promptkit/docs/roadmap/conversation-continuation.md

Conversation Continuation And Application-Directed Interaction

Status: continuation and coordinator behavioral decisions accepted; optional execution groups accepted as a separable addition with open decisions below. No implementation is authorized by this document. Exact public types and signatures remain for implementation planning.

North Star: Content-Neutral, Generally Useful Capabilities

Accepted direction: evaluate each capability for its usefulness across different downstream applications. PromptKit must be able to implement it and explain its value without assuming or inferring the meaning of the content passed through prompts and responses.

PromptKit can own conversation and execution mechanics: captured messages, continuations, independent branches, execution relationships, budgets, cancellation, observations, and usage. Applications own the interpretation of responses and their reasons for choosing the next action. Repair, analysis, clarification, and branching exploration are examples of consumer workflows that could use the same mechanisms.

Decision 3 selects an optional sequential coordinator built on independently usable continuation primitives. Its value is generally useful execution and conversation management, rather than implementation of Audita's repair loop. It executes application-directed interactions without needing to understand why the application requests another turn or stops. Applications retain orchestration across branches; that initial scope does not prohibit future capabilities.

Apply two questions to proposed public concepts and implementation machinery:

1. Can their value be explained using conversation and execution concepts alone?
2. Can different applications use them without translating their domain into repair-specific concepts?

Audita supplies concrete evidence of need and acceptance examples; its workflow does not define PromptKit's abstraction. Terms such as rejected candidate, accepted revision, and withdrawal belong to the consuming application. This direction concerns the new capabilities and does not rename or expand PromptKit's existing structural-output repair functionality.

Purpose And Scope

This roadmap records PromptKit's design questions arising from Audita's feature request. That request owns the consumer workflow and requested acceptance scenarios; this document owns the proposed PromptKit direction, discussion order, and eventual implementation sequencing.

The proposed capabilities are:

• Capture a rendered request and its final assistant content, then continue it using ordinary configured prompts without reconstructing the original request.
• Allow independent continuations from one captured exchange.
• Provide an optional sequential coordinator for bounded looping using application-supplied next actions and follow-up requests.
• Provide optional execution groups for explicitly selected runs and summaries of their available execution facts, independently of conversation continuation.

Continuation has direct value by removing consumer capture, reconstruction, prefix comparison, and history assembly. The coordinator centralizes turn progression, cancellation, and budgets while reusing ordinary execution observations. Captured exchanges supply the minimal parent relationship selected in Decision 5. Its value comes from removing repeated lifecycle work, beyond merely replacing a consumer's loop counter. Audita would continue owning candidate selection, domain validation, feedback meaning, branch scheduling, accepted sibling proposals, and transcript changes. Its domain-specific repair code would not move wholesale into PromptKit.

The request needs in-process use. Persistent sessions, restartable workflows, conversation storage, and batch-item reconciliation are outside its scope. The existing prepared execution handle remains a single-use execution capability; continuation needs its own semantics.

Decisions In Discussion Order

1. What Does A Captured Exchange Contain?

Status: accepted; exact capture API remains for implementation planning. Capture represents the logical conversation. Structural-output repair attempts and corrective instructions are internal execution details, not conversation turns.

Accepted approach: capture the complete rendered request, preserving message content, roles, and cache metadata, followed by the final assistant content returned by the run. When structural repair succeeds, retain only the final corrected assistant content, excluding intermediate candidates and corrective instructions from the logical conversation. This supplies reusable context for subsequent turns while keeping transport attempts and structural repair separate from application turns. It also matches Audita's requested history.

Accepted capture availability: when capture is requested, follow the ordinary distinction between a completed result and an operational error:

Execution outcome	Capture behavior
Completed response that satisfies its output contract	Capture the logical exchange.
Completed response with failed structural validation, including repair exhaustion	Capture the final returned response; preserve the failed validation status in the ordinary result.
Operational error with no completed result	Produce no new exchange; the preceding exchange remains usable.
Completed response followed by cancellation before coordinator callback delivery	Preserve the undelivered response under Decision 4's accepted optional-result behavior.

Capture does not imply validation success or application acceptance. The coordinator delivers completed results with their ordinary validation status; the application decides whether to stop or continue, potentially using another output contract. Failed structural validation is not reclassified as an operational error. Only the final returned content enters the logical history, never intermediate structural-repair responses or corrective instructions.

