Google Vertex AI API

本文按 Vertex AI 官方文档整理，完整展示 Vertex AI 上 Gemini 原生接口、Gemini OpenAI-compatible 入口、Claude partner model 入口、以及 OpenAI 风格 managed API 入口的请求地址、请求参数、请求示例、请求返回结果、请求返回示例。

鉴权方式

使用 Google Cloud Access Token。
常见 Header：Authorization: Bearer $(gcloud auth print-access-token)

请求地址

1. Vertex AI 上的 Gemini 原生接口

方法：POST
地址：https://aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/publishers/google/models/{model}:generateContent

2. Vertex AI 上的 Gemini OpenAI-compatible 入口

方法：POST
地址：https://aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/endpoints/openapi/chat/completions

3. Vertex AI 上的 Claude partner model 原生入口

方法：POST
地址：https://{location}-aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/publishers/anthropic/models/{model}:rawPredict

4. Vertex AI 上的 OpenAI 风格 managed API 入口

方法：POST
地址：https://{location}-aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/endpoints/openapi/chat/completions

请求参数

Gemini 原生接口请求参数

参数	类型	必填	说明
`contents`	array	是	对话内容。
`contents[].role`	string	否	`user` / `model`。
`contents[].parts`	array	是	文本、图片、文件等内容块。
`systemInstruction`	object	否	系统提示。
`tools`	array	否	工具定义。
`toolConfig`	object	否	工具配置。
`safetySettings`	array	否	安全设置。
`generationConfig`	object	否	生成配置，如 `temperature`、`topP`、`topK`、`maxOutputTokens`。
`cachedContent`	string	否	缓存上下文引用。

Gemini OpenAI-compatible 入口请求参数

参数	类型	必填	说明
`model`	string	是	例如 `google/gemini-2.0-flash-001`。
`messages`	array	是	OpenAI 风格消息数组。
`messages[].role`	string	是	`system`、`user`、`assistant` 等。
`messages[].content`	string \| array	是	文本或多模态内容。
`temperature`	number	否	采样温度。
`top_p`	number	否	nucleus sampling。
`max_tokens`	integer	否	最大输出 token 数。
`stream`	boolean	否	是否流式输出。
`tools`	array	否	工具定义。
`tool_choice`	string \| object	否	工具选择策略。
`response_format`	object	否	输出格式。
`reasoning_effort`	string	否	推理强度。
`extra_body.google.thinking_config`	object	否	Gemini 特定 thinking 配置。

Claude partner model 请求参数

参数	类型	必填	说明
`anthropic_version`	string	是	Anthropic 协议版本。
`messages`	array	是	Claude 消息数组。
`max_tokens`	integer	是	最大输出 token 数。
`system`	string \| array	否	系统提示。
`temperature`	number	否	采样温度。
`top_p`	number	否	nucleus sampling。
`top_k`	integer	否	候选 token 上限。
`tools`	array	否	工具定义。
`tool_choice`	string \| object	否	工具选择策略。

OpenAI 风格 managed API 请求参数

参数	类型	必填	说明
`model`	string	是	例如 `gpt-oss-120b-maas` 等 Vertex 托管模型 ID。
`messages`	array	是	OpenAI 风格消息数组。
`temperature`	number	否	采样温度。
`top_p`	number	否	nucleus sampling。
`max_tokens`	integer	否	最大输出 token 数。
`stream`	boolean	否	是否流式输出。
`tools`	array	否	工具定义。

请求示例

Gemini 原生接口

curl


curl -X POST \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
  "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/gemini-2.5-flash:generateContent" \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          {
            "text": "Explain GGPU Platform in one paragraph."
          }
        ]
      }
    ],
    "generationConfig": {
      "temperature": 0.7,
      "maxOutputTokens": 256
    }
  }'

TypeScript


const token = process.env.GOOGLE_ACCESS_TOKEN!;
 
const response = await fetch(
  "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/gemini-2.5-flash:generateContent",
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${token}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      contents: [
        {
          role: "user",
          parts: [
            {
              text: "Explain GGPU Platform in one paragraph.",
            },
          ],
        },
      ],
      generationConfig: {
        temperature: 0.7,
        maxOutputTokens: 256,
      },
    }),
  },
);
 
console.log(await response.json());

