Skip to content

Google Gemini

Autobot supports Google Gemini as an LLM provider via the OpenAI-compatible endpoint. This gives access to Gemini Pro, Flash, and other Google AI models.

Setup

1. Get an API key

Create an API key at aistudio.google.com.

2. Configure credentials

Add your API key to the .env file:

GEMINI_API_KEY=...

Or use the interactive setup:

autobot setup
# Select "Google Gemini" as provider

3. Configure the provider

In config.yml:

agents:
  defaults:
    model: "gemini/gemini-2.5-flash"

providers:
  gemini:
    api_key: "${GEMINI_API_KEY}"

4. Verify

autobot doctor
# Should show: ✓ LLM provider configured (gemini)

Model naming

Models use the gemini/ prefix followed by the Google model ID:

# Gemini 2.5
model: "gemini/gemini-2.5-pro"
model: "gemini/gemini-2.5-flash"

# Gemini 2.0
model: "gemini/gemini-2.0-flash"

# Gemini 1.5
model: "gemini/gemini-1.5-pro"
model: "gemini/gemini-1.5-flash"

The gemini/ prefix tells autobot to route to the Gemini API. It is stripped before sending to the API.

See the full model list in the Gemini docs.

Configuration reference

Field Required Default Description
api_key Yes — Google AI API key
api_base No https://generativelanguage.googleapis.com/v1beta/openai/chat/completions Custom API endpoint
extra_headers No — Additional HTTP headers for every request

How it works

Gemini uses Google's OpenAI-compatible endpoint, which follows the standard Chat Completions format:

  • Authorization: Bearer header for authentication
  • Standard message format with role and content fields
  • Function calling via tools array

Autobot detects Gemini models by the gemini keyword and routes to Google's OpenAI-compatible endpoint automatically. Tools, MCP servers, plugins, and all other features work the same as with other providers.

Note that Gemini's error responses may use an array-wrapped format ([{"error": {...}}]). Autobot handles both standard and array-wrapped error formats transparently.

Voice transcription

Gemini does not provide a Whisper-compatible transcription API. If you need voice message support, configure an additional Groq or OpenAI provider for Whisper-based transcription.

Known limitations

  • No streaming — Responses are returned in full after the model finishes generating.
  • Tool choice is always auto — There is no configuration to force a specific tool or disable tool use per-request.
  • Free tier limits — Google AI Studio has per-minute and per-day request limits on the free tier.

Troubleshooting

Enable debug logging to see request/response details:

LOG_LEVEL=DEBUG autobot agent -m "Hello"

Look for:

  • POST https://generativelanguage.googleapis.com/... model=... — confirms provider is active
  • Response 200 (N bytes) — confirms API response
  • HTTP 4xx/5xx: ... — API errors with details

Common issues

"No LLM provider configured" — Check that api_key is set and non-empty in config.yml.

"API error: API key not valid" — Invalid or expired API key. Verify at aistudio.google.com.

"API error: Resource has been exhausted" — Rate limit or quota exceeded. Check your usage and limits in Google AI Studio.

"API error: model not found" — Model ID is wrong or not available. Check the Gemini models page for current availability.