Overview

3 Releases published by 2 people

• v0.8.0 v0.8.0 (2025年08月26日)

  published Aug 26, 2025

• v0.8.1 v0.8.1 (2025年08月29日)

  published Aug 29, 2025

• v1.0.0b1 v1.0.0b1 (2025年08月30日)

  published Aug 30, 2025

50 Pull requests merged by 12 people

• add identifier field support to FileUrl and subclasses

  #2636 merged Sep 2, 2025

• Minor fixes: Agent.__aenter__ returns Self, use default instrumentation for MCP sampling model

  #2765 merged Sep 2, 2025

• Properly close VertexAI HTTP client at end of live test

  #2768 merged Sep 2, 2025

• Include project ID for Google service account auth

  #2755 merged Sep 1, 2025

• Reorganize the docs nav hierarchy

  #2748 merged Sep 1, 2025

• Close original response when retrying HTTP request

  #2753 merged Sep 1, 2025

• fix(examples): Check for BaseModel before serializing json

  #2729 merged Sep 1, 2025

• Fix error when streaming from Gemini includes only executable_code or code_execution_result

  #2719 merged Sep 1, 2025

• Add docstring_format, require_parameter_descriptions, schema_generator to FunctionToolset

  #2601 merged Sep 1, 2025

• Raise error if MCP server __aexit__ is called when _running_count is already 0

  #2696 merged Sep 1, 2025

• Update mcp server package version

  #2741 merged Sep 1, 2025

• Don't emit empty AG-UI thinking message events

  #2754 merged Sep 1, 2025

• Fix HITL tool approval docs example

  #2752 merged Sep 1, 2025

• Add beta release admonition

  #2744 merged Aug 30, 2025

• Fix newsletter signup form

  #2743 merged Aug 30, 2025

• Remove more todo comments

  #2739 merged Aug 30, 2025

• BREAKING CHANGE: Make many more dataclasses kw-only

  #2738 merged Aug 30, 2025

• V1: Remove cases and averages from eval span

  #2715 merged Aug 29, 2025

• Default InstrumentationSettings version to 2

  #2726 merged Aug 29, 2025

• Let almost all types used in docs examples be imported directly from pydantic_ai

  #2736 merged Aug 29, 2025

• V1: Remove errors when passing Retrying instead of RetryConfig to TenacityTransport

  #2717 merged Aug 29, 2025

• V1: Remove Temporal docs beta header

  #2727 merged Aug 29, 2025

• Restore DeferredToolCalls with deprecation warning

  #2735 merged Aug 29, 2025

• Include logfire with pydantic-ai package

  #2683 merged Aug 29, 2025

• Gracefully handle errors in evals

  #2295 merged Aug 29, 2025

• Add support for human-in-the-loop tool call approval

  #2581 merged Aug 29, 2025

• Deprecate OpenAIModelProfile.openai_supports_sampling_settings

  #2730 merged Aug 29, 2025

• Remove 'only on 3.10+' callouts

  #2733 merged Aug 29, 2025

• Use match statement now that we're dropping 3.9

  #2732 merged Aug 29, 2025

• Remove/update some TODO comments

  #2731 merged Aug 29, 2025

• Drop support for Python 3.9

  #2725 merged Aug 29, 2025

• Resolve TODO comment

  #2728 merged Aug 29, 2025

• V1: Rename ModelResponse.provider_request_id to provider_response_id

  #2710 merged Aug 29, 2025

• V1: Deprecate specifying a model name without a provider prefix, and the vertexai provider name

  #2711 merged Aug 28, 2025

• Remove unused non-user-code-prefixes stuff

  #2712 merged Aug 28, 2025

• Reenable vertexai docs test

  #2709 merged Aug 28, 2025

• ci: Fix pydantic-ai-slim job skipping all tests

  #2704 merged Aug 28, 2025

• Bump temporalio to 1.16.0

  #2703 merged Aug 27, 2025

• Add MockOpenAIResponses for use in tests

  #2702 merged Aug 27, 2025

• docs: Add message_history parameter documentation for CLI methods

  #2695 merged Aug 27, 2025

• Rename StreamedRunResult methods to be consistent with AgentStream

  #2692 merged Aug 27, 2025

• Add gen_ai.system_instructions attribute to agent run spans

  #2699 merged Aug 27, 2025

• Add v0.8.0 breaking changes to changelog

  #2694 merged Aug 26, 2025

• fix: Make FileUrl.media_type work correctly with inline-snapshot

  #2693 merged Aug 26, 2025

• fix: Handle missing token details in vLLM/OpenAI-compatible APIs

  #2669 merged Aug 26, 2025

• BREAKING CHANGE: Make AgentStreamEvent union of ModelResponseStreamEvent and HandleResponseEvent

  #2689 merged Aug 26, 2025

• docs: Add docstrings for EndStrategy, capture_run_messages, HandleResponseEvent to API docs

