
+	}
+	var attempts int
+	if err := node.Decode(&attempts); err != nil {
+		return nil, fmt.Errorf("%s structured_output_repair_attempts must be an integer: %w", context, err)
+	}
+	if attempts < 0 || attempts > 3 {
+		return nil, fmt.Errorf("%s structured_output_repair_attempts must be between zero and three", context)
+	}
+	return &attempts, nil
+}
+
+func cloneStructuredOutputRepairAttempts(attempts *int) *int {
+	if attempts == nil {
+		return nil
+	}
+	value := *attempts
+	return &value
+}
+
 func LoadFileConfig(path string) (FileConfig, error) {
 	data, err := os.ReadFile(path)
 	if err != nil {
@@ -349,6 +468,12 @@ func ParseFileConfigYAML(data []byte) (FileConfig, error) {
 	if err := decoder.Decode(&fileCfg); err != nil {
 		return FileConfig{}, fmt.Errorf("decode yaml: %w", err)
 	}
+	var trailing any
+	if err := decoder.Decode(&trailing); err == nil {
+		return FileConfig{}, fmt.Errorf("config must contain exactly one YAML document")
+	} else if err != io.EOF {
+		return FileConfig{}, fmt.Errorf("decode trailing yaml document: %w", err)
+	}
 	return fileCfg, nil
 }
 
@@ -513,6 +638,8 @@ func (c *Config) applyFileConfigWithLookup(fileCfg FileConfig, lookup func(strin
 		profile := pipeline.PipelineProfile{
 			ID:                             pipelineID,
 			LLMProfile:                     llmProfile,
+			StructuredOutputRepairAttempts: cloneStructuredOutputRepairAttempts(filePipeline.StructuredOutputRepairAttempts),
+			ValidationPolicy:               cloneValidationPolicyOverride(filePipeline.ValidationPolicy),
 			Input:                          filePipeline.Input.toPipelineBinding(),
 			Artifacts:                      make(map[string]pipeline.ArtifactLaneProfile, len(filePipeline.Artifacts)),
 			References:                     fileReferenceSourcesToPipeline(filePipeline.References),

internal/core/config/file_config_contract_test.go

@@ -50,6 +50,28 @@ func TestFileConfigMinimalVersion4AppliesOverDefaults(t *testing.T) {
 	}
 }
 
+func TestParseFileConfigYAMLRejectsAdditionalDocuments(t *testing.T) {
+	tests := []struct {
+		name    string
+		source  string
+		wantErr bool
+	}{
+		{name: "trailing whitespace", source: "version: 4\n\n  \t", wantErr: false},
+		{name: "trailing comment", source: "version: 4\n# trailing comment\n", wantErr: false},
+		{name: "second valid document", source: "version: 4\n---\nversion: 4\n", wantErr: true},
+		{name: "second empty document", source: "version: 4\n---\n", wantErr: true},
+		{name: "second malformed document", source: "version: 4\n---\nversion: [\n", wantErr: true},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			_, err := ParseFileConfigYAML([]byte(tt.source))
+			if (err != nil) != tt.wantErr {
+				t.Fatalf("ParseFileConfigYAML() error = %v, want error=%t", err, tt.wantErr)
+			}
+		})
+	}
+}
+
 func TestFilePipelineLLMProfileIsPresenceAwareAndDetached(t *testing.T) {
 	const pipelineYAML = `version: 4
 pipelines:
@@ -116,6 +138,168 @@ pipelines:
 	}
 }
 
