v0.2.0

What's Changed

• fix: clear and unref the embedding timeout timer by @nicholls73 in #5
• fix: warn and skip embeddings that no longer match the model by @nicholls73 in #4
• fix: open SQLite in WAL mode with a busy timeout by @nicholls73 in #2
• fix: reuse one embedder across index rebuild by @nicholls73 in #1
• fix: rank hybrid search with reciprocal rank fusion by @nicholls73 in #3
• Add memory metadata lifecycle by @nicholls73 in #6
• Notify agents when OpenBrain updates are available by @nicholls73 in #9
• Fix SQLite ABI mismatch recovery by @nicholls73 in #8
• feat: install Claude Code SessionStart hook in adapter by @nicholls73 in #10
• ci: add GitHub Actions workflow for build and test by @nicholls73 in #30
• chore: add Biome for lint and format by @nicholls73 in #31
• fix: rebuild index atomically on a single connection by @nicholls73 in #32
• fix: anchor excerpt on the numerically earliest query match by @nicholls73 in #33
• fix: prune episodes by their effective expiresAt, not filename age by @nicholls73 in #34
• fix: serialise config.json mutations and write it atomically by @nicholls73 in #35
• fix: give model load its own timeout and surface FTS-only degradation by @nicholls73 in #36
• perf: store embeddings as float32 blobs instead of JSON text by @nicholls73 in #37
• perf: reuse stored embeddings during index rebuild by @nicholls73 in #38
• fix: serialise embeds so timed-out inference does not pile up by @nicholls73 in #39
• fix: enforce config.embeddings.dimensions at the provider boundary by @nicholls73 in #40
• refactor: expose typed brain resolution via getBrainStatus by @nicholls73 in #41
• test: add opt-in integration test for the real embedding pipeline by @nicholls73 in #42
• feat: pin installer to verified releases and add a release workflow by @nicholls73 in #43

New Contributors

• @nicholls73 made their first contribution in #5

Full Changelog: https://github.com/nicholls73/openbrain/commits/v0.2.0