PydanticAI — The AI Toolchain

PydanticAI v2.36.0

Sat, 29 Aug 2026 01:25:08 GMT

PydanticAI v2.36.0 adds `@durable_operation` for third-party durable execution, stable `InstructionPart.id`, async-iterable audio input, and `--mcp-config` support in `clai`. • Adds `--mcp-config` flag to the `clai` CLI, enabling MCP server configuration from the command line; also adds tool-call streaming support to `clai`. • Introduces `@durable_operation` decorator with a required explicit operation name, plus a public backend API for integrating third-party durable execution engines. • Gives `InstructionPart` a stable `InstructionPart.id` field, making instruction parts addressable and stable across runs. • Accepts async iterables in RealtimeSession.send_audio(), enabling streaming microphone input from async generators rather than only discrete chunks.

PydanticAI v2.34.0

Tue, 25 Aug 2026 01:47:18 GMT

PydanticAI v2.34.0 adds GLM-5.3 support via `ZhipuModel` and a LangChain migration skill. • Adds a LangChain migration skill to help teams port existing LangChain agents to PydanticAI.

PydanticAI v2.32.0

Wed, 19 Aug 2026 03:51:35 GMT

PydanticAI v2.32.0 adds xAI attachment search lifecycle, OpenRouter web-search annotations, and instrumentation v6 with tool-role emissions. • Surfaces OpenRouter web-search sources in `provider_details["annotations"]` on message objects, making citation data accessible to downstream code. • Adds instrumentation version 6, emitting tool results under `role: 'tool'` for improved observability of tool call/result pairs in traces. • Supports xAI attachment search lifecycle, enabling attachment-based search flows via the xAI provider.

PydanticAI v2.31.0

Sat, 15 Aug 2026 03:04:46 GMT

PydanticAI v2.31.0 lets UIEventStream initialize without a run_input and gives AGUIEventStream its own thread_id/run_id. • Adds support for building a `UIEventStream` without a `run_input`, enabling stream construction before run context is available. • Gives `AGUIEventStream` its own `thread_id` and `run_id` fields for independent stream identification.

PydanticAI v2.31.0

Sat, 15 Aug 2026 03:04:46 GMT

UIEventStream can now be built without a run_input, and AGUIEventStream gets its own thread_id and run_id. • Allows `UIEventStream` to be constructed without a `run_input`, enabling more flexible event stream initialization. • Gives `AGUIEventStream` its own `thread_id` and `run_id` fields for independent stream identity.

PydanticAI v1.107.5

Fri, 14 Aug 2026 02:57:59 GMT

Adds `allowed_hosts` setting to the local dev web UI to prevent DNS rebinding attacks on Agent.to_web() / `clai web`. • Adds `allowed_hosts` setting to the local dev web chat UI (Agent.to_web() / `clai web`) to explicitly permit non-localhost hostnames, required for deployments reached under a real hostname. Breaking changes: • The local dev web chat UI (Agent.to_web(), `clai web`) now validates the `Host` header against localhost/loopback/LAN addresses by default; deployments served under a real (non-local) hostname will be blocked and must opt in with the new `allowed_hosts` setting.

PydanticAI v2.30.0

Fri, 14 Aug 2026 02:57:54 GMT

PydanticAI v2.30.0 adds `allowed_hosts` for the local web UI, OpenRouter web search, Gemini 3.7 Flash, and gRPC metadata on XaiProvider. • Adds `allowed_hosts` setting to Agent.to_web() and `clai web` to explicitly permit non-loopback hostnames when deploying the local dev web chat UI under a real hostname. • Adds support for `openrouter:web_search` as a web search model via the OpenRouter provider. • Adds `gemini-3.7-flash` to the supported Gemini model catalog. • Exposes gRPC `metadata` on `XaiProvider` for passing custom gRPC metadata to xAI endpoints. Breaking changes: • The local dev web chat UI (Agent.to_web(), `clai web`) now validates the `Host` header against localhost/loopback/LAN addresses by default; deployments reached under a real hostname will be blocked unless `allowed_hosts` is explicitly configured.

PydanticAI v2.30.0

Fri, 14 Aug 2026 02:57:54 GMT

PydanticAI v2.30.0 adds `allowed_hosts` for the web UI, OpenRouter web search, Gemini Flash 3.7, and xAI gRPC metadata support. • Adds `allowed_hosts` setting to Agent.to_web() and `clai web` to explicitly permit non-localhost hostnames, required for deployments served under a real hostname (introduced alongside a Host-header validation fix for GHSA-q2xc-rrxj-58x9). • Supports `openrouter:web_search` as a model string for built-in web search via OpenRouter. • Adds `gemini-3.7-flash` to the supported model catalog. • Exposes gRPC `metadata` on `XaiProvider` for passing custom gRPC metadata to xAI endpoints. Breaking changes: • The local dev web chat UI (Agent.to_web(), `clai web`) now validates the `Host` header against localhost/loopback/LAN addresses by default; deployments served under a real hostname will be blocked unless they opt in with the new `allowed_hosts` setting.

PydanticAI v2.29.0

Thu, 13 Aug 2026 04:02:53 GMT

PydanticAI v2.29.0 adds FastMCP 4 / MCP SDK v2 support and Azure AI Voice Live integration. • Adds `azure_voice_live` setting to enable Azure AI Voice Live realtime voice capabilities. • Supports FastMCP 4 and MCP SDK v2 in `MCPToolset` alongside the existing FastMCP 3 compatibility.

PydanticAI v2.29.0

Thu, 13 Aug 2026 04:02:53 GMT

PydanticAI v2.29.0 adds FastMCP 4 / MCP SDK v2 support in `MCPToolset` and Azure AI Voice Live via `azure_voice_live`. • Adds `azure_voice_live` setting to enable Azure AI Voice Live as a realtime voice backend. • Supports FastMCP 4 and MCP SDK v2 in `MCPToolset` alongside the existing FastMCP 3 compatibility.

PydanticAI v2.28.0

Wed, 12 Aug 2026 02:04:43 GMT

PydanticAI v2.28.0 adds real-time speech-to-speech via Agent.realtime() plus a new Crusoe provider • Adds Agent.realtime() method for real-time speech-to-speech interactions with an agent. • Adds browser WebRTC and server sideband support for real-time speech-to-speech sessions. • Adds Crusoe as a new LLM provider. • Adds `cerebras` optional dependency group.

PydanticAI v2.28.0

Wed, 12 Aug 2026 02:04:43 GMT

PydanticAI v2.28.0 adds real-time speech-to-speech via Agent.realtime(), WebRTC sideband support, and a Crusoe model provider. • Adds Agent.realtime() method for real-time speech-to-speech sessions, enabling live voice conversations with tool-calling support across OpenAI Realtime, Gemini Live, Azure, and xAI Grok Voice backends. • Adds browser WebRTC plus server sideband support for real-time speech-to-speech sessions initiated via Agent.realtime(). • Adds a `cerebras` optional dependency group for the Cerebras provider. • Adds Crusoe as a new model provider.

PydanticAI v2.27.0

Sat, 08 Aug 2026 03:51:40 GMT

PydanticAI v2.27.0 adds SnowflakeModel/SnowflakeProvider, xai_agent_count setting, and CompactionPart round-trip for Vercel AI and AG-UI adapters. • Adds `xai_agent_count` field to `XaiModelSettings` for controlling xAI agent concurrency. • Adds `SnowflakeModel` and `SnowflakeProvider` classes for integrating with Snowflake Cortex as an LLM backend. • Supports round-tripping `CompactionPart` through the Vercel AI and AG-UI adapters, preserving compaction state across adapter boundaries.

PydanticAI v2.27.0

Sat, 08 Aug 2026 03:51:40 GMT

PydanticAI v2.27.0 adds Snowflake Cortex support, xAI agent count control, and CompactionPart round-tripping across Vercel AI and AG-UI adapters. • Adds `xai_agent_count` field to `XaiModelSettings` to control the number of xAI agents used per request. • Adds `SnowflakeModel` and `SnowflakeProvider` classes for Snowflake Cortex LLM integration. • Supports round-tripping `CompactionPart` through the Vercel AI and AG-UI adapters, preserving compaction state across adapter boundaries.

PydanticAI v2.26.0

Fri, 07 Aug 2026 03:14:39 GMT

PydanticAI v2.26.0 adds run cancellation, hidden/revealed tools, DeepSeek V4 Flash support, and a public AgentRunEvents handle. • Adds AgentRun.cancel() and RunContext.cancel() for first-party run cancellation, raising `RunCancelled` to stop in-flight agent runs programmatically. • Adds Model.resolve_prompt_cache_retention() to resolve the effective prompt-cache retention setting from model settings. • Promotes the run_stream_events() iterator to a public `AgentRunEvents` handle exposing cancel() and run-state access. • Supports hiding function tools until revealed — via tool search, `load_capability`, or `ToolReturn.tools` — using each provider's native deferral/addition channel. • Covers DeepSeek V4 Flash via `OpenAIResponsesModel` and `DeepSeekProvider`.

PydanticAI v2.26.0

Fri, 07 Aug 2026 03:14:39 GMT

PydanticAI v2.26.0 adds first-party run cancellation, hidden/revealed tools, DeepSeek V4 Flash support, and a public AgentRunEvents handle. • Adds AgentRun.cancel() and RunContext.cancel() methods plus a `RunCancelled` exception for first-party run cancellation. • Adds Model.resolve_prompt_cache_retention() to resolve the effective prompt-cache retention from model settings. • Promotes run_stream_events() to a public `AgentRunEvents` handle with cancel() and run-state access. • Supports hiding function tools until revealed via tool search, `load_capability`, or `ToolReturn.tools`, using each provider's native deferral/addition channel. • Adds DeepSeek V4 Flash support via `OpenAIResponsesModel` and `DeepSeekProvider`.

PydanticAI v2.25.0

Thu, 06 Aug 2026 02:57:33 GMT

PydanticAI v2.25.0 forwards xAI FileSearchTool collection search options. • Forwards xAI `FileSearchTool` collections search options to the underlying API.

PydanticAI v2.25.0

Thu, 06 Aug 2026 02:57:33 GMT

PydanticAI v2.25.0 forwards xAI `FileSearchTool` collection search options to the xAI backend. • Forwards xAI `FileSearchTool` collections search options through to the xAI backend, enabling parameterized file-search collection queries.

PydanticAI v2.23.0

Tue, 04 Aug 2026 01:47:02 GMT

PydanticAI v2.23.0 adds cost tracking with `cost` and `cost_limit`, Bedrock `extra_headers`, and dynamic tool availability parts. • Adds `cost` field to `RunUsage` and `cost_limit` field to `UsageLimits` to track and cap monetary spend per agent run. • Adds `extra_headers` support in `ModelSettings` for Amazon Bedrock requests. • Adds `ToolAvailabilityDeltaPart` with native `tool_addition` and `additional_tools` rendering to represent dynamic tool availability changes in agent message streams.

PydanticAI v2.23.0

Tue, 04 Aug 2026 01:47:02 GMT

PydanticAI v2.23.0 adds cost tracking to RunUsage, a cost_limit to UsageLimits, ToolAvailabilityDeltaPart, and Bedrock extra_headers support. • Adds `cost` field to `RunUsage` and `cost_limit` to `UsageLimits` to track and cap monetary spend per agent run. • Adds `extra_headers` support to `ModelSettings` for Amazon Bedrock requests, matching parity with other providers. • Adds `ToolAvailabilityDeltaPart` with native `tool_addition` and `additional_tools` rendering for streaming tool-availability deltas.

PydanticAI v2.22.0

Sat, 01 Aug 2026 02:27:46 GMT

PydanticAI v2.22.0 adds RunContext.is_tool_available, MCP task-skipping via prefer_tasks, and Gemini VALIDATED tool mode by default. • Adds `RunContext.is_tool_available` method, letting tool code check at runtime whether another named tool is accessible in the current agent context. • Adds `prefer_tasks` parameter to `MCPToolset` clients, allowing optional MCP tasks to be skipped when the runtime does not support them. • Adds configurable `max_retries` to `ToolSearchToolset`, giving control over how many times a tool-search lookup is retried on failure. • Enables Gemini `VALIDATED` tool mode by default on supported models, improving structured tool-call reliability without manual configuration. • Sends mid-conversation system prompts as native `system` messages on Anthropic, aligning prompt delivery with Anthropic's native message format.

PydanticAI v2.22.0

Sat, 01 Aug 2026 02:27:46 GMT

PydanticAI v2.22.0 adds `RunContext.is_tool_available`, `MCPToolset` task-skipping via `prefer_tasks`, and Gemini `VALIDATED` tool mode by default. • Adds `RunContext.is_tool_available` method, letting tool code check at runtime whether another tool is currently available before attempting to call it. • Adds `prefer_tasks` parameter to `MCPToolset` clients, allowing optional MCP tasks to be skipped when not needed. • Adds configurable `max_retries` to `ToolSearchToolset` for controlling retry behavior on tool search failures. • Enables Gemini `VALIDATED` tool mode by default on supported models, improving structured tool-call reliability. • Sends mid-conversation system prompts as native `system` messages on Anthropic models instead of user-turn injections.

PydanticAI v2.21.0

Thu, 30 Jul 2026 01:59:06 GMT

PydanticAI v2.21.0 adds `per_request_input_tokens_limit` to `UsageLimits` for per-call token budgets. • Adds `per_request_input_tokens_limit` field to `UsageLimits` to cap input tokens on a per-request basis, independently of aggregate limits.

PydanticAI v2.21.0

Thu, 30 Jul 2026 01:59:06 GMT

PydanticAI v2.21.0 adds `per_request_input_tokens_limit` to `UsageLimits` for per-request token budgeting. • Adds `per_request_input_tokens_limit` field to `UsageLimits` to cap input tokens on a per-request basis, independently of cumulative session limits.

PydanticAI v2.20.0

Wed, 29 Jul 2026 03:26:46 GMT

PydanticAI v2.20.0 adds Claude Opus 5 and OpenAI Responses API `reasoning.context` support for GPT-5 families. • Adds `reasoning.context` support for the OpenAI Responses API, defaulting to `all_turns`, for the `gpt-5.4`, `gpt-5.5`, and `gpt-5.6` model families. • Adds `claude-opus-5` model support via the Anthropic provider.

PydanticAI v2.20.0

Wed, 29 Jul 2026 03:26:46 GMT

PydanticAI v2.20.0 adds Claude Opus 5 support and OpenAI Responses API `reasoning.context` for the gpt-5.4/5.5/5.6 families. • Adds `reasoning.context` support (default `all_turns`) in the OpenAI Responses API for the `gpt-5.4`, `gpt-5.5`, and `gpt-5.6` model families. • Adds support for `claude-opus-5` via the `anthropic:claude-opus-5` model string.

PydanticAI v2.19.0

Tue, 28 Jul 2026 03:27:21 GMT

PydanticAI v2.19.0 adds `headers` and `retry_after` fields to `ModelHTTPError` for richer HTTP error inspection. • Adds `headers` and `retry_after` attributes to `ModelHTTPError`, populated from all provider SDKs, enabling programmatic inspection of rate-limit and retry signals from HTTP errors.

PydanticAI v2.19.0

Tue, 28 Jul 2026 03:27:21 GMT

PydanticAI v2.19.0 adds `headers` and `retry_after` fields to `ModelHTTPError` across all provider SDKs. • Adds `headers` and `retry_after` attributes to `ModelHTTPError`, populated from all provider SDKs, giving callers direct access to HTTP response headers and rate-limit retry timing from a single exception type.

PydanticAI v2.18.0

Sat, 25 Jul 2026 01:01:03 GMT

PydanticAI v2.18.0 adds AdvisorTool for Anthropic/OpenRouter, BedrockMantleProvider, multi-region Google Cloud, and external web access for OpenAI WebSearchTool. • Adds `external_web_access` option to `WebSearchTool` for OpenAI Responses API, enabling built-in web search without a separate tool. • Adds `BedrockMantleProvider` for AWS Bedrock Mantle integration, with normalized response-scoped tool-call IDs. • Extends `AdvisorTool` support to Anthropic and OpenRouter providers. • Adds `'us'` and `'eu'` multi-region location values to `GoogleCloudProvider` location type.

PydanticAI v2.18.0

Sat, 25 Jul 2026 01:01:03 GMT

PydanticAI v2.18.0 adds AdvisorTool for Anthropic/OpenRouter, BedrockMantleProvider, multi-region Google Cloud, and external_web_access for WebSearchTool. • Adds `external_web_access` option to `WebSearchTool` for OpenAI Responses API, enabling web search grounding on OpenAI Responses-backed agents. • Adds `BedrockMantleProvider` for AWS Bedrock Mantle, with normalized response-scoped tool-call IDs. • Extends `AdvisorTool` support to Anthropic and OpenRouter providers. • Adds `'us'` and `'eu'` multi-region location values to `GoogleCloudProvider` location type.

PydanticAI v2.17.0

Fri, 24 Jul 2026 00:32:49 GMT

PydanticAI v2.17.0 adds arbitrary fields to usage types and caches OTel serialization to eliminate O(n²) instrumentation cost. • Adds support for arbitrary fields on `RequestUsage` and `RunUsage` to accommodate extended provider-specific pricing and metadata. • Caches per-message OpenTelemetry serialization, eliminating O(n²) instrumentation overhead for long conversation traces.

PydanticAI v2.17.0

Fri, 24 Jul 2026 00:32:49 GMT

RequestUsage and RunUsage now accept arbitrary fields; OTel serialization cached to eliminate O(n²) instrumentation cost. • Extends `RequestUsage` and `RunUsage` to accept arbitrary extra fields, enabling support for upcoming genai-prices metadata. • Caches per-message OpenTelemetry serialization to eliminate O(n²) instrumentation overhead on long runs.

PydanticAI v2.16.0

Thu, 23 Jul 2026 02:25:35 GMT

PydanticAI v2.16.0 adds ToolFailed, Model Armor, run_id support, Mistral caching, and OpenAI moderation surface. • Adds `mistral_prompt_cache_key` setting and passes `parallel_tool_calls` to the Mistral SDK via model settings. • Adds `openai_moderation` to `OpenAIChatModelSettings` and exposes Chat Completions moderation results in `provider_details`. • Adds Google Model Armor support for Google Cloud via `GoogleModelSettings`. • Adds optional `run_id=` parameter to agent runs, durable wrappers, and UI adapters for correlating runs. • Adds `ToolFailed` exception class for surfacing model-visible tool failures without triggering retries. • Adds `gemini-3.6-flash` and `gemini-3.5-flash-lite` as supported model identifiers.

