feat(health): report production or all code, and name a cause you can fix

[RaghavChamadiya]

Builds on #2153.

The health surfaces reported one number for the whole repository and called it "Defect risk". There was no way to ask what production code alone scores, a decline could not say which half of the score moved, and fix_first could name a file whose entire deficit is git history — an instruction with nothing to act on.

What changed

• scope=all|production on the health overview, files, map, trend and work-queue routes, on get_health, and on repowise health. The default stays all: tests score higher than production code, so narrowing lowers every figure without a defect having been found, and a caller should ask for that knowingly.
• The headline reads "Code health". Higher is better, and nothing prints an inverted copy of it.
• Trend alerts carry driver, structure_delta and history_delta, so a decline names the half that moved. Metric rows expose both halves and the unclamped score, which is the only number that moves for a file at the floor.
• fix_first names a cause an edit can remove, picking the highest-leverage file that has one. History-derived leads move to a watch field that says plainly there is nothing there to fix.
• non_code_files and average_health_code_only are gone; the population they described is no longer scored.

Behaviour

A narrowed response drops what it cannot honestly narrow rather than relabelling it. Snapshots record the production average, so a narrowed trend is weighted the same way as the headline; the hotspot figure, the worst performer and the per-file movement list have no narrowed copy and are omitted. Here, scope=all reads 7.17 over 3,776 files and scope=production reads 6.25 over 2,343.

Verification

Trend and KPI averages agree on a single response under both scopes, which they previously did not. The generated HTTP contract is regenerated. tests/unit/server/mcp/test_health.py covers the scope filter, the distinction between a file dropped by scope and one dropped by the exclude config, and the directive naming an actionable cause.

[repowise-bot]

🔍 2 things to check

1. get_health in .../mcp_server/tool_health.py signature changed. Review 46 callers outside this PR: tests/integration/test_mcp.py::test_mcp_health_accuracy, tests/integration/test_mcp.py::test_mcp_health_churn_complexity, tests/integration/test_mcp.py::test_mcp_health_default_is_lean (+43 more)
2. Run .../__tests__/health.test.ts, .../health/health-file-drawer.test.tsx, .../src/hosted.fixtures.test.ts (+19 more) first: they import the changed files

⚠️ Health of changed files: 4.8 → 4.7 (-0.1)
_{🚨 Change risk: high, riskier than 71% of this repo's commits.}

✅ Health gate: passed

---

📊 See the full report for this PR
_{Your repo map with this PR's blast radius lit up, every caller of the contracts it changes, and health before and after. No sign-in. · ⭐ Star Repowise · 📥 Install bot · Silence on a single PR with [skip repowise] in the title · Per-repo toggle on repowise.dev/settings?tab=bot · Updated 2026-09-07 13:51 UTC (since the last push: health 4.8 to 4.7)}