feat(phase5): epistemic confidence model

[salishforge]

Summary

Second of six PRs splitting #129. Ships Feature 1: Epistemic Confidence Model — calibrated uncertainty levels on warm-tier memories with auto-promotion via sleep cycle.

What lands

• New columns on warm_tier: epistemic_status TEXT NOT NULL DEFAULT 'provisional', evidence_count INTEGER NOT NULL DEFAULT 1, last_corroborated_at TIMESTAMPTZ. Index warm_tier_epistemic_idx on (agent_id, epistemic_status).
• Sleep cycle Phase 5.12 (Epistemic Promotion) — runs in the run() loop after deprecated decay, before drift snapshot. Promotes provisional → established when both evidence_count >= 3 AND positive retrievals come from ≥2 distinct namespaces (cross-context corroboration required). Stamps last_corroborated_at. Does NOT touch contested rows.
• Query filter: query({ epistemic: 'established' }) or array form filters retrieval by status.
• REST endpoint: GET /memory/:agentId/epistemic returns per-status counts.
• MCP tools: memforge_certainty (filtered query) and memforge_epistemic_profile (counts).
• SDK: TypeScript + Python clients both gain epistemic query param and epistemicProfile() method.

Migration: schema/migration-v3.9.sql.

Notes

• Fixed a bug from PR feat: Phase 5 — Autonomous Knowledge Architecture (7 features) #129: the original memforge_certainty MCP executor built epistemicOpts but passed a separate options object to client.query() that ignored the field. Corrected here so the tool actually filters.
• evidence_count defaults to 1 (not 0) — a newly consolidated row counts the consolidation event itself as one unit of evidence, matching PR feat: Phase 5 — Autonomous Knowledge Architecture (7 features) #129's design.

Part of the #129 split

• ✅ PR-A (feat(phase5): sentiment tagging + adaptive sleep intelligence #136) — Sentiment + Adaptive Sleep
• ✅ PR-B (this) — Epistemic Confidence
• ⏳ PR-C — Explainable Memory Ops (depends on these columns)
• ⏳ PR-D — Hierarchical Abstraction
• ⏳ PR-E — Causal Memory Graph
• ⏳ PR-F — Transfer Learning

Test plan

• `npm run type-check` — clean
• `npm run build` — clean
• `npm run lint` — clean
• `npm run test:epistemic-confidence` — 30/30 pass (7 unit + 12 integration + 5 e2e + 6 migration)
• `npm audit --omit=dev` — 0 vulnerabilities
• CI green on push

🤖 Generated with Claude Code