[codex] Improve TraceDecay follow-up diagnostics

[ScriptedAlchemy]

Summary

• add doctor reporting for registry/store manifest identity drift
• persist richer MCP analytics for transport/tool kind, failures, and context memory matches without fact bodies
• record TypeScript ignored-dependency candidates and surface MCP search hints for symbols behind ignored dependencies
• dedupe repeated pending session fact proposals and add project/session hint coverage
• warn when Cursor transcript rows contain literal ${workspaceFolder} placeholders that would break session recall
• make daemon status distinguish missing, stale, inaccessible, unreachable, and connectable sockets
• keep selected-project graph-reader search hints read-only when dispatching through registered project selectors

Verification

• cargo test --lib doctor::tests::registry_drift_findings_report_manifest_identity_mismatches -- --nocapture
• cargo test --test typescript_extraction_test test_ts_import_type_records_ignored_dependency_candidates -- --nocapture
• cargo test --test mcp_handler_test test_search_hints_at_ignored_dependency_candidates -- --nocapture
• cargo test --test automation_session_reflector_runner_test session_fact_proposals_dedupe_repeated_pending_facts_across_runs -- --nocapture
• cargo test --features test-transport --test mcp_server_test context_call_writes_memory_match_analytics_without_fact_bodies -- --nocapture
• cargo test --features test-transport --test mcp_server_test failed_tool_call_writes_mcp_runtime_analytics_event -- --nocapture
• cargo test --lib tool_hints -- --nocapture
• cargo test --test hooks_test test_build_codex_session_context -- --nocapture
• cargo test --test agent_test test_cursor_healthcheck_warns_on_literal_workspace_folder_transcript_path -- --nocapture
• cargo test --lib daemon::tests::service_status -- --nocapture
• cargo test --lib mcp::tools::handlers::tests::graph_reader_selector -- --nocapture
• cargo test --lib mcp::tools::handlers::tests::selected_project_retrieve_finds_selected_project_response_handle -- --nocapture
• cargo fmt --check
• cargo check --workspace
• tracedecay sync

[changeset-bot]

⚠️ No Changeset found

Latest commit: 2dfca29

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

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3d85f450e0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Derive memory analytics before rendering

For normal tracedecay_context calls, the output format defaults to markdown, so response.content[0].text will not parse as JSON; payload becomes None and the code below records match_count=0/fact_ids=[] even when context_memory_matches returned facts and rendered them. This corrupts the new analytics for default calls (and truncated JSON envelopes have the same problem); build the metadata from the structured context value instead of the rendered MCP response.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Avoid scanning unresolved refs for every search

This runs even when cg.search already returned matching symbols, and ignored_dependency_hint calls get_unresolved_refs(), which selects the entire unresolved_refs table. On large projects with one row per unresolved import/call, every tracedecay_search now pays an O(all refs) read just to maybe add an ignored-dependency hint; gate it to empty/low-confidence results or query only npm: candidates via the indexed reference_name.

Useful? React with 👍 / 👎.