+func TestStructuredOutputRepairAttemptsFileConfigurationPreservesPresenceAndOwnership(t *testing.T) {
+	const pipelineYAML = `version: 4
+pipelines:
+  main:
+    %s
+    input: input
+    artifacts:
+      lane:
+        extract:
+          module: extract
+          structured_output_repair_attempts: 2
+          validators:
+            - module: validator
+              structured_output_repair_attempts: 3
+`
+
+	t.Run("omitted pipeline value remains absent", func(t *testing.T) {
+		file := parseFileConfig(t, fmt.Sprintf(pipelineYAML, ""))
+		if got := file.Pipelines["main"].StructuredOutputRepairAttempts; got != nil {
+			t.Fatalf("file pipeline repair attempts = %v, want nil", *got)
+		}
+		cfg := Default()
+		if err := cfg.ApplyFileConfig(file); err != nil {
+			t.Fatal(err)
+		}
+		if got := cfg.Pipelines["main"].StructuredOutputRepairAttempts; got != nil {
+			t.Fatalf("pipeline repair attempts = %v, want nil", *got)
+		}
+		if got := cfg.Pipelines["main"].Input.StructuredOutputRepairAttempts; got != nil {
+			t.Fatalf("scalar input binding repair attempts = %v, want nil", *got)
+		}
+	})
+
+	t.Run("zero is explicit and survives configuration boundaries", func(t *testing.T) {
+		file := parseFileConfig(t, fmt.Sprintf(pipelineYAML, "structured_output_repair_attempts: 0"))
+		cfg := Default()
+		if err := cfg.ApplyFileConfig(file); err != nil {
+			t.Fatal(err)
+		}
+		profile := cfg.Pipelines["main"]
+		if profile.StructuredOutputRepairAttempts == nil || *profile.StructuredOutputRepairAttempts != 0 {
+			t.Fatalf("pipeline repair attempts = %v, want explicit zero", profile.StructuredOutputRepairAttempts)
+		}
+		lane := profile.Artifacts["lane"]
+		if lane.Extract.StructuredOutputRepairAttempts == nil || *lane.Extract.StructuredOutputRepairAttempts != 2 {
+			t.Fatalf("extract repair attempts = %v, want 2", lane.Extract.StructuredOutputRepairAttempts)
+		}
+		if lane.Extract.Validators.Validators[0].StructuredOutputRepairAttempts == nil || *lane.Extract.Validators.Validators[0].StructuredOutputRepairAttempts != 3 {
+			t.Fatalf("validator repair attempts = %v, want 3", lane.Extract.Validators.Validators[0].StructuredOutputRepairAttempts)
+		}
+
+		*file.Pipelines["main"].StructuredOutputRepairAttempts = 1
+		if got := *cfg.Pipelines["main"].StructuredOutputRepairAttempts; got != 0 {
+			t.Fatalf("applied config aliased file configuration: got %d, want 0", got)
+		}
+		fileProfile := file.Pipelines["main"]
+		fileLane := fileProfile.Artifacts["lane"]
+		*fileLane.Extract.StructuredOutputRepairAttempts = 1
+		*fileLane.Extract.Validators.Validators[0].StructuredOutputRepairAttempts = 1
+		fileProfile.Artifacts["lane"] = fileLane
+		file.Pipelines["main"] = fileProfile
+		configuredLane := cfg.Pipelines["main"].Artifacts["lane"]
+		if got := *configuredLane.Extract.StructuredOutputRepairAttempts; got != 2 {
+			t.Fatalf("configured extract aliased file configuration: got %d, want 2", got)
+		}
+		if got := *configuredLane.Extract.Validators.Validators[0].StructuredOutputRepairAttempts; got != 3 {
+			t.Fatalf("configured validator aliased file configuration: got %d, want 3", got)
+		}
+		cloned := cloneConfig(cfg)
+		*cloned.Pipelines["main"].StructuredOutputRepairAttempts = 1
+		if got := *cfg.Pipelines["main"].StructuredOutputRepairAttempts; got != 0 {
+			t.Fatalf("cloned config aliased source configuration: got %d, want 0", got)
+		}
+		redacted := cfg.Redacted()
+		*redacted.Pipelines["main"].StructuredOutputRepairAttempts = 1
+		if got := *cfg.Pipelines["main"].StructuredOutputRepairAttempts; got != 0 {
+			t.Fatalf("redacted config aliased source configuration: got %d, want 0", got)
+		}
+		summary := cfg.RedactedSummaryPayload().(Config)
+		if summary.Pipelines["main"].StructuredOutputRepairAttempts == nil || *summary.Pipelines["main"].StructuredOutputRepairAttempts != 0 {
