fix: reuse CR-settled thread summaries

[rianjs]

Summary

• add CR-settled inline thread summaries keyed by thread_id when GitHub provider resolution is unavailable
• reuse cached settled summaries in PR dossier discussion artifacts and reviewer selection context instead of re-summarizing full threads
• keep provider-resolved counts distinct from planned/failed resolve actions in cr respond output

Validation

• make tidy
• make build
• make lint
• make test
• go test ./internal/threadcontext ./internal/pipeline ./internal/cmd/respondcmd
• go test ./...
• built /tmp/cr-394 and dry-ran PR feat: add reusable thread lifecycle pipeline #359 with codex-rianjs-bot; verified a CR-settled cached summary from thread PRRT_kwDOSrNqjc6MjnWL was reused in dossier summary/final artifacts without leaking codereview:thread-summary marker metadata

Closes #394

[rianjs]

TDD review

STATUS: blockers=0 majors=0 minors=0 nits=0

The test coverage is acceptable to proceed.

Follow-up coverage added after the initial TDD pass:

• thread_id wins over mismatched anchor when decoding dossier discussion summaries.
• raw dossier cached_summary.source distinguishes provider_resolved and cr_settled.
• successful provider resolve counters are asserted in both text and JSON cr respond output.

Validation after coverage updates:

• rtk go test ./internal/pipeline ./internal/cmd/respondcmd ./internal/threadcontext
• rtk go test ./...
• rtk make lint
• rtk make build

[rianjs]

Final architect pass

STATUS: blockers=0 majors=0 minors=0 nits=0

Remaining merge-blocking concerns from the architecture/code pass: none identified. Draft status and the pending cr review loop remain procedural gates.

Focused verification run from the final architect pass:

• rtk go test ./internal/threadcontext ./internal/pipeline ./internal/cmd/respondcmd

[rianjs-bot]

Automated PR Review

Reviewed commit: ba495fa15c20
Profile: codex-rianjs-bot - Posting as: rianjs-bot[bot]

Summary

Reviewer	Findings
go:implementation-tests	1

go:implementation-tests (1 finding)

Major - internal/pipeline/pipeline.go:1024

When ThreadContext is present, this path rebuilds the prompt from normalized thread context but then clears knownThreadIDs before decoding the model response. buildSelectionPrompt still includes thread_id values in the selection input, and llm.DecodeSelection rejects any thread_actions whose IDs are not in KnownThreads, so a valid model response that selects a cached CR-settled thread will now fail as llm: unknown thread .... Keep the allowlist populated from req.ThreadContext (or derive it from the prompt threads) and add a focused selection-phase test that exercises a thread_actions response while ThreadContext is used.

Reviewer Coverage

Reviewer	Status	Inspected	Skipped	Constraints
go:implementation-tests	complete_broad	internal/cmd/respondcmd/respondcmd.go, internal/cmd/respondcmd/respondcmd_test.go, internal/pipeline/pipeline.go, internal/pipeline/pipeline_test.go, internal/threadcontext/threadcontext.go, internal/threadcontext/threadcontext_test.go	unavailable	Review limited to static inspection of the assigned Go files and nearby repo guidance in a read-only sandbox; I did not execute the test suite.
unassigned	incomplete_unassigned	unavailable	docs/architecture.md	changed files were not assigned to a selected reviewer

0 PR discussion threads considered. 0 summarized; 0 resolved.

---

Completed in 3m 52s | unavailable | gpt-5.4 | cr dev

Field	Value
Model	gpt-5.4
Reviewers	go:implementation-tests
Engine	codex_cli · gpt-5.4
Reviewed by	cr · rianjs-bot[bot]
Duration	3m 52s wall · 3m 32s compute
Cost	unavailable
Tokens	unavailable

Per-workstream usage

Workstream	Model	In	Out	Cache read	Cache create	Cost	Duration
orchestrator-selection	gpt-5.4	unavailable	unavailable	unavailable	unavailable	unavailable	17s
go:implementation-tests	gpt-5.4	unavailable	unavailable	unavailable	unavailable	unavailable	3m 05s
orchestrator-rollup	gpt-5.4	unavailable	unavailable	unavailable	unavailable	unavailable	9s

[rianjs-bot]

Automated PR review completed with outcome: request_changes.

[rianjs-bot]

Automated PR Review

Reviewed commit: de1425765c45
Profile: codex-rianjs-bot - Posting as: rianjs-bot[bot]

Summary

Reviewer	Findings
go:implementation-tests	0

Reviewer Coverage

Reviewer	Status	Inspected	Skipped	Constraints
go:implementation-tests	complete_constrained	internal/cmd/respondcmd/respondcmd.go, internal/cmd/respondcmd/respondcmd_test.go, internal/pipeline/pipeline.go, internal/pipeline/pipeline_test.go, internal/threadcontext/threadcontext.go, internal/threadcontext/threadcontext_test.go	unavailable	Review limited to the six assigned Go files and nearby repo-local context; docs/development.md was not present in this review scope.
unassigned	incomplete_unassigned	unavailable	docs/architecture.md	changed files were not assigned to a selected reviewer

0 PR discussion threads considered. 0 summarized; 0 resolved.

---

Completed in 3m 38s | unavailable | gpt-5.4 | cr dev

Field	Value
Model	gpt-5.4
Reviewers	go:implementation-tests
Engine	codex_cli · gpt-5.4
Reviewed by	cr · rianjs-bot[bot]
Duration	3m 38s wall · 3m 17s compute
Cost	unavailable
Tokens	unavailable

Per-workstream usage

Workstream	Model	In	Out	Cache read	Cache create	Cost	Duration
orchestrator-selection	gpt-5.4	unavailable	unavailable	unavailable	unavailable	unavailable	13s
go:implementation-tests	gpt-5.4	unavailable	unavailable	unavailable	unavailable	unavailable	2m 55s
orchestrator-rollup	gpt-5.4	unavailable	unavailable	unavailable	unavailable	unavailable	7s

[rianjs-bot]

Automated PR review completed with outcome: comment.