Accepted opt-in behavior: support capture per execution through both ordinary and prepared execution. Ordinary execution returns the captured exchange alongside its usual result. For prepared execution, retain the capture choice with the prepared handle and create the exchange when that exact execution completes. Preparation alone produces no completed exchange because it has no assistant response. A continuation operation implies capture so it can produce the next reusable exchange.

Construct capture from the request actually executed. Consumers must not need to reconstruct history or pair independently prepared data with a response. An ordinary request option is the preferred API direction; separate capture methods are also viable. Select the precise representation during API planning.

Accepted ownership and lifetime: a captured exchange is an immutable, reusable snapshot held by the caller. Continuing it appends the rendered follow-up messages and final response to produce a new exchange, leaving the source exchange and other branches unchanged. Independent branches may share their starting exchange. Inspection values cannot mutate retained history. Previously rendered history does not require the original input files or templates to be rendered again.

A completed exchange retains no backend admission lease, direct credential, or execution callback. Ordinary Go memory management handles its lifetime when references are released; it adds no Close or Discard lifecycle. The existing single-use prepared execution lifecycle remains separate.

Accepted engine independence: another engine instance in the same process may continue a captured exchange. The receiving engine resolves the new prompt, profile, output contract, and credentials under Decision 2's rules. It does not need the originating engine to rerender prior messages. This permits retaining conversations while an application rebuilds its engine with new configuration. It does not imply persistence, restartable sessions, serialization for later resumption, or cross-process portability.

Alternative considered, not selected: retain the corrective messages and responses that led to the final output. This provides more execution detail, but increases retained content and carries internal repair instructions into later model context. Fine-grained execution tracing is outside this feature's purpose. Existing attempt and usage observations remain separate from logical conversation content; this decision adds no requirement to capture internal repair text.

Additional alternatives considered, not selected:

• Capture only responses that satisfy their output contract. This would offer a validated-output condition but unnecessarily restrict continuing from otherwise available conversation content. Applications retain that choice.
• Bind exchanges to their originating engine. Prepared executions need ownership of execution-specific state, but captured conversation data has no equivalent requirement under the selected fresh-execution settings model.
• Use a mutable conversation object. This would make appending convenient but require explicit copying or forking to preserve independent branches. Immutable snapshots fit the selected branch behavior directly.

Implementation planning: select the opt-in API and exchange representation within these behaviors. The request's “full final assistant response” refers to content supported by PromptKit, without adding a raw provider-envelope API.

2. How Are Follow-Up Execution Settings Selected?

Status: accepted. A continuation supplies existing conversational context to a normally configured new execution. Conversation history does not introduce another source of execution-setting inheritance.

Accepted approach: preserve conversation content while resolving the new profile and output contract through the follow-up prompt and ordinary request overrides. The follow-up can therefore use a different model and response schema without modifying the original exchange. Explicit request overrides use ordinary precedence, including complete replacement of the output contract when supplied. Reuse existing selection and precedence rules.

In particular, prefer the follow-up prompt's default profile when no explicit profile is supplied, with the ordinary missing-profile behavior if neither is available. Do not retain the original direct credential merely to retain conversation history; resolve credential supply for the new execution using the ordinary credential rules.

Select the session ID using the new request's explicit value or the follow-up prompt's session template under the existing rules. Do not implicitly inherit the preceding execution's session ID. An application can explicitly supply the same session ID across turns. Conversation relationships are recorded separately and do not depend on a shared session ID.

Alternative considered, not selected: inherit the originating execution's profile and settings unless overridden. This is convenient for repeated turns against the same model, but adds another precedence source and can make a follow-up prompt's defaults surprising. Inheriting an old output contract is especially unsuitable when successive turns expect different response shapes. An explicit inheritance option is not selected for this feature; reconsider it only if ordinary selection leaves a demonstrated usability gap.

Implementation planning detail: specify message assembly order for captured history, follow-up definition messages, and caller-appended messages, preserving authored roles. This does not reopen the accepted execution-setting precedence.

3. Should PromptKit Coordinate Turns, And Where Does Application Work Run?

Status: accepted; exact public types remain for implementation planning. Provide independently usable continuation primitives and an optional sequential coordinator for one branch. The application selects each next request or decides to stop. PromptKit manages turn progression and execution bookkeeping under the caller's chosen additional-turn budget and context.

