[B1ZZ211]

Summary

This PR wires the [enrichment] extra (Jedi call-resolution) into the build pipeline and fixes the co-change benchmark mode that was always returning predicted_files = 0.

Changes

Core wiring (code_review_graph/incremental.py)

• Added _run_jedi_enrichment(store, repo_root) — best-effort wrapper matching the resolver pattern
• Called in full_build (always) and incremental_update (when Python files changed)
• Stats surfaced as result["jedi_enrichment"] (e.g. {"resolved": 5, "files": 1})
• Opt-out via CRG_DISABLE_ENRICHMENT=1 (keeps watch mode cheap on large repos)

Bug fix in code_review_graph/jedi_resolver.py

• Target paths now rebase onto the repo root as given to full_build (not force-resolved)
• Fixes /var vs /private/var symlink mismatches where enriched edges never matched stored node qualified names

Benchmark fix (code_review_graph/eval/benchmarks/impact_accuracy.py)

• Normalizes changed-file paths to the stored absolute form before lookups
• Co-change mode now emits honest non-zero predictions (was 0 on every commit)
• Graph-derived ground truth now includes real one-hop neighbors (recall drops from pinned 1.0 to measured 0.667 where neighbors exist)
• _graph_neighbor_files now counts File-node-sourced edges (IMPORTS_FROM, module-level calls) as neighbor files

Tests (tests/test_incremental.py, tests/test_eval.py)

• test_full_build_wires_jedi_enrichment — verifies wiring + stats surface
• test_full_build_jedi_enrichment_skips_when_disabled — opt-out works
• test_incremental_update_wires_jedi_enrichment — runs only on Python changes
• Mock regression test asserts predicted_files > 0 and true_positives >= 1 for co-change

Verification

• Full test suite: 2400 passed (1 pre-existing Windows test failure, unrelated)
• Build output: Jedi enrichment: resolved 5 calls in 1 files
• Co-change mode on real repos now reports honest numbers (previously always 0)
• All changes follow existing resolver patterns and conventions

Notes

• Requires [enrichment] extra (jedi>=0.19.2) — optional, skipped gracefully
• The published eval numbers in README are now stale; a fresh 6-repo capture would be needed to re-quote them

[tirth8205]

The wiring is close: the three new tests fail on main and pass here, the suite is green, and the /var vs /private/var rebase in jedi_resolver.py is correct. Two things need fixing.

incremental.py line 1360 runs the whole-repo pass on any Python change. enrich_jedi_calls re-reads and re-parses every .py file in the graph and retries goto on every unresolved call, so cost scales with repo size, not change size. A one-file update on this repo goes from about 0s to 1.1s with jedi installed; a few thousand Python files means tens of seconds per watch-mode save. Scope it to the changed files or keep it full-build only.

README.md line 263 claims the published predicted_files = 0 rows came from documentation seeds. The 2026-08-02 CSVs show code seeds at 0: code_review_graph/cli.py twice, gin context.go, tree.go, render/data.go. Those zeros were the path bug. Also mark the recall 1.000 table at line 261 as pre-fix, since the new ground truth cannot reproduce it.

The benchmark fix itself looks right.

[tirth8205]

Changes required: enabling Jedi enrichment makes a one-file Python update read and reparse every indexed Python file. With the [enrichment] extra installed, build a multi-file Python graph, edit one file and run code-review-graph update; enrich_jedi_calls still enumerates store.get_all_files() and reads and parses every Python file before selecting pending method calls. Limit incremental enrichment to changed files before claiming a coding-session efficiency improvement.

[tirth8205]

This no longer merges into staging. Conflicts in README.md, tests/test_incremental.py.

git fetch origin && git merge origin/staging
# resolve, then
uv run pytest tests/ -q
uv run ruff check code_review_graph/
uv run mypy code_review_graph/ --ignore-missing-imports --no-strict-optional
git push

When resolving:

• Rebase feat/jedi-enrichment-wiring onto current origin/staging and resolve the two conflicts before any gate or edge-case assessment can be made.

