report: count an all-file-node community as thin, not invisible (#3548)

[ayushcodes10]

Summary

Fixes #3548. The issue's own reported repro (shown_count overstating the number of rendered ### Community sections by counting all-file-node communities as shown) turned out to already be fixed on current v8 by #3148/#2129 — shown_count is now computed directly from the render condition rather than total - thin. But that earlier fix left a residual gap this issue's own invariant check still catches: the thin count kept a 0 < guard excluding all-file-node (zero-real-node) communities, so those communities — which the render loop skips for exactly the same reason as a genuinely thin one — went uncounted anywhere in the report. shown + thin no longer summed to the total community count, and the "N thin communities omitted" figure undercounted what was actually omitted.

Reproduced directly: a 3-community graph (one large, one genuinely thin, one all-file-node) reported "(1 shown, 1 thin omitted)" for 3 total communities — 1 + 1 = 2 ≠ 3. Both counting sites (thin_count_summary in the Summary section, thin_communities in the Knowledge Gaps section) drop the 0 < guard, exactly matching the issue's own suggested patch, applied against the current (already-partially-fixed) code shape rather than the older len(communities) - thin_count_summary form the issue was written against.

Test plan

• Three existing tests in tests/test_report_gap_thresholds.py pinned the old, undercounting figures — one explicitly, in its own docstring, asserting the exact shape this PR corrects. Updated all three to the now-correct counts, verified by hand against a direct generate() call before updating the assertions.
• Added a new test asserting the two invariants the issue itself suggested as a regression check (shown == rendered, shown + thin == total) at both thresholds already exercised in this file.
• Verified all 4 updated/new tests fail against the pre-fix code and pass with the fix.
• Full suite: python3 -m pytest -q — 5487 passed, only the pre-existing unrelated failures (test_ollama_retry_cap.py missing openai in this env, one flaky timing assertion in test_ts_import_type_arguments.py).
• python3 -m tools.skillgen --check — OK.

Note: the issue's separate "second, smaller issue" (the Knowledge Gaps section's hardcoded < 3 threshold disagreeing with min_community_size) was also already fixed on current v8 by #3148 — confirmed the code already reads min_community_size there, not a hardcoded 3.

🤖 Generated with Claude Code

https://claude.ai/code/session_017qfdzgbA5KedGEjD1AayNh

[graphify-labs]

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 1 advisory finding(s) below merit a look before merge.

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

Counts a zero-real-node (all-file) community as "thin" in generate, dropping the earlier 0 < guard from both thin_count_summary and thin_communities so a community the render loop skips is no longer omitted from every tally. This restores the shown + thin == len(communities) invariant, at the cost that any all-file community now shows as thin at every positive min_community_size. Tests assert the invariant and matching render count at thresholds 3 and 5.

Worth a look

• Thin/shown counts use a different file-node predicate than the render filter — graphify/report.py:175 · Escalate · medium
  • agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 353 functions depend on the 21 functions this change touches.

Health — this change adds coupling hotspots:

• new: _rebuild_code() — 116 callers, 51 callees
• new: dispatch_command() — 2 callers, 124 callees
• new: generate() — 34 callers, 7 callees
• new: run_pipeline() — 8 callers, 13 callees
• new: watch() — 5 callers, 7 callees
• new: load_learning_for_report() — 4 callers, 3 callees
• new: test_report_shows_avg_confidence_for_inferred() — 0 callers, 7 callees
• new: test_poisoned_manifest_is_healed() — 0 callers, 6 callees
• …and 1 more — each is listed as a finding

Verification — 353 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 178 function(s) in the blast radius were not formally verified this run

Test selection

Test selection

6 of 275 test file(s) selected (2%) via static blast radius.

• tests/test_confidence.py — impact
• tests/test_hypergraph.py — impact
• tests/test_pipeline.py — impact
• tests/test_report.py — impact
• tests/test_report_gap_thresholds.py — impact, changed-test
• tests/test_semantic_similarity.py — impact

Selection is safe under the controlled-regression assumption; always-run tests + a periodic full run are the backstops. Advisory — it never changes the check verdict.

Formal verification

Could not verify: Could not verify generate.

The verifier did not have enough to check generate, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 200 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly ValueError — names the real obstacle, not a sampling gap)

· 9 more finding(s) on lines outside this diff (see the check run).