Skip to main content

Use OpenAI models

LangExtract runs on Google Gemini by default. To use an OpenAI model instead, install the OpenAI extra, provide an OpenAI key, and pass an OpenAI model_id. Your prompt, examples, and result handling stay the same.

1. Install the OpenAI extra​

The OpenAI integration ships as an optional dependency:

pip install langextract[openai]

2. Provide your OpenAI key​

For OpenAI models, LangExtract reads OPENAI_API_KEY (falling back to LANGEXTRACT_API_KEY). See Supply API keys for all the options.

export OPENAI_API_KEY="your-openai-key"

3. Run an extraction with an OpenAI model​

Pass an OpenAI model_id such as gpt-4o. LangExtract selects the provider from the model id automatically. See Model backends.

import langextract as lx

examples = [
    lx.data.ExampleData(
        text="Maria felt hopeful as she opened the letter.",
        extractions=[
            lx.data.Extraction(extraction_class="person", extraction_text="Maria"),
            lx.data.Extraction(extraction_class="emotion", extraction_text="hopeful"),
        ],
    )
]

result = lx.extract(
    text_or_documents="Tom looked nervous before the interview.",
    prompt_description="Extract each person and the emotion they feel.",
    examples=examples,
    model_id="gpt-4o",
)

Batch mode​

For large jobs, the OpenAI provider can use the OpenAI Batch API. Enable it through language_model_params with a batch configuration; the threshold is the number of prompts at which batching starts.

result = lx.extract(
    text_or_documents=documents,
    prompt_description=prompt,
    examples=examples,
    model_id="gpt-4o",
    language_model_params={"batch": {"enabled": True, "threshold": 50}},
)

See also​