Release digest
Fixed
- CRYPTO-PDF-AI-FALLBACK-COPY (Evidence-Led Narrative Degradation) — Crypto PDF reports now render an evidence-led narrative fallback when AI analysis is unavailable. The fallback points readers to deterministic report sections such as Trust Score, Evidence Quality Matrix, Source Provenance, Security Readiness Packet, and evidence gaps instead of exposing provider errors, internal traces, or old raw-metrics copy.
Internal
- Added regression coverage for the AI analyst fallback path to verify it stays legal-safe, hides provider errors, avoids raw-metrics framing, and preserves the non-investment and non-audit boundary.
- Updated
docs/architecture/CRYPTO_REPORTS.mdandknowledge-base/docs/crypto/dashboard-ui.mdbecause this changes visible PDF narrative fallback behavior.