fix(cli): apply --no-stats to keep-marker stats line (#1706) by oddFEELING · Pull Request #1765 · abhigyanpatwari/GitNexus

oddFEELING · 2026-05-21T19:57:37Z

What

Closes #1706.

gitnexus analyze --no-stats is documented as "Omit volatile file/symbol counts from AGENTS.md and CLAUDE.md", but it did not strip the counts from the index-summary line when the block carries a  marker.

The regular (no keep marker) path already respects --no-stats via generateGitNexusContent (#1477). The keep-marker branch of upsertGitNexusSection, however, rebuilt the stats line itself and always re-injected (N symbols, N relationships, N execution flows) — noStats was never passed to that function. For teams that commit a trimmed AGENTS.md / CLAUDE.md with a keep marker, every analyze rewrote that line with different numbers, producing recurring no-value merge conflicts between branches.

How

Thread noStats into upsertGitNexusSection.
In the keep-marker branch, under --no-stats the stats line becomes Indexed as **<name>** — no parenthetical counts.
The project name still refreshes, so a repo rename still propagates.
statsPattern's parenthetical is now optional, so a count-free line left by a prior --no-stats run still matches — the name keeps refreshing, and counts return if --no-stats is later dropped.
generateAIContextFiles passes options?.noStats into both upsertGitNexusSection calls.

Tests

Updated the #1508 review-F3 test: under noStats + keep marker it now asserts the volatile counts are stripped (previously it expected them rewritten), while keeping its Always-Do-tuple corruption guards.
Added a test confirming the project name still refreshes under --no-stats + keep when the existing line has no parenthetical to match.

Full test/unit suite is green; prettier --check and eslint clean on the touched files.

Risk / rollback

Low. upsertGitNexusSection is a private function with only the two in-file callers updated here. No public CLI/MCP contract changes — the --no-stats help text already promised this behavior. Revert is a single-commit revert.

…#1706) The keep-marker branch of upsertGitNexusSection rebuilt the index-summary line on every analyze and always re-injected the volatile counts, ignoring --no-stats. For teams that commit a trimmed AGENTS.md/CLAUDE.md with a gitnexus:keep marker, that produced recurring no-value merge conflicts — exactly what --no-stats exists to prevent. Thread noStats into upsertGitNexusSection. Under --no-stats the keep-path stats line becomes "Indexed as **<name>**" with no (N symbols, ...) parenthetical; the project name still refreshes so renames propagate. The statsPattern parenthetical is now optional so a count-free line left by a prior --no-stats run still matches.

vercel · 2026-05-21T19:57:40Z

Someone is attempting to deploy a commit to the NexusCore Team on Vercel.

A member of the Team first needs to authorize it.

github-actions · 2026-05-21T21:10:09Z

Claude finished @magyargergo's task in 5m 14s —— View job

PR #1765 Production-Readiness Review

Tasks

Review bar for this PR

--no-stats must consistently suppress volatile stats counts in every relevant AI-context generation path, including keep-marker updates, without corrupting custom sections, losing suffix prose, freezing repo renames, reintroducing prior keep-marker bugs, or creating misleading CLI file-status output.

Problem being solved

gitnexus analyze --no-stats already stripped counts from freshly-generated AGENTS.md/CLAUDE.md via generateGitNexusContent (non-keep path). The keep-marker path in upsertGitNexusSection was not receiving noStats — it always rebuilt the stats line with live counts, producing count churn on every analyze for teams that commit a lean keep-section. This PR wires noStats into upsertGitNexusSection and makes the parenthetical optional in the matching regex so count-free lines remain updatable.

Current PR state

Open. Branch: oddFEELING:fix/no-stats-keep-section-1706 → main.
2 commits: d1b22a8 (fix), cb6abfd (merge-from-main).
Changed files: gitnexus/src/cli/ai-context.ts (+28/-6), gitnexus/test/unit/ai-context.test.ts (+44/-8).
CI: 3/6 workflows completed. Gitleaks ✅, Dependency Review ✅, PR Autofix ✅. CodeQL, Docker Build, and CI are still in progress at review time. The unit-test-and-typecheck pass (CI workflow) is not yet confirmed.

Merge status and mergeability

CI is in progress. Cannot confirm merge-readiness until the CI workflow completes. Vercel authorization is preview-deployment-only and is not a merge gate. No unresolved review threads visible.

