Baseten Integration

Call models served on Baseten — both model APIs and your own dedicated deployments — from Fused UDFs. Your API key is stored securely as a Fused secret, no keys need to be hardcoded.

Prerequisites

1. A Fused account on a paid plan (execution environment required).
2. A Baseten API key.

Setup

Add your API key

Open Settings > Integrations & Secrets, find the Baseten section, and paste your API key. It is stored as a Fused secret under the name BASETEN_API_KEY.

Inside UDFs, retrieve it with fused.secrets["BASETEN_API_KEY"].

Quick start

Baseten exposes an OpenAI-compatible inference endpoint at https://inference.baseten.co/v1. The easiest way to call it is with the openai SDK, which is pre-installed in the Fused runtime.

@fused.udf()
def udf(prompt: str = "Summarize what GeoParquet is in one sentence."):
    import fused
    from openai import OpenAI

    client = OpenAI(
        api_key=fused.secrets["BASETEN_API_KEY"],
        base_url="https://inference.baseten.co/v1",
    )

    response = client.chat.completions.create(
        model="deepseek-ai/DeepSeek-V3.1",
        messages=[{"role": "user", "content": prompt}],
        max_tokens=256,
    )
    return response.choices[0].message.content

Common operations

Chat completion

@fused.udf()
def udf():
    import fused
    from openai import OpenAI

    client = OpenAI(
        api_key=fused.secrets["BASETEN_API_KEY"],
        base_url="https://inference.baseten.co/v1",
    )

    response = client.chat.completions.create(
        model="moonshotai/Kimi-K2-Instruct",
        messages=[
            {"role": "system", "content": "You are a concise geospatial assistant."},
            {"role": "user", "content": "Which projection should I use for area calculations in California?"},
        ],
        max_tokens=512,
    )
    return response.choices[0].message.content

Streaming

@fused.udf()
def udf():
    import fused
    from openai import OpenAI

    client = OpenAI(
        api_key=fused.secrets["BASETEN_API_KEY"],
        base_url="https://inference.baseten.co/v1",
    )

    stream = client.chat.completions.create(
        model="deepseek-ai/DeepSeek-V3.1",
        messages=[{"role": "user", "content": "Write a haiku about S3."}],
        stream=True,
    )

    chunks = []
    for chunk in stream:
        delta = chunk.choices[0].delta.content or ""
        chunks.append(delta)
    return "".join(chunks)

Batched calls in a DataFrame

@fused.udf()
def udf():
    import fused
    import pandas as pd
    from openai import OpenAI

    client = OpenAI(
        api_key=fused.secrets["BASETEN_API_KEY"],
        base_url="https://inference.baseten.co/v1",
    )

    prompts = [
        "Define remote sensing in one sentence.",
        "Define spatial join in one sentence.",
        "Define raster tiling in one sentence.",
    ]

    rows = []
    for p in prompts:
        r = client.chat.completions.create(
            model="deepseek-ai/DeepSeek-V3.1",
            messages=[{"role": "user", "content": p}],
            max_tokens=80,
        )
        rows.append({"prompt": p, "response": r.choices[0].message.content})
    return pd.DataFrame(rows)

Call a dedicated model deployment

For models you have deployed to your own Baseten workspace, call the model-specific endpoint directly with the API key in the Authorization header.

@fused.udf()
def udf():
    import fused
    import requests

    api_key = fused.secrets["BASETEN_API_KEY"]
    model_id = "abcd1234"  # your Baseten model ID

    response = requests.post(
        f"https://model-{model_id}.api.baseten.co/production/predict",
        headers={"Authorization": f"Api-Key {api_key}"},
        json={"prompt": "Hello, world!"},
        timeout=60,
    )
    response.raise_for_status()
    return response.json()

Disconnecting

To revoke the integration, go to Settings > Integrations & Secrets, find the Baseten section, and remove your API key. You should also revoke the key in the Baseten dashboard if it is no longer needed.