Skip to main content
The google-ai provider supports Google’s generateContent and streamGenerateContent endpoints for Gemini models.
BAML uses the v1beta endpoint to access both existing v1 models and additional v1beta-exclusive models, following Google’s SDK conventions.

​
Quick Start

client<llm> MyClient {
  provider "google-ai"
  options {
    model "gemini-2.5-flash"
  }
}

​
Authentication

Set your Google API key as an environment variable: 
export GOOGLE_API_KEY="your-api-key-here"
Or specify it explicitly in your BAML configuration: 
client<llm> MyClient {
  provider "google-ai"
  options {
    api_key env.MY_GOOGLE_KEY
    model "gemini-2.5-flash"
  }
}

​
Configuration Options

​
BAML-Specific Options

These options modify the API request sent to Google AI.
​
api_key
string
default:"env.GOOGLE_API_KEY"
Passed as the x-goog-api-key header.
​
base_url
string
The base URL for the Google AI API.
​
headers
object
Additional headers to send with requests.
client<llm> MyClient {
  provider "google-ai"
  options {
    model "gemini-2.5-flash"
    headers {
      "X-My-Header" "my-value"
    }
  }
}

​
Supported Models

​
model
string
default:"gemini-2.5-flash"
The Gemini model to use.
ModelUse CaseContextKey Features
gemini-2.5-proComplex tasks, coding, STEM1MAdaptive thinking, multimodal
gemini-2.5-flashProduction apps, balanced performance1MBest price/performance
gemini-2.5-flash-liteHigh-volume, cost-sensitive1MLowest cost, fastest
You can specify any model name - BAML won’t validate whether it exists.See the Google Model Documentation for the latest models.

​
Model Parameters

Parameters like temperature are specified in the generationConfig object. See the Google AI documentation for details.
client<llm> MyClient {
  provider "google-ai"
  options {
    model "gemini-2.5-flash"
    generationConfig {
      temperature 0.5
      maxOutputTokens 1024
      topP 0.8
      topK 40
    }
  }
}
Common generationConfig parameters:
  • temperature - Controls randomness (0-2)
  • maxOutputTokens - Maximum tokens to generate
  • topP - Nucleus sampling parameter
  • topK - Top-k sampling parameter
  • stopSequences - Array of sequences that stop generation
For all options, see the Google Gemini API documentation.

​
Media Handling

Google AI uses send_base64_unless_google_url by default for images, which preserves Google Cloud Storage URLs (gs://) while converting other URLs to base64.

​
Features

  • Streaming: Automatically uses streamGenerateContent when you call the streaming interface
  • Multimodal: Supports text, image, audio, and video inputs
  • Large Context: Up to 1M tokens context window
  • Function Calling: Native support for tool use

​
Do Not Set

​
contents
DO NOT USE
BAML automatically constructs this from your prompt.

Build docs developers (and LLMs) love

Get started for freeTalk to us