Applications can create several independent continuations or coordinators from the same exchange. Scheduling branches, joining results, and deciding whether a branch failure should stop other work remain application responsibilities in this initial scope. Consumers can also use continuation primitives in their own loops without adopting the coordinator.

The current run flow retains logical admission through structural validation and repair. Releasing an active generation permit alone does not release that admission lease.

Accepted execution lifecycle: compose completed executions. Finish a turn and release its logical admission lease before invoking application work. Each later turn receives ordinary admission. The coordinator holds no backend capacity while awaiting application work. Callbacks can inspect responses, consult application data, wait for application input, or make model calls against the same engine and backend, including at capacity one. PromptKit need not interpret the purpose of that work.

Alternatives considered, not selected:

• Provide only continuation and observation primitives. This would keep the public API smaller, but leave repeated lifecycle and accounting coordination downstream. Independent primitive use remains supported alongside the selected coordinator.
• Provide a coordinator for an entire branching tree. This could manage more work, but introduces scheduling, concurrency, joining, and failure-policy questions without demonstrated shared answers across consumers. Applications can build those workflows from the selected branch-level capabilities.

Implementation planning: select exact request/stop and outcome types under Decision 4's accepted response delivery, turn-budget consumption, zero-turn behavior, cancellation, and minimal error/result semantics. Decision 5 records the selected execution association and accounting approach. The sketches below illustrate the selected responsibility split; they do not settle the remaining public contracts.

Illustrative Consumer Sketches

These are pseudocode, not implemented Go APIs or selected exported names. coordinate, Continue, Stop, and callback state fields are illustrative. request represents an ordinary PromptKit request, with the selection rules accepted in Decision 2. Each sketch starts from an already captured exchange. The callback can keep application state in a closure; PromptKit need not define a universal domain-state or feedback schema.

Analysis loop: the application examines the latest response and selects a question or decides that its investigation is complete.

analysis = application.start_analysis(initial_exchange)

choose_next(ctx, turn):
    if turn.new_result exists:
        analysis.record(turn.new_result)

    question = application.choose_next_question(analysis)
    if question is absent:
        return Stop()

    return Continue(request(
        prompt_id = "research/investigate",
        vars = {"question": question}
    ))

termination, error = coordinate(
    ctx, start = initial_exchange,
    additional_turns = 3, next = choose_next
)
application.present_analysis(analysis, termination, error)

PromptKit preserves the history and performs each selected execution. It does not decide whether the response answered the question, whether the evidence is sufficient, or what to investigate next. The application callback may consult other sources or call the engine before returning its next action.

Audita branch: the application starts with one already-assessed correction and its rejection findings. It interprets and assesses each new response; the coordinator uses the same generic controls as the analysis loop.

branch = audita.branch(candidate, existing_findings)

choose_next(ctx, turn):
    if turn.new_result exists:
        response = audita.decode_response(turn.new_result)
        if audita.is_withdrawal(response):
            branch.mark_withdrawn()
            return Stop()

        branch.update_candidate(response)
        assessment = audita.assess(ctx, engine, branch.candidate)
        branch.record(assessment)
        if audita.should_stop(assessment):
            return Stop()

    return Continue(request(
        prompt_id = branch.configured_follow_up_prompt,
        vars = audita.follow_up_variables(branch)
    ))

termination, error = coordinate(
    ctx, start = initial_exchange,
    additional_turns = 3, next = choose_next
)
audita.record_branch_outcome(branch, termination, error)

The first callback uses existing findings, so initial batch validation is not repeated. audita.assess may make model calls after the turn's admission lease has been released. Only Audita decides that a correction is accepted, withdrawn, or unavailable, and retains those meanings in its own state. Further corrections can start separate coordinators from the same initial exchange; accepted sibling corrections and other branches remain with Audita.

Both sketches omit application-specific error handling, including checks for failed structural validation before schema-specific decoding, and depict normal turn progression. The final permitted response follows the same callback path as earlier responses, as accepted in Decision 4. That decision also defines zero-turn budgets, cancellation around callback delivery, and a minimal termination result. The sketches do not imply an archive of responses in that result: applications already receive each completed response through the callback. Completed work must not be mistaken for application acceptance of its content. The concrete interface should let both examples use the same coordination mechanism without manual message-history assembly or per-turn lifecycle bookkeeping. Calls made inside callbacks retain independent observations; applications explicitly group them as selected in Decision 5.

