What's Changed
- fix: guard pg_search update with skip_embedding by @Dounx in #778
- feat/im optimization by @voidkey in #774
- feat(dev): fix initial dev environment setup issues by @Dounx in #779
- fix: add OpenRouter embedding provider support by @Dounx in #780
- feat: expose builtin parser engine in settings by @Dounx in #781
- fix: allow temperature 0, default if negative by @tsukiga-kirei in #777
- fix: use milvus result set length for empty search results by @Dounx in #782
- fix: respect global default storage engine when creating knowledge base (#772) by @lyingbug in #784
- fix: use service name instead of UUID for MCP tool naming (#715) by @lyingbug in #783
- fix(chat): preserve image markdown during sanitize by @qiangfeidog in #789
- chore(client): remove redeclared
UserInfoin client package by @chengjoey in #792 - fix(frontend): reset page counter when reloading knowledge files by @ChenRussell in #790
- fix(chat): support native reasoning for vllm by bypassing go-openai limitations by @tsukiga-kirei in #791
- feat: add Novita AI as LLM provider by @Alex-wuhu in #794
- Feat/im new platform by @voidkey in #796
- fix(rerank): filter empty passages to prevent Aliyun API error by @ChenRussell in #806
- feat(i18n): add missing locale keys for new features added after #724 by @ochanism in #804
- fix: use COSINE metric instead of IP for Milvus vector search by @Dounx in #802
- fix: add missing common.saveSuccess and common.saveFailed i18n keys by @tsukiga-kirei in #801
- fix(tenant): prevent api_key encryption loss on tenant settings update by @ochanism in #800
- fix(frontend): fix markdown code block rendering error by @manx98 in #797
- feat(agent): auto-describe MCP tool result images via VLM by @ochanism in #808
- feat(agent): add GetSuggestedQuestions API for retrieving agent-specific question suggestions by @lyingbug in #809
- chore(go.mod): update dependencies to include tiktoken-go and golang.… by @lyingbug in #810
- feat(security): enhance SSRF protection in RemoteAPIChat by @lyingbug in #811
- feat: upgrade MiniMax models from M2.1 to M2.7 by @octo-patch in #812
- fix(app): prevent UTF-8 truncation in summary fallback by @ChenRussell in #813
- feat(prompt): enhance context and intent classification in templates by @lyingbug in #814
- Add IM channel Mattermost by @rakaustin in #820
- fix(agent): prevent empty response when no Knowledge Base is configured by @ochanism in #819
- fix: marked usage errors by @manx98 in #816
- chore(release): update version to 0.3.5 by @lyingbug in #822
New Contributors
- @Alex-wuhu made their first contribution in #794
- @octo-patch made their first contribution in #812
- @rakaustin made their first contribution in #820
Full Changelog: v0.3.4...v0.3.5