Google Gemini Adapter¶

TokenPak supports Google Gemini models via the OpenAI-compatible endpoint.

Setup¶

export GEMINI_API_KEY=your-gemini-api-key

Python (via openai SDK)¶

import openai

client = openai.OpenAI(
    base_url="http://localhost:8766/v1",
    api_key="your-gemini-api-key",
)

response = client.chat.completions.create(
    model="gemini-1.5-flash",
    messages=[{"role": "user", "content": "Hello from Gemini!"}],
)
print(response.choices[0].message.content)

Configuration¶

Variable	Default	Description
`GEMINI_API_KEY`	—	Google AI Studio or Vertex AI key
`TOKENPAK_MODE`	`hybrid`	Compression mode (same as other providers)

Notes¶

Gemini API uses an OpenAI-compatible /v1/chat/completions format via Google AI Studio
For Vertex AI, set TOKENPAK_VERTEX_PROJECT and TOKENPAK_VERTEX_LOCATION
Model names: gemini-1.5-flash, gemini-1.5-pro, gemini-2.0-flash-exp