4. How Are Application Next Actions And Execution Outcomes Represented?

Status: accepted; exact public types remain for implementation planning. The coordinator uses content-neutral next actions and distinguishes application-directed completion from exhaustion.

Accepted normal flow: deliver each completed response once through the ordinary application callback, including the last permitted response. If the callback returns Stop, report application-directed completion. If it requests another turn and none remain, report exhaustion without executing that request. Using the entire budget does not itself imply exhaustion; exhaustion means the application requested additional work that the budget would not permit.

This gives consumers one response-processing path, including any application validation, without a separate path for the final response. Internal transport retries and structural repairs do not produce additional conversation-response callbacks. Cancellation and operational errors can interrupt normal delivery as specified below. Stop and other API spellings remain illustrative until public types are selected.

Alternative considered, not selected: stop immediately when the turn budget is used and expose the final response only in the report. This avoids another callback but requires separate consumer processing of the final response.

Accepted zero-turn behavior: with an active context, invoke the initial callback once even when the additional-turn budget is zero. The callback receives the starting exchange and knows that no additional turns remain; this is not delivery of a newly generated response. It can use existing application state without repeating earlier assessment.

• Stop produces application-directed completion.
• Requesting another turn produces exhaustion without executing the request.
• Returning an error produces an operational failure, subject to the cancellation checks below.

Zero bounds coordinator-driven continuation executions, not work inside the application callback. The callback can still consult data or make model calls. Applications wanting to disable all such work can skip invoking the coordinator. The alternative of treating zero as an immediate no-op with no callback is not selected: it would introduce special behavior and would not distinguish an already-complete interaction from one requesting more work.

Accepted turn-budget consumption: at most N additional continuation executions may be started. The first operational error ends coordination. Starting a continuation consumes one turn even if preparation or generation subsequently fails. Callback invocations do not consume turns, and internal transport retries and structural repairs remain within the same turn. The starting exchange is outside the additional-turn budget. Requesting a turn when none remain produces exhaustion without starting an execution.

For example, with a three-turn budget, a preparation failure in the second continuation returns that error and ends coordination. It does not start a third continuation or report exhaustion. The budget does not measure provider calls, work performed inside application callbacks, or cost.

Counting only completed continuations was considered and not selected. Because operational errors already stop coordination, that alternative would primarily change bookkeeping. Use the straightforward executions-started interpretation without introducing public counters, budget refunds, or recovery machinery for this purpose.

Accepted cancellation behavior: check the caller's context immediately before callback invocation, including the initial callback. If cancellation is already observed, do not invoke the callback; return the context error with its identity preserved. If the coordinator already obtained a completed response that has not been delivered to the callback, preserve that single response in the optional result described below. This does not imply application processing or acceptance, or require recovering a response that a failed underlying execution did not return.

Cancellation can race with callback invocation. The callback receives the same context and is responsible for cooperating with cancellation during its work. PromptKit checks the context again after the callback returns and before starting another execution. It does not forcibly terminate callback code. Thus ordinary response delivery is once per completed turn during normal progression, while cancellation can interrupt that progression. Preserving available completed work on interruption does not create a separate final-response processing path during normal budget exhaustion.

The alternative of invoking the callback after cancellation with a canceled context is not selected. It could help callbacks that only record results, but callbacks may perform substantial application work or model calls; PromptKit cannot assume they are merely bookkeeping.

Accepted operational-failure behavior: each continuation is an ordinary execution supplied with captured conversation history. Applications receive completed responses through the callback before another continuation begins. An execution or callback error stops coordination and propagates to the caller, preserving useful Go error identities. PromptKit does not automatically replay the failed continuation or callback. Existing transport retries and structural repair remain within ordinary execution. Earlier callback deliveries and application work remain intact; applications decide whether the failure affects other branches or the overall workflow.

Accepted minimal return contract: provide only the following information:

• An ordinary Go error for execution failures, callback failures, and cancellation, preserving errors.Is/errors.As behavior.
• A small outcome value distinguishing application-directed stopping from budget exhaustion on normal termination. These outcomes are not errors.
• An optional completed response when cancellation prevented its delivery to the callback. Reuse existing response types where practical.