PydanticAI v2.16.0

Thu, 23 Jul 2026 02:25:35 GMT

PydanticAI v2.16.0 adds ToolFailed, run_id, Model Armor, Mistral cache keys, OpenAI moderation, and two new Gemini models. • Adds `mistral_prompt_cache_key` setting to `MistralModelSettings` and passes `parallel_tool_calls` through to the Mistral SDK. • Hoists `openai_moderation` into `OpenAIChatModelSettings` and exposes Chat Completions moderation results in `provider_details`. • Adds Model Armor support for Google Cloud via `GoogleModelSettings`. • Adds optional `run_id=` parameter to agent runs, durable wrappers, and UI adapters for stable run identification. • Adds `ToolFailed` exception class for signalling model-visible tool failures without triggering retries. • Adds `gemini-3.6-flash` and `gemini-3.5-flash-lite` as supported model identifiers.

PydanticAI v2.15.0

Wed, 22 Jul 2026 03:43:44 GMT

PydanticAI v2.15.0 adds per-run tool-retry budget overrides, OpenAI moderation, and DynamicCapability toolset support in durable execution. • Adds `openai_moderation` setting to expose OpenAI Responses API moderation results in `provider_details`. • Supports overriding the tool-retry budget at `run`, `iter`, and `override` time, giving per-invocation control over retry limits. • Supports `DynamicCapability` toolsets in durable execution, wrapping `DynamicToolset` in DBOS steps and Prefect tasks. • Adds explicit prompt caching support for `gpt-5.6` in `OpenAIModel`. • Inlines text-like files in `MistralModel` prompts for cleaner multimodal input handling. • Introduces `ExaSearch` capability in Pydantic AI Harness as the successor to the Exa search common tools.

PydanticAI v2.15.0

Wed, 22 Jul 2026 03:43:44 GMT

PydanticAI v2.15.0 adds per-run tool-retry budget overrides, OpenAI moderation settings, and DynamicCapability support in durable execution. • Adds `openai_moderation` setting to expose OpenAI Responses API moderation results in `provider_details`. • Supports overriding the tool-retry budget at `run`, `iter`, and `override` time. • Supports `DynamicCapability` toolsets in durable execution and wraps `DynamicToolset` in DBOS steps and Prefect tasks. • Adds `register_legacy_workflows` to `DBOSDurability` for clean `DBOSAgent` migration. • Supports explicit prompt caching for `gpt-5.6` in `openai` provider. • Inlines text-like files in `MistralModel` prompts. • Adds `ExaSearch` capability in Pydantic AI Harness as the replacement for the deprecated Exa search common tools.

PydanticAI v2.14.0

Tue, 21 Jul 2026 01:15:50 GMT

PydanticAI v2.14.0 adds Mistral reasoning_effort support and three new durability capability classes. • Adds `reasoning_effort` support to the Mistral provider via thinking settings, enabling control over model reasoning depth. • Adds `TemporalDurability`, `DBOSDurability`, and `PrefectDurability` capability classes to replace the deprecated durability wrapper agents.

PydanticAI v2.13.0

Sat, 18 Jul 2026 02:28:48 GMT

PydanticAI v2.13.0 adds instrumentation controls, content-filter error raising, cache-hit ratio tracking, and new capability hooks. • Adds `include_model_request_parameters` instrumentation setting to control whether the `model_request_parameters` span attribute is included in traces. • Adds `RaiseContentFilterError` capability to raise an error when a non-empty content filter response is returned by the model. • Adds `cache_hit_ratio` property to `RequestUsage` and `RunUsage` for tracking cache efficiency across requests and runs. • Adds `get_model`, `resolve_model_id`, and `for_agent` capability hooks for customising model resolution and agent binding.

PydanticAI v2.12.0

Fri, 17 Jul 2026 02:37:27 GMT

PydanticAI v2.12.0 adds Kimi-K3 model support and two new agent stream events for deferred tools and enqueued messages. • Adds `DeferredToolCallEvent` and `DeferredToolResultEvent` to `AgentStreamEvent`, enabling stream-level visibility into deferred tool call lifecycle. • Emits `EnqueuedMessagesEvent` when previously enqueued messages are delivered into a run, making message-replay observable in the event stream. • Adds Moonshot AI `kimi-k3` model support.

PydanticAI v2.11.0

Thu, 16 Jul 2026 02:20:09 GMT

PydanticAI v2.11.0 exports `HistoryProcessor` and adds actionable hints to usage-limit and tool-retry errors. • Exports `HistoryProcessor` from the public API, making it directly importable for custom conversation-history handling. • Adds actionable hint messages to usage-limit and tool-retry errors, surfacing guidance at the point of failure.

PydanticAI v2.10.0

Wed, 15 Jul 2026 01:57:12 GMT

PydanticAI v2.10.0 adds automatic message-history repair and OpenAI background mode plus Anthropic pause-turn support. • Supports OpenAI background mode and handles Anthropic `stop_reason=pause_turn` in agent runs.

PydanticAI v2.9.0

Sat, 11 Jul 2026 03:07:10 GMT

PydanticAI v2.9.0 adds a `/usage` CLI command, GPT-5.6 + reasoning mode support, and `usage_limits` exposure on `RunContext`. • Exposes `usage_limits` on `RunContext` so tools and capabilities can inspect the current run's token/request limits at call time. • Adds `/usage` slash command to the `clai` CLI to display cumulative token usage across a session. • Adds GPT-5.6 models and reasoning mode support to the OpenAI provider.

PydanticAI v2.8.0

Fri, 10 Jul 2026 01:54:17 GMT

PydanticAI v2.8.0 lets to_cli() accept a model override and bumps the bundled chat UI to 2.0.0. • Adds `model` parameter to to_cli() so agents defined without a model can have one supplied at CLI invocation time. • Bumps bundled chat UI to `2.0.0` and targets `sdk_version=7` in Agent.to_web().

PydanticAI v2.7.0

Thu, 09 Jul 2026 01:56:41 GMT

PydanticAI v2.7.0 adds azure-responses shorthand and xAI grok-4.5 model support. • Supports `azure-responses:[model-id]` shorthand for specifying Azure Responses API models. • Adds xAI `grok-4.5` model support.

PydanticAI v2.6.0

Wed, 08 Jul 2026 01:22:35 GMT

PydanticAI v2.6.0 adds time-to-first-token tracking, file uploads to CodeExecutionTool, and new Bedrock model profiles. • Adds `files` support to `CodeExecutionTool` for Anthropic and OpenAI providers, enabling file uploads alongside code execution requests. • Records time-to-first-token for streaming model requests, exposing a new latency metric for streaming runs. • Adds Bedrock model profiles for Writer, Z.AI, and Moonshot AI, and refreshes `LatestBedrockModelNames` with current model listings.

PydanticAI v2.5.0

Sat, 04 Jul 2026 00:35:33 GMT

PydanticAI v2.5.0 adds `sanitize_messages` for message-history hardening and multimodal tool-return round-trips in AG-UI and Vercel AI adapters. • Adds `sanitize_messages` for inbound message-history hardening, closing a dangling-tool-call re-exposure on the Agent.to_ag_ui() / `AGUIAdapter` serving path. • Supports round-trip multimodal tool returns through the AG-UI and Vercel AI adapters, covering both history and streaming paths.

PydanticAI v2.4.0

Fri, 03 Jul 2026 01:24:57 GMT

PydanticAI v2.4.0 adds GEval, five agentic span evaluators, and splits file-upload security controls into two distinct parameters. • Splits `preserve_file_data` into `allow_uploaded_files` (inbound security control) and a separate AG-UI representation opt-in parameter, giving finer-grained control over uploaded file handling. • Adds `GEval` evaluator and standard quality metric rubrics for `LLMJudge`, enabling criteria-driven LLM-as-judge scoring. • Adds five agentic span-based evaluators — `ToolCorrectness`, `TrajectoryMatch`, `ArgumentCorrectness`, `MaxToolCalls`, and `MaxModelRequests` — for evaluating agent execution traces. Breaking changes: • The `preserve_file_data` parameter is split into `allow_uploaded_files` and an AG-UI opt-in; code referencing `preserve_file_data` will break on upgrade.

PydanticAI v2.3.0

Thu, 02 Jul 2026 02:09:16 GMT

PydanticAI v2.3.0 adds a native Z.AI (Zhipu AI) provider with thinking support. • Adds native Z.AI (Zhipu AI) provider with thinking support.

PydanticAI v2.2.0

Wed, 01 Jul 2026 00:01:01 GMT

PydanticAI v2.2.0 adds Claude Sonnet 5 support, retry options for GoogleProvider, factory functions for Dataset.evaluate, and OpenRouter cost fields. • Adds `retry_options` parameter to `GoogleProvider` for configurable retry behavior. • Adds `prompt` and `completions` cost fields to OpenRouter model responses. • Supports `claude-sonnet-5` as a new model identifier for Anthropic Claude Sonnet 5. • Allows factory functions as the `lifecycle` argument in `Dataset.evaluate`, enabling dynamic lifecycle object creation per evaluation run. • Adds a TwelveLabs Pegasus video-understanding integration example.

PydanticAI v2.1.0

Mon, 29 Jun 2026 09:39:33 GMT

PydanticAI v2.1.0 adds Anthropic web tools with server-tool replay, TypeAdapter for EvaluatorContext, and improved instrumentation serialization. • Adds `TypeAdapter` support for `EvaluatorContext`, enabling structured validation and serialization of evaluator context objects. • Adds Anthropic `_20260209` web tools with server-tool replay support for the Anthropic provider. • Serializes instrumentation message attributes using `to_json` instead of `json.dumps` for more robust OpenTelemetry attribute handling.

PydanticAI v2.0.0

Tue, 23 Jun 2026 15:38:17 GMT

PydanticAI v2.0 stable: capabilities primitive, gemini-embedding-2, AG-UI deferred tools, and new model settings. • Adds `xai_max_turns` to `XaiModelSettings` to cap the number of turns for xAI model runs. • Adds `gemini-embedding-2` embedding model support. • Adds `google_task` text-prefix conditioning for `gemini-embedding-2` embeddings to tune retrieval, classification, and other task types. • Maps AG-UI interrupts to `DeferredTools` in `AGUIAdapter`, enabling human-in-the-loop interrupt handling in AG-UI workflows. • Introduces V2 stable with capabilities as a core composable primitive, bundling an agent's tools, hooks, instructions, and model settings into a single unit. • Adds `cerebras_clear_thinking` setting and emits `reasoning_effort='none'` for Cerebras to suppress chain-of-thought output.

PydanticAI v1.107.0

Wed, 10 Jun 2026 14:40:24 GMT

PydanticAI v1.107.0 adds known_model_names(), OpenRouter prompt caching, and two new Claude model aliases. • Adds known_model_names() function to programmatically enumerate all `KnownModelName` members at runtime. • Adds `CachePoint` and prompt caching support for OpenRouter models. • Adds `claude-fable-5` and `claude-mythos-5` as supported model name aliases.

PydanticAI v1.106.0

Fri, 05 Jun 2026 01:16:38 GMT

PydanticAI v1.106.0 adds `api_host`, `timeout`, and `seed` support to `XaiProvider` • Adds `api_host` and `timeout` parameters to `XaiProvider`, enabling custom endpoint and timeout configuration for xAI connections. • Maps the base `seed` setting to xAI via `XaiProvider`, enabling reproducible xAI model outputs.

PydanticAI v1.105.0

Tue, 02 Jun 2026 06:05:25 GMT

PydanticAI v1.105.0 adds on-demand deferred loading for agent capabilities and Grok 4.3 `reasoning_effort` support. • Adds `reasoning_effort` support for Grok 4.3 via xAI model settings, along with updated current xAI model names. • Introduces on-demand (deferred loading) capabilities, allowing instructions, tools, model settings, and hooks to be loaded lazily at runtime.

PydanticAI v1.104.0

Fri, 29 May 2026 04:40:33 GMT

PydanticAI v1.104.0 adds Claude Opus 4.8 model support. • Adds support for Claude Opus 4.8 as a usable model.

PydanticAI v1.103.0

Wed, 27 May 2026 02:37:05 GMT

PydanticAI v1.103.0 adds MCP prompt listing, Vercel timestamp round-tripping, and OpenRouter eager streaming support. • Adds `list_prompts` and `get_prompt` methods to `McpServer`, enabling MCP clients to discover and retrieve prompts from a server. • Supports `anthropic_eager_input_streaming` in `OpenRouterModel`, bringing eager input streaming parity with the native Anthropic model. • Round-trips message timestamps through `VercelAIAdapter`'s `UIMessage.metadata`, preserving original message timing across the adapter boundary.

PydanticAI v1.101.0

Fri, 22 May 2026 04:49:44 GMT

PydanticAI v1.101.0 adds a pending message queue, MCP background tasks, model-agnostic XSearch, and top_k support across three model providers. • Adds `ctx.enqueue` and `agent_run.enqueue` for a pending message queue, enabling mid-run message injection into agent execution. • Adds `top_k` model setting support to `GoogleModel`, `AnthropicModel`, and `CohereModel`. • Adds MCP background task support (SEP-1686) via the `MCPServer` integration. • Makes `XSearch` capability model-agnostic through a subagent fallback, removing the previous model-specific constraint.

PydanticAI v1.100.0

Thu, 21 May 2026 03:52:43 GMT

PydanticAI v1.100.0 adds Bedrock native JSON output and strict tool calls support. • Adds support for Bedrock native JSON output and strict tool calls via the Bedrock integration.

PydanticAI v1.99.0

Wed, 20 May 2026 01:19:19 GMT

PydanticAI v1.99.0 adds support for the `gemini-3.5-flash` model. • Adds `gemini-3.5-flash` as a supported model.

PydanticAI v1.98.0

Tue, 19 May 2026 00:56:30 GMT

PydanticAI v1.98.0 adds OpenAI Responses token counting and a unified `retries` parameter on `Agent`. • Adds `OpenAIResponsesModel.count_tokens` method to count input tokens for OpenAI Responses model calls before sending them. • Replaces `Agent` parameters `tool_retries=` and `output_retries=` with a single `retries: int | AgentRetries` parameter, enabling unified retry control across tools and outputs. Breaking changes: • The `Agent` constructor parameters `tool_retries=` and `output_retries=` are replaced by `retries: int | AgentRetries`; code passing either removed keyword argument will break on upgrade.

PydanticAI v1.97.0

Fri, 15 May 2026 22:15:48 GMT

PydanticAI v1.97.0 adds MCPToolset, OnlineEvaluator error opt-in, streaming state tracking, and splits GoogleProvider into two classes. • Adds `OnlineEvaluator.run_on_errors` flag to opt into running evaluations on failed (errored) agent calls, not just successful ones. • Adds `MCPToolset` (backed by `fastmcp-slim[client]`) as the new MCP integration class, replacing the deprecated `MCPServer*` and `FastMCPToolset`. • Splits GoogleProvider(vertexai=True|False) into two separate classes: `GoogleProvider` (formerly `google-gla:`, now provider ID `google:`) and `GoogleCloudProvider` (formerly `google-vertex:`, now provider ID `google-cloud:`). • Sets `ModelResponse.state` to `incomplete` while a response is still streaming, enabling callers to distinguish in-progress from finished responses. • Promotes `pydantic_graph.beta` API out of beta into the stable namespace. • Adds stream_response() (singular) as the replacement for stream_responses(); the new method yields `ModelResponse` directly instead of a (ModelResponse, is_last) tuple. • Replaces the bundled `fasta2a` A2A integration with an external `fasta2a.pydantic_ai` adapter (requires `fasta2a` v0.6.1+), following DataLayer's adoption of the project. Breaking changes: • The `google-gla:` provider ID is renamed to `google:` and `google-vertex:` is renamed to `google-cloud:`; old names are deprecated and will be removed in v2. • stream_responses() is deprecated in favor of stream_response(); the new singular form yields `ModelResponse` instead of (ModelResponse, is_last), so any code unpacking the tuple will break when migrated. • Agent.to_a2a() and the bundled `fasta2a` integration are deprecated; users must switch to `fasta2a.pydantic_ai` (requires `fasta2a` v0.6.1) from the external package. • The `pydantic_graph.beta` module is deprecated; import paths that relied on the `.beta` namespace must be updated to the stable API.

PydanticAI v1.95.0

Wed, 13 May 2026 02:17:50 GMT

PydanticAI v1.95.0 adds native Tool Search on Anthropic/OpenAI, an `Instrumentation` capability class, and Gemini 3 structured-output + tool combinations. • Adds native Tool Search support for Anthropic and OpenAI providers, with custom search strategies available on any provider. • Introduces the `Instrumentation` capability class; the existing Agent(instrument=...) parameter is now deprecated in favour of `capabilities=[Instrumentation(...)]`. • Renames 'built-in tools' to 'native tools'; native tools are now registered via `capabilities=[NativeTool(...)]`; old fields are deprecated ahead of v2. • Adds `local=` opt-in parameter for provider-adaptive capability fallback; auto-fallback is deprecated. • Supports combining structured output and tool use together for Gemini 3 models via the Google provider.

PydanticAI v1.94.0

Tue, 12 May 2026 06:52:00 GMT

PydanticAI v1.94.0 adds `openai_chat_supports_multiple_system_messages` profile flag for OpenAI chat configuration. • Adds `openai_chat_supports_multiple_system_messages` profile flag to control whether multiple system messages are supported in OpenAI chat requests. Breaking changes: • The `mistralai` package is no longer installed as a dependency of `pydantic-ai`; installations that relied on it being pulled in transitively must now declare it explicitly.

PydanticAI v1.93.0

Sat, 09 May 2026 00:10:13 GMT

PydanticAI v1.93.0 adds `tool_choice` setting and new output tool call events for structured agent control. • Adds `tool_choice` setting to control which tool the model selects during agent runs. • Introduces `OutputToolCallEvent` and `OutputToolResultEvent` stream events for output tool calls, replacing deprecated function-tool events for failing output tool calls.

