Skip to content

gemini.md

Latest commit

 

History

History
72 lines (54 loc) · 2.1 KB

gemini.md

File metadata and controls

72 lines (54 loc) · 2.1 KB
Raw

Gemini

📁 GitHub Repository: HanaokaYuzu/Gemini-API

Credentials

  1. Go to https://gemini.google.com and sign in
  2. Open DevTools (F12) → Network tab → reload → click any request to gemini.google.com
  3. In Request Headers, find the Cookie: line and copy the entire value (or use Application → Cookies to get individual cookie values)
Cookie name Env var Required
__Secure-1PSID part of GEMINI_COOKIE ✅
__Secure-1PSIDTS part of GEMINI_COOKIE ❌ (account-dependent)

Cookies expire. Re-extract them when requests start failing.

Environment

Env var Required Description
GEMINI_COOKIE ✅ Full cookie string: __Secure-1PSID=xxx; __Secure-1PSIDTS=yyy

Endpoints

Endpoint Method Description
/v1/gemini/models GET List available models
/v1/gemini/chat/completions POST Chat completions (streaming supported)

Chat Completions — Additional fields

Beyond the common OpenAI-compatible fields, Gemini chat completions support:

Field Type Required Description
files array ❌ File paths to attach (images, PDFs, etc.)

Thinking models return reasoning_content in choices[].message.

Examples

# Basic chat
curl -s http://localhost:8000/v1/gemini/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-flash",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

# Streaming
curl -s http://localhost:8000/v1/gemini/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-flash",
    "messages": [{"role": "user", "content": "Hello!"}],
    "stream": true
  }'

# Attach files
curl -s http://localhost:8000/v1/gemini/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-flash",
    "messages": [{"role": "user", "content": "What is in this image?"}],
    "files": ["/path/to/image.jpg"]
  }'

# List models
curl -s http://localhost:8000/v1/gemini/models