Delivery means the callback was invoked with the response; it does not imply the application finished processing or accepted it. If that callback returns an error, the response is not classified as undelivered or automatically delivered again. The coordinator returns no archive of previously delivered responses. A small optional result accommodates the undelivered-response case; exact Go declarations remain for implementation planning.

Failure-stage fields, additional run-ID references, a separate failure classification hierarchy, and aggregate reporting are not part of this minimal termination contract. They require a demonstrated consumer need before addition. This does not remove fields from existing response values or observation APIs. Decision 5 reuses ordinary run accounting without expanding this result. Decision 6 exposes captured-exchange details without expanding the coordinator's termination result.

Audita retains its own meanings for acceptance, rejection, withdrawal, and assessment unavailability. A normal stop does not assert any of those outcomes.

Alternative considered, not selected for normal termination: use callback errors to signal ordinary termination as well as operational failure. This can reduce the number of result types, but makes normal control flow harder to distinguish from failure. Normal stopping uses an explicit action, while failures use ordinary Go errors. A broader structured termination report was also considered and not selected: its additional metadata is unnecessary for the demonstrated minimal contract.

Implementation planning: select exact types for next actions, outcomes, and the optional result under these accepted behaviors. As selected in Decision 1, a completed run with failed structural validation produces a reusable exchange when capture is requested and retains its ordinary validation result. That result follows the normal response callback path, subject to cancellation; the operational-error behavior does not reclassify it.

5. How Are Related Executions And Usage Accounted For?

Status: accepted; exact placement of relationship metadata remains for implementation planning. Reuse ordinary per-run accounting, expose a minimal parent relationship for continuations, and let applications select accounting membership explicitly. The separable execution-group addition below supplies optional summaries without expanding the coordinator. This applies to both standalone continuations and coordinated turns.

The current observation boundary keeps nested invocations independent. A callback's arbitrary model calls cannot simply be assumed to belong to the enclosing interaction's totals.

Accepted per-run accounting: each continuation reports its own usage and ordinary execution observations, including the existing distinctions between transport attempts, structural repairs, and available failure-path diagnostics. Keep existing usage presence/completeness semantics. The coordinator adds no second accounting system or cumulative totals to its termination result.

An initial exchange A that produces branches B and C contributes its usage only to A. B and C do not inherit that usage. Applications reporting the combined work can count A, B, and C once each. Usage reported for sending captured history again in B or C belongs to those new executions and still counts. Reusing content does not eliminate newly reported usage. Existing observation totals must not be added to the attempt events they already summarize.

Accepted relationship metadata: retain the originating run identity with each captured exchange and expose the immediate parent relationship in continuation metadata. Use existing run identifiers without introducing mandatory conversation, branch, or sequence identifiers. This supplies information that PromptKit knows directly without requiring consumers to reconstruct which exchange each continuation extends. Its exact public placement remains for implementation planning and need not enlarge the coordinator's termination result.

Accepted application boundary: model calls made inside a callback remain independently observed and accounted for. PromptKit does not infer that they belong to a continuation or aggregate merely because they execute inside its callback. Applications explicitly select which runs belong together, including concurrent calls, and may use the optional execution groups below to collect and summarize those facts. This revises the earlier choice to leave all aggregation to applications; membership and domain interpretation remain application-owned. Audita retains assembly of its overall usage report and mappings to its domain identifiers. PromptKit supplies execution facts without owning an application-wide reporting model or promising complete cost accounting.

Alternatives considered, not selected:

• Automatically group nested calls. This could simplify obtaining an overall total, but requires rules about which calls belong together and how observation context propagates, and could include unrelated work.
• Add no relationship metadata. This would keep the API smaller, but observation consumers would have to reconstruct continuation relationships themselves.

Implementation planning: place the minimal parent reference and preserve ordinary per-request observation behavior for every continuation. Execution groups address the separately demonstrated need for explicit accounting across related calls; broader reporting remains outside this scope. The additional-turn budget continues to bound continuation executions, not all provider calls performed inside those executions or application callbacks.