PydanticAI v1.92.0

Fri, 08 May 2026 01:18:52 GMT

PydanticAI v1.92.0 adds Anthropic task budget support and a runtime `output_retries` override for agents. • Adds runtime `output_retries` override on agent runs, allowing per-call control of output retry counts without reconfiguring the agent; `retries` is now deprecated in favour of `output_retries`. • Adds Anthropic task budget support, enabling token/compute budget constraints on Anthropic-backed agent calls.

PydanticAI v1.91.0

Thu, 07 May 2026 00:15:33 GMT

PydanticAI v1.91.0 adds gpt-image-2 options for OpenAI and support for deepseek-v4-flash and deepseek-v4-pro models. • Supports `gpt-image-2` model options via the OpenAI provider. • Adds `deepseek-v4-flash` and `deepseek-v4-pro` to the DeepSeek provider.

PydanticAI v1.90.0

Tue, 05 May 2026 00:39:51 GMT

PydanticAI v1.90.0 adds OpenAI Conversations API state support and typed OTel metadata for tool call syntax highlighting. • Adds `OpenAIResponsesModelSettings.openai_conversation_id` to persist conversation state across turns using the OpenAI Conversations API. • Adds typed OpenTelemetry metadata for code tool call syntax highlighting, enabling richer tracing of tool invocations.

PydanticAI v1.89.1

Fri, 01 May 2026 18:57:59 GMT

PydanticAI v1.89.1 adds bundled Library Skills for improved coding-agent support. • Adds bundled Library Skills (library-skills.io) to improve coding-agent support and tool discovery.

PydanticAI v1.89.0

Fri, 01 May 2026 03:08:53 GMT

PydanticAI v1.89.0 adds cross-run conversation correlation, dynamic model capabilities, and builtin-tool overrides. • Adds `conversation_id` to enable cross-run correlation, linking multiple agent runs into a single logical conversation. • Adds `builtin_tools` parameter to agent.override(), allowing builtin tools to be overridden at runtime. • Supports dynamic model capabilities via callables in the capabilities list, enabling runtime-evaluated capability flags.

PydanticAI v1.88.0

Wed, 29 Apr 2026 04:31:42 GMT

PydanticAI v1.88.0 adds output validate/process hooks, cross-provider `service_tier`, Anthropic fast mode, and new UI sanitization APIs. • Adds `prepare_output_tools` hook alongside `prepare_tools` — `prepare_tools` is now scoped to function tools only, while output validate/process hooks give fine-grained control over output tool execution. • Adds cross-provider `service_tier` model setting with support for Anthropic, Gemini API, and Vertex Priority PayGo. • Adds `fast` speed mode for Anthropic Opus 4.6. • Adds `UIAdapter.sanitize_messages` and `allowed_file_url_schemes` to the UI adapter for controlling which file URL schemes are permitted in messages. • Supports OpenAI Responses `phase` field on assistant messages. Breaking changes: • `prepare_tools` is now scoped to function tools only; callers relying on it to prepare output tools must migrate to the new `prepare_output_tools` hook.

PydanticAI v1.87.0

Sat, 25 Apr 2026 00:49:36 GMT

PydanticAI v1.87.0 adds deferred tool call handling, event stream processing capability, and GPT-5.5 thinking support. • Adds `HandleDeferredToolCalls` capability and `handle_deferred_tool_calls` hook for handling deferred tool calls in agent workflows. • Adds `ProcessEventStream` capability for processing event streams from model responses. • Supports the thinking setting for GPT-5.5 models.

PydanticAI v1.86.0

Thu, 23 Apr 2026 04:01:15 GMT

PydanticAI v1.86.0 adds `UIAdapter.manage_system_prompt` and `ReinjectSystemPrompt` for dynamic system prompt control in UI adapters. • Adds `UIAdapter.manage_system_prompt` method and `ReinjectSystemPrompt` capability, enabling UI adapters to control and reinject system prompts at runtime.

PydanticAI v1.85.0

Tue, 21 Apr 2026 16:53:23 GMT

PydanticAI v1.85.0 adds online evaluation surfaced through OpenTelemetry events. • Adds online evaluation via OpenTelemetry events, enabling real-time assessment of agent runs as telemetry data.

PydanticAI v1.84.0

Fri, 17 Apr 2026 00:19:40 GMT

PydanticAI v1.84.0 adds Claude Opus 4.7 support, stateful compaction for OpenAI, and a dedicated OllamaModel subclass. • Adds `OllamaModel` subclass with corrected Ollama capability flags, enabling reliable structured output on Ollama Cloud. • Adds stateful compaction mode to `OpenAICompaction` for managing conversation context across long runs. • Adds support for the Claude Opus 4.7 model.

PydanticAI v1.83.0

Thu, 16 Apr 2026 01:23:25 GMT

PydanticAI v1.83.0 adds xAI tool support, FastMCP metadata injection, Bedrock/Anthropic prompt caching, and a graceful parallel-tool end strategy. • Adds `XSearchTool` and `FileSearch` support for the xAI provider. • Adds metadata injection per tool call via `FastMCPToolset`. • Adds prompt cache TTL support for the Bedrock provider. • Adds automatic prompt caching support for Anthropic. • Adds a `'graceful'` end strategy for parallel tool calls. • Supports Agent(output_type=str | None) for optional agent output.

PydanticAI v1.80.0

Fri, 10 Apr 2026 23:20:38 GMT

PydanticAI v1.80.0 adds capability ordering, hooks ordering, and server-side context compaction for OpenAI and Anthropic • Adds `CapabilityOrdering` with relationship descriptors `innermost`, `outermost`, `wraps`, `wrapped_by`, and `requires` to control how capabilities compose and resolve ordering. • Adds an ordering parameter to `Hooks` and supports instance references in `wraps`/`wrapped_by` for finer control over hook execution order. • Adds `OpenAICompaction` and `AnthropicCompaction` capability classes to enable server-side context window compaction for those providers.

PydanticAI v1.79.0

Fri, 10 Apr 2026 01:11:29 GMT

PydanticAI v1.79.0 adds AG-UI 0.1.13/0.1.15 support, a new async HTTP client factory, and apply() on capability classes. • Adds `create_async_http_client` context manager to replace the internal HTTP client cache, giving callers explicit lifecycle control over async HTTP clients. • Adds apply() method to `AbstractCapability`, `CombinedCapability`, and `WrapperCapability`, enabling capabilities to be applied directly. • Adds full AG-UI 0.1.13 and 0.1.15 support, including reasoning, multi-modal messaging, and `dump_messages`.

PydanticAI v1.78.0

Wed, 08 Apr 2026 05:10:46 GMT

PydanticAI v1.78.0 adds `return_schema`, `function_signature`, and `SetToolMetadata` to `ToolDefinition`, plus OTel cached token span attributes. • Adds `return_schema` and `function_signature` fields to `ToolDefinition`, exposing richer tool metadata for inspection and downstream use. • Adds `SetToolMetadata` capability, enabling dynamic mutation of tool metadata at runtime. • Adds cached token span attributes to OTel traces per the OpenTelemetry specification, improving observability of token usage.

PydanticAI v1.77.0

Fri, 03 Apr 2026 02:07:16 GMT

PydanticAI v1.77.0 adds a local WebFetch tool, deferred tool loading, a ThreadExecutor capability, and smart Anthropic/Bedrock instruction caching. • Adds `defer_loading` parameter to tools and toolsets, enabling lazy/deferred tool loading to support tool search workflows. • Adds Agent.using_thread_executor() method and a `ThreadExecutor` capability for running agents in thread executors. • Adds a local `WebFetch` tool that activates automatically when a provider lacks built-in web-fetch support, extending `WebFetch` capability to more providers. • Adds smart instruction caching for Anthropic and Bedrock providers — automatically inserts a cache boundary at the static/dynamic instruction split. • Adds support for `server_message_id` in `VercelAIEventStream`.

PydanticAI v1.76.0

Thu, 02 Apr 2026 00:16:24 GMT

PydanticAI v1.76.0 adds agent self-reference in RunContext and automatic image-generation fallback via subagent. • Adds `agent` attribute to `RunContext`, giving tools and callbacks direct access to the running agent instance. • Adds automatic fallback for `ImageGeneration`: when the main model lacks image-generation capability, PydanticAI transparently delegates to a subagent running a dedicated imagegen model. • Updates the Mistral integration to support mistralai SDK v2.

PydanticAI v1.75.0

Wed, 01 Apr 2026 00:28:57 GMT

PydanticAI v1.75.0 adds Gemini embedding types/limits and Flex PayGo support for Vertex AI. • Adds types and limits for `gemini-embedding-2-preview` in the embeddings module. • Implements support for Flex PayGo billing mode with the Vertex AI provider.

PydanticAI v1.74.0

Tue, 31 Mar 2026 03:05:41 GMT

PydanticAI v1.74.0 adds online evaluation infrastructure, `TextContent` metadata, time-sortable run IDs, and MCP Server instructions support. • Adds `AbstractToolset.get_instructions` method and `include_instructions` argument to MCP Servers, enabling toolsets to surface dynamic instructions to agents. • Adds `TextContent` class for user prompts, supporting a `metadata` field that is attached to the content object but not sent to the model. • Introduces online evaluation infrastructure for `pydantic-evals`, enabling live/online evaluation workflows. • Makes agent run IDs time-sortable and propagates agent name and run ID as span attributes on all agent run child spans, improving observability.

PydanticAI v1.73.0

Fri, 27 Mar 2026 03:40:10 GMT

PydanticAI v1.73.0 adds CaseLifecycle hooks to Dataset.evaluate and lets hooks swap models or trigger retries. • Adds `CaseLifecycle` hooks to `Dataset.evaluate` for lifecycle callbacks around each evaluation case. • Allows before/wrap model request hooks to swap the active model via `ModelRequestContext`. • Allows hooks to raise `ModelRetry` to control retry flow from within hook logic.

PydanticAI v1.72.0

Thu, 26 Mar 2026 01:33:19 GMT

PydanticAI v1.72.0 adds Anthropic eager input streaming, sync tool prep functions, and implicit MCP URLs • Adds `anthropic_eager_input_streaming` to `AnthropicModelSettings` to control eager streaming behaviour for Anthropic models. • Supports synchronous tool preparation functions alongside existing async ones, removing the requirement to define `async def` for tool prep. • Removes the requirement to specify an explicit `url=` argument on the `MCP` capability, both in Python and in `AgentSpec` configuration.

PydanticAI v1.71.0

Tue, 24 Mar 2026 21:50:06 GMT

PydanticAI v1.71.0 adds Capabilities, AgentSpec, Hooks, Thinking, provider-adaptive tools, and two new OpenAI models. • Adds `Agent.from_file` for loading agents from YAML/JSON files, with templated instructions that reference deps via `TemplateStr`. • Adds `AbstractToolset.for_run` and `for_run_step` methods for per-run and per-step state isolation in toolsets. • Adds `Capabilities`: composable, reusable units of agent behavior that bundle tools, lifecycle hooks, instructions, and model settings into a single class pluggable into any agent. • Adds `AgentSpec` for declarative agent definitions loadable from YAML/JSON. • Adds `Hooks` capability for defining lifecycle hooks using decorators. • Adds `Thinking` capability and a cross-provider `thinking` model setting. • Adds provider-adaptive tool capabilities `WebSearch`, `WebFetch`, `MCP`, and `ImageGeneration` that automatically fall back from builtin (provider) tools to local tools. • Adds `openai:gpt-5.4-mini` and `openai:gpt-5.4-nano` model identifiers.

PydanticAI v1.70.0

Wed, 18 Mar 2026 04:15:30 GMT

PydanticAI v1.70.0 adds `bedrock_inference_profile` to Bedrock model and embedding settings. • Adds `bedrock_inference_profile` field to `BedrockModelSettings` and `BedrockEmbeddingSettings`, enabling inference profile selection for AWS Bedrock model and embedding calls.

PydanticAI v1.69.0

Tue, 17 Mar 2026 02:44:07 GMT

PydanticAI v1.69.0 adds agent descriptions for tracing, multimodal tool results, and response-based FallbackModel support. • Adds response-based fallback support to `FallbackModel`, enabling fallback logic driven by the model response rather than only on errors. • Sends multimodal tool results to APIs directly as a single part instead of splitting them into user parts.

PydanticAI v1.67.0

Fri, 06 Mar 2026 22:31:21 GMT

PydanticAI v1.67.0 adds GPT-5.4 support, WebSearchTool for OpenRouter, a Tavily search overhaul, and native structured output for Ollama. • Supports `WebSearchTool` for `OpenRouterModel` via OpenRouter plugins, enabling web search through the OpenRouter provider. • Enables native structured output support for the Ollama provider. • Adds GPT-5.4 model support. • Rehauled `TavilySearchTool` with updated internals and capabilities.

PydanticAI v1.66.0

Thu, 05 Mar 2026 00:45:49 GMT

PydanticAI v1.66.0 adds native structured output for Qwen 3.5 models and support for the Gemini image-preview model. • Enables native structured output for Qwen 3.5 models. • Adds support for `gemini-3.1-flash-image-preview` (Nano Banana 2) as a supported model.

PydanticAI v1.65.0

Tue, 03 Mar 2026 23:37:18 GMT

PydanticAI v1.65.0 adds provider-uploaded file support via `UploadedFile` and the `gemini-3.1-flash-lite-preview` model. • Adds `UploadedFile` object to support files uploaded directly to providers, enabling agents to work with provider-hosted file references. • Adds `gemini-3.1-flash-lite-preview` as a supported model identifier.

PydanticAI v1.64.0

Mon, 02 Mar 2026 23:45:27 GMT

PydanticAI v1.64.0 adds `template=False` on `PromptedOutput` and `NativeOutput` to suppress schema prompts. • Adds `template=False` parameter to `PromptedOutput` and `NativeOutput` to disable automatic schema prompt injection when you want full control over the model prompt.

PydanticAI v1.63.0

Mon, 23 Feb 2026 17:47:53 GMT

PydanticAI v1.63.0 adds `args_validator` for tools, Gemini 2.5 Pro Preview support, and Gemini logprob output. • Adds `args_validator` parameter to tool definitions for pre-execution argument validation before a tool runs. • Adds logprob support for Gemini models.

PydanticAI v1.62.0

Thu, 19 Feb 2026 04:47:51 GMT

PydanticAI v1.62.0 adds tool approval for Vercel AI, plus LinePlot, ROCAUCEvaluator, and KolmogorovSmirnovEvaluator. • Adds `LinePlot` analysis type, `ROCAUCEvaluator`, and `KolmogorovSmirnovEvaluator` for evaluating model outputs with statistical analysis. • Adds tool approval integration for the Vercel AI adapter, enabling human-in-the-loop approval flows for tool calls.

PydanticAI v1.61.0

Wed, 18 Feb 2026 01:32:10 GMT

PydanticAI v1.61.0 adds Python 3.14 support and Claude Sonnet 4.6 via updated Anthropic SDK. • Supports Python 3.14. • Adds Claude Sonnet 4.6 model availability via Anthropic SDK upgrade to 0.80.0.

PydanticAI v1.60.0

Tue, 17 Feb 2026 00:21:38 GMT

PydanticAI v1.60.0 adds video URL support to OpenRouterModel and upgrades OTel instrumentation for multimodal input. • Adds `video_url` support to `OpenRouterModel`, enabling video content to be passed as multimodal input through OpenRouter. • Upgrades instrumentation to version 4 to align with OTel GenAI semantic conventions for multimodal input.

PydanticAI v1.59.0

Sat, 14 Feb 2026 00:09:25 GMT

PydanticAI v1.59.0 adds Model.model_id, aggregated usage flag, BaseModel support in Contains, and Vercel AI metadata injection • Adds `Model.model_id` property that returns the model identifier in `provider:model` format. • Adds opt-in flag for aggregated usage attribute names. • Enhances `Contains` evaluator to support `pydantic.BaseModel` instances as expected values. • Allows `BaseChunk`s to be injected into the Vercel AI adapter through `ToolReturnPart.metadata`.

PydanticAI v1.58.0

Wed, 11 Feb 2026 01:34:25 GMT

PydanticAI v1.58.0 adds report-level evaluators, multi-run aggregation, and extra_headers for Google provider • Adds `extra_headers` support to the Google model provider, enabling custom HTTP headers on requests. • Adds a `repeat` parameter to pydantic-evals for multi-run aggregation, enabling statistical analysis across repeated experiment runs. • Introduces report-level evaluators and experiment-wide analyses to pydantic-evals, enabling summary metrics across all cases in an evaluation run.

PydanticAI v1.56.0

Fri, 06 Feb 2026 01:01:31 GMT

PydanticAI v1.56.0 adds Claude Opus 4.6 support, adaptive thinking, and new Anthropic model settings fields. • Adds `anthropic_effort` and `anthropic_thinking.type='adaptive'` to Anthropic model settings, enabling adaptive extended thinking for Claude models. • Adds `anthropic_betas` field to `AnthropicModelSettings`, allowing opt-in to Anthropic beta features per request. • Adds support for Claude Opus 4.6 as a new model option.

PydanticAI v1.54.0

Wed, 04 Feb 2026 15:32:52 GMT

PydanticAI v1.54.0 adds concurrency limiting for Agents and Models • Adds concurrency limiting for Agents and Models to cap parallel executions and prevent resource exhaustion.

PydanticAI v1.53.0

Wed, 04 Feb 2026 11:15:06 GMT

PydanticAI v1.53.0 automatically infers the gateway base URL from the token region. • Automatically infers the gateway base URL from the token region, removing the need to manually specify it.

PydanticAI v1.52.0

Tue, 03 Feb 2026 00:38:59 GMT

PydanticAI v1.52.0 adds OpenAI data-retention control, retry counts in run context, and reasoning-content passthrough. • Adds `openai_store` setting to `OpenAIChatModel` to control whether OpenAI retains request/response data. • Exposes the number of output-validation retries in the agent's run context, making retry count available to tool and result handlers. • Makes `OpenAIChatModel` return reasoning content via the same field it was received in, preserving round-trip fidelity of reasoning tokens.

PydanticAI v1.51.0

Sat, 31 Jan 2026 01:55:09 GMT

PydanticAI v1.51.0 adds `html_source` parameter to customize the Chat UI source. • Adds `html_source` parameter to the Chat UI to allow customization of the HTML source rendered by the chat interface.

