Repository navigation
What's Changed
- Let the Anthropic provider load the anthropic gem 2.x by @TonsOfFun in #667
- Show the ways an agent arrives on the empty agent list by @TonsOfFun in #696
- Docs: a guide to running evals in the test suite by @TonsOfFun in #697
- Docs: getting started with the hosted dashboard by @TonsOfFun in #691
- fix(anthropic): record token usage from streamed responses by @pcbeingused333 in #698
- fix(requesty): implement api_embed_execute so embed works by @pcbeingused333 in #688
- fix(model-capabilities): strip top_k for thinking-first models by @pcbeingused333 in #669
- fix(usage): sum duration_ms when adding Usage objects by @pcbeingused333 in #690
- Cancel a run only while it is still in progress by @yorzi in #672
- Keep an agent's own tool observer under Capture by @yorzi in #673
- fix(usage): include cache tokens in Anthropic total_tokens by @pcbeingused333 in #675
- feat: add Runcrate as an OpenAI-compatible provider by @heydryft in #689
- 1.9.8 follow-ups: changelog placement for the merged fixes, and the cancel race test on Rails 7.2 by @TonsOfFun in #699
- Keep a catalog product's named criteria lists by @yorzi in #701
- Keep a catalog's pass bars through dashboard import by @yorzi in #702
- Show a run's per-group pass gates in the dashboard by @yorzi in #703
- fix: preserve Responses streaming error details by @pcbeingused333 in #706
- Fail evals run on a missed per-group pass gate by @yorzi in #704
- Make the RubyLLM provider work with ruby_llm 2.2 by @TonsOfFun in #710
- Release 1.9.8 by @TonsOfFun in #711
New Contributors
Full Changelog: v1.9.7...v1.9.8