Skip to content

python_v0.1.85

Latest

Choose a tag to compare

@Tomas2D Tomas2D released this 28 Sep 18:43
python_v0.1.85
0304ca1

Bug Fixes

  • tools: generate typed schemas for keyword-only parameters (#1721)
  • propagate Python stream errors and cancellation (#1711)
  • adapters: preserve LangChain image content (#1715)
  • adapters: apply LiteLLM proxy response cost reported via _hidden_params (#1699)
  • tools: preserve cache and middleware when cloning file and shell tools (#1706)
  • internals: carry the abort state through AbortController.clone (#1704)
  • backend: copy the caller's settings dict instead of aliasing it (#1701)
  • agents: render the task template so backstory reaches the model (#1700)
  • internals: avoid blocking the event loop during confirmation (#1705)
  • agents: clone independent tool instances in LiteAgent and RequirementAgent (#1696)
  • tools: stop reporting a full result set as truncated in GlobTool (#1692)
  • cache: isolate sliding cache clones (#1683)
  • deps: resolve torch from PyPI on Apple Silicon and bump off 2.7.1 (#1673)
  • emitter: clean up child pipe listeners on destroy to prevent memory leak (#1670)
  • tools: stop reporting a full result set as truncated in GrepTool (#1668)
  • internals: apply the default retry backoff factor when the config leaves it unset (#1667)
  • tools: parse a YAML OpenAPI schema with yaml.safe_load, not yaml.parse (#1666)
  • internals: register the two field validators the decorator order disabled (#1664)
  • backend: keep cached token counts when merging chunks (#1663)
  • internals: restore default I/O in new contexts (#1662)
  • internals: prevent extra AbortError after handled tool failures (#1659)
  • tools: isolate grep subprocess stdin (#1661)
  • tools: preserve leading-hyphen grep patterns (#1660)
  • backend: add default 600s request timeout for LiteLLM chat model (#1647)
  • adapters: use constant-time API key comparison and anchored Bearer strip (#1656)

New Contributors

Full Changelog: python_v0.1.84...python_v0.1.85