Python


import os
import requests
 
response = requests.post(
    "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/gemini-2.5-flash:generateContent",
    headers={
        "Authorization": f"Bearer {os.environ['GOOGLE_ACCESS_TOKEN']}",
        "Content-Type": "application/json",
    },
    json={
        "contents": [
            {
                "role": "user",
                "parts": [
                    {
                        "text": "Explain GGPU Platform in one paragraph."
                    }
                ],
            }
        ],
        "generationConfig": {
            "temperature": 0.7,
            "maxOutputTokens": 256,
        },
    },
)
 
print(response.json())

Java


import java.io.IOException;
import java.net.URI;
import java.net.http.HttpClient;
import java.net.http.HttpRequest;
import java.net.http.HttpResponse;
 
public class VertexGeminiNativeExample {
  public static void main(String[] args) throws IOException, InterruptedException {
    String token = System.getenv("GOOGLE_ACCESS_TOKEN");
    String body = """
      {
        "contents": [
          {
            "role": "user",
            "parts": [
              {
                "text": "Explain GGPU Platform in one paragraph."
              }
            ]
          }
        ],
        "generationConfig": {
          "temperature": 0.7,
          "maxOutputTokens": 256
        }
      }
      """;
 
    HttpRequest request = HttpRequest.newBuilder()
        .uri(URI.create("https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/gemini-2.5-flash:generateContent"))
        .header("Authorization", "Bearer " + token)
        .header("Content-Type", "application/json")
        .POST(HttpRequest.BodyPublishers.ofString(body))
        .build();
 
    HttpResponse<String> response = HttpClient.newHttpClient()
        .send(request, HttpResponse.BodyHandlers.ofString());
    System.out.println(response.body());
  }
}

Go


package main
 
import (
	"bytes"
	"fmt"
	"net/http"
	"os"
)
 
func main() {
	body := []byte(`{
  "contents": [
    {
      "role": "user",
      "parts": [
        {
          "text": "Explain GGPU Platform in one paragraph."
        }
      ]
    }
  ],
  "generationConfig": {
    "temperature": 0.7,
    "maxOutputTokens": 256
  }
}`)
 
	req, err := http.NewRequest(http.MethodPost, "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/gemini-2.5-flash:generateContent", bytes.NewBuffer(body))
	if err != nil {
		panic(err)
	}
 
	req.Header.Set("Authorization", "Bearer "+os.Getenv("GOOGLE_ACCESS_TOKEN"))
	req.Header.Set("Content-Type", "application/json")
 
	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()
 
	buf := new(bytes.Buffer)
	_, _ = buf.ReadFrom(resp.Body)
	fmt.Println(buf.String())
}

Gemini OpenAI-compatible 入口

curl


curl -X POST \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
  "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions" \
  -d '{
    "model": "google/gemini-2.0-flash-001",
    "messages": [
      {
        "role": "user",
        "content": "Explain GGPU Platform in one paragraph."
      }
    ],
    "temperature": 0.7,
    "max_tokens": 256
  }'

TypeScript


const token = process.env.GOOGLE_ACCESS_TOKEN!;
 
const response = await fetch(
  "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions",
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${token}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      model: "google/gemini-2.0-flash-001",
      messages: [
        {
          role: "user",
          content: "Explain GGPU Platform in one paragraph.",
        },
      ],
      temperature: 0.7,
      max_tokens: 256,
    }),
  },
);
 
console.log(await response.json());

Python


import os
import requests
 
response = requests.post(
    "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions",
    headers={
        "Authorization": f"Bearer {os.environ['GOOGLE_ACCESS_TOKEN']}",
        "Content-Type": "application/json",
    },
    json={
        "model": "google/gemini-2.0-flash-001",
        "messages": [
            {
                "role": "user",
                "content": "Explain GGPU Platform in one paragraph.",
            }
        ],
        "temperature": 0.7,
        "max_tokens": 256,
    },
)
 
print(response.json())

Java


import java.io.IOException;
import java.net.URI;
import java.net.http.HttpClient;
import java.net.http.HttpRequest;
import java.net.http.HttpResponse;
 
public class VertexGeminiOpenAIExample {
  public static void main(String[] args) throws IOException, InterruptedException {
    String token = System.getenv("GOOGLE_ACCESS_TOKEN");
    String body = """
      {
        "model": "google/gemini-2.0-flash-001",
        "messages": [
          {
            "role": "user",
            "content": "Explain GGPU Platform in one paragraph."
          }
        ],
        "temperature": 0.7,
        "max_tokens": 256
      }
      """;
 
