Skip to content

Centralize language support and generalize agent packet retrieval#27

Merged
TheGreenCedar merged 51 commits into
mainfrom
autoresearch-review/codex-restore-ast-first-retrieval/01-current-final-tree
Jun 14, 2026
Merged

Centralize language support and generalize agent packet retrieval#27
TheGreenCedar merged 51 commits into
mainfrom
autoresearch-review/codex-restore-ast-first-retrieval/01-current-final-tree

Conversation

@TheGreenCedar

@TheGreenCedar TheGreenCedar commented Jun 12, 2026

Copy link
Copy Markdown
Owner

Summary

This PR aligns language-support claims across the codebase and removes benchmark-specific shortcuts from production retrieval.

  • Shared language registrycodestory-contracts now owns support tiers; the indexer, runtime, CLI, and docs consume that registry instead of maintaining separate claim maps.
  • Parser-backed expansion — Kotlin, Swift, Dart, and Bash move to fidelity-gated graph coverage with tree-sitter rules and fixture tests. Go, Ruby, PHP, and C# fidelity suites expand around symbols, imports, member ownership, and receiver/call resolution.
  • Generalized packet retrieval — production packet search no longer steers on benchmark-family terms. scripts/lint-retrieval-generalization.mjs guards against benchmark literals in retrieval and indexing paths.
  • Sidecar diagnostics — unresolved packet candidates are surfaced to callers and can block packet sufficiency when sidecar evidence exists but cannot be resolved.
  • CLI files counts — output separates whole-index, filtered, visible, and truncated totals so filtered views do not look like repository-wide counts.

Receiver/call resolution and agent packet quality remain bounded claims; see docs/architecture/language-support.md for the supported boundary.

Test plan

  • cargo fmt --check and cargo check --all-targets
  • node scripts/lint-retrieval-generalization.mjs
  • cargo test -p codestory-indexer --test fidelity_regression
  • cargo test -p codestory-indexer --test tictactoe_language_coverage
  • cargo test -p codestory-runtime packet_sufficiency -- --nocapture
  • cargo test -p codestory-cli --test cli_golden_path -- --nocapture
  • cargo test -p codestory-cli --test onboarding_contracts -- --nocapture
  • cargo build --release -p codestory-cli
  • Ignored repo-scale e2e stats (CODESTORY_ALLOW_SKIP_REAL_REPO_DRILL_CASES=1)

Repo-scale indexing stats are logged in docs/testing/codestory-e2e-stats-log.md (full-sidecar proof tier, 68.25s index, 238 files, 89,726 nodes).

Notes

  • Real-repo drill cases were skipped locally; treat the e2e stats row as indexing evidence, not drill proof.
  • HTML, CSS, and SQL remain structural collectors, not semantic code-navigation languages.

Packages the current branch tree as the review unit, not older kept commits, when commit-level kept evidence is stale or incomplete.
@TheGreenCedar TheGreenCedar changed the title Restore AST-first retrieval and strict CodeStory A/B harness Restore AST-first retrieval with evidence-gated agent benchmarks Jun 13, 2026
@TheGreenCedar TheGreenCedar changed the title Restore AST-first retrieval with evidence-gated agent benchmarks Make language support and retrieval benchmarks evidence-backed Jun 13, 2026
@TheGreenCedar TheGreenCedar changed the title Make language support and retrieval benchmarks evidence-backed Remove benchmark steering and expose retrieval evidence gaps Jun 13, 2026
@TheGreenCedar TheGreenCedar changed the title Remove benchmark steering and expose retrieval evidence gaps Tighten language support and retrieval evidence contracts Jun 13, 2026
@TheGreenCedar TheGreenCedar changed the title Tighten language support and retrieval evidence contracts Centralize language support and generalize agent packet retrieval Jun 14, 2026
Co-authored-by: Cursor <cursoragent@cursor.com>
@TheGreenCedar TheGreenCedar marked this pull request as ready for review June 14, 2026 18:30
@TheGreenCedar TheGreenCedar merged commit e8bf691 into main Jun 14, 2026
1 check passed
@TheGreenCedar TheGreenCedar deleted the autoresearch-review/codex-restore-ast-first-retrieval/01-current-final-tree branch June 14, 2026 18:39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant