promptfoo/site/docs/providers/palm.md

sidebar_position
43

Google AI Studio

The google provider is compatible with Google AI Studio (formerly known as PaLM), which offers access to Gemini models.

You can use it by specifying one of the available models. Currently, the following models are supported:

• google:gemini-pro
• google:gemini-pro-vision
• google:aqa (attributed question answering)
• google:chat-bison-001

:::tip If you are using Google Vertex, see the vertex provider. :::

Supported environment variables:

• GOOGLE_API_KEY (required) - Google AI Studio/PaLM API token
• GOOGLE_API_HOST - used to override the Google API host, defaults to generativelanguage.googleapis.com

The PaLM provider supports various configuration options such as safetySettings, stopSequences, temperature, maxOutputTokens, topP, and topK that can be used to customize the behavior of the model like so:

providers:
  - id: google:gemini-pro
    config:
      temperature: 0
      maxOutputTokens: 1024

You can also pass in a responseSchema file:

providers:
  - id: google:gemini-pro
    config:
      responseSchema: file://test.json

If you want more fine tuned control over the generationConfig:

providers:
  - id: google:gemini-1.5-flash-002
    config:
      generationConfig:
        response_mime_type: application/json
        response_schema:
          type: object
          properties:
            foo:
              type: string