Rust SQLite Lookup And Retrieval Tag Audit¶

Scope¶

This audit records two follow-up changes on top of the upstream repository-benchmark broadening that landed on origin/master in the same session:

the repository benchmark summaries now expose per-tag rollups and exclude repo-local .codex skill surfaces from the benchmark corpus
the Rust wrapper now exposes a SQLite index / lookup path so agents can do cheap tracked-file discovery before escalating to DSPy-backed synthesis

Executed successfully in this turn:

make hooks-install
uv run python -m compileall src tests
cargo build --manifest-path rust-cli/Cargo.toml
make rust-lookup-index
make rust-lookup QUERY='dspy training'
uv run pytest tests/test_utilities.py tests/test_project_surfaces.py tests/test_cli_and_dspy.py tests/test_repository_rag_bdd.py
uv run repo-rag smoke-test
uv run repo-rag verify-surfaces
uv run repo-rag retrieval-eval --root . --top-k 4 --top-k-sweep 1,2,4,8
uv run pytest tests/test_utilities.py tests/test_repository_rag_bdd.py
make quality

Observed and then corrected in this turn:

make rust-lookup QUERY='dspy training' failed with failed to initialize SQLite schema: database is locked when it was launched in parallel with make rust-lookup-index. The serialized rerun above passed, which is the supported workflow.

cargo build --manifest-path rust-cli/Cargo.toml: passed
make rust-lookup-index: passed and wrote artifacts/sqlite/repo-file-index.sqlite3 with indexed=204, skipped_binary=3, skipped_large=2
make rust-lookup QUERY='dspy training': passed; the top ranked hits were src/repo_rag_lab/notebook_scaffolding.py, README.DSPY.MD, src/repo_rag_lab/training_samples.py, and src/repo_rag_lab/dspy_training.py
uv run pytest tests/test_utilities.py tests/test_project_surfaces.py tests/test_cli_and_dspy.py tests/test_repository_rag_bdd.py: passed, 39 tests
uv run repo-rag smoke-test: passed with answer_contains_repository: true, mcp_candidate_count: 1, and manifest_path: artifacts/azure/repo-rag-smoke.json
uv run repo-rag verify-surfaces: passed with issue_count: 0
uv run repo-rag retrieval-eval --root . --top-k 4 --top-k-sweep 1,2,4,8: passed and reported:
benchmark_count: 8
default top_k: 4
pass_rate: 1.0
fully_covered_rate: 1.0
average_source_recall: 1.0
average_source_precision: 0.46875
average_reciprocal_rank: 1.0
best_pass_rate_top_k: 4
non-empty tag_summaries for benchmark tags such as agents, api, azure, docs, notebooks, publication, rag, and utilities
uv run pytest tests/test_utilities.py tests/test_repository_rag_bdd.py: passed, 13 tests
make quality: passed with 120 tests and 88.02% total coverage

Configured and verified in this turn:

Compile checks: present and passed through uv run python -m compileall src tests
Rust wrapper build: present and passed through cargo build --manifest-path rust-cli/Cargo.toml
Rust SQLite index and lookup path: present and passed through make rust-lookup-index and the serialized make rust-lookup QUERY='dspy training'
Utility, project-surface, CLI, and BDD pytest coverage for the changed user-facing surfaces: present and passed through the targeted uv run pytest ... slice above
Repository smoke test: present and passed through uv run repo-rag smoke-test
Repository-surface verification: present and passed through uv run repo-rag verify-surfaces
Retrieval-quality evaluation utility: present and passed through uv run repo-rag retrieval-eval --root . --top-k 4 --top-k-sweep 1,2,4,8
Baseline utility and BDD pytest slice: present and passed through uv run pytest tests/test_utilities.py tests/test_repository_rag_bdd.py
Full repository quality gate: present and passed through make quality

Still absent or not exercised in this turn:

UI or browser tests: none found in repository configuration
Full notebook batch execution: notebook lint and surface checks passed through make quality, but make notebook-report was not rerun end-to-end in this turn
Live Azure OpenAI and Azure AI Inference probes: not rerun in this turn
Post-push GitHub Actions evidence: not yet available before the push for this change set

The upstream broadened benchmark corpus from origin/master was kept intact in this turn; the local retrieval change is the reporting layer around it, not a replacement of those 8 training examples.
src/repo_rag_lab/benchmarks.py now excludes .codex from the benchmark corpus so repo-local skill instructions do not inflate retrieval-eval results.
summarize_benchmark_results(...) now includes tag_summaries, which flow through the CLI, utility helpers, and notebook scaffolding without changing the answer-generation path itself.
The Rust lookup path is intentionally lightweight and local-first: it builds an ignored SQLite FTS index over tracked UTF-8 files and prints ranked path/snippet hits before any DSPy step.