    HttpRequest request = HttpRequest.newBuilder()
        .uri(URI.create("https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions"))
        .header("Authorization", "Bearer " + token)
        .header("Content-Type", "application/json")
        .POST(HttpRequest.BodyPublishers.ofString(body))
        .build();
 
    HttpResponse<String> response = HttpClient.newHttpClient()
        .send(request, HttpResponse.BodyHandlers.ofString());
    System.out.println(response.body());
  }
}

Go


package main
 
import (
	"bytes"
	"fmt"
	"net/http"
	"os"
)
 
func main() {
	body := []byte(`{
  "model": "google/gemini-2.0-flash-001",
  "messages": [
    {
      "role": "user",
      "content": "Explain GGPU Platform in one paragraph."
    }
  ],
  "temperature": 0.7,
  "max_tokens": 256
}`)
 
	req, err := http.NewRequest(http.MethodPost, "https://aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions", bytes.NewBuffer(body))
	if err != nil {
		panic(err)
	}
 
	req.Header.Set("Authorization", "Bearer "+os.Getenv("GOOGLE_ACCESS_TOKEN"))
	req.Header.Set("Content-Type", "application/json")
 
	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()
 
	buf := new(bytes.Buffer)
	_, _ = buf.ReadFrom(resp.Body)
	fmt.Println(buf.String())
}

Claude partner model 入口

curl


curl -X POST \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
  "https://us-east5-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-east5/publishers/anthropic/models/claude-sonnet-4-5@20250929:rawPredict" \
  -d '{
    "anthropic_version": "vertex-2023-10-16",
    "max_tokens": 256,
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "type": "text",
            "text": "Explain GGPU Platform in one paragraph."
          }
        ]
      }
    ],
    "temperature": 0.7
  }'

TypeScript


const token = process.env.GOOGLE_ACCESS_TOKEN!;
 
const response = await fetch(
  "https://us-east5-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-east5/publishers/anthropic/models/claude-sonnet-4-5@20250929:rawPredict",
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${token}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      anthropic_version: "vertex-2023-10-16",
      max_tokens: 256,
      messages: [
        {
          role: "user",
          content: [
            {
              type: "text",
              text: "Explain GGPU Platform in one paragraph.",
            },
          ],
        },
      ],
      temperature: 0.7,
    }),
  },
);
 
console.log(await response.json());

Python


import os
import requests
 
response = requests.post(
    "https://us-east5-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-east5/publishers/anthropic/models/claude-sonnet-4-5@20250929:rawPredict",
    headers={
        "Authorization": f"Bearer {os.environ['GOOGLE_ACCESS_TOKEN']}",
        "Content-Type": "application/json",
    },
    json={
        "anthropic_version": "vertex-2023-10-16",
        "max_tokens": 256,
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": "Explain GGPU Platform in one paragraph.",
                    }
                ],
            }
        ],
        "temperature": 0.7,
    },
)
 
print(response.json())

Java


import java.io.IOException;
import java.net.URI;
import java.net.http.HttpClient;
import java.net.http.HttpRequest;
import java.net.http.HttpResponse;
 
public class VertexClaudeExample {
  public static void main(String[] args) throws IOException, InterruptedException {
    String token = System.getenv("GOOGLE_ACCESS_TOKEN");
    String body = """
      {
        "anthropic_version": "vertex-2023-10-16",
        "max_tokens": 256,
        "messages": [
          {
            "role": "user",
            "content": [
              {
                "type": "text",
                "text": "Explain GGPU Platform in one paragraph."
              }
            ]
          }
        ],
        "temperature": 0.7
      }
      """;
 
    HttpRequest request = HttpRequest.newBuilder()
        .uri(URI.create("https://us-east5-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-east5/publishers/anthropic/models/claude-sonnet-4-5@20250929:rawPredict"))
        .header("Authorization", "Bearer " + token)
        .header("Content-Type", "application/json")
        .POST(HttpRequest.BodyPublishers.ofString(body))
        .build();
 
    HttpResponse<String> response = HttpClient.newHttpClient()
        .send(request, HttpResponse.BodyHandlers.ofString());
    System.out.println(response.body());
  }
}