+			t.Fatalf("redacted summary pipeline repair attempts = %v, want explicit zero", summary.Pipelines["main"].StructuredOutputRepairAttempts)
+		}
+
+		data, err := json.Marshal(cfg)
+		if err != nil {
+			t.Fatal(err)
+		}
+		var roundTripped Config
+		if err := json.Unmarshal(data, &roundTripped); err != nil {
+			t.Fatal(err)
+		}
+		roundTrippedProfile := roundTripped.Pipelines["main"]
+		if roundTrippedProfile.StructuredOutputRepairAttempts == nil || *roundTrippedProfile.StructuredOutputRepairAttempts != 0 {
+			t.Fatalf("round-tripped pipeline repair attempts = %v, want explicit zero", roundTrippedProfile.StructuredOutputRepairAttempts)
+		}
+		roundTrippedLane := roundTrippedProfile.Artifacts["lane"]
+		if roundTrippedLane.Extract.StructuredOutputRepairAttempts == nil || *roundTrippedLane.Extract.StructuredOutputRepairAttempts != 2 {
+			t.Fatalf("round-tripped extract repair attempts = %v, want 2", roundTrippedLane.Extract.StructuredOutputRepairAttempts)
+		}
+		if roundTrippedLane.Extract.Validators.Validators[0].StructuredOutputRepairAttempts == nil || *roundTrippedLane.Extract.Validators.Validators[0].StructuredOutputRepairAttempts != 3 {
+			t.Fatalf("round-tripped validator repair attempts = %v, want 3", roundTrippedLane.Extract.Validators.Validators[0].StructuredOutputRepairAttempts)
+		}
+	})
+}
+
+func TestStructuredOutputRepairAttemptsFileConfigurationRejectsInvalidValues(t *testing.T) {
+	const pipelineYAML = `version: 4
+pipelines:
+  main:
+    input: input
+    artifacts:
+      lane:
+        extract: extract
+    %s
+`
+	const bindingYAML = `version: 4
+pipelines:
+  main:
+    input:
+      module: input
+      %s
+    artifacts:
+      lane:
+        extract: extract
+`
+
+	for _, tt := range []struct {
+		name   string
+		source string
+		want   string
+	}{
+		{name: "null pipeline value", source: "structured_output_repair_attempts: null", want: "pipeline profile structured_output_repair_attempts must not be null"},
+		{name: "fractional pipeline value", source: "structured_output_repair_attempts: 1.5", want: "pipeline profile structured_output_repair_attempts must be an integer"},
+		{name: "quoted pipeline value", source: "structured_output_repair_attempts: '1'", want: "pipeline profile structured_output_repair_attempts must be an integer"},
+		{name: "out of range pipeline value", source: "structured_output_repair_attempts: 4", want: "pipeline profile structured_output_repair_attempts must be between zero and three"},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			_, err := ParseFileConfigYAML([]byte(fmt.Sprintf(pipelineYAML, tt.source)))
+			if err == nil || !strings.Contains(err.Error(), tt.want) {
+				t.Fatalf("ParseFileConfigYAML() error = %v, want %q", err, tt.want)
+			}
+		})
+	}
+
+	for _, tt := range []struct {
+		name   string
+		source string
+		want   string
+	}{
+		{name: "null binding value", source: "structured_output_repair_attempts: null", want: "module binding structured_output_repair_attempts must not be null"},
+		{name: "noninteger binding value", source: "structured_output_repair_attempts: true", want: "module binding structured_output_repair_attempts must be an integer"},
+		{name: "out of range binding value", source: "structured_output_repair_attempts: -1", want: "module binding structured_output_repair_attempts must be between zero and three"},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			_, err := ParseFileConfigYAML([]byte(fmt.Sprintf(bindingYAML, tt.source)))
+			if err == nil || !strings.Contains(err.Error(), tt.want) {
+				t.Fatalf("ParseFileConfigYAML() error = %v, want %q", err, tt.want)
+			}
+		})
+	}
+}
+
 func TestFileModuleBindingRejectsExplicitEmptyLLMProfile(t *testing.T) {
 	const configYAML = `version: 4
 pipelines: