v0.3.3

What's Changed

• Utils for supplying images to LLMs by @ZmeiGorynych in #114

Full Changelog: v0.3.1...v0.3.3

v0.3.2

What's Changed

• Default LLM is now GPT-4.1 instead of GPT-4o
• Allow not specifying agent invoke() input by @whimo in #108
• Set the streaming mode of LLM in ReActToolCallingMotleyAgent by @ZmeiGorynych in #111
• Change motleycrew wrappers to omit temperature setting by @ZmeiGorynych in #109
• Let ReActToolCallingMotleyAgent accept a List[BaseMessage] as input by @ZmeiGorynych in #110
• Various helpful utilities around images (specify image inputs, detect if an image contains a chart, parse chart data from image) by @ZmeiGorynych in #112
• Fix signatures of StructuredPassthrough and AgenticValidationLoop by @whimo in #113

Full Changelog: v0.3.0...v0.3.2

v0.3.0

What's Changed

• Add structured passthrough tool by @ZmeiGorynych in #101
• New PythonREPLTool implementation
• Agent interface improvements regarding prompts by @whimo in #106
  to avoid confusion, prompt_prefix argument was removed, prompt should be used instead which supports placeholder variables
• Refactor exception handling control in MotleyTool by @whimo in #107
  again to avoid confusion, exceptions_to_reflect argument was renamed to handle_exceptions, which also supports boolean values, where True means "return any unhandled exception message as the return value"

Full Changelog: v0.2.5...v0.3.0

v0.2.5

What's Changed

• Add KC cache to agent by @ZmeiGorynych in #99

Full Changelog: v0.2.4...v0.2.5

v0.2.4

What's Changed

• Update Multi-step research agent.ipynb with local embeddings example by @iSevenDays in #87
• Support AzureOpenAI LLMs & fix research agent defaulting to OpenAI by @whimo in #94

New Contributors

• @iSevenDays made their first contribution in #87

Full Changelog: v0.2.3...v0.2.4

v0.2.3

What's Changed

• Support agent app & async tools by @whimo in #89
• Customer support app & event driven workflows docs by @whimo in #92
• Fixes for async tools & doc updates by @whimo in #93

Full Changelog: v0.2.2...v0.2.3

v0.2.2

What's Changed

• Fixes for asynchronous crew execution by @whimo in #81
• Event-driven orchestration demo with Faust by @whimo in #85
• Support custom LLMs in research agent by @whimo in #86
• Upgrade Langchain to v0.3 by @whimo in #82
• Retry mechanism in MotleyTool by @whimo in #88

Full Changelog: v0.2.1...v0.2.2

v0.2.1

What's Changed

• Support format strings in prompt_prefix by @ViStefan in #77
• populate version by @ViStefan in #78
• Fix CrewAI delegation by @whimo in #79
• Remove setuptools explicit dependency by @whimo in #80

New Contributors

• @ViStefan made their first contribution in #77

Full Changelog: v0.2.0...v0.2.1

v0.2.0

What's Changed

In v0.2, some breaking changes were introduced. Agents no more have a distinct output_handler argument. Output handlers are now regular tools with return_direct set to True. Multiple output handlers are supported.

Agents now have a force_output_handler argument, which will block direct output.

• Unify output handlers and regular tools, improve exception management, major refactor by @whimo in #76

v0.1.15

What's Changed

• MotleyTool LlamaIndex compatibility fix by @whimo in #74
• Support various LLM providers (Ollama, Groq, Together...) + docs by @whimo in #75

Full Changelog: v0.1.14...v0.1.15