Google Gemini¶
Google's Gemini models via the official Google Generative AI API.
Configuration¶
PROVIDER=gemini
API_KEY=your-google-ai-api-key
MODEL=gemini-1.5-flash
TEMPERATURE=0.3
TOP_P=1.0
MAX_TOKENS=1024
Key Parameters¶
| Parameter | Description | Default |
|---|---|---|
API_KEY |
Google AI API key (required) | - |
MODEL |
Model to use | - |
TEMPERATURE |
Randomness (0.0-1.0) | 0.3 |
TOP_P |
Nucleus sampling | 1.0 |
TOP_K |
Top-k sampling | - |
MAX_TOKENS |
Max response tokens | 1024 |
TIMEOUT |
Request timeout (seconds) | 60 |
PRESENCE_PENALTY |
Presence penalty | - |
FREQUENCY_PENALTY |
Frequency penalty | - |
SEED |
Random seed for deterministic output | - |
INCLUDE_THOUGHTS |
Include reasoning | true |
THINKING_BUDGET |
Max tokens for reasoning | 1000 |
BASE_URL |
Custom API endpoint | https://generativelanguage.googleapis.com/v1beta |
API_VERSION |
API version to use | - |
Features¶
- ✅ Streaming responses
- ✅ Function calling
- ✅ MCP support
- ✅ Explicit reasoning via
ThinkingConfig - ✅ Multimodal capabilities
- ✅ Deterministic output with seed parameter
Important Notes¶
- API key requires Google AI Studio account
- System prompts are supported
- Message history is properly maintained
- Uses the
google-genaiclient library - Pay-as-you-go billing through Google Cloud
- Free tier with generous limits available