Pydantic AI (Snowflake Cortex) connection¶

The pydanticai_snowflake connection type configures access to Snowflake Cortex’s OpenAI-compatible chat endpoint via the pydantic-ai framework. It backs PydanticAISnowflakeHook, the dedicated subclass of PydanticAIHook for Snowflake Cortex.

Unlike using a plain pydanticai connection with a snowflake: model and pydantic-ai’s own SNOWFLAKE_ACCOUNT / SNOWFLAKE_TOKEN environment variables, this connection type sources credentials from an existing Snowflake connection instead (the same connection your other Snowflake hooks use): the credential lives in one place, goes through Airflow’s secrets backend, can be rotated in one place, and – for key-pair JWT authentication – is refreshed automatically as the token nears expiry, which a static environment variable cannot do.

Install with:

pip install 'apache-airflow-providers-snowflake[common.ai]'

Default Connection IDs¶

The PydanticAISnowflakeHook uses pydanticai_snowflake_default by default.

Configuring the Connection¶

This connection type needs two connections: this one, and an existing snowflake connection it points at. All fields below are extra (JSON) fields on the pydanticai_snowflake connection; Schema, Port, Login, Host, and Password are hidden in the connection form because credentials and host come from the Snowflake connection instead.

Model

Cortex model identifier (e.g. snowflake:claude-4-sonnet).

A bare name is automatically resolved to snowflake:<name> – Snowflake Cortex is this connection type’s own platform.

Fallback Connections

Other connection IDs to fail over to, in order, while this provider is unavailable. Stored in extra["fallback_conn_ids"]. Entries may name any pydanticai connection type, so one chain can span vendors. See Provider fallback.

Snowflake Connection ID

Connection ID of an existing Snowflake connection to source credentials, account, and host from. Stored in extra["snowflake_conn_id"]. Also settable via the PydanticAISnowflakeHook(snowflake_conn_id=...) constructor argument, which takes precedence over the extra field. One of the two is required.

Authentication¶

Credentials are not configured on this connection – they come from whichever authentication method the referenced Snowflake connection uses, set the same way as for SnowflakeSqlApiHook and SnowflakeCortexAgentHook:

OAuth: set authenticator to oauth in the Snowflake connection’s extra, and configure a refresh token, a client credentials grant, or azure_conn_id.
PAT (Programmatic Access Token): set authenticator to programmatic_access_token and put the PAT value in the Snowflake connection’s Password field.
Key-pair JWT: the default when neither of the above is set. Configure private_key_file or private_key_content (optionally with a passphrase in Password), and set Login to the Snowflake user name the key is registered to.

See the Snowflake connection page for the full field reference.

Example¶

Two connections: the Snowflake connection holding credentials, and this one pointing at it.

{
    "conn_type": "snowflake",
    "extra": "{\"account\": \"myorg-myaccount\", \"authenticator\": \"programmatic_access_token\"}",
    "password": "<programmatic_access_token>"
}

{
    "conn_type": "pydanticai_snowflake",
    "extra": "{\"model\": \"snowflake:claude-4-sonnet\", \"snowflake_conn_id\": \"snowflake_default\"}"
}

Model family support¶

Support varies by model family, as defined by pydantic-ai’s SnowflakeProvider.model_profile (check it for the version you have installed):

Model family	Tools	Structured output
Claude (`claude*`)	Yes	Native (JSON schema)
OpenAI (`openai-*`)	Yes	Native
Others (`llama`, `snowflake-llama`, `mistral`, `mixtral`, `deepseek*`, and any unlisted family)	No	Prompted fallback

For a tool-using agent (AgentOperator, or any agent with toolsets), use a Claude or OpenAI family model.