  #2688 merged Aug 26, 2025

• Add message_history parameter to agent.to_cli()

  #2674 merged Aug 26, 2025

• Fix docs for MCP elicitation callback

  #2685 merged Aug 26, 2025

• fix: Properly deserialize complex tool arguments with Temporal

  #2686 merged Aug 26, 2025

10 Pull requests opened by 9 people

• Load MCP servers from file

  #2698 opened Aug 27, 2025

• Use output_type in generate_dataset

  #2716 opened Aug 28, 2025

• [Draft] Example for sequential tool calls

  #2718 opened Aug 28, 2025

• Support Python 3.14

  #2745 opened Aug 30, 2025

• Reorganize the tools docs

  #2749 opened Aug 31, 2025

• add support for previous_response_id from Responses API

  #2756 opened Sep 1, 2025

• add response id for streaming response

  #2757 opened Sep 1, 2025

• Docs: update `MCPServerStreamableHTTP` usage example to use `url` argument

  #2758 opened Sep 2, 2025

• Fix bug related to Azure OpenAI model streaming response

  #2763 opened Sep 2, 2025

• Add `operation.cost` span attribute to model request spans

  #2767 opened Sep 2, 2025

30 Issues closed by 8 people

• Support serializable object as input to agent run

  #2329 closed Sep 2, 2025

• `McpError` when `run-mcp-python` code sandbox handles concurrent requests

  #2655 closed Sep 2, 2025

• `genai-prices` is not a dev dependency

  #2760 closed Sep 2, 2025

• Change out storage type on pydantic a2a worker

  #2239 closed Sep 2, 2025

• Minor sandbox security issues in MCP-Run-Python

  #2332 closed Sep 2, 2025

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

  #2070 closed Sep 2, 2025

• Google Service Account authentication still tries to use Application Default Credentials

  #2746 closed Sep 1, 2025

• `AsyncTenacityTransport` does not release connections on retries

  #2724 closed Sep 1, 2025

• Assertion error when using Gemini Models in streaming mode ALONG with CodeExecutionTool()

  #2626 closed Sep 1, 2025

• GPT-5/o4-mini empty summary responses and AG-UI integration

  #2687 closed Sep 1, 2025

• Recursively merge model settings

  #2578 closed Sep 1, 2025

• Human-in-the-Loop behaviour

  #2747 closed Sep 1, 2025

• Drop 3.9 support and make dataclasses and functions with many arguments keywords-only

  #2720 closed Aug 31, 2025

• Use new OpenTelemetry GenAI chat span attribute conventions

  #2348 closed Aug 29, 2025

• Handle errors more gracefully in pydantic-evals

  #2612 closed Aug 29, 2025

• Parallel tool call containing both ToolReturn and deferred tool does not work

  #2500 closed Aug 29, 2025

• Human-in-the-loop approval of tool calls

  #1995 closed Aug 29, 2025

• Deprecate `OpenAIModelProfile.openai_supports_sampling_settings` in favor of `openai_unsupported_model_settings`

  #2721 closed Aug 29, 2025

• Message History is not working with redis.

  #2664 closed Aug 29, 2025

• `pydantic-graph`: persistence across `GraphRun`

  #1361 closed Aug 29, 2025

• Builtin tools for Anthropic models?

  #2722 closed Aug 29, 2025

• Questions about message mechanism

  #2654 closed Aug 28, 2025

• Providing API Key programmatically

  #2706 closed Aug 28, 2025

• Function tools don't support datetime arguments

  #2682 closed Aug 28, 2025

• Agent with MCP in google Colab

  #2697 closed Aug 28, 2025

• Incorrect PartStart classification in some OpenAI models

  #2580 closed Aug 27, 2025

• AttributeError: 'NoneType' object has no attribute 'reasoning_tokens' with vLLM +OpenAIResponsesModel

  #2598 closed Aug 26, 2025

• Could we add message history to the agent.to_cli() ?

  #2650 closed Aug 26, 2025

• Tool function with `BaseModel` argument does not work with Temporal

  #2666 closed Aug 26, 2025

• openai.BadRequestError: Invalid type for 'messages[1].content': expected one of a string or array of objects, but got an object instead.

  #2653 closed Aug 26, 2025

15 Issues opened by 15 people

• Streaming token usage shows doubled input tokens vs non-streaming

  #2770 opened Sep 2, 2025

• Recommended way to view full requests to model (without a Logfire account)?

  #2769 opened Sep 2, 2025

• AG UI: Validation errors when using reasoning models with Ollama (gpt-oss and qwen3)

  #2766 opened Sep 2, 2025

• `NativeOutput` for groq Agents using supported models

  #2764 opened Sep 2, 2025

• UrlContextTool: '400 Tool use with function calling is unsupported' if `output_type` is specified

  #2761 opened Sep 2, 2025

• Groq models: Handling of invalid tool-calls

  #2759 opened Sep 2, 2025

• Can POML be used in PydanticAI?

