[Copilot]

Summary

• simplify(lcm): replace JSON-in-route summary metadata with a typed input #4296 fully met — typed provided-summary route/extraction input; parser/envelope removed; persisted metadata pinned by provided_summary_metadata_rows_are_byte_stable.
• simplify(lcm): remove the parallel Fake summarizer path from production #4297 fully met — production Fake mode removed; fixtures use validated Provided input; rejection covered by retired_fake_mode_is_rejected_at_deserialization.
• simplify(lcm): carry typed compression outcomes instead of string-and-Option states #4298 fully met — borrowed closed compression outcomes, guarded deferral, malformed status rejection, and durable pressure-recovery coverage.
• simplify(session-runtime): delete the test-only LcmDaemonStore abstraction and absent-store state #4300 fully met — test-only store trait/optional absent state removed; authority owns the registered lease and calls inherent compact/status/doctor methods.
• simplify(lcm): define compression overrides once instead of copying eleven fields between requests #4305 fully met — one flattened LcmCompressionOverrides owner preserves the eleven existing flat request keys and null/duplicate semantics.

The cutover preserves public response fields, token-budget algorithms, route normalization, source/frontier transaction behavior, and metadata bytes.

Verification

• Updated from master before validation. The daemon authority's compression response import and direct typed deferral return are restored.
• The shared integration fixture accepts both borrowed and owned summary text. Metadata expectations keep the pinned route/extraction bytes and pass them through the canonical privacy authority before comparing the entire persisted string, including all receipt bindings. The fixture also asserts its exact source count and expansion hint.
• Native Bazel session integration target: all 72 LCM compression tests passed (349 unrelated cases filtered), covering summary/frontier publication, source identity, overflow, replay, tool transactions, rollback, and exact protected metadata bytes.
• Strict Bazel Clippy passed on the full session integration target and its dependencies with -D warnings, clearing the 54 reported fixture type errors.
• The earlier session-runtime run passed all 14 selected authority and unavailable-summary tests, including grant denial, cancellation, deadlines, pressure compaction, override preservation, and pending-frontier behavior.
• Pinned rustfmt, diff checks, and commitlint passed.
• The fleet on 7b96709c451e3b69ff10dd754269809686f2e9ad passed all 421 session integration tests (90.68s) and 167 of 169 native targets. Its remaining failures were the host-handoff fixture's stale generation baseline and the shared heavy-refresh timeout tracked in fix(memory): size and charge text builds in one admission #4393.
• After updating from master, the host-handoff fixture waits for the decoded graph admission work to settle before pinning its initial generation. All original generation, freshness, broker-debt, and handoff assertions remain. Both focused handoff tests passed locally (38.59s); strict Bazel Clippy passed on the daemon unit-test target and dependencies.
• On 6c70575ac6a627ea8f864202bb802b1cc0f02d1d, Clippy, the shipped Linux CLI build, repository gates, Dashboard, and feature gates passed; the Linux fleet was still running when the required-check notice arrived. Updated the branch from master (44d8265c13a5feb70e7ea834c1d11ed413f3f2af) to 1ed4a0ab365d2a48bae2d232f1237b9bb3018951. Full hosted CI, including the Linux fleet and aggregate, must finish on this new head before merge.

Fixes #4296
Fixes #4297
Fixes #4298
Fixes #4300
Fixes #4305

[changeset-bot]

⚠️ No Changeset found

Latest commit: 21cb484

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[ScriptedAlchemy]

Cross-referenced #4296, #4297, #4298, #4300 and #4305. The typed input/policy migration has one concrete outcome-preservation regression: the existing forced-overflow producer emits a successful response that the new take_outcome rejects. This needs correction before merging; the precise producer and retained transaction path are described inline. This finding is from source/caller tracing, not a completed native test run.

[ScriptedAlchemy]

Pushed e5aae67 to address the pressure-recovery regression found in review. take_outcome() now accepts completed ok and best_effort responses carrying critical_pressure_catch_up, while malformed authoritative-summary states still fail. Added full wire-response round-trip coverage and a registered retained-pressure recovery regression that reopens the store and verifies summaries were durably committed for both statuses.

Pinned CI rustfmt and git diff --check pass. Native tests have not run locally because this environment exhausted memory on the repository's native dependency build; CI is the execution authority. Marked this PR ready for review because draft PRs explicitly skip the heavy native lane. Relevant coverage is in the LCM unit tests and session-runtime lcm_effects tests. This is an implementation update, not a merge or approval.

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.

[ScriptedAlchemy]

Pushed a3d1c10 for the remaining #4298 outcome-conversion hole. Unknown decoded statuses previously became Complete when no request/retry marker was present; the conversion now accepts only the supported ok, best_effort, and needs_summary combinations. The critical-pressure completion forms remain supported. Added invalid-status and complete-wire round-trip regressions. Pinned rustfmt, whitespace and commitlint pass; native behavior is now with CI.