feat(cli,report): add graphify rebuild + surface pre-#1504 collision banner

[JunoLee1]

Closes #4200.

요약

pre-#1504 node-ID scheme (same-name file/method collision) 전수 재추출을 명시적 subcommand 로 노출. graphify rebuild <path> 는 manifest 백업/삭제 후 전체 재추출 트리거. semantic cache 는 보존 → docs/papers/images replay, code 만 re-AST. GRAPH_REPORT.md 상단에 collision 배너 자동 추가.

변경

• graphify/cli.py — 신규 rebuild <path> subcommand. manifest.json 을 .manifest-backup-<ts>.json 으로 백업 후 삭제, 기존 _rebuild_code 호출. cache/ 디렉토리는 untouched 라 semantic 재추출 비용 0.
• graphify/__main__.py — help text 에 rebuild 등록.
• graphify/report.py — _label_collision_percent() helper + 상단 ⚠️ 배너. duplicate label with different source_file 가 structural node 의 0.5% 이상일 때 활성. concept node (empty source_file) 와 label 없는 node 는 제외.

테스트

• tests/test_rebuild_cmd.py (신규, 2 tests): manifest 백업 + _rebuild_code delegation, no-manifest 환경에서도 실행
• tests/test_report_collision_banner.py (신규, 6 tests): collision percent helper (unique/colliding/concept/missing-field), banner 출력/미출력 양방향
• 기존 test_cluster.py/test_analyze.py 89/89 pass + 2 skipped

Motivation (English)

Full re-extract today requires users to delete manifest.json by hand — an unobvious operation hidden behind a one-line note that only appears on query/path/explain output, never on GRAPH_REPORT.md. In practice full rebuild is cheap: the semantic cache is content+prompt hashed, so docs/papers/images replay from disk; only code re-ASTs (seconds, no tokens). The gap is purely in surface area.

This PR:

1. Gives users a one-liner (graphify rebuild) that does the right thing (back up → delete manifest → re-extract), with no risk of losing the semantic cache.
2. Puts the collision warning in GRAPH_REPORT.md so readers who never run query still see it, and the remediation is in the same paragraph.

Backward compatibility

• Existing update subcommand unchanged.
• No schema/output format changes except the optional banner block at the top of GRAPH_REPORT.md — only renders when a collision is actually detected.
• No new dependencies.

Honesty-rule compliance: the banner is a diagnostic, not an edit. Nodes/edges in graph.json are untouched; the fix (if the user runs it) rebuilds with path-qualified IDs through the normal pipeline.

🤖 Generated with Claude Code

[github-actions]

Thanks for the pull request, @JunoLee1. A maintainer will review it soon.

Want to talk it through while it is in review? Come join us on our Discord server. For longer-form discussion there is also GitHub Discussions.

A couple of things that speed up review: make sure the test suite passes on Python 3.10 and 3.13, and that the change keeps extraction deterministic.

[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. PR-changed functions: 1/3 verified (0 proven, 1 may-equivalent, 0 distinguished) · 2 not verified (2 vacuous).

Not verified on this run: dispatch\_command (vacuous: never exercised), generate (vacuous: never exercised).

---

Graphify review — findings

Adds a graphify rebuild [path] command that backs up manifest.json to a timestamped .manifest-backup-*.json and re-runs the code rebuild. Every file re-extracts, but the semantic cache replays from disk and only code is re-ASTed; with no manifest it builds from scratch, and with no path it uses the saved .graphify_root. GRAPH_REPORT.md now opens with a warning banner pointing at graphify rebuild when _label_collision_percent finds at least 0.5% of nodes sharing a label with a node from a different source_file; concept nodes and unlabeled nodes are ignored.

Worth a look

• collision detector flags already path-qualified distinct nodes — graphify/report.py:64 · 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 — 579 functions depend on the 114 functions this change touches.

Health — this change adds coupling hotspots:

• new: _rebuild_code() — 158 callers, 56 callees
• new: dispatch_command() — 5 callers, 129 callees
• new: generate() — 39 callers, 9 callees
• new: main() — 102 callers, 3 callees
• new: run_pipeline() — 8 callers, 13 callees
• new: _refresh_stale_skills() — 25 callers, 4 callees
• new: _run_cli() — 6 callers, 7 callees
• new: _stale_graph_sources() — 7 callers, 6 callees
• …and 7 more — each is listed as a finding

Verification — 579 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: 572 function(s) in the blast radius were not formally verified this run

Test selection

Test selection

36 of 346 test file(s) selected (10%) via static blast radius.

• tests/test_affected_cli.py — impact
• tests/test_agents_platform.py — impact
• tests/test_codebuddy.py — impact
• tests/test_confidence.py — impact
• tests/test_dedup_shrink_refuses_force_write.py — impact
• tests/test_devin.py — impact
• tests/test_explain_cli.py — impact
• tests/test_extract_cli.py — impact
• tests/test_global_add_tag_inference.py — impact
• tests/test_god_nodes_cli.py — impact
• tests/test_hollow_chunks_arm_shrink_guard.py — impact
• tests/test_hook_guard_token_match.py — impact
• tests/test_hook_out_of_project_paths.py — impact
• tests/test_hook_strict.py — impact
• tests/test_hypergraph.py — impact
• tests/test_incomplete_build_guard.py — impact
• tests/test_install.py — impact
• tests/test_install_references.py — impact
• tests/test_install_version_warning.py — impact
• tests/test_merge_chunks_validation.py — impact
• tests/test_multigraph_diagnostics.py — impact
• tests/test_no_dedup_flag.py — impact
• tests/test_partial_cache.py — impact
• tests/test_path_cli.py — impact
• tests/test_pipeline.py — impact
• tests/test_query_cli.py — impact
• tests/test_query_induced_edges.py — impact
• tests/test_rebuild_cmd.py — impact, changed-test
• tests/test_report.py — impact
• tests/test_report_collision_banner.py — impact, changed-test
• tests/test_report_gap_thresholds.py — impact
• tests/test_semantic_similarity.py — impact
• tests/test_skill_auto_refresh.py — impact
• tests/test_skill_version_warning.py — impact
• tests/test_stale_prune.py — impact
• tests/test_unverified_semantic_shrink.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

No difference found (not proven): No behavior difference found in \_run\_cli (not a proof).

The verifier ran both versions of \_run\_cli on many inputs and saw identical behavior every time. Strong evidence the change is safe, but evidence, not a proof.

Guarantee: Empirical: differential testing (both versions run on many generated inputs). A divergence on an untested input remains possible, so this is 'no counterexample found', not 'proven equivalent'.

Note: An input the sampler did not try could still differ.

Could not verify: Could not verify dispatch\_command.

The verifier did not have enough to check dispatch\_command, 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: no capturable inputs from the test suite; property tier: not verifiable: all 40 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly SystemExit — names the real obstacle, not a sampling gap)

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: no capturable inputs from the test suite; property tier: not verifiable: all 264 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)

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