跳转到主要内容
Qhaigc supports the native Gemini API format through the /v1beta/models/{model}:generateContent path. Send requests exactly as you would to the Google AI API — Qhaigc proxies and authenticates them for you using your Qhaigc API key, so you do not need a separate Google API key. Endpoint: POST https://api.qhaigc.net/v1beta/models/{model}:generateContent Replace {model} in the path with the model name, such as gemini-2.5-flash or gemini-2.5-pro.
You can also pass "style": "gemini" in the body of POST /v1/chat/completions to receive a Gemini-format response from the unified endpoint. The native path shown here is recommended when you are migrating existing Gemini SDK code.

​
Path Parameter

​
model
string
必填
The Gemini model name. Examples: gemini-2.5-flash, gemini-2.5-pro.

​
Request Parameters

​
contents
array
必填
The conversation contents. Each element has a role ("user" or "model") and a parts array.
​
generationConfig
object
Generation parameters.
​
safetySettings
array
Safety filter configuration. Pass an array of category/threshold objects to adjust content filtering.
​
tools
array
Function calling tool definitions. Pass an array of tool objects to enable function calling.

​
Response Fields

​
candidates
array
Array of generated response candidates.
​
usageMetadata
object
Token usage for this request.

​
Code Examples

import requests

url = "https://api.qhaigc.net/v1beta/models/gemini-2.5-flash:generateContent"
headers = {
    "Authorization": "Bearer sk-your-api-key-here",
    "Content-Type": "application/json"
}

payload = {
    "contents": [
        {
            "role": "user",
            "parts": [{"text": "Explain how RAG works in three concise steps."}]
        }
    ],
    "generationConfig": {
        "temperature": 0.4,
        "maxOutputTokens": 512
    }
}

response = requests.post(url, headers=headers, json=payload)
data = response.json()
print(data["candidates"][0]["content"]["parts"][0]["text"])

​
Example Response

{
  "candidates": [
    {
      "content": {
        "role": "model",
        "parts": [
          {
            "text": "RAG works in three steps:\n\n1. **Retrieve** — search a knowledge base for documents relevant to the user's query...\n2. **Augment** — prepend those documents to the prompt as context...\n3. **Generate** — the LLM produces a grounded answer using the retrieved context."
          }
        ]
      },
      "finishReason": "STOP"
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 18,
    "candidatesTokenCount": 76,
    "totalTokenCount": 94
  }
}

​
Multi-Turn Conversation

To send a multi-turn conversation, include prior turns in the contents array, alternating "user" and "model" roles: 
payload = {
    "contents": [
        {"role": "user", "parts": [{"text": "What is a vector database?"}]},
        {"role": "model", "parts": [{"text": "A vector database stores embeddings..."}]},
        {"role": "user", "parts": [{"text": "Which ones are most popular?"}]}
    ],
    "generationConfig": {"temperature": 0.5, "maxOutputTokens": 256}
}