  #2751 opened Sep 1, 2025

• Request: Review auto-generated MCP permission manifest for Python_Sandbox

  #2750 opened Sep 1, 2025

• Support MCP tool call response `structuredContent`

  #2742 opened Aug 30, 2025

• Z.ai models

  #2723 opened Aug 29, 2025

• Magistral - handling of mistralai.models.thinkchunk.ThinkChunk

  #2713 opened Aug 28, 2025

• Use a better OTel attribute structure for evals case data

  #2708 opened Aug 28, 2025

• Enable: output_type + stream_text

  #2707 opened Aug 28, 2025

• Send ThinkingPart back for OpenAIChatModel models which require reasoning content

  #2701 opened Aug 27, 2025

• Potential MCP issue (asyncio.exceptions.CancelledError)

  #2700 opened Aug 27, 2025

35 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.

• Proposition MVP integration with Outlines

  #2623 commented on Sep 1, 2025 • 29 new comments

• [Draft] Add an AgentToolset

  #2589 commented on Sep 2, 2025 • 21 new comments

• Add id and finish_reason to ModelResponse

  #2590 commented on Sep 2, 2025 • 20 new comments

• Add tool calls limit to `UsageLimits`

  #2633 commented on Sep 2, 2025 • 19 new comments

• Add LiteLLM provider for OpenAI API compatible models

  #2606 commented on Sep 2, 2025 • 18 new comments

• Support DBOS Agent by wrapping run as DBOS workflows and moving model requests and MCP to DBOS steps

  #2638 commented on Sep 1, 2025 • 15 new comments

• Add basic support for UploadedFile UserContent

  #2611 commented on Sep 1, 2025 • 10 new comments

• Add Bedrock/Anthropic prompt caching

  #2560 commented on Sep 1, 2025 • 9 new comments

• Add citations

  #2657 commented on Aug 27, 2025 • 1 new comment

• Fix StructuredDict with nested JSON schemas using $ref

  #2570 commented on Sep 1, 2025 • 1 new comment

• add XML attributes when formatting Pydantic models in prompts

  #2313 commented on Aug 29, 2025 • 1 new comment

• Support gpt-oss directly via Harmony format

  #2439 commented on Aug 26, 2025 • 0 new comments

• Missing Initial Tokens After Tool Call in Streaming Output (But Present in Final Output) with Gemini 2.5 Flash Model Using agent.iter()

  #2659 commented on Sep 2, 2025 • 0 new comments

• Expose Span Customization / Event Iteration for handle_ag_ui_request and other AG-UI functions

  #2639 commented on Sep 2, 2025 • 0 new comments

• Added MCP elicitation support for tool injection in mcp-run-python

  #2258 commented on Sep 2, 2025 • 0 new comments

• Add `on_complete` callback to AG-UI functions to get access to `AgentRunResult`

  #2429 commented on Sep 1, 2025 • 0 new comments

• Fixes #2513: Add support for Free-Form Function Calling and Context Free Grammar constraints over tools

  #2572 commented on Sep 2, 2025 • 0 new comments

• Add `cost` to `RunUsage`

  #2684 commented on Sep 1, 2025 • 0 new comments

• How to use streaming in sync ?

  #1502 commented on Aug 26, 2025 • 0 new comments

• Add a llama-cpp Model?

  #1801 commented on Aug 27, 2025 • 0 new comments

• Gemini unable to stream structured output

  #1237 commented on Aug 27, 2025 • 0 new comments

• ChatCompletionService giving out error when using hugging face router with providers

  #2499 commented on Aug 28, 2025 • 0 new comments

• Loss of Final Tokens After Tool Call When Using kimi-k2 Model with Streaming Output

  #2660 commented on Aug 28, 2025 • 0 new comments

• OpenAI Responses: `previous_response_id` support.

  #2663 commented on Aug 28, 2025 • 0 new comments

• Run Tools Sequentially

  #2628 commented on Aug 28, 2025 • 0 new comments

• Failed to yield ToolReturnPart streaming with structured output type via tool call

  #2640 commented on Aug 28, 2025 • 0 new comments

• How do I handle pydantic_ai.exceptions.UnexpectedModelBehavior

  #1380 commented on Aug 29, 2025 • 0 new comments

• Fix naming of MCP Client & Server

  #2496 commented on Aug 29, 2025 • 0 new comments

• Ability to Persist Messages in External Stores

  #530 commented on Aug 30, 2025 • 0 new comments

• Pydantic AI V1

  #1974 commented on Aug 31, 2025 • 0 new comments

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

  #797 commented on Sep 1, 2025 • 0 new comments

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

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

• MCPServerStdio initialization exception not propagated

  #2607 commented on Sep 1, 2025 • 0 new comments

• Add `id` and `finish_reason` to `ModelResponse`

  #886 commented on Sep 1, 2025 • 0 new comments

• Endless Tool Loop (Vertex Gemini)

  #765 commented on Sep 2, 2025 • 0 new comments