feat: profile precision enhancement v2 — Beta confidence model with 44 fixes

[rookie136]

Summary

Comprehensive upgrade from linear confidence model to Beta posterior model, plus injection quality improvements, semantic deduplication, Thompson sampling isolation, and validator extensions.

Core Changes

Beta Confidence Model

• syncConfidence(item): confidence = alpha/(alpha+beta) × min(timeFactor, trendMultiplier)
• Alpha/beta accumulation replaces direct confidence manipulation
• lazyMigrateAlpha: detects legacy Thompson prior (0.5, 1.5) and migrates to evidence-based alpha
• Fills missing fields: weakAlpha, weakBeta, lastMatchTime, firstSeen, pendingValidation

Merge Method Constraints

• combineItems → mergeConfirmedMatch: only called by exact/strong/weak-upgrade
• Forced merge and cross-validation use direct field manipulation (avoid F5-9/F5-10 alpha desync)
• combineThree: alpha=a1+a2+1, beta=b1+b2, frequency=f1+f2+1

Thompson Sampling (Isolated)

• Weak hits use isolated weakAlpha/weakBeta
• Cumulative limit > 7 triggers forced judgment (mean ≥ 0.45 upgrades)
• Upgrade: total alpha +1.0 (mergeConfirmedMatch +0.5 + bonus +0.5)

Decay & Staleness

• Removed linear decay and conf < 0.3 hard threshold
• Stale condition: alpha ≤ 2 && age > 30 days
• Removed decayWeakHits, applyConfidenceDecay, capConfidences

Semantic Deduplication

• deduplicateItems: LLM checks same-category pairs with cos ≥ 0.50
• Max 5 LLM calls per cycle, dedupCheckedCache LRU (1000 limit)
• Merged items accumulate alpha/beta/weakAlpha/weakBeta

pendingValidation

• New items start with pendingValidation = true
• Confirmed next cycle: alpha += 1.0 + syncConfidence

Validator (5 verdicts)

• confirmed/contradicted/no_evidence/inaccurate/oversimplified
• Covers preferences (top-5) + patterns (top-3)
• oversimplified triggers evolveAndUpdate when evidence ≥ 3

Injection Quality

• dedupByCategory + scoreByRecency (conf×0.7 + exp(-age/90)×0.3)
• Workflow context anchor in analysis prompt

Bug Fixes

• P3: retry path now calls decayInMemory
• P10: evolve evidence < 3 no longer adopts

Deployment Results (2026-07-03)

Metric	Before	After
conf saturation (≥0.98)	7/22	2/20
conf=1.0	6 items	0
conf range	0.50~0.998	0.42~0.99
Dedup merges	0	2 pairs
Forced merge conf→1	16/cycle	0
Decay false positives	12/cycle	0
Build	—	0 errors

[tickernelz]

Thanks for this substantial PR. I attempted to merge it as part of a batch release, but I'm holding it back because it does not pass the test suite on its own.

Merging #152 cleanly onto the current main (no git conflicts) still fails 3 tests:

• tests/config.test.ts › CONFIG defaults › should have user profile settings as numbers
• tests/profile-write.test.ts › merges a new explicit preference into an existing profile without clobbering other prefs
• tests/profile-write.test.ts › deduplicates when the same explicit preference is written twice, boosting confidence

Example failure:

expect(conf).toBeGreaterThan(0.9)
Expected: > 0.9
Received: 0.7142857142857143

The new Beta-confidence model changes the resulting confidence values (e.g. ~5/7 ≈ 0.714), but tests/profile-write.test.ts in this PR still asserts the previous clamped-near-1.0 behavior, so the PR contradicts itself. Please reconcile the model and the assertions (update the tests to match the intended Beta model, and fix the config.test.ts numeric-defaults expectation) so bun test is green. Once it passes, please also rebase onto the latest main and I'll merge it in the next release. I did not modify your tests myself — that's your design call to make.

[rookie136]

Fixed all 3 test failures:

• config.test.ts: updated userProfileMax* → userProfileDisplay* field names
• profile-write.test.ts: confidence assertion changed from >0.9 to >0.6 (Beta model ~0.714)
• profile-write.test.ts: explicit writes now bypass cold buffer (previously buffered items were silently eaten)
• Added frequency field to test preference items

All 4 checks pass. Ready for review.

[rookie136]

Fixed all 3 test failures:

• config.test.ts: updated userProfileMax* → userProfileDisplay* field names
• profile-write.test.ts: confidence assertion changed from >0.9 to >0.6 (Beta model ~0.714)
• profile-write.test.ts: explicit writes now bypass cold buffer (previously buffered items were silently eaten)
• Added frequency field to test preference items

All 4 checks pass. Ready for review.

[tickernelz]

Merged into main as part of the v2.19.0 batch release. Thanks for aligning the test assertions with the Beta confidence model (fix: align test assertions with Beta confidence model) — the suite is now green (197 pass / 0 fail) with typecheck and build clean on the fully-merged tree. Appreciate the follow-up.