Google Gemini proxy

POST /v1/generateContent— calls the Google Generative Language API through mintoken. The request/response shape matches Google's own API.

Basic example

curl https://api.mintoken.in/v1/generateContent \
  -H "Authorization: Bearer mt_live_xxxxx" \
  -H "X-Provider-Key: <your-google-ai-key>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-1.5-flash",
    "contents": [
      {
        "parts": [
          {"text": "Explain connection pooling"}
        ]
      }
    ]
  }'

Model field

Unlike OpenAI/Anthropic (model goes in the body), Google's native API puts the model in the URL path (/models/gemini-1.5-flash:generateContent). To keep mintoken's interface consistent, mintoken reads the model from the request body's model field and constructs the upstream path for you.

System instructions

Gemini's equivalent of a system prompt is systemInstruction. Mintoken prepends its compression rules to the first text part — if you don't set one, we create it for you.

Supported features

• Tool use (function calling) via tools
• Multi-modal input (images, audio, video via base64/URL parts)
• Structured output via responseMimeType
• Safety settings passed through as-is

Common models

• gemini-1.5-flash — default pick; cheap and fast
• gemini-1.5-pro — higher quality, larger context
• gemini-2.0-flash / gemini-2.5-pro — newer generations, check availability on your project