What's Changed
- feat: unified recall merges shared surface results by @achrllrogia45 in #181
- fix: improve recall precision by @WXBR in #179
- feat: add memory audit log for mutation traceability by @achrllrogia45 in #182
- feat: collaborative memory validation with attestation chain by @achrllrogia45 in #183
- feat(embeddings): explicit opt-in for non-OpenAI OpenRouter embedding models by @val-spc in #187
- feat: query cache and synonym expansion by @AxDSan in #188
- feat: temporal parser with schema migration and auto-extraction by @AxDSan in #189
- feat: weibull decay, MMR re-ranking, and query intent classification by @AxDSan in #190
- fix: prepend /no_think to disable Qwen3 thinking mode by @ether-btc in #192
- feat(embeddings): add MNEMOSYNE_EMBEDDING_API_URL/KEY as preferred env var names by @flooryyyy in #194
- fix: irrelevant context injection — strict fact match, entity prefix guard, single-token support by @AxDSan in #199
New Contributors
- @val-spc made their first contribution in #187
- @flooryyyy made their first contribution in #194
Full Changelog: v3.1.0...v3.1.2