Gemini对话

POST

https://api.aicombo.com/v1/publishers/google/models/{model_name}:{action}

Vertex AI 原生协议的 Gemini 内容生成接口。
通过 {action} 参数统一覆盖非流式与流式两种调用模式：

generateContent → 非流式，等待完整响应后一次性返回

streamGenerateContent / streamGenerateContent?alt=sse→ 流式，以 SSE data 帧逐步返回生成内容

认证方式支持两种（二选一）：

Header: x-goog-api-key

Query: ?key=

请求参数

Path 参数

Header 参数

Body 参数application/json必填

{
  "contents": [
    {
      "role": "user",
      "parts": [
        {
          "text": "reply in one word: capital of France?"
        }
      ]
    }
  ]
}

请求示例代码

Shell

JavaScript

Java

Swift

PHP

Python

HTTP

Objective-C

Ruby

OCaml

Dart

curl --location 'https://api.aicombo.com/v1/publishers/google/models/:' \
--header 'x-goog-api-key: ' \
--header 'Content-Type: application/json' \
--data '{
  "contents": [
    {
      "role": "user",
      "parts": [
        {
          "text": "reply in one word: capital of France?"
        }
      ]
    }
  ]
}'

返回响应

🟢200Success

application/json

以下成功响应结构是 action=GenerateContent 时。
当 action=streamGenerateContent 时，每帧结构与本 Schema 相同，HTTP 传输时每帧前缀为 "data: "，帧间以空行分隔。

Bodyapplication/json

{
  "candidates": [
    {
      "content": {
        "role": "model",
        "parts": [
          {
            "text": "Paris"
          }
        ]
      },
      "finishReason": "STOP",
      "avgLogprobs": -13.343818664550781
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 9,
    "candidatesTokenCount": 1,
    "totalTokenCount": 121,
    "trafficType": "ON_DEMAND",
    "promptTokensDetails": [
      {
        "modality": "TEXT",
        "tokenCount": 9
      }
    ],
    "candidatesTokensDetails": [
      {
        "modality": "TEXT",
        "tokenCount": 1
      }
    ],
    "thoughtsTokenCount": 111
  },
  "modelVersion": "gemini-2.5-pro",
  "createTime": "2026-05-14T03:18:16.481821Z",
  "responseId": "-D4Fap20HYiQj8wP08ua4AU"
}

Gemini对话

原生Claude格式