PydanticAI v1.50.0

Thu, 29 Jan 2026 23:51:01 GMT

PydanticAI v1.50.0 exposes usage limits and model settings to CLI users and adds OpenAI raw text annotation access. • Adds `usage_limits` and `model_settings` parameters accessible to users running agents with to_cli(), enabling runtime control of limits and settings from the command line. • Adds a setting to include OpenAI raw text annotations in `TextPart.provider_details`, surfacing provider-level annotation data to callers.

PydanticAI v1.49.0

Thu, 29 Jan 2026 01:08:52 GMT

PydanticAI v1.49.0 adds BedrockEmbeddingModel for Nova/Cohere/Titan and parallel tool calls in DBOSAgent. • Adds `BedrockEmbeddingModel` class supporting AWS Bedrock embedding endpoints for Nova, Cohere, and Titan models. • Enables parallel tool call execution in `DBOSAgent`. • Updates Vercel AI SDK type definitions to match AI SDK v6.

PydanticAI v1.48.0

Tue, 27 Jan 2026 23:58:15 GMT

PydanticAI v1.48.0 adds domain allowlisting for WebSearchTool, continuous usage stats for OpenAI, and model_settings support for Mistral streaming. • Adds `allowed_domains` parameter to `WebSearchTool` to restrict OpenAI web searches to specific domains. • Adds `continuous_usage_stats` model setting for OpenAI to receive token usage statistics on every streamed chunk. • Applies `model_settings` to Mistral streaming JSON mode, enabling per-request model configuration during Mistral streaming.

PydanticAI v1.47.0

Sat, 24 Jan 2026 00:33:18 GMT

PydanticAI v1.47.0 preserves thought signatures and provider metadata through Vercel AI frontend round trips. • Thought signatures and other provider metadata now survive a round trip through a Vercel AI frontend.

PydanticAI v1.46.0

Thu, 22 Jan 2026 23:55:39 GMT

PydanticAI v1.46.0 adds a native xAI SDK model class, replacing the OpenAI-compatible Grok provider. • Adds `XaiModel` class that integrates with the xAI SDK natively, replacing the deprecated `GrokProvider` which relied on the OpenAI-compatible API. Breaking changes: • `GrokProvider` is deprecated; callers should migrate to `XaiModel` which uses the xAI SDK directly.

PydanticAI v1.45.0

Thu, 22 Jan 2026 01:36:58 GMT

PydanticAI v1.45.0 adds VoyageAI embeddings support. • Adds VoyageAI embeddings support as a new integration.

PydanticAI v1.44.0

Sat, 17 Jan 2026 01:15:16 GMT

PydanticAI v1.44.0 adds Exa search tools integration and AWS Bedrock Nova 2.0 Code Interpreter support. • Adds Exa search tools integration, enabling agents to perform web search via the Exa API as a built-in tool. • Adds support for the AWS Bedrock Nova 2.0 built-in `Code Interpreter` tool, allowing agents backed by Bedrock Nova 2.0 to execute code natively.

PydanticAI v1.43.0

Thu, 15 Jan 2026 23:57:08 GMT

PydanticAI v1.43.0 adds support for Google embedding models. • Supports Google embedding models, enabling text embedding workflows via the Google provider.

PydanticAI v1.42.0

Wed, 14 Jan 2026 01:10:16 GMT

PydanticAI v1.42.0 adds SambaNova provider, ContentFilterError for empty responses, and OTel GenAI semantic attributes. • Raises `ContentFilterError` consistently when a model returns an empty response due to a content filter, giving callers a typed exception to catch. • Adds SambaNova as a supported provider. • Adds OpenTelemetry GenAI semantic convention attributes to telemetry output.

PydanticAI v1.41.0

Sat, 10 Jan 2026 02:38:46 GMT

PydanticAI v1.41.0 adds YAML/TOML media type support in BinaryContent and metadata for DeferredToolResults. • Adds YAML and TOML media type support to `BinaryContent.from_path`, enabling those file types to be loaded as binary content. • Adds `metadata` support for `DeferredToolResults`, allowing metadata to be attached to deferred tool result objects.

PydanticAI v1.40.0

Wed, 07 Jan 2026 01:25:13 GMT

PydanticAI v1.40.0 adds human-readable Temporal activity summaries and configurable retries for nonexistent tool calls. • Agents now retry calls to nonexistent tools up to the `Agent` `retries` limit, matching the existing retry behavior for real tools. • Sets human-readable activity summaries for Temporal activities, improving observability in Temporal-based agent workflows.

PydanticAI v1.39.0

Wed, 24 Dec 2025 03:19:28 GMT

PydanticAI v1.39.0 adds embedding model support, agent run metadata on results and spans, and a new BedrockModelSettings tier field. • Adds `bedrock_service_tier` setting to `BedrockModelSettings` for controlling AWS Bedrock service tier per agent run. • Adds agent and agent run metadata, exposed on result objects and OpenTelemetry span attributes. • Introduces embedding model support via new embedding model classes and APIs. • Supports `ThinkingPart` in MCP Sampling, enabling reasoning-aware model responses over the Model Context Protocol. • Allows system prompt functions to return `None`, treating it as a no-op rather than an error.

PydanticAI v1.38.0

Tue, 23 Dec 2025 00:34:08 GMT

PydanticAI v1.38.0 adds local timestamps to request/response models and typed RunContext support in TextOutput signatures. • Adds typed `RunContext[Deps]` support in `TextOutput` function signatures, enabling dependency-injected output handlers. • Adds local timestamps to request and response models, with provider timestamps surfaced in `provider_details`. • Supports `VideoUrl.vendor_metadata` for GCS URIs on the Google Vertex provider.

PydanticAI v1.37.0

Sat, 20 Dec 2025 02:00:36 GMT

PydanticAI v1.37.0 adds runtime model switching and DynamicToolset for TemporalAgent, plus Vertex AI image output controls. • Adds `output_compression` and `output_format` parameters to `ImageGenerationTool` for Vertex AI Gemini image models. • Enables `TemporalAgent` to switch model at `agent.run`-time, allowing per-run model selection. • Adds `DynamicToolset` support in Temporal, enabling runtime-defined tool sets for Temporal workflows. • Adds a model profile flag for APIs that support native output but still require JSON schema in instructions. • Updates known Groq model names to add production/preview variants and remove deprecated entries. • Sets a configurable message on `ToolRetryError` for clearer retry error reporting.

PydanticAI v1.35.0

Wed, 17 Dec 2025 23:59:40 GMT

PydanticAI v1.35.0 adds FileSearchTool, DashScopeProvider, AG-UI multimodal messages, and Gemini 3 Flash support. • Adds `FileSearchTool` class with support for OpenAI and Google backends, enabling file-search capabilities within PydanticAI agents. • Adds `DashScopeProvider` for Alibaba Cloud, plus audio input support for Qwen Omni models. • Adds `size` parameter to `ImageGenerationTool` for Gemini image models, controlling generated image dimensions. • Supports OpenAI reasoning summary option `'auto'` for reasoning-capable models. • Adds Gemini 3 Flash model support. • Supports AG-UI multi-modal messages, enabling richer message types in AG-UI integrations. • Sets timestamps on AG-UI events for improved event traceability.

PydanticAI v1.34.0

Tue, 16 Dec 2025 23:44:51 GMT

PydanticAI v1.34.0 adds a Web Chat UI launchable via `clai web` or Agent.to_web(), plus `FileUrl.force_download` support for Anthropic and OpenAI Responses models. • Adds `clai web` CLI command and Agent.to_web() method to launch a Web Chat UI for any agent. • Supports `FileUrl.force_download` in `AnthropicModel` and `OpenAIResponsesModel` for forced file downloads. • Makes `OpenRouterProvider` and `DeepSeekProvider` `__init__` overloads less restrictive, broadening valid initialization patterns.

PydanticAI v1.33.0

Tue, 16 Dec 2025 00:16:39 GMT

PydanticAI v1.33.0 adds native s3:// URL support in BedrockConverseModel and broadens instructions support across models. • Passes `s3://` file URLs directly to the API in `BedrockConverseModel`, enabling S3-hosted files to be referenced without pre-fetching. • Inserts agent `instructions` after `system_prompt`s for models that don't natively support instructions, broadening the operational surface of the `instructions` field across providers.

PydanticAI v1.32.0

Sat, 13 Dec 2025 00:26:32 GMT

PydanticAI v1.32.0 adds tool timeouts, multi-agent Temporal workflow registration, and OTel log-based observability. • Adds tool timeout support, allowing individual tools to be given a maximum execution duration. • Allows `TemporalAgent`s to be registered to a Temporal workflow via the `__pydantic_ai_agents__` field on a workflow class. • Extends `end_strategy` to apply to output tools in addition to regular tools, giving consistent early-exit behaviour across both tool types. • Replaces OpenTelemetry events with OTel logs for agent observability, aligning with the OTel logging data model. Breaking changes: • OTel events emitted by PydanticAI are replaced with OTel logs; any pipeline or backend that consumes the old event format will no longer receive those signals.

PydanticAI v1.31.0

Fri, 12 Dec 2025 00:31:36 GMT

PydanticAI v1.31.0 adds prompt caching for AWS Bedrock, Agent.output_json_schema(), custom MCP clientInfo, and GPT-5.2 support. • Adds Agent.output_json_schema() method to retrieve the JSON schema for an agent's output type programmatically. • Adds `provider_url` field to `ModelResponse`, used by cost() to route cost calculations correctly across providers. • Adds prompt caching support for AWS Bedrock, reducing latency and token costs on repeated prompts. • Allows custom `clientInfo` when connecting to MCP servers, enabling clients to self-identify to MCP endpoints. • Allows model to be passed as a plain string in `LLMJudge`, simplifying evaluation setup. • Adds support for GPT-5.2 and bumps the OpenAI dependency to v2.11.0.

PydanticAI v1.30.0

Thu, 11 Dec 2025 00:28:38 GMT

PydanticAI v1.30.0 adds CerebrasModel, prompt caching options for OpenAI, and multi-modal output in LLMJudge. • Adds `CerebrasModel` as a new supported LLM provider integration. • Adds prompt caching options to `OpenAIChatModelSettings` for controlling OpenAI prompt cache behavior. • Supports multi-modal output in `LLMJudge` evaluations.

PydanticAI v1.29.0

Wed, 10 Dec 2025 00:23:36 GMT

PydanticAI v1.29.0 adds aspect ratio support for Gemini image generation and passes `container_id` to the Anthropic API. • Passes `container_id` back to the Anthropic API, enabling container-scoped tool interactions. • Adds aspect ratio support for Gemini image generation. • Removes the requirement for the `anthropic` dependency when using an Anthropic model through a third-party provider.

PydanticAI v1.28.0

Tue, 09 Dec 2025 00:28:57 GMT

PydanticAI v1.28.0 adds structured output for claude-haiku-4-5 and multi-character Bedrock geo-prefix support. • Adds native structured output support for `claude-haiku-4-5`. • Supports `us-gov.` and other multi-character AWS Bedrock geo prefixes.

PydanticAI v1.27.0

Fri, 05 Dec 2025 03:07:19 GMT

PydanticAI v1.27.0 adds dynamic built-in tool config via RunContext, MCP tool/resource caching, CoT reasoning support, and VercelAIAdapter message conversion. • Adds VercelAIAdapter.dump_messages() method to convert PydanticAI messages to Vercel AI message format. • Supports tool and resource caching for MCP servers that emit change notifications, reducing redundant round-trips. • Enables dynamic runtime configuration of built-in tools via `RunContext`, allowing per-run tool behavior without rebuilding agents. • Supports raw Chain-of-Thought (CoT) reasoning output from LM Studio and other OpenAI Responses-compatible APIs. • Uses a custom reasoning field for OpenRouter to surface model reasoning traces.

PydanticAI v1.26.0

Wed, 03 Dec 2025 01:26:12 GMT

PydanticAI v1.26.0 adds Grok models, custom OpenAI reasoning fields, Deepseek JSON output, and gateway model name support. • Adds custom reasoning field support to OpenAI model profiles, enabling configuration of reasoning behaviour for compatible models. • Adds `gateway/...:...` pattern to known model names, allowing gateway-routed models to be referenced by name without custom setup. • Supports JSON object output for the Deepseek provider, enabling structured response parsing from Deepseek models. • Adds latest Grok (xAI) models to the supported model list. • Automatically omits `TTL` from `cache_control` when `AnthropicModel` is used with a Bedrock client, preventing unsupported-field errors.

PydanticAI v1.25.0

Fri, 28 Nov 2025 04:49:33 GMT

PydanticAI v1.25.0 adds support for `gemini-3-pro-image-preview` and improved Google tool error reporting. • Adds support for the `gemini-3-pro-image-preview` model. • Returns tool errors to Google in the `error` key, enabling structured error feedback in Google model integrations.

PydanticAI v1.24.0

Thu, 27 Nov 2025 00:04:54 GMT

PydanticAI v1.24.0 adds native JSON output for Anthropic, Pydantic validation context, logprobs from Responses API, and instructions-only agent runs. • Supports instructions-only agent runs with `OpenAIResponsesModel`, enabling prompts without user message content. • Adds native JSON output and strict tool calls for Anthropic models. • Supports Pydantic validation context, allowing contextual data to be passed into validators during model output parsing. • Supports logprobs output from the OpenAI Responses API.

PydanticAI v1.23.0

Wed, 26 Nov 2025 01:22:41 GMT

PydanticAI v1.23.0 adds Anthropic WebFetchTool, cache-message settings, HITL user prompts, and Gemini 3 Pro via OpenRouter. • Adds `anthropic_cache_messages` model setting for Anthropic models, with automatic stripping of cache points that exceed the provider limit. • Adds support for Anthropic's built-in `WebFetchTool`, enabling web-fetch capability natively through the Anthropic provider. • Allows `user_prompt` to be supplied in Human-in-the-Loop (HITL) interactions. • Adds Gemini 3 Pro support to `OpenRouterModel`. • Ensures the `openrouter_reasoning` model setting is correctly forwarded to the OpenRouter API.

PydanticAI v1.22.0

Sat, 22 Nov 2025 00:22:41 GMT

PydanticAI v1.22.0 adds OpenRouterModel, broadens FallbackModel error handling, and extends Anthropic caching support. • Adds `OpenRouterModel` as an `OpenAIChatModel` subclass with additional feature support for the OpenRouter API. • Expands `FallbackModel` to fall back on all model API errors, not only HTTP 4xx+ status responses. • Adds `document` to the allowed `cacheable_types` for Anthropic, enabling document caching.

PydanticAI v1.21.0

Fri, 21 Nov 2025 01:24:36 GMT

PydanticAI v1.21.0 adds MCP client resource support, server instructions exposure, and a BinaryContent path loader. • Adds `BinaryContent.from_path` convenience method for loading binary content directly from a file path. • Exposes MCP server instructions via the `MCPServer.instructions` property. • Adds MCP client Resources support, enabling agents to read resources from MCP servers. • Enforces that message history always starts with a user message. • Always strips Markdown fences from structured output, improving reliability of parsed responses. Breaking changes: • Message history that does not start with a user message is now rejected — any existing code passing histories beginning with a non-user message will break.

PydanticAI v1.20.0

Wed, 19 Nov 2025 00:36:05 GMT

PydanticAI v1.20.0 adds Gemini 3 Pro support, metadata fields on model messages, TTL for Anthropic cache, and enhanced Gemini JSON Schema features. • Adds `ModelRequest.metadata` and `ModelResponse.metadata` fields for attaching arbitrary metadata to model messages. • Adds `ttl` field to `CachePoint` and Anthropic caching model settings, enabling control over cache entry lifetime. • Adds support for Gemini 3 Pro via `GoogleModel`. • Supports Gemini enhanced JSON Schema features when using `GoogleModel`. • Makes `RunContext.usage` available in Temporal workflow contexts. • Wraps `google.genai.errors.APIError` in `ModelHTTPError` so `GoogleModel` errors are handled correctly by `FallbackModel`. • Extracts Google model usage metrics using genai-prices for more accurate token cost tracking.

PydanticAI v1.19.0

Tue, 18 Nov 2025 00:47:31 GMT

PydanticAI v1.19.0 adds metadata passthrough to deferred tool exceptions and Anthropic token-counting support. • Adds `count_tokens` method to `AnthropicModel` for explicit token counting. • Adds support for `UsageLimits.count_tokens_before_request` with `AnthropicModel`, enabling pre-flight token budget checks. • Allows `metadata` to be passed to `CallDeferred` and `ApprovalRequired` exceptions, propagating it onto `DeferredToolRequests`.

PydanticAI v1.18.0

Fri, 14 Nov 2025 23:54:10 GMT

PydanticAI v1.18.0 adds Anthropic prompt caching support and recognizes GPT-5.1 as a known model name. • Adds Anthropic prompt caching support. • Adds `gpt-5.1` to the list of known OpenAI model names; bumps `openai` dependency to v2.8.0 (v1 still supported). • Bumps `temporalio` to v1.19.0 and adopts `SimplePlugin`.

PydanticAI v1.17.0

Fri, 14 Nov 2025 00:26:01 GMT

PydanticAI v1.17.0 adds Temporal support for FastMCPToolset and environment variable expansion in mcp.json • Supports environment variable expansion inside `mcp.json` when loading servers via load_mcp_servers(). • Enables `FastMCPToolset` to work with `Temporal` for durable, workflow-based MCP tool execution.

PydanticAI v1.15.0

Thu, 13 Nov 2025 00:53:41 GMT

PydanticAI v1.15.0 adds run IDs across run/message classes and token-counting support for BedrockConverseModel. • Adds `BedrockConverseModel.count_tokens` method, enabling `UsageLimits.count_tokens_before_request` to work with Bedrock-backed agents. • Adds a unique `run_id` field to run, run result, and message (request and response) classes for correlating events across a single agent run. • Wraps `BedrockConverseModel` errors in `ModelHTTPError`, making Bedrock failures handled correctly when used inside a `FallbackModel`.

PydanticAI v1.14.0

Mon, 10 Nov 2025 23:30:14 GMT

PydanticAI v1.14.0 allows custom provider factories to be passed into `infer_model`. • Supports passing a custom provider factory into `infer_model`, enabling user-defined model resolution logic.

