v0.51.0
What's Changed
- fix: parse literal characters in LLM function calling arguments by @tyaroshko in #734
- fix: update calculation of cached tokens by @tyaroshko in #731
- fix: apply custom default reasoning effort per model for OpenAI by @tyaroshko in #733
- feat: add missing cancellation logic for nodes by @TrachukT in #728
- feat: add checkpoints on input timeout by @maksymbuleshnyi in #730
- fix: fix response format streaming by @mykhailobuleshnyi in #740
- chore: update models with default reasoning_effort=none by @tyaroshko in #741
- fix: update saving of function calling messages to memory by @tyaroshko in #739
- fix: extended summarization by @mykhailobuleshnyi in #742
- fix: bedrock for structured output and xml stopsequence by @olbychos in #737
- feat: add splitters by @olbychos in #711
- build: dynamiq v0.51.0 by @tyaroshko in #743
Full Changelog: v0.50.0...v0.51.0