Go


package main
 
import (
	"bytes"
	"fmt"
	"net/http"
	"os"
)
 
func main() {
	body := []byte(`{
  "anthropic_version": "vertex-2023-10-16",
  "max_tokens": 256,
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Explain GGPU Platform in one paragraph."
        }
      ]
    }
  ],
  "temperature": 0.7
}`)
 
	req, err := http.NewRequest(http.MethodPost, "https://us-east5-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-east5/publishers/anthropic/models/claude-sonnet-4-5@20250929:rawPredict", bytes.NewBuffer(body))
	if err != nil {
		panic(err)
	}
 
	req.Header.Set("Authorization", "Bearer "+os.Getenv("GOOGLE_ACCESS_TOKEN"))
	req.Header.Set("Content-Type", "application/json")
 
	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()
 
	buf := new(bytes.Buffer)
	_, _ = buf.ReadFrom(resp.Body)
	fmt.Println(buf.String())
}

OpenAI 风格 managed API 入口

curl


curl -X POST \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
  "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions" \
  -d '{
    "model": "gpt-oss-120b-maas",
    "messages": [
      {
        "role": "user",
        "content": "Explain GGPU Platform in one paragraph."
      }
    ],
    "temperature": 0.7,
    "max_tokens": 256
  }'

TypeScript


const token = process.env.GOOGLE_ACCESS_TOKEN!;
 
const response = await fetch(
  "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions",
  {
    method: "POST",
    headers: {
      Authorization: `Bearer ${token}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      model: "gpt-oss-120b-maas",
      messages: [
        {
          role: "user",
          content: "Explain GGPU Platform in one paragraph.",
        },
      ],
      temperature: 0.7,
      max_tokens: 256,
    }),
  },
);
 
console.log(await response.json());

Python


import os
import requests
 
response = requests.post(
    "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions",
    headers={
        "Authorization": f"Bearer {os.environ['GOOGLE_ACCESS_TOKEN']}",
        "Content-Type": "application/json",
    },
    json={
        "model": "gpt-oss-120b-maas",
        "messages": [
            {
                "role": "user",
                "content": "Explain GGPU Platform in one paragraph.",
            }
        ],
        "temperature": 0.7,
        "max_tokens": 256,
    },
)
 
print(response.json())

Java


import java.io.IOException;
import java.net.URI;
import java.net.http.HttpClient;
import java.net.http.HttpRequest;
import java.net.http.HttpResponse;
 
public class VertexOpenAIManagedExample {
  public static void main(String[] args) throws IOException, InterruptedException {
    String token = System.getenv("GOOGLE_ACCESS_TOKEN");
    String body = """
      {
        "model": "gpt-oss-120b-maas",
        "messages": [
          {
            "role": "user",
            "content": "Explain GGPU Platform in one paragraph."
          }
        ],
        "temperature": 0.7,
        "max_tokens": 256
      }
      """;
 
    HttpRequest request = HttpRequest.newBuilder()
        .uri(URI.create("https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions"))
        .header("Authorization", "Bearer " + token)
        .header("Content-Type", "application/json")
        .POST(HttpRequest.BodyPublishers.ofString(body))
        .build();
 
    HttpResponse<String> response = HttpClient.newHttpClient()
        .send(request, HttpResponse.BodyHandlers.ofString());
    System.out.println(response.body());
  }
}

Go


package main
 
import (
	"bytes"
	"fmt"
	"net/http"
	"os"
)
 
func main() {
	body := []byte(`{
  "model": "gpt-oss-120b-maas",
  "messages": [
    {
      "role": "user",
      "content": "Explain GGPU Platform in one paragraph."
    }
  ],
  "temperature": 0.7,
  "max_tokens": 256
}`)
 
	req, err := http.NewRequest(http.MethodPost, "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/endpoints/openapi/chat/completions", bytes.NewBuffer(body))
	if err != nil {
		panic(err)
	}
 
	req.Header.Set("Authorization", "Bearer "+os.Getenv("GOOGLE_ACCESS_TOKEN"))
	req.Header.Set("Content-Type", "application/json")
 
	resp, err := http.DefaultClient.Do(req)
	if err != nil {
		panic(err)
	}
	defer resp.Body.Close()
 
	buf := new(bytes.Buffer)
	_, _ = buf.ReadFrom(resp.Body)
	fmt.Println(buf.String())
}

Omni HOT

Vertex AI 的 Interactions 视频生成目前暂不支持 gemini-omni-1.1-flash，请使用 vertex-ai/gemini-omni-flash-preview。以下示例使用平台创建的访问令牌，并将令牌放在 GGPU_API_KEY 环境变量中。

