eric/promptkit
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Remove superseded configuration loaders and consolidate runtime tests

Browse Source
This commit is contained in:
Eric Rakestraw
2026-09-30 04:18:27 +00:00
parent f17a38f60b
commit 48343300e2
98 changed files with 817 additions and 12180 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
docs/policy/architecture.md
View File

@@ -51,23 +51,22 @@ The [internal component overview](../internal/overview.md) owns the inventory
of implemented packages and their concrete responsibilities. This policy owns
the boundaries and dependency rules those components must preserve.
The internal engine assembles one immutable backend registry, one capacity manager,
the internal repositories, renderer, validator, outbound client, and use-case
runner. The registry contains validated maintained definitions plus engine-scoped
consumer additions. The root facade translates public values and errors at the
library boundary.
The internal engine constructs the capacity manager from the registry's immutable
policy snapshot, wraps the selected built-in or injected model client, and
supplies bounded admission to the runner. The defaults and renderer depend on
the domain model. Prompt-definition and profile repositories use the domain
model, file catalog, and YAML decoder. The catalog adapter validates imported
immutable maintained data, including its reserved namespace, before internal
engine assembly supplies it to the profile package. Artifact reading uses the
domain model and application-neutral defaults. Validation uses the domain model,
file catalog, and JSON Schema implementation. The model client uses the domain
model and application-neutral defaults, and an injected or standard-library HTTP
client. The use-case runner depends on the narrow interfaces owned by each
internal component, including backend lookup and run admission.
The internal engine composes selected catalog and caller ConfigKit inputs once,
then assembles the captured definitions, capacity manager, renderer, validator,
outbound client, and use-case runner. The configuration adapter interprets and
validates only the final effective values, including presence, null, and
resource context. It retains source history for explicit operator inspection,
not for validation or a second resolution pass. The root facade translates
public values and errors at the library boundary.
The engine constructs capacity policy from final backend definitions, wraps
the selected built-in or injected model client, and supplies bounded admission
to the runner. Runtime/domain components consume execution-owned values, not
ConfigKit types. Artifact reading and validation use the domain model and
application-neutral defaults; validation also uses file-catalog helpers and
JSON Schema. The model client uses the domain model, defaults, and an injected
or standard-library HTTP client. The runner coordinates narrow lookup and
collaborator interfaces without rereading configuration.
The current implementation follows this dependency direction:
@@ -84,9 +83,10 @@ downstream consumers, including Scriptorium
narrow injected abstractions
```
The backend registry depends on the domain model and shared JSON-value
validation, has no mutation API after construction, and consumes the
OpenAI-compatible reserved request-field rule owned by the model client. The
The backend index depends on the domain model and shared JSON-value ownership
and has no mutation API after construction. The configuration adapter applies
the OpenAI-compatible reserved request-field rule owned by the model client
when it validates final provider maps. The
capacity component depends on the domain model and the narrow internal
model-client boundary, not on provider transport implementation. The model
client does not depend on registry or capacity configuration. The facade