v0.0.27 2025-02-26

What's Changed

• Add FallbackModel support by @sydney-runkle in #894
• fix: send the right data format to gemini image input by @Kludex in #991

Full Changelog: v0.0.26...v0.0.27

v0.0.26 2025-02-25

What's Changed

• Support multimodal inputs by @Kludex in #971
• Fix agent graph types by @dmontagu in #983

Full Changelog: v0.0.25...v0.0.26

v0.0.25 2025-02-24

What's Changed

• Add InstrumentedModel by @alexmojaki in #892
• Fix JSON streaming unicode issues with gemini by @dmontagu in #917
• Remove stale name methods from openai and mistral models by @sydney-runkle in #925
• Add placeholder API key for OpenAI compatible models by @sydney-runkle in #929
• Add request_stream to InstrumentedModel by @alexmojaki in #922
• Add GraphRun object to make use of next more ergonomic by @dmontagu in #833
• Use raw OTel and actual event loggers in InstrumentedModel by @alexmojaki in #945
• fix: anthropic parallel tool call results. by @ioga in #653

New Contributors

• @ioga made their first contribution in #653

Full Changelog: v0.0.24...v0.0.25

v0.0.24 2025-02-12

What's Changed

• Remove now-invalid reference to graph extra from docs by @dmontagu in #873
• New gemini 2.0 models for production by @mvaldi in #880
• Remove async from _get_model by @alexmojaki in #893
• Populate ModelResponse.model_name from responses by @alexmojaki in #883
• Fix OpenAI Model for working with local Ollama without passing an API key by @mageos in #874
• Use abstract properties for otel semantic attributes on models and stream responses by @sydney-runkle in #904

New Contributors

• @mvaldi made their first contribution in #880
• @jexp made their first contribution in #876
• @mageos made their first contribution in #874

Full Changelog: v0.0.23...v0.0.24

v0.0.23 2025-02-07

What's Changed

• Fix error in evals example by @samuelcolvin in #852
• Add o3 support for OpenAIModel by @sydney-runkle in #853
• Remove the AgentModel class by @dmontagu in #800
• Support reasoning_effort param for OpenAIModel by @sydney-runkle in #854
• Fix OpenAI env var tests following by @samuelcolvin in #855
• Replace search UI with Algolia backed one by @petyosi in #845
• Add support for gemini safety settings by @dAIsySHEng1 in #790
• Fix: missing graph extra by @mplemay in #863
• Remove unnecessary project_id == creds_id check for vertexai by @sydney-runkle in #862
• Using model_name and system model properties by @sydney-runkle in #865

New Contributors

• @mplemay made their first contribution in #863

Full Changelog: v0.0.22...v0.0.23

v0.0.22 2025-02-04

What's Changed

• Port pydantic_ai.Agent to using pydantic_graph by @dmontagu in #725
• PR previews by @samuelcolvin in #815
• Various docs improvements by @sydney-runkle in #809
• Adding new gemini experimental models by @sydney-runkle in #811
• Various minor fixes by @dmontagu in #834
• Build docs for python<3.12 by @samuelcolvin in #841
• Weather Agent Sample: Only include tool_call_id to chatbot.append if it's set by @chug2k in #771
• Support locally served models which do not require an api key by @mikeedjones in #814

New Contributors

• @chug2k made their first contribution in #771
• @mikeedjones made their first contribution in #814

Full Changelog: v0.0.21...v0.0.22

v0.0.21 2025-01-30

What's Changed

• Update index.md by @kauabh in #759
• Add codespell support (config, workflow to detect/not fix) and make it fix few typos by @yarikoptic in #757
• Ensure custom args are used when retried by @jlowin in #692
• update AgentDeps refs to AgentDepsT by @zzstoatzz in #767
• Adding subclasses of ModelSettings to support specialized model requests by @sydney-runkle in #766
• Remove _utils.Either by @dmontagu in #768
• Remove ArgsDict and ArgsJson by @dmontagu in #769
• Fix typo in Google VertexAIModel documentation (mkdocs) by @chris-dare in #772
• Fix CI by removing references to shutdown groq models by @dmontagu in #780
• Updated Docs by @kauabh in #773
• Clean up known model names by @dmontagu in #781
• Update Function tools docs with usage of docstring_format and require_parameter_descriptions by @chris-dare in #776
• Disable running gemini tests in test_live.py by @dmontagu in #788
• Update graph.md by @kauabh in #791
• Minor clean up in preparation of graph agent by @dmontagu in #779
• Fix o1 usage with tool calls by @sydney-runkle in #764
• added documentation to visualize a graph within a jupyter-notebook by @jonathanbouchet in #786
• Remove OllamaModel in favor of usage with OpenAIModel by @sydney-runkle in #805
• Docs fix: use model provider prefix in all examples by @sydney-runkle in #806
• Add Cohere docs and additional (live) tests by @sydney-runkle in #810
• uprev to 0.0.21 by @samuelcolvin in #813

