2 Releases published by 1 person

• v0.1.9 v0.1.9 (2025-05-02)

  published May 2, 2025

• v0.1.10 v0.1.10 (2025-05-06)

  published May 6, 2025

6 Pull requests merged by 6 people

• Add thinking_config to GeminiModel

  #1609 merged May 6, 2025

• Add extra_headers to ModelSettings

  #1644 merged May 6, 2025

• Allow setting temperature to 0 on bedrock.

  #1631 merged May 2, 2025

• Remove note about unavailable streaming capability of the model

  #1630 merged May 2, 2025

• Add base_url to Mistral Provider

  #1617 merged May 1, 2025

• Store additional usage details from Anthropic

  #1549 merged May 1, 2025

11 Pull requests opened by 10 people

• improve instrumentation docs

  #1625 opened May 1, 2025

• Add Cloudflare Provider

  #1627 opened May 1, 2025

• Support structured and manual JSON output_type modes in addition to tool calls

  #1628 opened May 2, 2025

• Add AWS Profile Feature

  #1634 opened May 2, 2025

• Refactor type guards in Agent class to use TypeIs instead of TypeGuard

  #1637 opened May 2, 2025

• Add newline separation between Anthropic system prompts

  #1638 opened May 3, 2025

• Expose run_cli() method on Agent

  #1642 opened May 4, 2025

• Adding `ModelResponse.usage`

  #1647 opened May 5, 2025

• gemini: support int enum

  #1653 opened May 6, 2025

• Corrected version of PR #1600

  #1656 opened May 7, 2025

• Allow definition of Model Settings for LLMJudge

  #1662 opened May 8, 2025

9 Issues closed by 5 people

• How to structure multi-message agents to provide few shot examples as separate user-assistant message pairs?

  #1652 closed May 7, 2025

• Possible extra_headers interface similar to extra_body?

  #1573 closed May 6, 2025

• Error while importing pydantic_ai

  #1635 closed May 5, 2025

• new_messages not always including tool-call information

  #1640 closed May 3, 2025

• Missing Support for `mistral-ocr-latest` in pydantic_ai

  #1633 closed May 2, 2025

• Obsolete note in AnthropicModel documentaion

  #1611 closed May 2, 2025

• New output_type not as flexible as the deprecated result_type / result_tool_name / result_tool_description

  #1629 closed May 2, 2025

• Exception when bedrock no response

  #1518 closed May 1, 2025

• Streaming and awaiting `get_output` leads to duplicated messages

  #1620 closed May 1, 2025

16 Issues opened by 15 people

• ModelRetry not support stream output

  #1663 opened May 8, 2025

• Incorrect ValueRenderer selection for expected_output when building eval report table

  #1661 opened May 7, 2025

• Evaluators: Return LLMJudge score

  #1660 opened May 7, 2025

• Erratic performance when using nested schemas

  #1659 opened May 7, 2025

• Can pydantic-ai support non LLM models (for example text2image models)

  #1658 opened May 7, 2025

• evals progress bar

  #1657 opened May 7, 2025

• model request timeout

  #1655 opened May 7, 2025

• QWEN3-32B failed to call tool without argument in streaming

  #1654 opened May 7, 2025

• Is it possible to trace agents, tool call/ expected tool calls in Pydantic Eval?

  #1650 opened May 6, 2025

• Support Tool Calling with Llama 3.3 on Bedrock

  #1649 opened May 5, 2025

• Update Agent Signature in lieu of PEP-747

  #1648 opened May 5, 2025

• Dynamic System prompts not being included on fresh runs with message_history

  #1646 opened May 5, 2025

• How to refine function tool arguments before calling tool itself

  #1643 opened May 5, 2025

• Setup for Cursor to understand Pydantic AI

  #1641 opened May 3, 2025

• Import Agent throws a TypeError: unhashable type: 'list'

  #1639 opened May 3, 2025

• Streamable HTTP Implementaion

  #1632 opened May 2, 2025

40 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.

