OpenAIEmbeddingOperator¶

Use the OpenAIEmbeddingOperator to interact with the OpenAI API to create embeddings for given text.

Using the Operator¶

The OpenAIEmbeddingOperator requires the input_text as an input to embedding API. Use the conn_id parameter to specify the OpenAI connection to use to connect to your account.

An example of using the operator:

tests/system/openai/example_openai.py[source]

    OpenAIEmbeddingOperator(
        task_id="embedding_using_xcom_data",
        conn_id="openai_default",
        input_text=task_to_store_input_text_in_xcom(),
        model="text-embedding-3-small",
    )

    OpenAIEmbeddingOperator(
        task_id="embedding_using_callable",
        conn_id="openai_default",
        input_text=input_text_callable(
            "input_arg1_value",
            "input2_value",
            input_kwarg1="input_kwarg1_value",
            input_kwarg2="input_kwarg2_value",
        ),
        model="text-embedding-3-small",
    )
    OpenAIEmbeddingOperator(
        task_id="embedding_using_text",
        conn_id="openai_default",
        input_text=texts,
        model="text-embedding-3-small",
    )

OpenAIResponseOperator¶

Use the OpenAIResponseOperator to generate a model response with the OpenAI Responses API, OpenAI’s recommended interface for text generation and tool use. The operator returns the response’s aggregated output text.

Using the Operator¶

The OpenAIResponseOperator requires the input_text prompt. Use the conn_id parameter to specify the OpenAI connection to use, and response_kwargs to pass through options such as tools, conversation or previous_response_id.

tests/system/openai/example_openai.py[source]

    OpenAIResponseOperator(
        task_id="openai_response",
        conn_id="openai_default",
        input_text="Write a haiku about data pipelines.",
        response_kwargs={"instructions": "You are a helpful assistant."},
    )

Using the OpenAIHook for Responses and Conversations¶

The OpenAIHook exposes the Responses and Conversations APIs directly for use inside @task functions or custom operators:

Responses: create_response, get_response, delete_response and cancel_response (the last cancels a response created with background=True).
Conversations: create_conversation, get_conversation, update_conversation and delete_conversation. Pass the conversation id to create_response (via the operator’s response_kwargs or the hook) to persist state across responses.

For example, to create a conversation and continue it across responses:

hook = OpenAIHook()
conversation = hook.create_conversation()
hook.create_response(input="Hello", conversation=conversation.id)

Note

The Assistants/Threads hook methods (create_assistant, create_thread, create_run and related) are deprecated, mirroring OpenAI’s deprecation of the Assistants API. Migrate to the Responses and Conversations methods above.

OpenAITriggerBatchOperator¶

Use the OpenAITriggerBatchOperator to interact with the OpenAI API to trigger a batch job. This operator is used to trigger a batch job and wait for the job to complete.

Using the Operator¶

The OpenAITriggerBatchOperator requires the prepared batch file as an input to trigger the batch job. Provide the file_id and the endpoint to trigger the batch job, and use the conn_id parameter to specify the OpenAI connection to use.

An example of using the operator:

tests/system/openai/example_trigger_batch_operator.py[source]

    from airflow.providers.openai.operators.openai import OpenAITriggerBatchOperator

    batch_id = OpenAITriggerBatchOperator(
        task_id="batch_operator_deferred",
        conn_id=OPENAI_CONN_ID,
        file_id=batch_file_id,
        endpoint="/v1/chat/completions",
        deferrable=True,
    )