PydanticAI v1.13.0

Mon, 10 Nov 2025 16:56:37 GMT

PydanticAI v1.13.0 adds AgentRun message accessors and new gateway config fields for API type, profile, and routing group. • Adds AgentRun.all_messages(), AgentRun.new_messages(), AgentRun.all_messages_json(), and AgentRun.new_messages_json() methods to retrieve accumulated or incremental messages from an agent run. • Adds `api_type` support to the gateway integration, enabling selection of the backend API type via gateway config. • Adds `profile` and `routing_group` support to the gateway integration, enabling fine-grained routing control. • Expands known model lists for Cerebras and Heroku providers.

PydanticAI v1.11.1

Thu, 06 Nov 2025 00:34:40 GMT

PydanticAI v1.11.1 adds FallbackModel support for Native output mode and ModelProfile.default_structured_output_mode • Adds `ModelProfile.default_structured_output_mode` support to `FallbackModel`, enabling native output mode control when falling back across models.

PydanticAI v1.11.0

Wed, 05 Nov 2025 03:56:49 GMT

PydanticAI v1.11.0 adds runtime instructions to agent.run() and partial_output access in output validators. • Adds `instructions` parameter to agent.run(), allowing additional instructions to be injected at call time without reconfiguring the agent. • Adds `partial_output` field to `RunContext` supplied to output validators, exposing the partially-constructed output during validation.

PydanticAI v1.10.0

Tue, 04 Nov 2025 04:13:46 GMT

PydanticAI v1.10.0 adds synchronous streaming via `Agent.run_stream_sync`, `application/msword` file detection, and `OpenAIResponsesModel.base_url`. • Adds `Agent.run_stream_sync` method and synchronous convenience methods on `StreamedRunResult` for consuming streamed agent runs without an async runtime. • Implements `OpenAIResponsesModel.base_url` property, exposing the configured base URL on the responses model. • Adds support for detecting and handling `application/msword` files as agent inputs.

PydanticAI v1.9.1

Fri, 31 Oct 2025 02:29:32 GMT

PydanticAI v1.9.1 adds AsyncAnthropicVertex support and makes AG-UI frontend state readable from on_complete handlers. • Supports `AsyncAnthropicVertex` as the value for `AnthropicProvider.anthropic_client`, enabling async Anthropic Vertex AI usage. • Sets AG-UI frontend state directly on the provided `deps` object so it can be read from the `on_complete` handler.

PydanticAI v1.9.0

Wed, 29 Oct 2025 22:50:46 GMT

PydanticAI v1.9.0 adds support for the Vercel AI Data Stream Protocol. • Supports the Vercel AI Data Stream Protocol, enabling PydanticAI agents to stream responses in a format compatible with Vercel AI SDK consumers.

PydanticAI v1.8.0

Wed, 29 Oct 2025 15:21:05 GMT

PydanticAI v1.8.0 adds experiment metadata support and honors `openai_supports_tool_choice_required` in OpenAI Responses models. • Adds `openai_supports_tool_choice_required` model profile setting support in `OpenAIResponsesModel`, enabling correct tool-choice behavior for OpenAI-compatible endpoints that do or don't support required tool choice. • Adds experiment metadata support via the new experiment metadata API.

PydanticAI v1.7.0

Tue, 28 Oct 2025 00:43:29 GMT

PydanticAI v1.7.0 adds OutlinesModel for running local LLMs via Transformers, Llama.cpp, MLXLM, SGLang, and vLLM • Adds `OutlinesModel` class to run local models through the Outlines library, supporting Transformers, Llama.cpp, MLXLM, SGLang, and vLLM backends.

PydanticAI v1.6.0

Fri, 24 Oct 2025 22:40:10 GMT

PydanticAI v1.6.0 adds FastMCPToolset and a vLLM Responses API compatibility flag for OpenAI model profiles. • Adds `OpenAIModelProfile.openai_responses_requires_function_call_status_none` flag to enable compatibility with vLLM's Responses API. • Adds `FastMCPToolset` for integrating FastMCP tools into PydanticAI agents. • Auto-generated output tool names are now sanitized to support generic types.

PydanticAI v1.5.0

Fri, 24 Oct 2025 15:37:17 GMT

PydanticAI v1.5.0 introduces a beta graph API and improves OTel span naming for non-default backends. • Introduces a new graph API in beta for building agent execution graphs. • Pre-formats run graph and node span names for compatibility with non-Logfire OTel backends.

PydanticAI v1.4.0

Fri, 24 Oct 2025 03:12:15 GMT

PydanticAI v1.4.0 adds native MCP server support for OpenAI and Anthropic via the built-in `MCPServerTool`. • Adds `MCPServerTool` built-in tool to support OpenAI and Anthropic native MCP (Model Context Protocol) server integration. • Raises a clear error when a Google content filter produces an empty response, making content-filter failures visible instead of silent.

PydanticAI v1.3.0

Wed, 22 Oct 2025 23:56:50 GMT

PydanticAI v1.3.0 adds AWS Bedrock gateway support, OVHcloud provider, IncompleteToolCall errors, and expanded OTel attributes. • Raises `IncompleteToolCall` exception when a token limit is reached mid-generation of a tool call, giving callers a typed signal to handle truncated tool invocations. • Adds `http_client` option to `GoogleProvider` and adds `api_key` support for Vertex AI; uses PydanticAI's cached httpx client by default. • Uses `gateway/:` as the provider name prefix for Gateway model references. • Adds AWS Bedrock support to the PydanticAI Gateway. • Adds OVHcloud AI Endpoints as a new provider. • Makes `AbstractBuiltinTool` serializable and compatible with durable execution workflows. • Includes eval report averages in OpenTelemetry span attributes. • Includes all usage fields (beyond token counts) in OpenTelemetry span attributes. • Ensures toolset spans (e.g. MCP sampling) are nested under the agent run span in traces.

PydanticAI v1.2.0

Mon, 20 Oct 2025 12:08:51 GMT

PydanticAI v1.2.0 adds Claude Haiku 4.5 support and genai-prices-based OpenAI usage extraction. • Adds Claude Haiku 4.5 as a supported model. • Extracts OpenAI usage data via the `genai-prices` library for more accurate token cost reporting. • Includes `final_result` as an agent span attribute after streaming completes, improving observability in traces.

PydanticAI v1.1.0

Wed, 15 Oct 2025 15:26:58 GMT

PydanticAI v1.1.0 adds Prefect durable execution support and a `description` argument for tool decorators. • Adds `description` argument to tool function decorators, allowing inline documentation of tools without relying solely on docstrings. • Adds durable execution support with Prefect, enabling fault-tolerant, resumable agent runs orchestrated via Prefect workflows.

PydanticAI v1.0.18

Mon, 13 Oct 2025 15:54:04 GMT

PydanticAI v1.0.18 adds Nebius AI Studio provider, EvaluationReport.render(), and ToolCallPart.id for OpenAI Responses. • Adds `ToolCallPart.id` field to carry tool-call identifiers from the OpenAI Responses API. • Adds `render` method to the `EvaluationReport` class for displaying evaluation results. • Adds Nebius AI Studio as a supported model provider. • Adds `anyio` and `httpcore` to Temporal passthrough modules, enabling those libraries to work correctly in Temporal workflows.

PydanticAI v1.0.17

Thu, 09 Oct 2025 16:40:39 GMT

PydanticAI v1.0.17 lets you pass `builtin_tools` at agent run time instead of only at agent definition time. • Allows `builtin_tools` to be specified at agent run time, enabling per-run control over which built-in tools are available without redefining the agent.

PydanticAI v1.0.16

Wed, 08 Oct 2025 16:21:36 GMT

PydanticAI v1.0.16 adds datetime.time/timedelta XML formatting, contextual agent name overrides, and FileUrl force-download support. • Respects `FileUrl.force_download` flag in OpenAI Chat and Responses model integrations. • Supports `datetime.time` and `timedelta` types in `format_as_xml`, enabling richer XML serialization of time-based fields. • Allows agent name to be overridden contextually at runtime, without changing the agent's definition. • Accepts `Sequence[ModelMessage]` instead of `list` for method argument types, broadening compatibility with any sequence type. • Validates `FileUrl` and `BinaryContent` objects without an `identifier` as valid inputs.

PydanticAI v1.0.15

Fri, 03 Oct 2025 23:40:35 GMT

PydanticAI v1.0.15 adds image generation support, streaming events API, and new ModelResponse convenience methods. • Adds `AgentRunResult.response` convenience method to retrieve the latest model response from a completed agent run. • Adds `ModelResponse.text`, `ModelResponse.thinking`, `ModelResponse.files`, `ModelResponse.images`, `ModelResponse.tool_calls`, and `ModelResponse.builtin_tool_calls` convenience methods for accessing parts of a model response. • Adds Agent.run_stream_events() convenience method as a shorthand wrapper around run(event_stream_handler=...). • Supports image generation and image output with Google and OpenAI providers. • Adds content (e.g. files) returned by a tool to `FunctionToolResultEvent`, making tool output accessible in event streams. • Sets `MCPServer` `id` and `tool_prefix` attributes automatically in `load_mcp_servers`. • Adds `gemini-2.5-flash` and `gemini-2.5-flash-lite` model names and aliases for the Google Gemini provider. • Supports enums in `format_as_xml` for structured XML formatting of enum values.

PydanticAI v1.0.13

Thu, 02 Oct 2025 01:12:03 GMT

PydanticAI v1.0.13 adds contextual agent instruction overrides, exposes MCPServer.server_info, and upgrades OTel instrumentation. • Exposes `server_info` on `MCPServer` instances, giving access to MCP server metadata at runtime. • Supports contextually overriding agent instructions at runtime, enabling dynamic per-request instruction customization. • Upgrades OpenTelemetry instrumentation to version 3 with updated eval attributes for improved observability.

PydanticAI v1.0.12

Wed, 01 Oct 2025 00:14:32 GMT

PydanticAI v1.0.12 adds Anthropic built-in memory tool support, OpenAI document URL/binary content for text/JSON/XML/YAML, and evals cost metrics. • Adds `retry` args to `pydantic_evals.Dataset.evaluate_sync` for configurable retry behavior in evaluation runs. • Adds cost metric to `pydantic-evals` output, giving visibility into token spend per evaluation. • Supports Anthropic's built-in memory tool, enabling agents to persist and recall information across turns via the provider-native mechanism. • Supports text, JSON, XML, and YAML `DocumentUrl` and `BinaryContent` on OpenAI, expanding the range of document types agents can process. • Prefers `structuredContent` in MCP tool results when present, enabling richer structured data from MCP tool calls. • Exposes `.messages` and `.toolsets` types in the top-level `pydantic_ai` namespace to improve IDE auto-import discovery. • Broadens the type of `common_tools` to work with agents of any deps type, removing a previous type-narrowing restriction. • Handles Gemini responses with more than one candidate without raising an error. • Handles Ollama responses that omit `finish_reason` and adds documentation for Ollama Cloud.

PydanticAI v1.0.11

Tue, 30 Sep 2025 00:25:29 GMT

PydanticAI v1.0.11 adds OpenAI image detail via vendor_metadata, operation.cost metrics, and makes OutputObjectDefinition public. • Supports OpenAI image detail level on `ImageUrl` and `BinaryContent` via the `vendor_metadata` parameter, enabling fine-grained vision API control. • Adds `operation.cost` metric to instrumented models, exposing per-call cost data through OpenTelemetry instrumentation. • Makes `OutputObjectDefinition` publicly importable from `pydantic_ai.output`, enabling programmatic construction of output schemas. • Supports callable classes (not just functions) as history processors, broadening the composition options for message-history pipelines. • Adds `claude-sonnet-4-5` to the list of known model name strings recognized by the library.

PydanticAI v1.0.10

Sat, 20 Sep 2025 00:08:09 GMT

PydanticAI v1.0.10 adds model class names as XML tags and field-level metadata options to `format_as_xml`. • Adds option to include field titles and descriptions as attributes in `format_as_xml`, and uses model class names as XML tags by default.

PydanticAI v1.0.9

Thu, 18 Sep 2025 23:37:26 GMT

PydanticAI v1.0.9 adds RunContext retry introspection and streams built-in tool calls from OpenAI, Google, and Anthropic. • Adds `RunContext.max_retries` and `RunContext.last_attempt` so tool functions can inspect retry limits and detect whether the current invocation is the final attempt. • Streams built-in tool calls from OpenAI, Google, and Anthropic and returns them on the next request, enabling support for OpenAI reasoning models. • Includes built-in tool calls and their results in OpenTelemetry (OTel) messages for full observability of tool interactions.

PydanticAI v1.0.8

Wed, 17 Sep 2025 01:14:18 GMT

PydanticAI v1.0.8 lets tools emit AG-UI events independently from the result returned to the model. • Tools can now return AG-UI events separately from the result sent to the model, enabling richer streaming side-effects without coupling event emission to the model's input.

PydanticAI v1.0.7

Mon, 15 Sep 2025 23:37:58 GMT

PydanticAI v1.0.7 adds MCP metadata filtering, FunctionToolset defaults, and improved RunContext prompt access. • Adds `ToolDefinition.metadata` field to carry MCP metadata and annotations, enabling filtering of MCP tools by metadata. • Adds support for default values for `strict`, `sequential`, `requires_approval`, and `metadata` parameters on `FunctionToolset`, reducing per-tool boilerplate. • When a run starts with a message history ending in a `ModelRequest`, its content is now available in `RunContext.prompt`. • Removes the requirement to install `mcp` or `logfire` extras when using Temporal or DBOS integrations. • Combines consecutive AG-UI user and assistant messages into a single model request/response.

PydanticAI v1.0.6

Fri, 12 Sep 2025 23:12:17 GMT

PydanticAI v1.0.6 adds `previous_response_id` support for the Responses API and file-based MCP server loading. • Adds `previous_response_id` parameter support for the OpenAI Responses API, enabling stateful multi-turn conversations backed by server-side response chaining. • Enables MCP servers to be loaded from a file, allowing declarative configuration of MCP server definitions outside of Python code.

PydanticAI v1.0.4

Thu, 11 Sep 2025 17:37:07 GMT

PydanticAI v1.0.4 adds a Pydantic AI Gateway provider for routing and managing LLM calls. • Adds a Pydantic AI Gateway provider, enabling use of the Pydantic AI Gateway as an LLM backend.

PydanticAI v1.0.3

Wed, 10 Sep 2025 23:58:15 GMT

PydanticAI v1.0.3 adds sequential tool call context manager, AG-UI callbacks, and Google model seed support • Adds agent.sequential_tool_calls() context manager to enforce sequential (non-parallel) tool execution within an agent run. • Adds `on_complete` callback to AG-UI functions, providing access to `AgentRunResult` when a run finishes. • Supports `ModelSettings.seed` in `GoogleModel` for reproducible outputs. • Supports `NativeOutput` with `FunctionModel`, enabling native structured output in function-backed models. • Sends AG-UI thinking start and end events, surfacing model reasoning steps to AG-UI consumers. • Includes thinking parts in subsequent model requests to improve performance and cache hit rates. • Raises a clear error when `WebSearchTool` is used with `OpenAIChatModel` and an unsupported model, rather than failing silently. • Supports models that return output tool args as `{"response": ""}`, broadening compatibility with non-standard model response formats.

PydanticAI v1.0.2

Tue, 09 Sep 2025 00:33:49 GMT

PydanticAI v1.0.2 adds DBOS durable execution, sequential tool calling, and Google cached content support • Adds `GoogleModelSettings.google_cached_content` field to pass `cached_content` when calling Google models, enabling prompt caching. • Adds `ModelResponse.finish_reason` attribute and populates `provider_response_id` during streaming responses. • Adds support for `gen_ai.response.id` in OpenTelemetry instrumentation spans. • Adds support for durable execution with DBOS, enabling fault-tolerant, resumable agent workflows. • Adds support for sequential tool calling, allowing agents to invoke tools one at a time in order rather than in parallel.

PydanticAI v1.0.0

Fri, 05 Sep 2025 00:15:29 GMT

PydanticAI v1.0.0 adds human-in-the-loop tool approval, LiteLLM provider, tool-call usage limits, and NativeOutput for Groq. • Adds `tool_calls_limit` to `UsageLimits` and `tool_calls` to `RunUsage` to cap and track tool-call counts per run. • Adds `docstring_format`, `require_parameter_descriptions`, and `schema_generator` parameters to `FunctionToolset` for fine-grained tool schema control. • Adds `operation.cost` span attribute to model request spans; renames ModelResponse.price() to ModelResponse.cost(). • Adds `identifier` field to `FileUrl` and its subclasses. • Adds human-in-the-loop tool call approval support, enabling agents to pause and await human confirmation before executing tools. • Adds a LiteLLM provider for OpenAI-API-compatible models. • Supports `NativeOutput` with Groq models. • Bundles `logfire` with the `pydantic-ai` package so tracing is available without a separate install. • Allows most types used in documentation examples to be imported directly from `pydantic_ai`. • Defaults `InstrumentationSettings` `version` to `2`. Breaking changes: • Python 3.9 is no longer supported; the minimum supported version is now Python 3.10. • ModelResponse.price() is renamed to ModelResponse.cost(); call sites using .price() will break. • `OpenAIModelProfile.openai_supports_sampling_settings` is deprecated. • `mcp-run-python` has been moved to its own repository and is no longer part of this package.

PydanticAI v0.8.1

Fri, 29 Aug 2025 14:42:15 GMT

PydanticAI v0.8.1 adds system-instructions tracing to agent run spans and renames key streaming and response methods. • Adds `gen_ai.system_instructions` attribute to agent run spans, exposing system prompt content in OpenTelemetry traces. • Renames `StreamedRunResult` methods to be consistent with `AgentStream` (see breaking changes). • Renames `ModelResponse.provider_request_id` to `provider_response_id` (see breaking changes). Breaking changes: • `StreamedRunResult` methods are renamed to match `AgentStream` naming conventions — callers using the old method names will break on upgrade. • `ModelResponse.provider_request_id` is renamed to `provider_response_id` — any code referencing `provider_request_id` will break on upgrade. • Specifying a model name without a provider prefix is deprecated, as is the `vertexai` provider name — configurations using bare model names or `vertexai` will need to be updated.

