2 Releases published by 1 person

• v0.3.5 v0.3.5 (2025-06-30)

  published Jun 30, 2025

• v0.3.6 v0.3.6 (2025-07-04)

  published Jul 4, 2025

15 Pull requests merged by 12 people

• Fix GitHub Models docs links

  #2135 merged Jul 4, 2025

• Add GitHub Models provider

  #2114 merged Jul 4, 2025

• Add model_request_stream_sync to direct API

  #2116 merged Jul 4, 2025

• Make AgentStream.stream_output (available inside agent.iter) stream validated output data instead of raising validation errors

  #2134 merged Jul 4, 2025

• Revert "Use contextvars for tracking the MCP sampling model"

  #2132 merged Jul 4, 2025

• simplify weather example

  #2129 merged Jul 4, 2025

• Fix model parameters not being customized in fallback model request stream

  #2120 merged Jul 3, 2025

• Use contextvars for tracking the MCP sampling model

  #2117 merged Jul 2, 2025

• Use contextvars for agent overriding, rather than a local attribute

  #2118 merged Jul 2, 2025

• Record tool response in tool run span

  #2109 merged Jul 1, 2025

• Add support for predicted outputs in OpenAIModelSettings

  #2106 merged Jul 1, 2025

• Update client.md - Typo

  #2105 merged Jul 1, 2025

• Update starlette subdomain in docs

  #2099 merged Jul 1, 2025

• Indicate to the model that a RetryPromptPart not tied to a tool call contains validation feedback rather than a user message

  #2008 merged Jul 1, 2025

• Deprecate {FunctionToolCallEvent,FunctionToolResultEvent}.call_id in favor of tool_call_id

  #2028 merged Jul 1, 2025

12 Pull requests opened by 12 people

• Bugfix: avoid race condition when refreshing google token

  #2100 opened Jun 30, 2025

• feat: AG-UI adapter (toolsets)

  #2101 opened Jun 30, 2025

• Builtin tool

  #2102 opened Jun 30, 2025

• Upgrade fasta2a to A2A v0.2.3 and Enable Dependency Injection

  #2103 opened Jul 1, 2025

• Added support for google specific arguments for video analysis

  #2110 opened Jul 1, 2025

• feat: add HistoryProcessors wrapper

  #2124 opened Jul 3, 2025

• Add temporal as a dependency group

  #2126 opened Jul 3, 2025

• Add context dependencies to Tool.from_schema()

  #2130 opened Jul 4, 2025

• Add: structured output support via StructuredOutput type

  #2133 opened Jul 4, 2025

• Add `settings` to Model base class

  #2136 opened Jul 5, 2025

• Adding CountToken to Gemini

  #2137 opened Jul 5, 2025

• Fix/cli refactor

  #2138 opened Jul 6, 2025

8 Issues closed by 5 people

• sync method for direct api `model_request_stream`

  #2096 closed Jul 4, 2025

• Automatically retry Gemini requests that fail with a 500 INTERNAL

  #2115 closed Jul 3, 2025

• How to disable thinking for Qwen/Qwen3-32B (deployed by vllm) in Pydantic AI?

  #2121 closed Jul 3, 2025

• LLM misinterprets "Plain text responses are not permitted" message as coming from the user

  #1993 closed Jul 2, 2025

• Show tool response in the running tool span

  #2004 closed Jul 1, 2025

• Best Practice for Streaming Output from Nested Agents

  #2107 closed Jul 1, 2025

• Add support for predicted outputs in OpenAIModelSettings

  #2098 closed Jul 1, 2025

• Inconsistent naming between `FunctionToolCallEvent.call_id` and `FunctionToolResultEvent.tool_call_id`

  #2007 closed Jul 1, 2025

16 Issues opened by 14 people

• Support more audio format in Gemini models

  #2143 opened Jul 7, 2025

• Making File Downloading Model Profile Compatible

  #2142 opened Jul 7, 2025

• Greater introspection into the successful model when using FallbackModel

  #2141 opened Jul 7, 2025

• ModelResponsePart should support multimodal data

  #2140 opened Jul 7, 2025

• Add optional src or metadata field to BinaryContent

  #2139 opened Jul 7, 2025

• LLM thinking blocks displayed in Logfire message view

  #2131 opened Jul 4, 2025

• Support Anthropic tool result citations

  #2128 opened Jul 3, 2025

• Support OpenAI encrypted thinking tokens

  #2127 opened Jul 3, 2025

• [Suggestion] Documenting how to implement 12 Factor Agents

  #2125 opened Jul 3, 2025

• Agent tool calling works with mistral but fails with llama models (on AWS Bedrock)

  #2123 opened Jul 3, 2025

• Get dependencies in function Tool.from_schema()

  #2122 opened Jul 3, 2025

• FallbackModel to allow model_settings specific to each model

  #2119 opened Jul 2, 2025

• Feature Request: Realtime/Live API integration

  #2113 opened Jul 2, 2025

• Feature Request: Enhancements for Test Orchestration, Non-Determinism, and Extensibility