6. Where Should Continuation Inspection Live?

Status: accepted; exact method and type shape remain for implementation planning. Provide explicit access to the captured exchange's existing content and minimal identity metadata, following the general pattern of PreparedExecution.Details().

Accepted inspection scope: expose the logical message history, including the final assistant response, the exchange's originating run identity, and the immediate parent relationship selected in Decision 5. Existing execution results and preparation details continue supplying their ordinary information. Reuse existing types where useful rather than creating a second comprehensive execution report.

Details should be generically renderable through ordinary encoders so consumers can display them without field-by-field adapters. Reading or modifying a returned details value must not alter retained history or another branch's execution. Access reads captured values without rerendering templates, reloading inputs, or making provider calls. Captured message content remains separate from ordinary metadata-only execution observations.

The exchange contains logical conversation history because continuation needs it. This does not create a separate archive of execution results or expand the coordinator's minimal termination result. InspectPrompt, InspectProfile, and InspectBackend retain their configured-resource responsibilities.

Representation alternatives for implementation planning: a small details accessor on a retained exchange is the preferred pattern. A public data value may also satisfy the accepted behavior with fewer API elements, provided its ownership rules prevent caller mutations from affecting retained branch state. Choose the exact representation and serialization details during API planning.

Alternative considered, not selected: an engine-level conversation lookup and inspection API. Such an API could serve a searchable conversation registry, but this feature does not select that registry. Explicit access through an already-held exchange satisfies the present need. Richer reporting or registry capabilities require separate consumer justification.

Separable Addition: Optional Execution Groups

Status: accepted direction; the four decisions below remain open. Audita's follow-up request identifies a distinct need: a validator call can belong to the same logical operation as a continuation without extending its conversation. Conversation ancestry describes which exchange supplies history; accounting membership describes which executions the application chooses to report together. Neither relationship implies the other.

An application creates a group for a logical operation and explicitly associates selected ordinary runs or coordinated continuations with it. Calls made by a validator or other callback join only when the application chooses. Groups use existing run identities and observation/usage semantics to summarize available execution facts. They are usable without capture or a coordinator, and continuation and coordination remain usable without them. Selecting a group does not change execution scheduling, capacity, cancellation, retry, or validation behavior.

For example, an application may group a continuation and a separate validator run while keeping their conversations independent. It may also include the initial run shared by several branches. Replaying that run's captured history does not enroll it or count its usage again; newly reported usage for sending history in later executions still belongs to those later runs.

The group does not infer content meaning, automatically capture arbitrary nested calls, schedule work, determine application outcomes, or introduce a reporting framework. Applications retain domain identifiers, presentation, and any broader report assembly. Group summaries are obtained separately from the coordinator's minimal termination result. They preserve unavailable versus explicit zero usage and do not add terminal totals to the attempts they already summarize.

Group Decision 1: How Does An Application Select Membership?

Recommendation: first evaluate a group collector compatible with the existing observer API. Attaching it explicitly to a request would enroll the observed run; a coordinator could receive an explicit selection for its own continuation runs. Callback calls would still require their own selection. This reuses the existing source of execution facts and avoids a second accounting path.

The design must allow the application to keep its existing observer alongside the group. Specify ordinary and prepared execution attachment and coordinator propagation without inheriting membership into arbitrary nested calls. Preparation alone is not an executed member; discarded handles and failed claims do not currently emit execution observations.

Alternative: explicit group options on requests and the coordinator, backed by ordinary observation internally. This makes membership clearer and separates it from application event delivery, but adds public API plumbing. Choose after sketching the smallest workable interfaces for ordinary, prepared, coordinated, and callback-initiated calls. Do not select a context-based implicit scope.

Group Decision 2: What Is The Group's Lifetime?

Recommendation: expose summary snapshots from a caller-held group, with no mandatory close or wait operation. Applications already own the work and decide when to read or discard the group. Support concurrent selected executions and summary reads without imposing serial scheduling.

A snapshot must distinguish observed terminal runs from members whose terminal facts are not yet available. A missing terminal event can reflect ongoing work or interruption such as a panic; it is not evidence of zero usage or success. Do not promise terminal observations beyond the existing observation contract.

Alternative: an explicit seal/close or wait lifecycle could provide a fixed membership boundary or completion wait. It adds lifecycle rules and overlaps with application orchestration; defer unless a consumer demonstrates that need.