PydanticAI v0.8.0

Tue, 26 Aug 2025 23:32:08 GMT

PydanticAI v0.8.0 adds elicitation callbacks for MCP servers, message history in CLI agents, and a richer AgentStreamEvent union type. • Adds `message_history` parameter to agent.to_cli() to seed CLI sessions with prior conversation context. • Adds elicitation callback support to MCP servers, enabling agents to request additional input from users during tool execution. • Makes `AgentStreamEvent` a union of `ModelResponseStreamEvent` and `HandleResponseEvent`, expanding the event types available when streaming agent responses.

PydanticAI v0.7.6

Tue, 26 Aug 2025 12:57:56 GMT

PydanticAI v0.7.6 adds a Cerebras provider and renames the OpenAI model class. • Adds `CerebrasProvider` (Cerebras provider integration) for running agents against Cerebras-hosted models. • Replaces `all_messages_events` with `pydantic_ai.all_messages` span/event name under InstrumentationSettings(version=2). • Deprecates `OpenAIModel` in favor of the new `OpenAIChatModel` class. Breaking changes: • The tenacity retry implementation has changed behavior — existing retry logic built on the prior `AsyncTenacityTransport` semantics may behave differently on upgrade.

PydanticAI v0.7.5

Mon, 25 Aug 2025 08:02:45 GMT

PydanticAI v0.7.5 adds cost pricing on ModelResponse, span/trace IDs on EvaluationReport, and updated OpenTelemetry GenAI conventions. • Adds price() method to `ModelResponse` to retrieve cost information for a model response. • Adds `span_id` and `trace_id` fields to `EvaluationReport` for linking evaluations to distributed traces. • Updates OpenTelemetry instrumentation to use the new GenAI chat span attribute conventions. • Includes thoughts tokens in `output_tokens` accounting for Google models. • Allows proper typing on `AnthropicProvider` when using the Bedrock backend. Breaking changes: • OpenTelemetry span attributes for GenAI chat now follow the new GenAI conventions — any dashboards, alerts, or attribute-based queries built on the old attribute names will need to be updated.

PydanticAI v0.7.4

Wed, 20 Aug 2025 10:07:49 GMT

PydanticAI v0.7.4 adds `takes_ctx` to `Tool.from_schema` and supports Google's `url_context` built-in tool. • Adds `takes_ctx` argument to `Tool.from_schema`, letting callers control whether the generated tool receives the agent context. • Supports Google's `url_context` built-in tool via the new `UrlContextTool` class, now exported in `__all__`.

PydanticAI v0.7.3

Tue, 19 Aug 2025 06:43:51 GMT

PydanticAI v0.7.3 adds a CLI clipboard command, lets FallbackModel accept string names, and splits Usage into RequestUsage and RunUsage. • Adds `/cp` command to the CLI to copy the last response to the clipboard. • `FallbackModel` now accepts plain string model names in addition to model objects. • Moves `system_prompt_role` from `OpenAIModel` to `OpenAIModelProfile`, making it configurable at the profile level. • Introduces `RequestUsage` and `RunUsage` as replacements for the unified `Usage` class, providing finer-grained usage tracking. Breaking changes: • `Usage` is deprecated in favour of `RequestUsage` and `RunUsage`; code referencing `Usage` directly will need to migrate to the appropriate replacement class.

PydanticAI v0.7.2

Thu, 14 Aug 2025 22:55:59 GMT

PydanticAI v0.7.2 adds OllamaProvider, HuggingFace profile/settings, and max_uses for Anthropic WebSearchTool. • Adds `OllamaProvider` for connecting PydanticAI agents to locally hosted Ollama models. • Adds `profile` and `settings` parameters to `HuggingfaceModel` for finer control over HuggingFace inference. • Forwards `max_uses` parameter to Anthropic's `WebSearchTool`, allowing callers to cap the number of web searches per run. • Allows message history to end on a `ModelResponse` and automatically executes any pending tool calls, enabling richer conversation resumption. • Prompts the model to retry when it produces a response containing only thinking tokens (no text or tool calls), improving reliability with reasoning models. Breaking changes: • Removes the `anthropic-beta` default header previously set in `AnthropicModel`; integrations relying on that header being sent automatically will need to set it explicitly.

PydanticAI v0.7.1

Wed, 13 Aug 2025 22:20:19 GMT

PydanticAI v0.7.1 adds GPT-5 models, OpenAI verbosity support, pre-request token counting via Gemini, and a new model inference string. • Adds `UsageLimits.count_tokens_before_request` to count tokens using Gemini's `count_tokens` API before a request is sent, enabling proactive limit enforcement. • Supports the `"openai-responses"` model inference string for selecting the OpenAI Responses API via string-based model configuration. • Adds support for the OpenAI `verbosity` parameter in the Responses API. • Adds new OpenAI GPT-5 models to the supported model list.

PydanticAI v0.7.0

Tue, 12 Aug 2025 23:14:18 GMT

PydanticAI v0.7.0 adds Temporal workflow support, dynamic toolsets, event stream handlers, and new agent abstractions. • Adds `event_stream_handler` parameter to agent and run methods for subscribing to agent event streams. • Adds Agent.override(tools=...) to replace or inject tools into an existing agent at runtime. • Adds `AbstractAgent` and `WrapperAgent` base classes for building composable agent wrappers. • Enables running `Agent` inside a Temporal workflow by dispatching model requests, tool calls, and MCP as Temporal activities. • Supports dynamically building toolsets based on run context via the toolset API. • Adds a history processor API that replaces message history on each run, enabling custom context-window management.

PydanticAI v0.6.2

Thu, 07 Aug 2025 15:40:52 GMT

PydanticAI v0.6.2 adds `builtin_tools` parameter to the `Agent` class. • Adds `builtin_tools` parameter to `Agent`, enabling control over which built-in tools are available to an agent.

PydanticAI v0.6.1

Thu, 07 Aug 2025 14:59:24 GMT

PydanticAI v0.6.1 adds automatic OpenAI strict mode, Bedrock thinking parts, AWS bearer token support, and new Heroku models. • Supports `AWS_BEARER_TOKEN_BEDROCK` environment variable for authenticating with AWS Bedrock via bearer token. • Makes `InlineDefsJsonSchemaTransformer` part of the public API, allowing direct use in custom JSON schema transformations. • Automatically enables OpenAI strict mode for output types that are strict-compatible, removing the need for manual configuration. • Sends `ThinkingPart`s back to Anthropic when accessed through AWS Bedrock, enabling extended thinking round-trips. • Adds new Heroku models to the supported model list.

PydanticAI v0.6.0

Wed, 06 Aug 2025 16:03:13 GMT

PydanticAI v0.6.0 adds a new Anthropic model and removes a wave of long-deprecated APIs. • Adds a new Anthropic model (removes older deprecated Anthropic models in the same change). Breaking changes: • The next() method is removed from `Graph`. • The `data` attribute is removed from `FinalResult`. • The `get_data` and `validate_structured_result` methods are removed from `StreamedRunResult`. • The `format_as_xml` module is removed entirely. • The `result_type` parameter (and similar parameters) is removed from `Agent`. • Four months of accumulated deprecation warnings are now hard removals — any code that relied on those deprecated APIs will break.

PydanticAI v0.5.0

Mon, 04 Aug 2025 23:04:54 GMT

PydanticAI v0.5.0 expands OpenAI strict JSON mode compatibility and adds default values to tool argument schemas. • Enables more `BaseModel`s to use OpenAI strict JSON mode by defaulting `additionalProperties=False` automatically. • Supports string `format`, `pattern`, and related constraints within OpenAI strict JSON mode. • Includes default values in the JSON schema generated for tool arguments. Breaking changes: • The `EvaluationReport.print` and `EvaluationReport.console_table` methods now require most arguments to be passed by keyword. • The `source` field of `EvaluationResult` is now of type `EvaluatorSpec` instead of the actual `Evaluator` instance; existing code that accessed the live evaluator instance via `source` will break.

PydanticAI v0.4.11

Fri, 01 Aug 2025 23:53:18 GMT

PydanticAI v0.4.11 adds AG-UI convenience functions and custom thinking-tag support on model profiles. • Supports custom thinking tags specified on the model profile, letting callers control how chain-of-thought tokens are surfaced per model. • Adds convenience functions to handle AG-UI requests with request-specific dependencies.

PydanticAI v0.4.10

Wed, 30 Jul 2025 22:53:16 GMT

PydanticAI v0.4.10 adds `priority` `service_tier` to OpenAI settings and HTTP Referer header support for Vercel AI Gateway. • Adds `priority` `service_tier` option to `OpenAIModelSettings`, respected by `OpenAIResponsesModel`, enabling OpenAI priority-tier routing from model configuration. • Adds HTTP Referer request header support to the Vercel AI Gateway provider.

PydanticAI v0.4.8

Mon, 28 Jul 2025 14:52:16 GMT

PydanticAI v0.4.8 adds tenacity retry integration and thinking-part tracing in OpenTelemetry model response events. • Adds tenacity utilities and integration for improved retry handling in agent workflows. • Includes `ThinkingPart` in OpenTelemetry OTEL events emitted via `ModelResponse`, surfacing model reasoning in traces.

PydanticAI v0.4.7

Thu, 24 Jul 2025 21:36:58 GMT

PydanticAI v0.4.7 adds MoonshotAI, Vercel AI Gateway providers, Gemini Files API support, and MCP ResourceLink handling. • Renames `MCPServer` parameter `sse_read_timeout` to `read_timeout`, which is now passed through to `ClientSession`. • Adds `MoonshotAI` provider with Kimi-K2 model support. • Adds `Vercel AI Gateway` provider. • Supports passing files uploaded to the Gemini Files API and setting a custom media type. • Parses `` tags in streamed text as thinking parts (`ThinkingPart`). • Adds support for `MCP ResourceLink` returned from tools. Breaking changes: • The `MCPServer` parameter `sse_read_timeout` is renamed to `read_timeout`; any code passing `sse_read_timeout` by keyword will break on upgrade.

PydanticAI v0.4.6

Wed, 23 Jul 2025 14:42:40 GMT

PydanticAI v0.4.6 adds URL and binary PDF support for the Mistral provider. • Adds URL and binary PDF input support for the Mistral provider, enabling document-based prompts via URL or raw binary PDF. • Speeds up the internal `_estimate_string_tokens` function, improving throughput for token-heavy workloads.

PydanticAI v0.4.5

Tue, 22 Jul 2025 13:11:38 GMT

PydanticAI v0.4.5 adds streamable HTTP transport support to mcp-run-python and changes format_as_xml defaults. • Supports streamable HTTP transport in `mcp-run-python`, enabling streaming MCP server connections over HTTP. Breaking changes: • The default values for `format_as_xml` have changed; existing code relying on the previous defaults may produce different XML output after upgrading.

PydanticAI v0.4.4

Fri, 18 Jul 2025 21:41:13 GMT

PydanticAI v0.4.4 adds Toolsets, AG-UI protocol support, new OpenAI/Grok/Kimi models, and an `identifier` field on `BinaryContent`. • Adds `identifier` field to the `BinaryContent` class for tagging binary content objects. • Introduces Toolsets and Deferred Tools, enabling grouped and lazily-resolved tool registration on agents. • Supports the AG-UI protocol for frontend-agent communication. • Adds OpenAI models `o1-pro`, `o3-pro`, `o3-deep-research`, and `computer-use` as selectable models. • Adds `grok-4` and `kimi-k2` (via Groq) as selectable models. • Speeds up `AgentRunResult._set_output_tool_return` by ~18,798%, unlocking high-throughput agent run scenarios. Breaking changes: • Old Google models have been removed; any code referencing those model identifiers will break on upgrade.

PydanticAI v0.4.3

Wed, 16 Jul 2025 16:28:06 GMT

PydanticAI v0.4.3 adds Hugging Face provider support, output function tracing, and base64 encoding for tool returns. • Adds base64 encoding support to `tool_return_ta`, enabling binary data to be returned from tools. • Adds output function tracing, allowing agent output functions to be captured in traces. • Adds Hugging Face as a new model provider. Breaking changes: • The `duckduckgo-search` package dependency is renamed to `ddgs`; any install or import referencing `duckduckgo-search` will break.

PydanticAI v0.4.2

Thu, 10 Jul 2025 18:43:54 GMT

PydanticAI v0.4.2 adds StructuredDict for custom JSON schema outputs, model settings on model classes, and DeepSeek reasoning_content streaming support. • Adds `StructuredDict` class for defining structured outputs with a custom JSON schema, giving callers direct control over the schema shape returned by the model. • Allows model `settings` to be passed directly to model classes, enabling per-model configuration at instantiation time. • Supports DeepSeek `reasoning_content` field in streamed responses, surfacing chain-of-thought reasoning tokens from DeepSeek models during streaming. • Speeds up internal `_ensure_decodeable` function by 634%, unlocking higher-throughput decoding for workloads processing large volumes of model output. Breaking changes: • FastA2A has been dropped from the PydanticAI repository and is no longer available as part of the package.

PydanticAI v0.4.1

Thu, 10 Jul 2025 08:09:32 GMT

PydanticAI v0.4.1 adds sync task evaluation support and drops FastA2A as a transitive dependency. • Adds support for evaluating synchronous tasks in PydanticAI's evals framework, expanding coverage beyond async-only workflows. Breaking changes: • FastA2A is no longer a PydanticAI dependency; projects that relied on it being pulled in transitively must now declare it as a direct dependency.

PydanticAI v0.4.0

Tue, 08 Jul 2025 19:40:40 GMT

PydanticAI v0.4.0 adds broader Gemini audio support and makes ToolDefinition.description optional. • Makes `ToolDefinition.description` optional, removing the requirement to supply a description when defining tools. • Adds all Gemini-supported audio types to `AudioUrl`, expanding multimodal input coverage for Gemini models. • Retains default values in non-strict OpenAI schemas, preserving schema fidelity when targeting OpenAI backends. Breaking changes: • `EvaluationReport` and `ReportCase` are now generic dataclasses — any code that instantiates or type-annotates these without type parameters may require updates.

PydanticAI v0.3.7

Mon, 07 Jul 2025 20:49:10 GMT

PydanticAI v0.3.7 adds GitHub Models provider, ACI.dev Tools integration, sync streaming, and Google video analysis args. • Adds `model_request_stream_sync` to the direct API, enabling synchronous streaming of model requests. • Adds GitHub Models as a new provider via the GitHub Models provider integration. • Adds support for Google-specific arguments for video analysis in the Google provider. • Implements ACI.dev Tools integration, providing a convenient way to use ACI.dev tools in PydanticAI. • `AgentStream.stream_output` (available inside `agent.iter`) now streams validated output data instead of raising validation errors mid-stream.

PydanticAI v0.3.6

Fri, 04 Jul 2025 15:20:00 GMT

PydanticAI v0.3.6 adds predicted outputs to OpenAIModelSettings and records tool responses in trace spans. • Adds support for `predicted_outputs` in `OpenAIModelSettings`, enabling speculative/predicted output hints when calling OpenAI models. • Records tool response data in tool-run spans, enriching tracing and observability for agent tool calls. • Improves model communication by marking a `RetryPromptPart` not tied to a tool call as validation feedback rather than a user message, giving the model clearer signal on why a retry is occurring. • Switches agent overriding from a local attribute to `contextvars`, making agent context propagation safe across async/concurrent workloads.

PydanticAI v0.3.5

Mon, 30 Jun 2025 13:10:23 GMT

PydanticAI v0.3.5 lets tools return `ToolReturn` for richer model content and adds strict mode to `NativeOutput`. • Supports strict mode in `NativeOutput`, enabling stricter schema validation for native model outputs. • Allows tools to return a `ToolReturn` object to pass additional content to the model or attach metadata that is not forwarded to the model. • Sets `'us-central1'` as the default region on `GoogleProvider`, removing the need to configure it explicitly. • Moves `ThinkingPart` to precede `TextPart` in `OpenAIResponsesModel`, aligning reasoning output ordering. • Adds a progress bar during evaluation runs. Breaking changes: • The default region for `GoogleProvider` is now `'us-central1'`; existing setups that relied on no default region being set may route requests differently after upgrading.

PydanticAI v0.3.4

Thu, 26 Jun 2025 06:16:15 GMT

PydanticAI v0.3.4 adds sensitive-content scrubbing to agent pipelines. • Adds sensitive content scrubbing to redact or sanitize private data within agent interactions.

PydanticAI v0.3.3

Tue, 24 Jun 2025 17:31:48 GMT

PydanticAI v0.3.3 adds NativeOutput and PromptedOutput modes and captures more OpenAI-compatible usage fields. • Adds `NativeOutput` and `PromptedOutput` output modes alongside the existing `ToolOutput` mode, giving agents more control over how structured results are produced. • Captures additional usage fields returned by OpenAI-compatible APIs, surfacing richer token and cost details in `Usage` objects. • Makes `Edge` hashable, enabling graph edges to be stored in sets and used as dict keys.

PydanticAI v0.3.0

Wed, 18 Jun 2025 01:30:00 GMT

PydanticAI v0.3.0 adds `ThinkingPart` support, parsing provider thinking blocks into a dedicated message part type. • Adds `ThinkingPart` as a new message part type: provider-specific `...` blocks in text responses are now parsed and surfaced as structured `ThinkingPart` objects rather than raw text. Breaking changes: • `ThinkingPart`s are not sent back to the provider in subsequent turns — existing agents that relied on thinking content being echoed back in the message history will no longer include it, reducing costs but changing round-trip behavior.

PydanticAI v0.2.20

Wed, 18 Jun 2025 00:15:46 GMT

PydanticAI v0.2.20 adds a process_tool_call hook for MCP servers and RunContext support in history processors. • Adds `process_tool_call` hook to MCP servers, enabling interception and modification of tool arguments, metadata, and return values before and after MCP tool execution. • Adds `RunContext` support to history processors, giving them access to the full run context when processing conversation history. • Adds `ModelSettings.timeout` enforcement in `GoogleModel`, so timeout settings are now respected when calling Google models.

PydanticAI v0.2.19

Tue, 17 Jun 2025 00:29:16 GMT

PydanticAI v0.2.19 adds `history_processors` to Agent and surfaces events for unknown tool calls • Adds `history_processors` parameter to `Agent` for programmatic pre-processing of message history before each model call. • Yields streaming events for unknown tool calls instead of silently dropping them, enabling downstream handling of unrecognised tool responses. • Makes `infer_provider` more flexible, accepting a broader range of inputs when resolving provider from a model string. • Ignores dynamic instructions that return an empty string, preventing blank system-prompt entries from being appended to the message list. Breaking changes: • Anthropic `max_tokens` is now set to 4096 by default; any agent relying on the previous default behaviour may produce truncated responses or incur different token usage.