A large integration branch landed on staging today, which is why this drifted.

PRs now target staging, not main. Yours was retargeted already, so nothing to do there.

[tirth8205]

This no longer merges into staging. 3 files conflict and the branch is 313 commits behind:

• README.md
• code_review_graph/incremental.py
• tests/test_incremental.py

Worth knowing before you resolve:

• code_review_graph/incremental.py: staging added a build-state checkpoint at the tail of full_build and moved the Git timeout into constants.py.

git fetch origin
git merge origin/staging
# resolve, then
uv run pytest tests/ -q
uv run ruff check code_review_graph/
uv run mypy code_review_graph/ --ignore-missing-imports --no-strict-optional
git push

I have not reviewed the change itself yet. That comes once it merges and the checks run against the merged state, since staging has moved a long way and the result is what matters.

[tirth8205]

Thank you for this PR. I tested it on a small repo and on a merge with staging.

Before I can merge it, please fix these:

1. Each incremental update reads every Python file again. See code_review_graph/incremental.py:1360. enrich_jedi_calls loops over store.get_all_files(). Example: a repo has 122 Python files. You edit one file and run update. Jedi then reads all 122 files. Please scan only the changed files.
2. The new text at README.md:263 and README.md:283 is wrong. It says doc seeds caused the zeros. But code seeds were zero too, like code_review_graph/cli.py and gin context.go. See evaluate/results/*_impact_accuracy_2026-08-02.csv. The real cause was the path bug you fixed. Please also say the recall table is from before this fix.
3. Co-change rows now save a full local path in seed_file. See impact_accuracy.py:186. Example: /tmp/.../mock_repo/helper.py instead of helper.py. Please keep the path relative to the repo root.

Please update your branch with the latest staging. There are conflicts in README.md, code_review_graph/incremental.py and tests/test_incremental.py. On staging, _refresh_target_resolution(store) must run last. Please call the Jedi step before it.

After you push the fix, I will test it again.

[B1ZZ211]

Thanks for the review. All three are fixed and the branch is updated with staging:

1. Incremental updates now pass only the re-parsed Python files to Jedi; full builds still scan all.
2. README now attributes the zeros to the path bug (code seeds were zero too) and notes the recall table predates the fix.
3. seed_file is now repo-relative.

The Jedi step runs before _refresh_target_resolution, which stays last.

[Copilot]

🟡 Changes recommended

Jedi enrichment still skips unresolved calls and leaves dependent callers’ edges stale, while non-canonical benchmark roots cause lookup misses and empty co-change predictions.

3 open issues

• Canonicalize repo_path before deriving the benchmark root · New
• Refine unresolved parser edges before skipping Jedi resolution · New
• Refresh Jedi edges in affected Python dependents · New

What changed in this PR

This PR integrates optional Jedi call resolution into graph builds and corrects path handling in the impact-accuracy benchmark.

Changes:

• Adds best-effort Jedi enrichment, an environment opt-out, build statistics, and incremental scan scoping.
• Updates benchmark path normalization and neighbor detection while preserving relative seed paths in results.
• Adds regression tests and documents that published benchmark figures need a fresh capture.

File	Description
tests/​test_incremental.py	Tests Jedi wiring, opt-out behavior, statistics, and scan scope.
tests/​test_eval.py	Checks non-zero co-change predictions and relative seed reporting.
README.md	Explains the benchmark path bug and qualifies historical results.
code_review_graph/​jedi_resolver.py	Adds file-scoped scans and rebases resolved target paths.
code_review_graph/​incremental.py	Invokes Jedi enrichment during full and incremental builds.
code_review_graph/​eval/​benchmarks/​impact_accuracy.py	Normalizes lookup paths and includes File-node sources in neighbors.

_{Review effort: Balanced}