v0.13.0

0.13.0 (2026-02-21)

🚀 Features

• chat: add hardware-adaptive inference configuration (b67338e)
• chat: add Qwen3 model family with thinking mode support (fdb2dfa)
• db: add sqlite-vec partition keys and filtered vector search (d7e0230)
• frontend: runtime bootstrap UI, AI tool discovery, and drag fix (db58a10)
• frontend: sync types with Qwen3 backend, expand AG-UI + A2UI rendering (cd15e29)
• mcp: integrate Official MCP Registry (6,000+ servers) (336ffb1)
• protocols: add zero-config runtime bootstrap system (af027f3)
• protocols: expand AG-UI to 30+ events and add A2A agent card (493a4da)

🐛 Bug Fixes

• ci: resolve 3 clippy errors in runtime_bootstrap tests (ca20399)
• resolve clippy dead_code warnings and doc indentation (2187d09)

⚡ Performance

• add dist build profile and reduce re-index interval to 60min (c66cc0d)
• embeddings: real tensor batching + GPU device selection (ce6cca6)

📚 Documentation

• update CLAUDE and ROADMAP files for clarity and accuracy; bump ghost version to 0.12.0 (2b7827a)
• update core documents for Qwen3, sqlite-vec, and A2A additions (5c0b7b1)
• update README, ROADMAP, and CLAUDE.md for runtime bootstrap (03e2cd2)
• website: auto-sync content from source files [skip ci] (ad7284d)
• website: auto-sync content from source files [skip ci] (103b362)