Google AI Studio API Relay

Compatible with Google Generative Language API spec (Gemini / Veo / Lyria). Authenticate with an x-goog-api-key: YOUR_API_KEY or Authorization: Bearer YOUR_API_KEY header (never place keys in URL query strings; ?key= query authentication is disabled — the official Google SDK uses the header by default).

POST https://www.myai168.com/en/api/google/v1beta/models/{model}:generateContent

Gemini chat (non-streaming).

curl "https://www.myai168.com/en/api/google/v1beta/models/gemini-3.1-pro-preview:generateContent" \
  -H "x-goog-api-key: YOUR_DEV_KEY" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"parts":[{"text":"Hello"}]}]}'

import requests

r = requests.post(
    "https://www.myai168.com/en/api/google/v1beta/models/gemini-3.1-pro-preview:generateContent",
    headers={"x-goog-api-key": "YOUR_DEV_KEY"},
    json={"contents": [{"parts": [{"text": "Hello"}]}]},
)
print(r.json())

POST https://www.myai168.com/en/api/google/v1beta/models/{model}:streamGenerateContent

Gemini chat (SSE streaming). This relay always returns SSE (the server enforces alt=sse toward the upstream), so unlike calling the official Google API directly you do not need to add ?alt=sse to the URL (it is ignored if present).

curl "https://www.myai168.com/en/api/google/v1beta/models/gemini-3.1-pro-preview:streamGenerateContent" \
  -H "x-goog-api-key: YOUR_DEV_KEY" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"parts":[{"text":"Hello"}]}]}'

import requests

r = requests.post(
    "https://www.myai168.com/en/api/google/v1beta/models/gemini-3.1-pro-preview:streamGenerateContent",
    headers={"x-goog-api-key": "YOUR_DEV_KEY"},
    json={"contents": [{"parts": [{"text": "Hello"}]}]},
    stream=True,
)
for line in r.iter_lines():
    if line:
        print(line.decode("utf-8"))

POST https://www.myai168.com/en/api/google/v1beta/models/gemini-3-pro-image-preview:generateContent

Nano Banana Pro text-to-image (Gemini multimodal, returns inline_data base64 image).

curl "https://www.myai168.com/en/api/google/v1beta/models/gemini-3-pro-image-preview:generateContent" \
  -H "x-goog-api-key: YOUR_DEV_KEY" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"parts":[{"text":"A cat"}]}],"generationConfig":{"responseModalities":["TEXT","IMAGE"]}}'

import requests

r = requests.post(
    "https://www.myai168.com/en/api/google/v1beta/models/gemini-3-pro-image-preview:generateContent",
    headers={"x-goog-api-key": "YOUR_DEV_KEY"},
    json={"contents": [{"parts": [{"text": "A cat"}]}], "generationConfig": {"responseModalities": ["TEXT", "IMAGE"]}},
)
print(r.json())

POST https://www.myai168.com/en/api/google/v1beta/models/veo-{ver}:predictLongRunning

Veo video generation (LRO; first call returns operation name, poll for result).

curl "https://www.myai168.com/en/api/google/v1beta/models/veo-3.1-generate-preview:predictLongRunning" \
  -H "x-goog-api-key: YOUR_DEV_KEY" \
  -H "Content-Type: application/json" \
  -d '{"instances":[{"prompt":"A cat dancing"}],"parameters":{"durationSeconds":8}}'

import requests

r = requests.post(
    "https://www.myai168.com/en/api/google/v1beta/models/veo-3.1-generate-preview:predictLongRunning",
    headers={"x-goog-api-key": "YOUR_DEV_KEY"},
    json={"instances": [{"prompt": "A cat dancing"}], "parameters": {"durationSeconds": 8}},
)
print(r.json())

POST https://www.myai168.com/en/api/google/v1beta/models/lyria-{ver}:predict

Lyria music generation (returns base64 audio). Note: Google has moved Gemini API music generation to Lyria 3 (lyria-3-clip-preview / lyria-3-pro-preview via the new Interactions API), and the official docs no longer list the lyria-002 :predict interface — this relay keeps forwarding the legacy interface; if the upstream has dropped it you will get an upstream error (no credits deducted). Check Google's official announcements for availability.

curl "https://www.myai168.com/en/api/google/v1beta/models/lyria-002:predict" \
  -H "x-goog-api-key: YOUR_DEV_KEY" \
  -H "Content-Type: application/json" \
  -d '{"instances":[{"prompt":"Upbeat jazz"}],"parameters":{"sampleCount":1}}'

import requests

r = requests.post(
    "https://www.myai168.com/en/api/google/v1beta/models/lyria-002:predict",
    headers={"x-goog-api-key": "YOUR_DEV_KEY"},
    json={"instances": [{"prompt": "Upbeat jazz"}], "parameters": {"sampleCount": 1}},
)
print(r.json())

GET https://www.myai168.com/en/api/google/v1beta/operations/{name}

Query LRO operation status (used for Veo poll, query endpoint, no credits deducted).

curl "https://www.myai168.com/en/api/google/v1beta/operations/OP_NAME" \
  -H "x-goog-api-key: YOUR_DEV_KEY"

import requests

r = requests.get(
    "https://www.myai168.com/en/api/google/v1beta/operations/OP_NAME",
    headers={"x-goog-api-key": "YOUR_DEV_KEY"},
)
print(r.json())

Request parameters (generateContent / streamGenerateContent)

The request body is JSON, identical to the official Google spec. Authenticate with an x-goog-api-key or Authorization: Bearer header.

The model id goes in the URL path (models/{model}:generateContent); models not in the pricing table return "No pricing info! Please use other models!". Veo uses instances + parameters (durationSeconds / numberOfVideos / resolution) and Lyria uses instances + parameters.sampleCount (see the examples above).

Response format

Non-streaming: official JSON — text in candidates[0].content.parts[], usage in usageMetadata; response headers include X-Cost-Points (points charged) and X-Credits-Remaining (balance after the call). Streaming (:streamGenerateContent) is SSE; after the stream ends, SSE comment lines : cost=N / : credits-remaining=M carry the points charged and remaining balance. You can also query GET https://www.myai168.com/en/api/user/get_credits any time.

Error codes

No points are ever charged on errors (including when the vendor returns no usage info); polling LRO status (operations) is free.