Repository history considered

--no-stats flag has no effect (commander.js negation collides with options.noStats reads) #1477 / fix(cli): make --no-stats actually omit volatile counts (#1477) #1478 / fix(cli): translate Commander --no-stats into noStats option (#704) #1059: The Commander stats: false → noStats: true translation pre-exists in analyze.ts:926 and :1036. This PR does not touch that bridge. analyze-no-stats-bridge.test.ts guards it.
feat: gitnexus:keep marker preserves custom context sections (resubmit of #605) #1508 F1–F5: Guards verified present. See Lane E below.
Option to disable the amount of indexed files in CLAUDE.md and AGENTS.md #704: Original --no-stats feature request. CLI contract unchanged.
Bug: gitnexus analyze corrupts docs when marker comments appear inline in text #1041: findSectionMarkerIndex inline-prose-marker protection. Unchanged.

Branch hygiene assessment

The branch contains a merge-from-main commit (cb6abfd Merge branch 'main' into fix/no-stats-keep-section-1706). The merge introduces zero code changes itself; it only brings in main's history. The fix commit (d1b22a8) touches exactly two files, both in the CLI+test domain. No release churn, no lockfile, no unrelated domains. Structurally this is clean aside from the merge commit, which repo maintainers may require to be rebased to a linear history.

Understanding of the change

ai-context.ts:

upsertGitNexusSection gains a noStats?: boolean parameter (line 204).
In the keep-marker branch, statsLine is now conditional:
- noStats=true → Indexed as **${projectName}** (parenthetical dropped)
- noStats=false/undefined → Indexed as **${projectName}** (${newStatsInner}) (unchanged behavior)
statsPattern changes from /^(?:Indexed as|indexed by GitNexus as) \*\*[^*]+\*\* $[^)]+$/m to /^(?:Indexed as|indexed by GitNexus as) \*\*[^*]+\*\*(?: $[^)]+$)?/m — the parenthetical group becomes optional.
Both upsertGitNexusSection call sites in generateAIContextFiles now pass options?.noStats.

ai-context.test.ts:

Existing noStats + keep test (#1508 review F3) retitled to #1706 and updated: previously expected 42 symbols (was the wrong behavior, not actually testing the bug); now asserts no count parenthetical.
New test: noStats + keep marker: project name still refreshes when counts are stripped — seeds a count-free line, runs with noStats: true and a new name, asserts name updated, counts absent, suffix preserved.

Findings

F1 — No explicit test for counts returning after `--no-stats` is dropped

Risk: A prior --no-stats run leaves Indexed as **Name** (count-free). A later run without --no-stats should restore counts. If not, the flag becomes sticky. This is the "count-free existing line + noStats: false" scenario.

Evidence: Code analysis confirms the path is correct: the optional (?: $[^)]+$)? in statsPattern matches a count-free line (the optional group matches nothing), statsLine includes the parenthetical when noStats is falsy, and .replace(statsPattern, statsLine) appends the counts. The suffix (. Custom.) is preserved because the pattern has no $ anchor. No test exercises this path explicitly.

From the test suite: the rename test (line 674) seeds a count-free line with noStats: true — this confirms the optional group matches correctly. But no test seeds a count-free line and runs with noStats: false to assert counts return.

Recommended fix: Add one test: seed Indexed as **Name**. Custom., run generateAIContextFiles with noStats unset (or false), assert Indexed as **Name** (N symbols, ...). Custom. appears. Fix this →

Blocks merge: maybe. Code is correct; test coverage gap is the only issue.

F2 — New #1706 noStats+keep tests cover only CLAUDE.md, not AGENTS.md

Risk: Both call sites in generateAIContextFiles were updated symmetrically. A future asymmetry wouldn't be caught if tests only assert CLAUDE.md.

Evidence: The two new tests (noStats + keep marker: stats line drops the volatile counts at line 629 and noStats + keep marker: project name still refreshes at line 674) read only CLAUDE.md. The pre-existing AGENTS.md keep-path test at line 532 tests the normal (with-stats) update path for AGENTS.md, confirming the function works for both files. But no test explicitly asserts noStats=true + keep-marker on AGENTS.md.

Both call sites are clearly identical:

// AGENTS.md call site (line 402–408):
await upsertGitNexusSection(agentsPath, content, projectName, stats, options?.noStats);
// CLAUDE.md call site (line 413–419):
await upsertGitNexusSection(claudePath, content, projectName, stats, options?.noStats);

The function is the same; the only difference is the file path. This makes the risk low.

Recommended fix: Extend one new test to also assert AGENTS.md (add agentsPath write/read alongside claudePath in the same tempdir). Fix this →

Blocks merge: maybe. Code is correct; test parity is the gap.

F3 — Merge-from-main commit in branch

Risk: cb6abfd is a merge-from-main commit. The diff of the fix commit is clean (+72/-14, two files, one domain). No code change is in the merge commit itself.

Evidence: git log --oneline HEAD~5..HEAD shows cb6abfd Merge branch 'main' into fix/no-stats-keep-section-1706 as the HEAD commit, ahead of the actual fix commit. The PR was updated with a rebase-to-main merge to resolve any conflicts with recent main commits.

