Bill 00a9c4d976 Sprint 63: Learned Adapter Hints (Non-Authoritative) — Steps 859-868 ...

Implements the learned adapter hints layer with governance guardrails:
- HintFeatureExtractor: extract features from decision ledgers (step 859)
- HintModelInterface: pair-specific hint query interface (step 860)
- DeterministicFallback: safe fallback when hints unavailable (step 861)
- HintConfidencePacket: confidence bands and escalation logic (step 862)
- HintABHarness: A/B harness to measure hint effectiveness (step 863)
- HintRollbackControl: suppress and lift hint suppression per pair (step 864)
- whetstone_get_adapter_hints MCP tool (step 865)
- whetstone_set_hint_policy MCP tool (step 866)
- HintSafetyGuardrails: non-authoritative enforcement + audit (step 867)
- Sprint63IntegrationSummary (step 868)

All 10 steps passing (94 tests). Governance rule: learned hints may
suggest; deterministic policy decides.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

2026-02-22 18:40:31 -07:00

2.0 KiB

Raw Blame History

View Raw