PydanticAI v0.2.18

Fri, 13 Jun 2025 09:51:02 GMT

PydanticAI v0.2.18 adds MCP Streamable HTTP, OpenAI Responses API vendor ID, and reuses last message when no prompt is given. • Exposes the OpenAI Responses API response ID as `vendor_id` on the model response object. • Adds MCP Streamable HTTP transport implementation. • Reuses the last request from message history automatically when no user prompt is provided, enabling continuation flows without re-supplying context. • Switches Gemini inference to use `GoogleModel` instead of `GeminiModel`.

PydanticAI v0.2.17

Thu, 12 Jun 2025 14:06:27 GMT

PydanticAI v0.2.17 adds token usage to InstrumentedModel, service_tier for OpenAI, custom httpx clients for MCP, and Gemini direct file URL support. • Adds `service_tier` field to `OpenAIModelSettings` to control OpenAI service tier selection. • Adds token usage metrics to `InstrumentedModel` for observability of model calls. • Allows users to supply a custom `httpx.AsyncClient` in `MCPServerHTTP` for full control over HTTP transport. • Supports `fileData` field (direct file URL) for `GeminiModel` and `GoogleModel`, enabling direct URL-based file inputs. • Suppresses inapplicable sampling settings (`temperature`, `top_p`) when targeting OpenAI reasoning models.

PydanticAI v0.2.16

Sun, 08 Jun 2025 20:01:28 GMT

PydanticAI v0.2.16 adds HerokuProvider, stop_sequences for Google models, and LangChain community tool integration. • Adds `HerokuProvider` to connect agents to Heroku-hosted models. • Adds `stop_sequences` parameter support for Google models. • Adds a convenience method to use LangChain community tools directly within PydanticAI agents. • Improves output type inference when callables are provided as output types.

PydanticAI v0.2.13

Tue, 03 Jun 2025 17:18:56 GMT

PydanticAI v0.2.13 adds expected-output support to LLMJudge evaluations. • Adds option to pass expected output to `LLMJudge`, enabling reference-based LLM evaluation scoring.

PydanticAI v0.2.12

Thu, 29 May 2025 14:25:31 GMT

PydanticAI v0.2.12 adds function output types, ModelProfile config, Together/Fireworks/Grok providers, and Claude 4 on Bedrock. • Adds `ModelProfile` class to configure model-specific behaviors independently of the model class, enabling fine-grained control over provider quirks without subclassing. • Adds new provider classes for Together AI, Fireworks AI, and Grok with automatic model profile selection. • Adds `vendor_id` and `vendor_details.finish_reason` fields to Gemini/Google model response objects. • Supports functions as `output_type` in agents, including lists of functions mixed with other types. • Adds support for Claude 4 Sonnet and Opus models via the Bedrock provider. • Enhances Gemini usage tracking to collect comprehensive token data beyond basic prompt/completion counts.

PydanticAI v0.2.10

Tue, 27 May 2025 13:24:57 GMT

PydanticAI v0.2.10 adds Claude Sonnet 4 support, MCP Streamable HTTP transport, and MCP client init timeouts. • Adds support for Claude Sonnet 4 as a model target. • Adds MCP Streamable HTTP transport support, enabling HTTP-based MCP server connections alongside the existing stdio transport. • Adds a timeout for initializing MCP clients, preventing indefinite hangs during MCP server startup. • Updates supported Google models.

PydanticAI v0.2.9

Mon, 26 May 2025 07:40:01 GMT

PydanticAI v0.2.9 adds Vertex AI label support for Gemini/Google models and improves Agent CLI output handling. • Supports `labels` field for `GeminiModel` and `GoogleModel` on Vertex AI, enabling resource labeling for cost attribution and organization. • Non-textual responses in `Agent.to_cli` are now cast to `str`, allowing the CLI interface to handle structured or binary model outputs.

PydanticAI v0.2.7

Sat, 24 May 2025 15:34:59 GMT

PydanticAI v0.2.7 adds MCP tool_prefix namespacing, real-time Anthropic streaming, and a customizable prog_name for CLI agents. • Adds `tool_prefix` option to MCP servers to namespace tool names and raises an error on conflicting tool names across servers. • Makes `prog_name` customizable on CLI agents, allowing teams to brand or script against a consistent program name. • Removes the hardcoded `n` parameter from `OpenAIModel` requests, unlocking use of endpoints and deployments that reject that field. • Streams tool calls and structured output from Anthropic incrementally as tokens arrive instead of buffering the full response. • Supports streaming tool calls from models that pass `args` as `None` when a function has no parameters.

PydanticAI v0.2.6

Wed, 21 May 2025 08:12:40 GMT

PydanticAI v0.2.6 adds `prepare_tools` param to Agent and `'openrouter'` as a supported OpenAIModel provider. • Adds `prepare_tools` parameter to the `Agent` class, enabling dynamic control over which tools are presented to the model at runtime. • Supports `'openrouter'` as a valid string value for the `provider` parameter of `OpenAIModel`, enabling routing through OpenRouter.

PydanticAI v0.2.5

Tue, 20 May 2025 14:31:26 GMT

PydanticAI v0.2.5 adds OpenRouter and Google GenAI providers, logprobs support, and new instrumentation controls. • Adds `include_binary_content` flag to `InstrumentationSettings` to control whether binary content is captured in traces; renames the OTel attribute key from `content` to `binary_content` for `BinaryPart`s. • Adds `logprobs` to OpenAI model settings and response objects, exposing token-level log probability data. • Adds `vendor_id` field to model response objects. • Adds ability to specify the evaluation name for all built-in Evaluators. • Adds OpenRouter provider for routing requests across LLM backends. • Adds Google GenAI provider for direct integration with Google's generative AI APIs. • Makes `capabilities` a required field on `AgentCard` in the fasta2a integration. Breaking changes: • The OTel attribute key for `BinaryPart`s is renamed from `content` to `binary_content`; any dashboards, queries, or processors filtering on the old key will stop matching. • `capabilities` is now required on `AgentCard` in fasta2a; existing `AgentCard` instantiations that omit `capabilities` will raise a validation error.

PydanticAI v0.2.3

Tue, 13 May 2025 13:54:06 GMT

PydanticAI v0.2.3 adds an A2A server, a `direct` public API, and model-settings support for `LLMJudge`. • Adds `direct` public API for invoking models directly. • Adds an A2A (Agent-to-Agent) server, enabling agents to communicate via the A2A protocol. • Allows `ModelSettings` to be defined on `LLMJudge` to control model behavior during evaluations.

PydanticAI v0.2.2

Tue, 13 May 2025 09:50:11 GMT

PydanticAI v0.2.2 adds a to_cli() method to `Agent` for instant command-line interfaces. • Adds to_cli() method to the `Agent` class, enabling any agent to be exposed as a CLI application.

PydanticAI v0.2.1

Tue, 13 May 2025 09:27:38 GMT

PydanticAI v0.2.1 adds AWS Profile support, CLI config persistence, and OpenTelemetry BinaryContent tracing. • CLI now stores prompt history and configuration under `~/.pydantic-ai` for persistence across sessions. • OpenTelemetry integration now sends `BinaryContent` information in traces. • Adds AWS Profile support for authenticating with AWS-backed models. • Improves Agent.is_*_node() type narrowing by switching to `TypeIs` for more precise static analysis.

PydanticAI v0.2.0

Mon, 12 May 2025 15:32:13 GMT

PydanticAI v0.2.0 moves usage data into ModelResponse and adds non-string enum support for Gemini. • Adds `usage` field to `ModelResponse` (defaults to Usage() for backward-compatible deserialization), making token/cost usage directly accessible on every model response and in message history sequences. • Adds support for non-string enums in Gemini model integrations. Breaking changes: • The return type of `Model.request` changed from `tuple[ModelResponse, Usage]` to `ModelResponse` — callers that unpack the two-element tuple will break; usage is now accessed via `response.usage`.

PydanticAI v0.1.11

Sat, 10 May 2025 06:41:36 GMT

PydanticAI v0.1.11 renames the CLI entry point to `clai`. • Renames the CLI entry point to `clai`, replacing the previous command name. Breaking changes: • The CLI command is now `clai`; any scripts or aliases invoking the old CLI name will break on upgrade.

PydanticAI v0.1.10

Tue, 06 May 2025 14:40:59 GMT

PydanticAI v0.1.10 adds extra_headers to ModelSettings and thinking_config to GeminiModel • Adds `extra_headers` field to `ModelSettings` to pass custom HTTP headers to model API calls. • Adds `thinking_config` parameter to `GeminiModel` to control extended thinking behavior. • Allows setting `temperature` to `0` on `BedrockConverseModel` for deterministic outputs.

PydanticAI v0.1.9

Fri, 02 May 2025 10:06:45 GMT

PydanticAI v0.1.9 adds base_url support for Mistral, richer Anthropic usage details, and multi-modal MCP tool call responses. • Adds `base_url` parameter to the Mistral provider, enabling custom or self-hosted Mistral endpoint configuration. • Stores additional usage details returned by Anthropic in the response metadata. • Handles multi-modal and error responses from MCP tool calls, broadening the range of MCP tool outputs PydanticAI can process.

PydanticAI v0.1.8

Mon, 28 Apr 2025 20:23:08 GMT

PydanticAI v0.1.8 lets tools return multi-modal content such as images and audio alongside text. • Tools can now return multi-modal content (e.g. images, audio, binary data) directly from tool functions, not just text or structured data.

PydanticAI v0.1.7

Mon, 28 Apr 2025 12:18:10 GMT

PydanticAI v0.1.7 adds Gemini video support, multi-instruction agents, and attribute docstrings on tools by default. • Sets `use_attribute_docstrings=True` as the default on tools, so attribute-level docstrings are automatically used in tool schemas without explicit configuration. • Supports multiple `instructions` on an `Agent`, with correct concatenation when more than one instruction is provided. • Adds Gemini video support, enabling video content to be passed to Gemini models via the PydanticAI message API.

PydanticAI v0.1.6

Fri, 25 Apr 2025 12:18:48 GMT

PydanticAI v0.1.6 adds OpenTelemetry tracing for AudioUrl, VideoUrl, DocumentUrl, and ImageUrl content. • OpenTelemetry spans now include `AudioUrl`, `VideoUrl`, `DocumentUrl`, and `ImageUrl` content metadata, enabling full observability over multimodal model interactions.

PydanticAI v0.1.4

Thu, 24 Apr 2025 10:36:32 GMT

PydanticAI v0.1.4 adds MCP logging, o3/o4-mini support, and OpenAI document input types. • Supports `DocumentUrl` and `BinaryContent` document types for OpenAI provider inputs. • Adds support for OpenAI `o3` and `o4-mini` models. • Supports MCP logging and raises minimum MCP version requirement to `1.6.0`. • Makes agent and graph runs serializable, enabling persistence and resumption of run state. Breaking changes: • Minimum MCP version is now `1.6.0`; installations using an older MCP version will break.

PydanticAI v0.1.3

Fri, 18 Apr 2025 15:04:32 GMT

PydanticAI v0.1.3 adds `extra_body` to `ModelSettings`, OpenTelemetry instruction spans, and Gemini 2.5 Flash support. • Adds `extra_body` field to `ModelSettings` for passing arbitrary additional body parameters to model API requests. • Adds OpenTelemetry span events for `instructions`, making instruction content visible in traces. • Adds support for the `gemini-2.5-flash-preview-04-17` model.

PydanticAI v0.1.2

Thu, 17 Apr 2025 08:16:05 GMT

PydanticAI v0.1.2 exposes `StdioServerParameters.cwd` for controlling MCP server working directories. • Exposes `StdioServerParameters.cwd` parameter, allowing callers to set the working directory for stdio MCP server processes.

PydanticAI v0.1.0

Tue, 15 Apr 2025 21:35:45 GMT

PydanticAI v0.1.0 renames result→output, adds VideoUrl for Bedrock, spans on run results, and an instructions parameter. • Adds `VideoUrl` input support to `BedrockConverseModel` for passing video content to the Bedrock Converse API. • Adds additional configuration fields to `BedrockConverseModel` for the Bedrock Runtime API. • Adds `instructions` parameter to agents for supplying system-level instructions at call time. • Exposes `spans` as an attribute on agent/graph runs and run results for OpenTelemetry trace access. • Adds support for `gemini-2.5-pro-preview-03-25` (paid tier of Gemini 2.5 Pro). • Generalizes JSON schema transformations across model backends. Breaking changes: • The `result` field/attribute is renamed to `output` across agent runs and run results — any code referencing `.result` will break on upgrade. • `format_as_xml` has been moved to a new location — imports referencing the old module path will break on upgrade.

PydanticAI v0.0.55

Wed, 09 Apr 2025 15:08:03 GMT

PydanticAI v0.0.55 allows empty user prompts in streaming runs and adds a PydanticAI User-Agent header to outbound requests. • Adds a `PydanticAI` `User-Agent` header to all outbound HTTP requests, enabling easier identification of traffic in server logs and API dashboards. • Supports empty `user_prompt` values in `run_stream`, allowing streaming runs to be initiated with no user message.

PydanticAI v0.0.54

Wed, 09 Apr 2025 10:01:06 GMT

PydanticAI v0.0.54 adds `stop_sequences` to `ModelSettings`, optional `user_prompt`, and yields the initial graph node during iteration. • Adds `stop_sequences` field to `ModelSettings` to control where model output is terminated. • Makes `user_prompt` optional, allowing agent invocations without a required user-facing prompt. • Graph (and therefore Agent) iteration now yields the initial node, giving callers visibility into the full execution sequence from the start.

PydanticAI v0.0.53

Mon, 07 Apr 2025 08:45:35 GMT

PydanticAI v0.0.53 adds OpenAI strict mode support for structured output. • Adds OpenAI strict mode support, enabling stricter schema enforcement when using OpenAI models for structured output generation.

PydanticAI v0.0.52

Thu, 03 Apr 2025 14:04:17 GMT

PydanticAI v0.0.52 adds dependency injection support to the evals framework. • Enables passing dependencies into evals, bringing PydanticAI's dependency-injection model to the evaluation framework.

PydanticAI v0.0.51

Thu, 03 Apr 2025 10:02:18 GMT

PydanticAI v0.0.51 switches mcp-run-python to Deno and aligns OpenAI model strictness. • Switches the `mcp-run-python` server runtime from its previous backend to Deno. • Matches OpenAI models in strictness, aligning structured-output enforcement with OpenAI's strict mode behavior.

PydanticAI v0.0.49

Tue, 01 Apr 2025 16:20:19 GMT

PydanticAI v0.0.49 adds Gemini 2.5 Pro, OpenAI built-in tools, and new `OpenAIResponsesModelSettings` fields • Adds `generate_summary` and `truncation` fields to `OpenAIResponsesModelSettings` for controlling response summarization and context truncation. • Adds OpenAI built-in tools support, exposing OpenAI-native tool integrations through the PydanticAI interface. • Adds Gemini 2.5 Pro model support alongside CLI improvements. • Anthropic models now pass `ImageUrl` and `DocumentUrl` references directly without downloading content, enabling more efficient media handling.

PydanticAI v0.0.48

Mon, 31 Mar 2025 14:47:52 GMT

PydanticAI v0.0.48 adds support for the OpenAI Responses API. • Adds support for the OpenAI Responses API, enabling PydanticAI agents to use OpenAI's newer stateful response interface.

PydanticAI v0.0.47

Mon, 31 Mar 2025 11:04:30 GMT

PydanticAI v0.0.47 ships the new `pydantic-evals` package, read/connect timeouts for Bedrock, and OpenTelemetry spans around tool calls. • Adds `read_timeout` and `connect_timeout` settings to the Bedrock provider for finer-grained network control. • Introduces the `pydantic-evals` package, a new library for evaluating AI agent outputs. • Wraps every tool call in an OpenTelemetry span for deeper observability into agent execution. • Supports passing a plain `str` as the model argument, broadening how models can be specified at call sites. • Allows running under `PYTHONOPTIMIZE=1` (stripped assertions) without errors.

PydanticAI v0.0.46

Wed, 26 Mar 2025 17:22:20 GMT

PydanticAI v0.0.46 adds headers, timeout, and SSE read timeout to MCPServerHTTP • Adds `headers`, `timeout`, and `sse_read_timeout` parameters to `MCPServerHTTP` for fine-grained control over HTTP MCP server connections. • Uses different HTTP clients based on providers, enabling provider-specific HTTP client behaviour.

PydanticAI v0.0.45

Wed, 26 Mar 2025 10:42:43 GMT

PydanticAI v0.0.45 adds `user` mapping support in OpenAI chat completion requests. • Adds `user` mapping in OpenAI chat completion requests, allowing callers to pass a user identifier through to the OpenAI API.

PydanticAI v0.0.44

Tue, 25 Mar 2025 12:04:37 GMT

PydanticAI v0.0.44 adds a Cohere provider, exposes tool definitions on chat spans, and drops the `system` parameter from `OpenAIModel`. • Adds `model_request_parameters` attribute (containing tool definitions) to chat spans, making tool configurations observable in traces. • Adds a Cohere provider class for inference, enabling PydanticAI agents to target Cohere models via the provider pattern. • Migrates OpenAI models from `max_tokens` to `max_completion_tokens` in requests. • Adds function return docstrings to the generated tool description passed to models. Breaking changes: • The `system` parameter is removed from `OpenAIModel`; code that passes `system=` to `OpenAIModel` will break on upgrade.

PydanticAI v0.0.43

Fri, 21 Mar 2025 09:02:38 GMT

PydanticAI v0.0.43 adds a timestamp field to SystemPromptPart and auto-refreshes Google Vertex tokens on 401. • Adds `timestamp` field to `SystemPromptPart`, enabling precise tracking of when system prompts were created. • Automatically recreates the access token on HTTP 401 responses for the Google Vertex provider, enabling uninterrupted long-running sessions.