  #2112 opened Jul 2, 2025

• Support for Long running / Background Agents

  #2111 opened Jul 1, 2025

• Show output functions span in logfire

  #2108 opened Jul 1, 2025

38 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Toolsets

  #2024 commented on Jul 4, 2025 • 41 new comments

• Implemented a convenient way to use ACI.dev Tools in PydanticAI

  #2093 commented on Jul 4, 2025 • 22 new comments

• Update tests to be compatible with new OpenAI, MistralAI and MCP versions

  #2094 commented on Jul 5, 2025 • 6 new comments

• Add Hugging Face as a provider

  #1911 commented on Jul 2, 2025 • 5 new comments

• Anthropic extended thinking: invalid request error when using tools

  #2040 commented on Jul 3, 2025 • 0 new comments

• Pre-validation tool response modifications

  #1985 commented on Jul 3, 2025 • 0 new comments

• Add option to filter reasoning content from message history

  #1157 commented on Jul 3, 2025 • 0 new comments

• Remove deprecated preview models from LatestGeminiModelNames

  #2016 commented on Jul 4, 2025 • 0 new comments

• Extracting `new_messages()` in the middle of an `Agent.iter` run to support user-in-the-loop approval of tool calls

  #1995 commented on Jul 4, 2025 • 0 new comments

• Gemini unable to stream structured output

  #1237 commented on Jul 5, 2025 • 0 new comments

• Native temporal support

  #1975 commented on Jul 5, 2025 • 0 new comments

• Save `history_processors`'s result for next round model request

  #2095 commented on Jul 5, 2025 • 0 new comments

• Add Resources, Prompts to MCP Client

  #1558 commented on Jul 5, 2025 • 0 new comments

• Adding LiteLLM as model wrap just like how google-adk does it.

  #1496 commented on Jul 5, 2025 • 0 new comments

• Handoffs / sub-agent delegation

  #1978 commented on Jul 6, 2025 • 0 new comments

• Add a setting to remove prompts and completions from tracing

  #1571 commented on Jul 7, 2025 • 0 new comments

• Support image output

  #1130 commented on Jun 30, 2025 • 0 new comments

• Add `builtin_tools` to `Agent`

  #1722 commented on Jul 1, 2025 • 0 new comments

• Dockerise the MCP Run Python server (#1837)

  #2090 commented on Jun 30, 2025 • 0 new comments

• BinaryContent is naively parsed when include_input is used in LLMJudge

  #2089 commented on Jun 30, 2025 • 0 new comments

• Feature Request: Support Both JSON & YAML as Tool Output Formats

  #2074 commented on Jun 30, 2025 • 0 new comments

• Running many models on Bedrock with `output_type` fails with "This model doesn't support the toolConfig.toolChoice.any field."

  #2091 commented on Jun 30, 2025 • 0 new comments

• MALFORMED_FUNCTION finishReason in Gemini candidate

  #631 commented on Jun 30, 2025 • 0 new comments

• [OpenAI] Error Handling for Exceeding Maximum Token Limits

  #1098 commented on Jun 30, 2025 • 0 new comments

• Messages passed to history_processor functions include current run messages

  #2050 commented on Jun 30, 2025 • 0 new comments

• add the ability to set a list of allowed tools on an mcp server

  #1219 commented on Jun 30, 2025 • 0 new comments

• For list-like `result_models`, LLM tries to call `final_result` multiple times

  #1429 commented on Jul 1, 2025 • 0 new comments

• Feature Request: mcp-run-python support streamable http

  #2059 commented on Jul 1, 2025 • 0 new comments

• Gemini causes 'Event loop is closed' when running inside an async context

  #748 commented on Jul 1, 2025 • 0 new comments

• run_stream not working properly with tools in streaming mode

  #1007 commented on Jul 1, 2025 • 0 new comments

• Gemini 2.5 Pro Streamed response has no content field (just streamed thinking parts)

  #2097 commented on Jul 1, 2025 • 0 new comments

• Support `o3-deep-research` & `o4-mini-deep-research`

  #2086 commented on Jul 1, 2025 • 0 new comments

• Response parts are not timestamped

  #1364 commented on Jul 1, 2025 • 0 new comments

• Add media processing settings to GoogleModelSettings

  #2017 commented on Jul 2, 2025 • 0 new comments

• Azure OpenAI API Streaming Response Causes AttributeError in pydantic-ai

  #797 commented on Jul 2, 2025 • 0 new comments

• [Feature Request] Stateful Python Execution Environment for Agents in run_python MCP server (Jupyter-like Workspace)

  #2070 commented on Jul 3, 2025 • 0 new comments

• Feature Request: Error handling multiple MCP Servers

  #2083 commented on Jul 3, 2025 • 0 new comments

• support batch processing

  #1771 commented on Jul 3, 2025 • 0 new comments