Recommended fix: Rebase onto current main before merge (git rebase origin/main) to linearize history, if repo policy requires it. This is mechanical.

Blocks merge: depends on repo policy. Not a behavioral regression; purely hygiene.

F4 — CI `CI` workflow not yet confirmed green

Risk: The unit test suite, typecheck, and lint results are unknown at review time.

Evidence: CI summary shows in_progress for the main CI workflow (run #26253272756).

Recommended fix: Wait for CI to complete before merging.

Blocks merge: pending confirmation.

PR-specific assessment sections

Lane A — CLI option propagation: ✅ Confirmed correct. analyze.ts:926 and :1036 both translate options.stats === false to noStats: true. This bridge is untouched and guarded by analyze-no-stats-bridge.test.ts. Both generateAIContextFiles call sites in the keep path now receive options?.noStats. run-analyze.ts:956 passes options.noStats directly. The entire CLI chain is wired end-to-end.

Lane B — keep-marker stats-line replacement: ✅ Confirmed correct. statsLine construction is clean: conditional on noStats, using caller-provided projectName and stats (no fragile re-extraction). Suffix preservation is intact: no $ anchor means . Custom. and similar text after the matched span is untouched. The optional parenthetical (?: $[^)]+$)? makes a count-free line matchable without over-broadening: it requires the full (text) form (needs closing )) and is anchored to line-start. Manual trace confirms counts return correctly when noStats is dropped. Pattern cannot match the verbose-template "This project is indexed by GitNexus as" form — this is a pre-existing limitation predating this PR, not introduced here.

Lane C — generated-template no-stats behavior: ✅ Unchanged. generateGitNexusContent at line 134 correctly handles noStats with its existing ternary. This PR does not touch it.

Lane D — AGENTS.md / CLAUDE.md parity: ✅ Both call sites updated symmetrically. Minor test parity gap noted (F2).

Lane E — repository history regressions:

F1 (honest return value when no stats line matches): guarded at test line 704. ✅
F2 (structured projectName/stats, no fragile extraction): code uses caller-provided values. ✅
F3 (no Always-Do tuple corruption): explicitly guarded with /\(\{target:/ and /direction:\s*"upstream"/ assertions at line 660–661. ✅
F4 (regex scope): ^ + m flag prevents mid-paragraph matches. No $ preserves suffix. ✅
F5 (test coverage breadth): keep-scope, AGENTS parity, idempotency, CRLF, markdown-sensitive names, noStats+keep, preserved return semantics all covered. Count-return and AGENTS.md parity for noStats have minor gaps (F1, F2).

Lane F — CI / bidi / hygiene:

git diff --check on the fix commit: clean (no whitespace errors).
file reports both changed files as UTF-8 text with no encoding anomalies.
The GitHub bidi warning visible on the diff page is a rendering artifact for Unicode text files in general; no actual bidi control characters were found in the changed content.
CI: 3/6 complete (Gitleaks ✅, Dependency Review ✅, PR Autofix ✅). Main CI workflow still in progress.

Back-and-forth avoided by verifying

"This project is indexed by GitNexus as" not matched by statsPattern: Confirmed pre-existing; not introduced by this PR. Keep-marker sections use Indexed as **Name** form by convention (verified in all existing tests).
The Optional parenthetical causes zero-length match concerns: Confirmed safe — the optional group requires the full (...) balanced form; empty match is only the outer zero-repetition, which is correct behavior.
Merge-from-main commit: Confirmed it introduces no code changes; it is purely a history update.
Vercel authorization comment: Confirmed preview-deployment-only; not a merge gate.

Open questions

None unavoidable. The two test gaps (F1, F2) are clearly non-blocking given the code correctness is confirmed by trace. CI completion (F4) is the only hard gate.

Final verdict

production-ready with minor follow-ups

The core fix is correct end-to-end: noStats is threaded through all call sites, the optional regex correctly allows name refresh and count return on count-free lines, suffix text is preserved, #1508 regression guards are intact, and the CLI translation bridge is untouched. Two minor test-coverage gaps exist — no test for counts returning after --no-stats is dropped (F1), and the new noStats+keep regression test does not assert AGENTS.md parity (F2) — but the code is demonstrably correct for both cases. The merge-from-main commit should be rebased to linearize history per repo convention. CI must complete before merge.

| CI workflow

github-actions · 2026-05-21T21:27:26Z

CI Report

✅ All checks passed

Pipeline Status

Stage	Status	Details
✅ Typecheck	`success`	tsc --noEmit
✅ Tests	`success`	unit tests, 3 platforms
✅ E2E	`success`	gitnexus-web changes only

Test Results

Tests	Passed	Failed	Skipped	Duration
9461	9460	0	1	482s

✅ All 9460 tests passed

1 test(s) skipped — expand for details

buildTypeEnv > known limitations (documented skip tests) > Ruby block parameter: users.each { |user| } — closure param inference, different feature

