Changelog

New Features:

OpenAI File Support: Added support for File attached to prompts for agents with OpenAIChat models.

Llama API: Various improvements for Llama and LlamaOpenAI model classes including structured output and image input support
Async Custom Retriever: The retriever parameter can now be an async function to be used with agent.arun and agent.aprint_response.
Gemini Video URL Input: Added support for Video(url=...) for Gemini.

OpenAI Responses o3 / o4 Tools: Fixed broken tool use for advanced reasoning models on OpenAIResponses.
MCP on CLI Support: Fixed support for MCPTools usage while calling agent.acli_app.

llama-api-update-ag-3035 by @ysolanky in #3037
llama api fixes and updates by @kausmeows in #3038
Website kb + pinecone DB should respect namespace arg by @5739n4 in #2973
add support for async retriever by @kausmeows in #3026
Bug Fix [3041]: Gemni models always return tool_call_id as None and a… by @gauravdhiman in #3042
Make sure reasoning models are correctly set by @dirkbrnd in #3057
fix o3/o4-mini for tools by @kausmeows in #3023
Allow PDF / doc uploads for OpenAIChat by @Mustafa-Esoofally in #3052
feat(agent): add support for async cli app by @manuhortet in #3045
Allow new models for OpenAI Transcription by @Mustafa-Esoofally in #3040
fix(bedrock-model): propagate request_params to Bedrock Converse API by @leandrodamascena in #3035
delete_user_memory missing existence check by @jkatigb in #2999
gemini-video-url-ag-3017 by @ysolanky in #2980
fix(mcp): Apply checks for include_tools and exclude_tools during initialize. by @daavoo in #2985
[Fix] Fix issues in Workflows when loading sessions with new default memory by @dirkbrnd in #3056
llama-image-support-ag-3037 by @ysolanky in #3053
Fix schema handling in SQLTools for table inspection by @tiberiu92 in #3060
add better exception for website reader by @Ayush0054 in #2975
Fix: metrics-fix-ag-3045 by @ysolanky in #3062
Release 1.4.4 by @dirkbrnd in #3061

Full Changelog: v1.4.3...v1.4.4