New Contributors

• @kauabh made their first contribution in #759
• @yarikoptic made their first contribution in #757
• @chris-dare made their first contribution in #772
• @jonathanbouchet made their first contribution in #786

Full Changelog: v0.0.20...v0.0.21

v0.0.20 2025-01-23

What's Changed

• Add missing allow_model_requests check by @YanSte in #697
• Improve Algolia indexing by @petyosi in #721
• Add support for Cohere models by @rafidka in #203
• Add model_name to ModelResponse by @sydney-runkle in #701
• Adding 'deepseek-r1' to the list of Ollama Model Names by @izzyacademy in #735
• Test minimum versions by @samuelcolvin in #743
• Fix handling of OpenAI system prompts in order to support o1 by @sydney-runkle in #740
• Removing from_text and from_tool_call utilities that complicate snapshot testing by @sydney-runkle in #744
• Add support for controlling direction of state diagram generated by Mermaid code by @izzyacademy in #716
• Increase default IsNow() delta by @dmontagu in #746
• Auto-use the set_event_loop fixture by @dmontagu in #747
• Make non-required parameters not required by @dmontagu in #742
• Improve variance of classes by @dmontagu in #726
• Support parallel_tool_calls in ModelSettings by @sydney-runkle in #750
• Add support for user role system prompts o1-preview-2024-09-12 by @siavashg in #754
• Anthropic streaming support by @piercefreeman in #684
• Fix an issue with retry counting by @dmontagu in #749

New Contributors

• @rafidka made their first contribution in #203
• @hrahmadi71 made their first contribution in #753
• @siavashg made their first contribution in #754
• @piercefreeman made their first contribution in #684

Full Changelog: v0.0.19...v0.0.20

v0.0.19 2025-01-15

What's Changed

• Docs: add asyncio to all examples that need it. by @ME-Msc in #638
• fix: typo by @hwong557 in #641
• Add Gradio Demo for weather agent example by @freddyaboulton in #230
• Docs: fix freeze frame for gradio demo video by @sydney-runkle in #644
• Support docstring_format and require_parameter_descriptions on tools by @sydney-runkle in #643
• Generate llms.txt by @petyosi in #649
• Refactor streaming by @dmontagu in #468
• Adds phi4 to Ollama by @josead in #668
• Use Algolia for docs search by @petyosi in #669
• Fix agent run/ run_sync docs by @KranthiGV in #688
• docs: fix typos in testing-evals and RAG example by @lealre in #694
• Graph Support by @samuelcolvin in #528
• Upgrade "inline-snapshot" by @samuelcolvin in #657
• uprev to v0.0.19 by @samuelcolvin in #696

New Contributors

• @ME-Msc made their first contribution in #638
• @hwong557 made their first contribution in #641
• @freddyaboulton made their first contribution in #230
• @petyosi made their first contribution in #649
• @KranthiGV made their first contribution in #688

Full Changelog: v0.0.18...v0.0.19

v0.0.18 2025-01-07

What's Changed

• Added docs contribution docs, a troubleshooting item for docs changes by @andrewdmalone in #600
• Improve null handling for Gemini by @dmontagu in #608
• Improve string format handling for gemini by @dmontagu in #609
• Suppress griffe logging related to return type by @dmontagu in #606
• readme links by @samuelcolvin in #622
• Docs: Add screenshot for chat-app example. by @hengoren in #612
• docs: add OpenAI-compatible models section for Grok (xAI) and DeepSeek by @imfing in #613
• add diff coverage check by @samuelcolvin in #623
• set UV_FROZEN globally in CI by @samuelcolvin in #626
• Adds dynamic to system_prompt decorator, allowing reevaluation by @josead in #560
• Use --reinstall instead of uninstall/install for docs insiders packages by @T-256 in #627
• Custom result_type on a run by @samuelcolvin in #629
• Prefix all models with provider for consistency by @sydney-runkle in #593
• prepare for v0.0.18 by @samuelcolvin in #635

New Contributors

• @hengoren made their first contribution in #612
• @josead made their first contribution in #560
• @T-256 made their first contribution in #627

Full Changelog: v0.0.17...v0.0.18