Surface unclassified files instead of a silent successful run (#3511)

[ayushcodes10]

Summary

Fixes #3511. detect() already computes and returns unclassified — every file it saw but could not classify (no supported extension or shebang) — but nothing surfaced it. On a corpus in a language graphify has no extractor for (the report's example: a 260-file Lean 4 repo), graphify update . exits 0, prints a normal-looking "Rebuilt: N nodes..." summary, and GRAPH_REPORT.md's Corpus Check says "corpus is large enough that graph structure adds value" — describing the 4 shell scripts and 2 READMEs that happened to be classifiable, with the other 260 files invisible and unmentioned.

graphify extract already had this reporting, added by #1692 — the gap is specifically the update/watch rebuild path (graphify/watch.py), which never got it, plus GRAPH_REPORT.md's Corpus Check section, which had no field for it at all regardless of which command wrote the graph.

• graphify/report.py: adds an Unclassified: N file(s) not represented in the graph (top: .ext A, .ext B) line to the Corpus Check section whenever detection_result carries any.
• graphify/watch.py: _rebuild_code now prints the same "N file(s) not classified..." line extract already prints, and threads the count into the local detection summary it builds for GRAPH_REPORT.md (which previously only carried total_files/total_words, dropping unclassified even though detect() had already computed it).

Scoped to just the reporting gap, matching the issue's own "Suggested fix" section. Left graph.json's unclassified_files field (the issue's third suggested surface) out of this PR — threading it through would mean passing detection_result down through build/build_from_json/to_json's call chain, which is a much larger, more invasive change than the two display-only additions here; happy to follow up separately if wanted. Also left the issue's separate "should graphify parse Lean 4" question alone — the reporter raised it explicitly as a discussion point for a maintainer architecture call, not as part of this fix.

Test plan

• tests/test_report.py: new tests confirming the Corpus Check section includes the unclassified line (with per-extension breakdown) when present, and omits it entirely when absent (backward compatible).
• tests/test_watch.py: new test confirming _rebuild_code prints the unclassified count and filenames for a mixed .py/.lean corpus.
• Manually reproduced the issue's shape end to end (a .py file plus several .lean files, _rebuild_code called directly): confirmed both the console line and the GRAPH_REPORT.md Corpus Check line appear correctly.
• Full suite: python3 -m pytest -q — 5423 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.

🤖 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

Surfaces files that detect() saw but couldn't classify (no supported extension or shebang) in two places that previously dropped them silently: the corpus report now appends an Unclassified: N file(s) line with the top three extensions when any exist, and the watch/update _rebuild_code path prints a skip notice listing up to six filenames and threads unclassified through the detection dict it builds. When there are no unclassified files, both paths behave exactly as before.

Worth a look

• Unclassified files are hidden when corpus check has a warning — graphify/report.py:146 · 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 — 644 functions depend on the 437 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: make_inputs() — 16 callers, 5 callees
• new: watch() — 5 callers, 7 callees
• new: _reconcile_graph_html() — 6 callers, 5 callees
• new: load_learning_for_report() — 4 callers, 3 callees
• …and 5 more — each is listed as a finding

Verification — 644 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: 474 function(s) in the blast radius were not formally verified this run

Test selection

Test selection

9 of 268 test file(s) selected (3%) via static blast radius.

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

Could not verify: Could not verify \_rebuild\_code.

The verifier did not have enough to check \_rebuild\_code, 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: parameter `watch_path` is annotated `Path` — outside the synthesizable primitive/collection set

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