v0.2.0 — Trust Ledger

Highlights

Validation engine — 3 of 5 strategies now live

• Semantic Drift and Temporal Pattern strategies are wired into the validation orchestrator alongside Source-Linked
• Fact types are classified before validation; learned staleness patterns are looked up per tenant
• Contradicted memories are quarantined directly with proper reason mapping (contradicted vs stale)

Dashboard — complete "Trust Ledger" redesign

• New visual identity: warm paper theme, serif numerals, stamp-style status badges
• Grouped sidebar (Monitor / Operate / Configure) with live flagged/quarantined count badges
• ⌘K command palette for jumping to any page or filtered view
• Global health pill in the top bar; consistent numbered page headers

Docs & housekeeping

• README overhaul: worked validation example with real evidence JSON, positioning vs in-conversation invalidation, How It Works lifecycle, MCP connection setup for Claude Code/Desktop, and a contributor roadmap
• Honest strategy status table (Cross-Reference and Causal Chain marked as planned)
• Single consolidated .env.example with ports matching the actual compose mappings
• Fresh dashboard screenshots

Full Changelog: v0.1.1...v0.2.0