PydanticAI v0.0.42

Wed, 19 Mar 2025 15:26:30 GMT

PydanticAI v0.0.42 adds MCP server support, a Python sandbox MCP server, and customizable tool JSON schema generation. • Renames `MCPServerSSE` to `MCPServerHTTP` for connecting agents to MCP servers over HTTP (see breaking changes). • Adds support for MCP (Model Context Protocol) servers, allowing agents to connect to and use tools exposed via MCP. • Adds a built-in MCP server for running Python code in a sandbox environment. • Enables overriding JSON schema generation for tools, giving developers control over how tool parameters are described to the model. Breaking changes: • `MCPServerSSE` is renamed to `MCPServerHTTP`; any code referencing `MCPServerSSE` will break on upgrade.

PydanticAI v0.0.41

Mon, 17 Mar 2025 15:01:44 GMT

PydanticAI v0.0.41 adds Anthropic and Mistral provider classes. • Adds Anthropic provider classes for direct integration with Anthropic models. • Adds Mistral provider classes for direct integration with Mistral models.

PydanticAI v0.0.40

Sat, 15 Mar 2025 14:41:05 GMT

PydanticAI v0.0.40 adds AzureProvider, env-var base URL for OpenAIProvider, state persistence, and Anthropic PDF support. • Adds `AzureProvider` class for connecting PydanticAI agents to Azure-hosted OpenAI deployments. • Adds environment variable support for base URL configuration in `OpenAIProvider`, removing the need to hard-code endpoints. • Adds state persistence support, enabling agents to save and restore conversational state across runs. • Adds PDF document support to the Anthropic provider, allowing PDF content to be passed as model input.

PydanticAI v0.0.39

Thu, 13 Mar 2025 13:18:43 GMT

PydanticAI v0.0.39 adds Groq provider classes for LLM integration. • Adds Groq provider classes, enabling Groq-hosted models as a PydanticAI LLM backend.

PydanticAI v0.0.38

Thu, 13 Mar 2025 10:33:52 GMT

PydanticAI v0.0.38 adds DocumentUrl and BinaryContent document support for passing documents to models. • Adds `DocumentUrl` class to pass documents to models by URL. • Adds document support via `BinaryContent` for passing raw binary document data to models.

PydanticAI v0.0.37

Wed, 12 Mar 2025 11:49:19 GMT

PydanticAI v0.0.37 adds `base_url` to models, tool name override on decorators, and VertexAI pre-loaded service account support. • Adds `base_url` parameter to models, and populates `server.address` and `server.port` fields in OpenTelemetry spans for tracing. • Allows specifying a custom tool name when registering a function with the tool decorator. • Supports pre-loaded VertexAI service account info, removing the need to read credentials from disk at runtime. • Serializes `bytes` values as base64 automatically when converting to JSON.

PydanticAI v0.0.36

Fri, 07 Mar 2025 11:11:41 GMT

PydanticAI v0.0.36 adds AWS Bedrock Converse API support and expands VertexAI region coverage. • Adds support for the AWS Bedrock Converse API as a new model backend. • Expands `VertexAIRegion` `Literal` with updated region URLs for broader Vertex AI regional coverage.

PydanticAI v0.0.34

Wed, 05 Mar 2025 16:21:14 GMT

PydanticAI v0.0.34 adds Agent.instrument_all(), a `pai` CLI, and tool names in response events. • Adds Agent.instrument_all() class method to instrument all agents globally by default, without configuring each agent individually. • Adds `pai` CLI for interacting with PydanticAI from the command line. • Adds tool name to tool response events, making it easier to identify which tool produced a given response in streaming or event-driven workflows.

PydanticAI v0.0.33

Wed, 05 Mar 2025 09:55:15 GMT

PydanticAI v0.0.33 introduces a new Providers API for configuring model backends. • Adds a `Provider`s API for configuring and supplying model backends to agents.

PydanticAI v0.0.32

Tue, 04 Mar 2025 20:51:52 GMT

PydanticAI v0.0.32 adds an `instrument` param to Agent, supports Claude Sonnet 3.7 and Gemini 2.0 Pro Exp. • Adds `instrument` param to `Agent` to opt into OpenTelemetry tracing explicitly, replacing the previous always-on auto-instrumentation. • Adds support for `claude-sonnet-3-7` model. • Adds support for `gemini-2.0-pro-exp-02-05` model. Breaking changes: • OpenTelemetry instrumentation is now DISABLED by default; agents that relied on automatic tracing must now pass `instrument=True` to Agent(...) explicitly to restore telemetry.

PydanticAI v0.0.31

Mon, 03 Mar 2025 12:02:20 GMT

PydanticAI v0.0.31 adds recursive return types, async graph iteration, and improved OpenTelemetry instrumentation. • Supports recursive objects in `return_type`, enabling agents to return self-referential data structures. • Makes `Graph.iter` an async context manager, enabling asynchronous iteration over graph execution. • Replaces the `model request` span with `InstrumentedModel` for more structured OpenTelemetry tracing. • Replaces `all_messages` in the agent span with `all_messages_events`, aligning its format with the `InstrumentedModel` span. Breaking changes: • `HandleResponseNode` is renamed to `CallToolsNode` — any code referencing `HandleResponseNode` by name will break. • `Graph.iter` is now an async context manager — code using it as a sync context manager will break. • The `model request` span is replaced by `InstrumentedModel` — any telemetry pipelines filtering on the `model request` span name will no longer receive it. • The `all_messages` field in the agent span is replaced by `all_messages_events` — any telemetry pipelines reading `all_messages` from agent spans will no longer find it.

PydanticAI v0.0.30

Fri, 28 Feb 2025 13:00:55 GMT

PydanticAI v0.0.30 adds GPT-4.5 support, attributes mode for InstrumentedModel, and richer TestModel content inputs. • Adds `gpt-4.5-preview` as a supported model name for `OpenAIModel`. • Adds attributes mode to `InstrumentedModel` for OpenTelemetry instrumentation. • Supports different content input types in `TestModel` for richer test scenarios. • Replaces the existing streaming implementation with the .iter() API.

PydanticAI v0.0.29

Thu, 27 Feb 2025 16:15:47 GMT

PydanticAI v0.0.29 adds `max_results` parameter to the DuckDuckGo search tool. • Adds `max_results` parameter to the DuckDuckGo search tool to control the number of results returned per query.

PydanticAI v0.0.28

Thu, 27 Feb 2025 11:43:07 GMT

PydanticAI v0.0.28 adds DuckDuckGo and Tavily search tools, exposes `tool_call_id` on `RunContext`, and broadens Anthropic image MIME support. • Adds `tool_call_id` field to `RunContext`, giving tool implementations access to the specific call ID during execution. • Adds `DuckDuckGoSearch` built-in tool for agent web search without an API key. • Adds `TavilySearch` built-in tool for agent web search via the Tavily API. • Broadens accepted MIME types for `ImageUrl` when using Anthropic models, enabling a wider range of image formats.

PydanticAI v0.0.27

Wed, 26 Feb 2025 09:27:51 GMT

PydanticAI v0.0.27 adds FallbackModel for automatic model failover • Adds `FallbackModel` class to enable automatic failover across multiple LLM backends when a model call fails.

PydanticAI v0.0.26

Tue, 25 Feb 2025 08:15:39 GMT

PydanticAI v0.0.26 adds multimodal input support for agents. • Adds multimodal input support, enabling agents to accept non-text content (e.g. images, audio) alongside text messages.

PydanticAI v0.0.25

Mon, 24 Feb 2025 10:21:24 GMT

PydanticAI v0.0.25 adds InstrumentedModel with OTel/streaming support and a new GraphRun object for ergonomic agent graph traversal. • Adds `InstrumentedModel` class to wrap any model with OpenTelemetry instrumentation, using raw OTel and actual event loggers. • Adds `request_stream` support to `InstrumentedModel`, enabling streaming calls alongside standard instrumented requests. • Adds `GraphRun` object to make use of `next` more ergonomic when iterating agent graph execution. • Adds placeholder API key support for OpenAI-compatible models, easing integration with local or third-party OpenAI-compatible endpoints. Breaking changes: • The `name` methods are removed from OpenAI and Mistral model classes; any code calling those methods will break on upgrade.

PydanticAI v0.0.24

Wed, 12 Feb 2025 22:47:02 GMT

PydanticAI v0.0.24 adds Gemini 2.0 production models and populates ModelResponse.model_name from live responses. • Populates `ModelResponse.model_name` automatically from actual model responses, giving agents visibility into which model variant handled a request. • Adds new Gemini 2.0 models for production use.

PydanticAI v0.0.23

Fri, 07 Feb 2025 17:39:26 GMT

PydanticAI v0.0.23 adds o3 model support, OpenAI reasoning_effort param, and Gemini safety settings. • Supports `reasoning_effort` parameter for `OpenAIModel`, enabling control over reasoning depth on compatible OpenAI models. • Adds `o3` model support to `OpenAIModel`. • Adds Gemini safety settings support for configuring content safety thresholds on Gemini models. Breaking changes: • The `AgentModel` class has been removed.

PydanticAI v0.0.22

Tue, 04 Feb 2025 21:58:56 GMT

PydanticAI v0.0.22 adds new Gemini experimental models and support for locally served models without API keys. • Supports locally served models that do not require an API key, enabling use of self-hosted LLM endpoints. • Adds new Gemini experimental models to the supported model list. • Ports `pydantic_ai.Agent` internals to use `pydantic_graph` as its execution backend.

PydanticAI v0.0.21

Thu, 30 Jan 2025 09:04:59 GMT

PydanticAI v0.0.21 adds model-specific ModelSettings subclasses, drops OllamaModel in favor of OpenAIModel, and adds Cohere support. • Adds subclasses of `ModelSettings` to support specialized, per-model request parameters beyond the base settings common to all models. • Removes `OllamaModel` — Ollama is now used via `OpenAIModel` with the appropriate base URL, consolidating provider support. • Adds `Cohere` model support with documentation and live tests. Breaking changes: • `OllamaModel` has been removed; existing code using `OllamaModel` must be migrated to use `OpenAIModel` with Ollama's OpenAI-compatible endpoint. • `ArgsDict` and `ArgsJson` have been removed from the public API. • `AgentDeps` type alias is renamed to `AgentDepsT`.

PydanticAI v0.0.20

Fri, 24 Jan 2025 01:48:42 GMT

PydanticAI v0.0.20 adds Cohere model support, Anthropic streaming, parallel tool calls, and DeepSeek-R1 via Ollama. • Adds `parallel_tool_calls` field to `ModelSettings` to control whether the model may invoke multiple tools simultaneously. • Adds `model_name` field to `ModelResponse`, exposing which model produced each response. • Adds support for Cohere models as a new model provider integration. • Adds `'deepseek-r1'` to the recognized Ollama model name list, enabling typed use of DeepSeek-R1 via Ollama. • Adds Anthropic streaming support, enabling streamed responses from Anthropic models. • Adds support for `user`-role system prompts for `o1-preview-2024-09-12` to work around that model's system-prompt restrictions. • Adds direction control for Mermaid state diagram generation. Breaking changes: • Removes `from_text` and `from_tool_call` utilities, which will break any code that imports or calls these methods.

PydanticAI v0.0.19

Wed, 15 Jan 2025 20:17:00 GMT

PydanticAI v0.0.19 adds graph support, tool docstring controls, streaming refactor, and phi4 on Ollama. • Adds `docstring_format` and `require_parameter_descriptions` parameters to tool definitions, giving callers explicit control over how tool schemas are generated from Python docstrings. • Introduces graph support via `pydantic_ai.graph`, enabling stateful, multi-step agent workflows modelled as explicit graphs. • Adds `phi4` model support to the Ollama provider. • Refactors streaming internals, improving the reliability and composability of streamed agent responses.

PydanticAI v0.0.18

Tue, 07 Jan 2025 23:42:36 GMT

PydanticAI v0.0.18 adds dynamic system prompts, per-run custom result types, and provider-prefixed model names. • Adds `dynamic` parameter to the `system_prompt` decorator, enabling system prompts to be re-evaluated on each agent run rather than computed once at definition time. • Supports custom `result_type` overrides on individual .run() calls, allowing the expected output type to be set per-run without changing the agent definition. • All model names are now prefixed with their provider (e.g. `openai:gpt-4o`) for consistency across providers. Breaking changes: • All models are now prefixed with their provider name for consistency — any hardcoded unprefixed model strings passed to agents may need to be updated to the new provider-prefixed format.

PydanticAI v0.0.17

Fri, 03 Jan 2025 12:20:45 GMT

PydanticAI v0.0.17 defaults AgentDeps to None and adds formatting examples support • Adds formatting examples support to improve how examples are structured and displayed. • `AgentDeps` now defaults to `None`, simplifying agent definitions that do not require explicit dependency typing.

PydanticAI v0.0.16

Mon, 30 Dec 2024 20:31:11 GMT

PydanticAI v0.0.16 adds multi-agent support, extends RunContext, and brings Ollama API key config and nested capture_run_messages. • Adds `capture_run_messages` support for nested agent calls, enabling message capture across multi-agent workflows. • Extends `RunContext` with additional fields/methods to expose more context inside tool and result functions. • Adds Ollama API key configuration support for authenticating against Ollama endpoints. • Introduces multi-agent usage patterns, allowing agents to delegate to or call other agents. • Adds support for `X | None = None` optional type annotations with the Gemini provider.

PydanticAI v0.0.15

Mon, 23 Dec 2024 16:56:53 GMT

PydanticAI v0.0.15 adds `capture_run_messages` for message capture and optimizes Mistral model support. • Adds `capture_run_messages` to replace `last_run_messages` for capturing run messages. • Adds a default to `ResultData` so agents no longer require an explicit result-type argument. • Optimizes Mistral model integration for improved performance. • Tool calls are now prioritized over eager text responses when a model returns both. Breaking changes: • `last_run_messages` is removed; replace all uses with `capture_run_messages`.

PydanticAI v0.0.14

Thu, 19 Dec 2024 18:35:29 GMT

PydanticAI v0.0.14 adds usage limits, renames Cost to Usage, and supports the openai:o1 model. • Renames `Cost` to `Usage` across the library — callers must update any references to the old name. • Adds support for usage limits via the new `Usage` tracking infrastructure. • Adds `openai:o1` model support. Breaking changes: • `Cost` is renamed to `Usage` — any code referencing `Cost` will break on upgrade.

PydanticAI v0.0.13

Mon, 16 Dec 2024 21:57:14 GMT

PydanticAI v0.0.13 adds Mistral and Anthropic support, new ModelSettings, Gemini 2.0 Flash, and a reworked message format. • Adds `base_url` kwarg to `OpenAIModel` to point the client at any OpenAI-compatible endpoint (e.g. local or self-hosted inference servers). • Adds `messages` field to `RunContext` so tool functions can inspect the full conversation history mid-run. • Adds `ToolReturnPart` message part to `pydantic_ai.messages`, emitted for every tool call result and included in the message stream. • Adds basic `ModelSettings` class for passing model-level configuration (temperature, etc.) to agent runs. • Adds Mistral model support as a new first-class provider. • Adds non-streaming Anthropic model support. • Adds `gemini-2.0-flash-exp` to the supported Gemini model names. • Adds `llama-3.3-70b-versatile` to `GroqModelName`. • Supports tool calling when a structured result type is also provided, allowing both to be used simultaneously. • Reformats message history as a simple `list[ModelRequest | ModelResponse]`, unifying request and response representations across all providers. • Streamed response messages are now captured and included in the message history. Breaking changes: • The message format has changed significantly; existing stored or serialized messages are incompatible with the new `list[ModelRequest | ModelResponse]` structure. • `ToolReturnPart` is now emitted for every tool call, adding more message parts than previous releases — code that iterates or counts message parts will see different results. • The field `tool_id` has been renamed to `tool_call_id` across message types.

PydanticAI v0.0.12

Mon, 09 Dec 2024 00:19:43 GMT

PydanticAI v0.0.12 adds Ollama support, dynamic tools, and tool-result generation for structured outputs. • Adds Ollama as a supported model provider, enabling local LLM inference within PydanticAI agents. • Introduces dynamic tools, allowing tool definitions to be resolved or modified at runtime rather than statically at agent construction. • Enables tool-result generation when using structured result types, so structured-output workflows now produce proper tool result messages alongside the response.

PydanticAI v0.0.10

Fri, 06 Dec 2024 12:54:12 GMT

PydanticAI v0.0.10 adds Agent.name for identifying agents in traces and logs. • Adds `Agent.name` attribute to label agent instances, enabling identification in observability output.

PydanticAI v0.0.9

Wed, 04 Dec 2024 11:06:43 GMT

PydanticAI v0.0.9 lets you register tools at Agent construction time and return any type from tool functions. • Adds `tools` parameter to Agent(tools_=[...]) constructor, allowing tools to be registered at instantiation rather than only via decorators. • Allows tool functions to return `Any` type, removing the previous restriction that tool return values had to be a specific type.

PydanticAI v0.0.6

Mon, 25 Nov 2024 22:28:10 GMT

PydanticAI v0.0.6 adds Vertex AI and Groq provider support, plus a slim install option. • Adds `VertexAI` model support, including recognition of Vertex AI models in `infer_model`. • Adds Groq client support as a new LLM provider. • Introduces `pydantic-ai-slim` as a minimal install target via uv workspaces, with OpenAI now an optional dependency.

PydanticAI v0.0.3

Tue, 19 Nov 2024 00:24:42 GMT

PydanticAI v0.0.3 adds streamed responses, dependency override support, and expanded Gemini model coverage. • Adds streamed response support, enabling agents to consume model output incrementally as it arrives. • Allows overriding dependencies at runtime (e.g. in testing) via the new dependency override mechanism. • Changes `Agent` initialization to accept a deps type rather than a deps instance, decoupling agent definition from runtime dependencies. • Expands Gemini model coverage with additional test and integration support. • Adds a chat application example with streaming support to the examples library. Breaking changes: • The `Agent` constructor now takes a deps type instead of a deps instance — existing code passing a deps object directly will need to be updated. • `ToolCall` has been renamed to `Structured` in most places — code referencing `ToolCall` by name will break.

PydanticAI v0.0.2

Wed, 30 Oct 2024 14:05:30 GMT

PydanticAI v0.0.2 adds timezone support and `TypeAliasType` union handling. • Supports `TypeAliasType` unions, allowing type aliases defined with Python's `TypeAliasType` to be used in agent result schemas and tool signatures. • Adds timezone support to datetime handling within agents.