Release notes

0.3.0

Breaking Changes

  • Remove obj support (#85)

0.2.2

New Features

  • Make full messages available after streaming (#84)

0.2.0

Breaking Changes

  • toolloop now returns an iterator of every message, including each tool request and result

New Features

  • Add tool call results to toolloop result (#83)
  • show_thk param for contents (#82)

0.1.11

New Features

  • Use updated text editor tool (#80)
  • Skip hidden directories in view()

0.1.10

New Features

  • Support Claude 4 (#79)

0.1.9

New Features

  • Tool loop now continues when function calls return an error (#78), thanks to @erikgaas
  • feat: add text editor tool implementation (#71), thanks to @ncoop57

0.1.8

New Features

  • Add exhausted tool loop warning (#75)
  • Text editor tool implementation (#71), thanks to @ncoop57
  • Async tool loop (#70)
  • Pre-serialized funcs in tool calling (#67), thanks to @erikgaas
  • Extended Thinking (#66)

0.1.7

Bugs Squashed

  • Bump required msglm version

0.1.6

Bugs Squashed

  • Bump required anthropic version

0.1.5

New Features

  • add extended thinking (#65), thanks to @comhar
  • Make Sonnet 3.7 the default sonnet model (#63)
  • Add model capabilities attributes has_streaming_models, has_temperature_models, and has_system_prompt_models (#57), thanks to @austinvhuang

Bugs Squashed

0.1.3

New Features

  • add caching to async client too (#55), thanks to @bclavie

0.1.2

New Features

  • Add continuously-updated cache support to Chat and Client (#54), thanks to @bclavie
  • Enhance AsyncChat with improved tool support and message handling (#52), thanks to @ncoop57
  • Add support for user defined types in tool calling functions (#51), thanks to @austinvhuang
  • Add detailed cost breakdown and improve content handling (#49), thanks to @ncoop57

0.1.1

New Features

  • add structured to async (#48), thanks to @hamelsmu
  • add msglm (#46), thanks to @comhar
  • Add support for new Claude 3.5 Haiku model (#44), thanks to @ncoop57
  • trace history instead of chat response in toolloop (#39), thanks to @comhar

0.1.0

Breaking Changes

  • tool_choice is no longer a Chat instance variable; instead it is a parameter to Chat.__call__

New Features

  • Add temp param to Chat (#38)

Bugs Squashed

  • pr included but not used (#37)
  • fix tool use bug (#35), thanks to @comhar

0.0.10

New Features

  • Add Client.structured (#32)
  • Use dict2obj (#30)
  • Store tool call result without stringifying (#29)

0.0.9

New Features

  • Async support (#21)

0.0.7

New Features

  • Prompt caching (#20)
  • add markdown to doc output (#19)
  • Support vscode details tags (#18)
  • Add a cont_pr param to Chat as a “default” prompt #15), thanks to @tom-pollak

Bugs Squashed

  • Explicit tool_choice causes chat() to call tool twice. (#11)

0.0.6

New Features

  • Default chat prompt & function calling refactor (#15), thanks to @tom-pollak

0.0.5

New Features

  • Better support for stop sequences (#12), thanks to @xl0

0.0.3

New Features

  • Amazon Bedrock and Google Vertex support (#7)

Bug Fixes

  • Update model paths for non-beta tool use (#2), thanks to @sarahpannn

0.0.1

  • Initial release