refactor(health): resolve current auto-sprint hotspot

[aimasteracc]

Summary

• extract class-diagram assembly from UMLExporter.class_diagram into the existing focused builder module
• preserve the public UML API while removing one major long-function finding
• push both caller and callee literal-prefix preconditions into the architectural-constraint SQL query
• keep full regex, exception, import-reachability, and deduplication checks unchanged

Closes #1224

Why

The daily backlog's first two snapshots were stale (uml_state.py and cache/extraction.py now grade A). The next live candidate, uml_export.py, remained grade C with major long-function findings. During the required quick gate, the quarantined 500 ms constraint evaluator budget failed twice under xdist contention. The callee-prefix SQL pushdown removes impossible rows before the Python hot loop without weakening the budget or changing matching semantics.

Evidence

• uml_export.py: health 73.6 → 77.8; refactor suggestions 4 → 3
• constraints/evaluator.py: health 78.8/C → 82.9/B
• focused UML pre/post gate: 324 passed
• focused coverage gate: 89 passed, 1 skipped
• patch coverage: no added executable misses
• default quick gate: 1316 passed, 1 skipped in 33.31s
• staged change-impact: REVIEW; focused-then-default verification completed successfully

Safety

• no public API signatures changed
• no pytest budget or runtime defaults weakened
• no benchmark cells or model callbacks executed
• unrelated untracked documentation and generated lockfile drift excluded

[github-actions]

🔍 Dogfood PR Check

Layer	Status
Change Impact	✅ passed
Claim Invariant Suite	✅ passed

Change impact: change_impact changed=3 risk=low pytest_required=True
Verification: uv run pytest -q
Claim summary: ====================== 17 passed, 12 deselected in 1.48s =======================

Claim invariant test output (last 30 lines)

============================= test session starts ==============================
collected 29 items / 12 deselected / 17 selected

tests/benchmarks/claims/test_bm25_ranking_invariant.py ....              [ 23%]
tests/benchmarks/claims/test_reactive_push_e2e.py .......                [ 64%]
tests/benchmarks/claims/test_safety_layers_smoke.py ......               [100%]

- generated xml file: /home/runner/work/tree-sitter-analyzer/tree-sitter-analyzer/dogfood-claims.xml -
====================== 17 passed, 12 deselected in 1.48s =======================

What these tests guard:

• 390× cross-language mis-wire claim
• BM25 ranked search with relevance_score
• 5 safety layers (safe/guard/constraints/impact/verdicts)
• Reactive push subscribe/unsubscribe (RFC-0001)
• Index build speed regression

Full results: workflow run

---

Dogfood PR Check — docs/superpowers/specs/2026-07-08-svcd-dogfood-system-design.md

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 80dc8f7263

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!