Group Decision 3: May Groups Overlap?

Recommendation: allow a run to belong to more than one explicitly selected group, counting each run identity once within each group. This accommodates both an operation summary and a broader application summary without requiring a hierarchical group model. Overlapping group totals are not necessarily additive.

Alternative: require one group per run. This simplifies membership plumbing, but restricts useful accounting views and makes applications rebuild the broader aggregation themselves. There is no demonstrated need for that restriction. Do not add group nesting, automatic parent membership, or group scheduling merely to support overlap.

Group Decision 4: Which Facts Does A Summary Expose And Aggregate?

Recommendation: start with member run identities, terminal-observation status, and a small selection of existing execution counters and presence-aware usage totals. Select the exact fields during API design; avoid a comprehensive new report schema or new outcome hierarchy. Summaries describe execution facts, not application acceptance or monetary cost. Summed run durations, if included, must not be presented as elapsed group time when runs overlap.

Prefer aggregating one terminal event per run. Terminal observations already summarize retries, structural repairs, reported usage, and failure-path facts, including executions that return no ordinary result. This is simpler than reconstructing those totals from attempt events. Preserve each usage field's known/completeness distinctions and expose missing terminal coverage separately, so a total over finished members cannot imply complete accounting for all members.

Alternative: accumulate attempt events for live partial totals, reconciling terminal events without counting the same work twice. This exposes more facts before completion or after a missing terminal event, but needs additional reconciliation rules. Prefer terminal-based totals unless live attempt accounting has demonstrated value. Whichever representation is selected, absent evidence must not become a known zero, and terminal summaries must never be added to their already-counted attempts.

Proposed Delivery And Validation

The six continuation/coordinator behavioral decisions are accepted. Their remaining work is API and staged implementation planning: exact types and signatures, capture opt-in syntax, message assembly details, placement of relationship metadata, and the concrete inspection representation. Execution groups have four open decisions above and a separate delivery track; resolving or implementing them is not a prerequisite for continuation. The recommended continuation delivery sequence is:

1. Deliver capture and independent continuation through the public facade, allowing Audita to remove its reconstruction and prefix-verification work.
2. Build the optional sequential coordinator on that capability, reusing ordinary execution, validation, capacity, and observation machinery. Refine both consumer sketches against the selected public contracts to verify that the same mechanism removes useful repeated work in each application.

Plan execution groups independently after resolving their open decisions, reusing ordinary observations and preserving existing application observers. They can ship separately without changing the coordinator termination contract.

Use Audita's acceptance scenarios as consumer checks, alongside analysis and branching-conversation examples that exercise the same APIs without repair-specific concepts. Additional proposed coverage should establish:

• The selected conversation history when the initial run needed structural repair, including use of its full final assistant content.
• Capture of the final returned response after structural validation fails, preserving its ordinary validation status through the coordinator callback. Operational failures with no completed result create no new exchange.
• Capture opt-in through ordinary and prepared execution, using the executed snapshot and producing no completed exchange during preparation alone.
• Continuation through another engine instance in the same process, retaining prior content while applying the receiving engine's follow-up configuration.
• Continuation after the original input files change or disappear, without altering the captured history.
• Independent branches retaining their own history and outcomes.
• Assessment calls against the same backend at capacity one, without capacity retained by the coordinator blocking them.
• Per-run usage and failure diagnostics available through ordinary observations, including incomplete usage. Consumer examples explicitly select accounting membership and count shared initial work once, without coordinator-provided totals; optional groups may supply those summaries when available.
• Immediate parent relationships distinguish successive turns and independent branches using existing run identities.
• Explicit exchange details expose captured logical history and minimal identity metadata through ordinary encoding. Mutating returned details does not affect retained branch state, and inspection requires no rendering or external reads.

Execution-group validation should separately cover explicit ordinary, prepared, coordinator, and callback membership; unrelated nested calls remaining outside the group; independent ancestry and membership; concurrent collection alongside an existing application observer; failure-path terminal facts; unavailable and explicit-zero usage; missing terminal observations; and no attempt/terminal double counting. If overlapping groups are selected, verify once-per-run counting within each group without implying that overlapping totals can be added.

These are planning targets, not new implemented guarantees. Keep additional mechanisms and constraints tied to the selected decisions and demonstrated consumer needs.