异步生成

设置 "background": true 后，接口会立即创建后台任务并返回平台任务信息。请保存返回的 id，再使用统一任务接口获取状态和结果。


curl --location --request POST "https://api.ggpu.ai/v1beta/interactions" \
  --header "x-goog-api-key: $GGPU_API_KEY" \
  --header "Content-Type: application/json" \
  --data-raw '{
    "model": "vertex-ai/gemini-omni-flash-preview",
    "input": "小狗跳舞",
    "response_format": {
      "type": "video",
      "aspect_ratio": "16:9"
    },
    "generation_config": {
      "video_config": {
        "task": "text_to_video"
      }
    },
    "background": true,
    "store": true
  }'

使用返回的任务 ID 查询异步任务状态：


curl --location --request GET "https://api.ggpu.ai/v1/tasks/<task_id>" \
  --header "Authorization: Bearer $GGPU_API_KEY"

同步生成

省略 background（或显式设置为 false）时，接口按同步 interaction 处理，请求会等待生成结果返回。


curl --location --request POST "https://api.ggpu.ai/v1beta/interactions" \
  --header "x-goog-api-key: $GGPU_API_KEY" \
  --header "Content-Type: application/json" \
  --data-raw '{
    "model": "vertex-ai/gemini-omni-flash-preview",
    "input": "小狗跳舞",
    "response_format": {
      "type": "video",
      "aspect_ratio": "16:9"
    },
    "generation_config": {
      "video_config": {
        "task": "text_to_video"
      }
    },
    "store": true
  }'

请求返回结果

Gemini 原生接口返回结果

参数	类型	说明
`candidates`	array	候选输出。
`candidates[].content`	object	生成内容。
`candidates[].finishReason`	string	停止原因。
`promptFeedback`	object	Prompt 反馈。
`usageMetadata`	object	token 用量。
`modelVersion`	string	实际模型版本。

Gemini / OpenAI 风格 chat completions 返回结果

参数	类型	说明
`id`	string	completion ID。
`object`	string	对象类型。
`created`	integer	创建时间。
`model`	string	实际模型。
`choices`	array	completion 候选。
`choices[].message`	object	assistant 消息。
`choices[].finish_reason`	string	停止原因。
`usage`	object	token 用量。

Claude partner model 返回结果

参数	类型	说明
`id`	string	消息 ID。
`type`	string	通常为 `message`。
`role`	string	通常为 `assistant`。
`content`	array	输出内容块。
`model`	string	实际模型。
`stop_reason`	string	停止原因。
`usage`	object	token 用量。

请求返回示例

Gemini 原生接口返回示例


{
  "candidates": [
    {
      "content": {
        "role": "model",
        "parts": [
          {
            "text": "GGPU Platform helps teams design, execute, and scale AI workflows on top of Vertex AI and external model providers."
          }
        ]
      },
      "finishReason": "STOP"
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 13,
    "candidatesTokenCount": 21,
    "totalTokenCount": 34
  },
  "modelVersion": "gemini-2.5-flash"
}

Gemini / OpenAI 风格 chat completions 返回示例


{
  "id": "chatcmpl-vertex-01",
  "object": "chat.completion",
  "created": 1741572201,
  "model": "google/gemini-2.0-flash-001",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "GGPU Platform helps teams design, execute, and scale AI workflows on top of Vertex AI and external model providers."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 13,
    "completion_tokens": 20,
    "total_tokens": 33
  }
}

Claude partner model 返回示例


{
  "id": "msg_vertex_01",
  "type": "message",
  "role": "assistant",
  "model": "claude-sonnet-4-5@20250929",
  "content": [
    {
      "type": "text",
      "text": "GGPU Platform helps teams design, execute, and scale AI workflows on top of Vertex AI and external model providers."
    }
  ],
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": {
    "input_tokens": 15,
    "output_tokens": 20
  }
}

说明

Vertex AI 的 Omni Interactions 视频生成暂不支持 gemini-omni-1.1-flash，请使用 vertex-ai/gemini-omni-flash-preview。
background: true 表示异步任务，创建后使用 /v1/tasks/<task_id> 轮询；省略或设为 false 时按同步 interaction 处理。
Vertex AI 不是单一的一套接口，Gemini、Claude partner models、OpenAI-compatible 入口各自有不同 schema。
如果文档目标是“官方原生展示”，应按入口分别呈现，而不是合并成单一最小示例。
更细的 streaming、function/tool calling、多模态 part 类型，请以 Vertex AI 官方文档为准。