Installation¶

The provider needs Airflow 3.0 or later. Install it with the extra that matches the model vendor your connection will point at:

pip install "apache-airflow-providers-common-ai[openai]"

Quote the package name so the brackets survive your shell. Swap openai for anthropic, google or bedrock as needed, or combine several extras with a comma. The minimum versions of apache-airflow and pydantic-ai-slim are listed in the requirements table on the landing page.

Choosing extras¶

The provider’s extras split into a few groups:

Model providers (openai, anthropic, google, bedrock, typesafe): pick the one matching your llm_conn_id connection (Supported model providers maps vendors to extras, prefixes and connection types). typesafe differs from the rest in kind: it installs a classifier model that answers typed questions and cannot write text (see Classifier models). The first four mirror the identically named pydantic-ai-slim optional dependency groups, and typesafe adds the typesafe-sdk the built-in adapter talks to; pydantic-ai supports more model providers than these, each under its own extra name, so check the pydantic-ai install docs for the full list.
Agent tooling (mcp, skills, code-mode, shields, modal): MCP servers, Agent Skills, code-mode tool execution, shield capabilities (input/output guards, tool guards, cost tracking), and the hosted Modal backend for sandboxed execution.
Document loading (pdf, docx, avro, parquet): file formats for document pipelines.
Retrieval / SQL (sql, common.sql, langchain, llamaindex): RAG and SQL-schema tooling.
Git-backed content (git): pulling Agent Skills or documents from a git connection.

The Optional dependencies table on the landing page lists the exact package each extra installs.

Features gated on the Airflow version¶

The provider runs on Airflow 3.0, but some features need a newer core:

Feature	Needs
Approval gates and HITL review	Airflow 3.1
Retry policies	Airflow 3.3
Durable execution without configuring `[common.ai] durable_cache_path` (the task state store)	Airflow 3.3

Next steps¶

Quick start creates a connection and runs a first @task.llm.
Pydantic AI connection covers every connection field and the vendor-specific connection types.
Installing from sources covers verifying and installing a release artifact by hand.