Gemini compatibility

Gemini model IDs are encoded in the request path. SayGM supports generateContent and SSE streamGenerateContent.

Python

import os
from google import genai
from google.genai import types

client = genai.Client(
    api_key=os.environ["GM_API_KEY"],
    http_options=types.HttpOptions(base_url="https://api.saygm.com"),
)

response = client.models.generate_content(
    model="gemini-3.5-flash",
    contents="Say hello.",
)
print(response.text)

REST

curl "https://api.saygm.com/v1beta/models/gemini-3.5-flash:generateContent" \
  -H "Authorization: Bearer $GM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{"role": "user", "parts": [{"text": "Say hello."}]}]
  }'

For streaming, use :streamGenerateContent?alt=sse. SayGM requires alt=sse and does not support Gemini’s legacy JSON-array streaming format.

Usage and cost

Every successful response carries settled cost inside usageMetadata, alongside the normal token counts, using Gemini’s camelCase convention:

{
  "usageMetadata": {
    "promptTokenCount": 12,
    "candidatesTokenCount": 9,
    "totalTokenCount": 21,
    "cost": 0.000140000,
    "costNanoUsd": 140000
  }
}

cost is USD, rendered as a JSON number with exactly nine fractional digits — the amount actually settled against the account’s prepaid balance, not an estimate from /v1/models rates. costNanoUsd is the same amount as an integer in nano-USD (1 nUSD = 10⁻⁹ USD), matching GET /v1/credits. Error responses carry neither field. See per-request cost for how to use the figure.

The Google GenAI SDK discards fields it does not recognize, so response.usage_metadata does not expose the cost. Read the raw HTTP response when you need it.