• Add `low_level` public API

  #1599 commented on May 2, 2025 • 32 new comments

• feat: check mcp tool name and add tool_prefix

  #1266 commented on May 7, 2025 • 16 new comments

• Add A2A server

  #1537 commented on May 7, 2025 • 15 new comments

• feat: adds logprobs to OpenAi response

  #1238 commented on May 2, 2025 • 4 new comments

• [Issue-1363] Allow overriding tools

  #1616 commented on May 2, 2025 • 1 new comment

• Changes to get reasoning to work with the responses API

  #1587 commented on May 8, 2025 • 0 new comments

• Add support for non-string enums in Gemini

  #1564 commented on May 6, 2025 • 0 new comments

• Fix for Bedrock Parallel Tool Call User message issue for certain models

  #1600 commented on May 7, 2025 • 0 new comments

• Add prepare_tools param to Agent class

  #1474 commented on May 7, 2025 • 0 new comments

• Add support for callable output tools

  #1463 commented on May 6, 2025 • 0 new comments

• Add provider for Anthropic's Vertexai Client

  #1392 commented on May 5, 2025 • 0 new comments

• Fall back on guessed MIME type when requested multi-modal content doesn't have Content-Type header

  #1613 commented on May 6, 2025 • 0 new comments

• Support field `fileData` (direct file URL) for Gemini models

  #1136 commented on May 6, 2025 • 0 new comments

• Do not add system prompt part when dynamic system prompt function returns empty value

  #864 commented on May 4, 2025 • 0 new comments

• GoogleSearchTool for GeminiModel and VertexAIModel

  #840 commented on May 8, 2025 • 0 new comments

• Allow definition of Model Settings for LLMJudge

  #1624 commented on May 8, 2025 • 0 new comments

• Token Counter

  #1106 commented on May 8, 2025 • 0 new comments

• Intermittent 400 Errors with Tool Calling in Groq API (qwen-qwq-32b)

  #1233 commented on May 7, 2025 • 0 new comments

• (openai.APITimeoutError) OpenAI requests time out, but Gemini/Anthropic requests work fine

  #774 commented on May 7, 2025 • 0 new comments

• Agents using Gemini models throw error if `result_type` pydantic model contains a `Field` with `examples` defined

  #1167 commented on May 7, 2025 • 0 new comments

• Human in the loop with Agent in Pydantic Graph

  #1607 commented on May 1, 2025 • 0 new comments

• Structured outputs as an alternative to Tool Calling

  #582 commented on May 2, 2025 • 0 new comments

• Support for AWS Profile Names

  #1604 commented on May 2, 2025 • 0 new comments

• Prompt caching support

  #138 commented on May 2, 2025 • 0 new comments

• Notes on structured output strategies

  #1615 commented on May 2, 2025 • 0 new comments

• Add integration to stripe-agent-toolkit

  #1606 commented on May 3, 2025 • 0 new comments

• Expose Agent as a CLI for testing during development

  #1595 commented on May 5, 2025 • 0 new comments

• Amazon Nova (Bedrock) limitations with tool schema

  #1623 commented on May 5, 2025 • 0 new comments

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

  #748 commented on May 6, 2025 • 0 new comments

• [BUG] Error responses are not handled correctly for google openai/openrouter

  #527 commented on May 6, 2025 • 0 new comments

• Gemini Agent run from cached context

  #664 commented on May 6, 2025 • 0 new comments

• Support for returning response directly from tool

  #1189 commented on May 6, 2025 • 0 new comments

• Using Mistral Agents with tool and message_history leads to "Unexpected role 'user' after role 'tool'"

  #1210 commented on May 6, 2025 • 0 new comments

• Reasoning response support

  #907 commented on May 6, 2025 • 0 new comments

• Question on Batching (for Anthropic Models)

  #1144 commented on May 7, 2025 • 0 new comments

• Disable thinking mode in Gemini 2.5 flash

  #1555 commented on May 7, 2025 • 0 new comments

• Adding custom metadata attributes to result

  #981 commented on May 7, 2025 • 0 new comments

• Streaming final output while iterating over agent graph

  #1449 commented on May 7, 2025 • 0 new comments

• Support direct file URL for Gemini models on Vertex AI and GLA

  #1134 commented on May 7, 2025 • 0 new comments

• Add hook to modify messages before they're sent to LLM

  #1596 commented on May 7, 2025 • 0 new comments