Code Coverage

Tests

Metric	Coverage	Covered	Base	Delta	Status
Statements	78.21%	31230/39927	N/A%	—	🟢 ███████████████░░░░░
Branches	66.76%	19900/29804	N/A%	—	🟢 █████████████░░░░░░░
Functions	82.24%	3183/3870	N/A%	—	🟢 ████████████████░░░░
Lines	81.58%	28167/34523	N/A%	—	🟢 ████████████████░░░░

_{📋 View full run · Generated by CI}

magyargergo · 2026-05-21T21:29:09Z

@oddFEELING can you please look into claude's actionable findings?

oddFEELING · 2026-05-21T21:31:42Z

@oddFEELING can you please look into claude's actionable findings?

Giving it a look now

…p path Addresses review findings F1 and F2 on PR abhigyanpatwari#1765: - F1: add a test that counts RETURN when --no-stats is dropped after a prior count-free run — guards against the flag becoming sticky. - F2: extend the noStats+keep "drops the volatile counts" test to assert AGENTS.md alongside CLAUDE.md, so a future asymmetry between the two upsertGitNexusSection call sites is caught.

oddFEELING · 2026-05-22T05:38:54Z

Addressed the actionable findings in 3726182a:

F1: Added noStats + keep marker: counts return when --no-stats is dropped after a count-free run — seeds a count-free keep-section line, runs analyze without --no-stats, and asserts the (N symbols, …) parenthetical comes back. Confirms the flag isn't sticky.
F2: Extended the noStats + keep marker: stats line drops the volatile counts test to seed and assert both AGENTS.md and CLAUDE.md (looped, with per-file assertion messages), so a future asymmetry between the two upsertGitNexusSection call sites is caught.

F3 (merge-from-main commit): left as-is for now since the repo squash-merges PRs, so it collapses on merge. Happy to rebase to a linear history if you'd prefer.

F4: CI was already green (9459/9459) on the prior run.

ai-context.test.ts is now 27/27 locally; typecheck, prettier, and eslint all clean.

magyargergo · 2026-05-22T06:04:03Z

Thank you for your PR!

Merge branch 'main' into fix/no-stats-keep-section-1706

cb6abfd

Merge branch 'main' into fix/no-stats-keep-section-1706

07d2ead

magyargergo merged commit 954b184 into abhigyanpatwari:main May 22, 2026
32 of 33 checks passed

magyargergo mentioned this pull request May 27, 2026

feat(review): add PR reviewer swarm agents #1851

Merged

7 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix(cli): apply --no-stats to keep-marker stats line (#1706)#1765

fix(cli): apply --no-stats to keep-marker stats line (#1706)#1765
magyargergo merged 4 commits into
abhigyanpatwari:mainfrom
oddFEELING:fix/no-stats-keep-section-1706

oddFEELING commented May 21, 2026 •

edited

Loading

Uh oh!

vercel Bot commented May 21, 2026

Uh oh!

github-actions Bot commented May 21, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented May 21, 2026 •

edited

Loading

Uh oh!

magyargergo commented May 21, 2026

Uh oh!

oddFEELING commented May 21, 2026

Uh oh!

oddFEELING commented May 22, 2026 •

edited

Loading

Uh oh!

Uh oh!

magyargergo commented May 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

oddFEELING commented May 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

How

Tests

Risk / rollback

Uh oh!

vercel Bot commented May 21, 2026

Uh oh!

github-actions Bot commented May 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR #1765 Production-Readiness Review

Review bar for this PR

Problem being solved

Current PR state

Merge status and mergeability

Repository history considered

Branch hygiene assessment

Understanding of the change

Findings

F1 — No explicit test for counts returning after --no-stats is dropped

F2 — New #1706 noStats+keep tests cover only CLAUDE.md, not AGENTS.md

F3 — Merge-from-main commit in branch

F4 — CI CI workflow not yet confirmed green

PR-specific assessment sections

Back-and-forth avoided by verifying

Open questions

Final verdict

Uh oh!

github-actions Bot commented May 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CI Report

Pipeline Status

Test Results

Code Coverage

Tests

Uh oh!

magyargergo commented May 21, 2026

Uh oh!

oddFEELING commented May 21, 2026

Uh oh!

oddFEELING commented May 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

magyargergo commented May 22, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

oddFEELING commented May 21, 2026 •

edited

Loading

github-actions Bot commented May 21, 2026 •

edited

Loading

F1 — No explicit test for counts returning after `--no-stats` is dropped

F4 — CI `CI` workflow not yet confirmed green

github-actions Bot commented May 21, 2026 •

edited

Loading

oddFEELING commented May 22, 2026 •

edited

Loading