feat(miner): AMS min-rank calibration loop over the local event ledger

[JSONbored]

Summary

Closes #8184. Closes #8185. Closes #8186. Closes #8187. (The full #8172 arc as one coherent change, per the fewer-larger-PRs direction.)

Transposes the ORB backtest discipline onto the miner's own ledgers, reusing the engine primitives untouched:

• Corpus + advisory backtest (ams-calibration: backtest a proposed min-rank skip threshold against the node's own selection history #8184): buildAmsRankCorpus/runAmsMinRankBacktest in the engine (fixed-seed split, the symmetric compareBacktestScores Pareto floor per the issue's required pattern; a CLOSED take labels reversed — skipping it would have been right — a MERGED take confirmed, rank score replayed as classifier confidence). pr_outcome rows now carry the claimed issueNumber so the corpus can join discovery-time rank records to realized outcomes; pre-pairing rows simply never join. CLI: loopover-miner calibration backtest-threshold --candidate <x> prints the shared renderer's report for both splits and persists an ams_threshold_backtest_run ledger event; exit code never reflects the verdict.
• Track record (ams-calibration: aggregate backtest track record into the miner calibration report + snapshot event #8185): the calibration report aggregates every persisted run's comparisons via computeRegressedVerdictTrackRecord (the same engine aggregation ORB uses, zero new math), with an explicit no-runs line; the Phase-7 snapshot payload gains a tolerant optional backtestTrackRecord section.
• Proposals surface (ams-calibration: surface backtest-cleared knob proposals in miner doctor/status output (report-only) #8186): report + doctor list current backtest-cleared proposals (latest run per candidate inside the 90-day lookback, visible improved + held-out non-regressed) with full evidence and the explicit nothing-applies-automatically line.
• Double-gated self-adjustment (ams-calibration: bounded, explicitly-approved self-adjustment of the min-rank threshold (mirrors #8121 discipline) #8187): .loopover-ams.yml gains minRankAutotuneEnabled (default OFF, tolerant parser) — gate one; calibration apply-min-rank --candidate <x> --approve is gate two, refuses without a recent cleared run for that exact candidate, validates the hard bounds (0, 0.5] on every read AND write, and records evidence-carrying apply/revert ledger events. calibration revert-min-rank --approve is the one-command revert. Consumption happens at exactly one point (discover's enqueue minRankScore), fail-open to shipped, and the dry-run preview reflects the same threshold. Docs note added to the miner calibration page.

Validation

• Full local gate npm run test:ci: every check green except one unrelated suite (test/unit/mcp-cli-telemetry.test.ts) that hit the known subprocess-spawn 15s timeout while two full gates ran in parallel on this machine — it passes 6/6 standalone in this worktree, and no MCP file is touched here. npm audit --audit-level=moderate clean.
• 100% line+branch on every changed range across engine + all eight touched miner lib files (verified via lcov against the exact diff hunks, with the dual .js/.ts include-glob discipline vitest.config documents).
• Rebased onto current main post-feat(review): capture per-reviewer AI votes with swap-proof attribution #8265/feat(calibration): unify the tightening loop under the backtest evidence standard #8267/fix(scripts): don't reject miner package source files merely named with "secret" #8266; post-rebase typecheck clean and the new suites pass.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	loopover-ui	66f414d	Commit Preview URL Branch Preview URL	Jul 23 2026, 09:46 PM

[codecov]

Codecov Report

❌ Patch coverage is 92.00000% with 18 lines in your changes missing coverage. Please review.
✅ Project coverage is 92.16%. Comparing base (590befd) to head (66f414d).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
packages/loopover-miner/lib/ams-calibration.ts	88.07%	0 Missing and 13 partials ⚠️
packages/loopover-miner/lib/calibration-cli.ts	92.42%	0 Missing and 5 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8270      +/-   ##
==========================================
- Coverage   92.16%   92.16%   -0.01%     
==========================================
  Files         786      788       +2     
  Lines       78941    79162     +221     
  Branches    23840    23926      +86     
==========================================
+ Hits        72757    72960     +203     
  Misses       5062     5062              
- Partials     1122     1140      +18     

Flag	Coverage Δ
shard-1	54.83% <79.55%> (-1.88%)	⬇️
shard-2	51.84% <37.33%> (+1.66%)	⬆️
shard-3	55.59% <7.11%> (+0.41%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
packages/loopover-engine/src/ams-policy-spec.ts	100.00% <100.00%> (ø)
...loopover-engine/src/calibration/ams-rank-corpus.ts	100.00% <100.00%> (ø)
packages/loopover-miner/lib/calibration-run.ts	100.00% <100.00%> (ø)
packages/loopover-miner/lib/cli.ts	87.50% <ø> (ø)
packages/loopover-miner/lib/discover-cli.ts	99.53% <100.00%> (+0.01%)	⬆️
packages/loopover-miner/lib/loop-cli.ts	100.00% <ø> (ø)
packages/loopover-miner/lib/pr-outcome.ts	100.00% <100.00%> (ø)
packages/loopover-miner/lib/status.ts	100.00% <100.00%> (ø)
packages/loopover-miner/lib/calibration-cli.ts	95.53% <92.42%> (-4.47%)	⬇️
packages/loopover-miner/lib/ams-calibration.ts	88.07% <88.07%> (ø)