问答 Agent 对话（流式）

基于 ReAct（Reasoning + Acting）推理引擎的流式对话接口。LLM 会根据用户问题自主决定是否调用工具（知识检索、查询重写等），并通过 SSE 实时推送思考过程、工具调用和最终回答。点击 Try it 按钮可快速试用。

POST /zrag/agent/chat

问答 Agent 对话（流式）

cURL

curl --request POST \
  --url https://open.bigmodel.cn/api/zrag/agent/chat \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "messages": [
    {
      "role": "user",
      "content": "公司的年假制度是什么？"
    }
  ],
  "model": "glm-5v-turbo",
  "temperature": 0.2,
  "max_steps": 10,
  "retrieval": {
    "know_ids": [
      "123"
    ],
    "top_k": 8,
    "top_n": 10,
    "enable_rerank": false
  }
}
'

import requests

url = "https://open.bigmodel.cn/api/zrag/agent/chat"

payload = {
    "messages": [
        {
            "role": "user",
            "content": "公司的年假制度是什么？"
        }
    ],
    "model": "glm-5v-turbo",
    "temperature": 0.2,
    "max_steps": 10,
    "retrieval": {
        "know_ids": ["123"],
        "top_k": 8,
        "top_n": 10,
        "enable_rerank": False
    }
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

response = requests.post(url, json=payload, headers=headers)

print(response.text)

const options = {
  method: 'POST',
  headers: {Authorization: 'Bearer <token>', 'Content-Type': 'application/json'},
  body: JSON.stringify({
    messages: [{role: 'user', content: '公司的年假制度是什么？'}],
    model: 'glm-5v-turbo',
    temperature: 0.2,
    max_steps: 10,
    retrieval: {know_ids: ['123'], top_k: 8, top_n: 10, enable_rerank: false}
  })
};

fetch('https://open.bigmodel.cn/api/zrag/agent/chat', options)
  .then(res => res.json())
  .then(res => console.log(res))
  .catch(err => console.error(err));

HttpResponse<String> response = Unirest.post("https://open.bigmodel.cn/api/zrag/agent/chat")
  .header("Authorization", "Bearer <token>")
  .header("Content-Type", "application/json")
  .body("{\n  \"messages\": [\n    {\n      \"role\": \"user\",\n      \"content\": \"公司的年假制度是什么？\"\n    }\n  ],\n  \"model\": \"glm-5v-turbo\",\n  \"temperature\": 0.2,\n  \"max_steps\": 10,\n  \"retrieval\": {\n    \"know_ids\": [\n      \"123\"\n    ],\n    \"top_k\": 8,\n    \"top_n\": 10,\n    \"enable_rerank\": false\n  }\n}")
  .asString();

package main

import (
	"fmt"
	"strings"
	"net/http"
	"io"
)

func main() {

	url := "https://open.bigmodel.cn/api/zrag/agent/chat"

	payload := strings.NewReader("{\n  \"messages\": [\n    {\n      \"role\": \"user\",\n      \"content\": \"公司的年假制度是什么？\"\n    }\n  ],\n  \"model\": \"glm-5v-turbo\",\n  \"temperature\": 0.2,\n  \"max_steps\": 10,\n  \"retrieval\": {\n    \"know_ids\": [\n      \"123\"\n    ],\n    \"top_k\": 8,\n    \"top_n\": 10,\n    \"enable_rerank\": false\n  }\n}")

	req, _ := http.NewRequest("POST", url, payload)

	req.Header.Add("Authorization", "Bearer <token>")
	req.Header.Add("Content-Type", "application/json")

	res, _ := http.DefaultClient.Do(req)

	defer res.Body.Close()
	body, _ := io.ReadAll(res.Body)

	fmt.Println(string(body))

}

<?php

$curl = curl_init();

curl_setopt_array($curl, [
  CURLOPT_URL => "https://open.bigmodel.cn/api/zrag/agent/chat",
  CURLOPT_RETURNTRANSFER => true,
  CURLOPT_ENCODING => "",
  CURLOPT_MAXREDIRS => 10,
  CURLOPT_TIMEOUT => 30,
  CURLOPT_HTTP_VERSION => CURL_HTTP_VERSION_1_1,
  CURLOPT_CUSTOMREQUEST => "POST",
  CURLOPT_POSTFIELDS => json_encode([
    'messages' => [
        [
                'role' => 'user',
                'content' => '公司的年假制度是什么？'
        ]
    ],
    'model' => 'glm-5v-turbo',
    'temperature' => 0.2,
    'max_steps' => 10,
    'retrieval' => [
        'know_ids' => [
                '123'
        ],
        'top_k' => 8,
        'top_n' => 10,
        'enable_rerank' => false
    ]
  ]),
  CURLOPT_HTTPHEADER => [
    "Authorization: Bearer <token>",
    "Content-Type: application/json"
  ],
]);

$response = curl_exec($curl);
$err = curl_error($curl);

curl_close($curl);

if ($err) {
  echo "cURL Error #:" . $err;
} else {
  echo $response;
}

200

{
  "sessionId": "<string>",
  "messageId": "<string>",
  "data": "<string>",
  "usage": {
    "prompt_tokens": 123,
    "completion_tokens": 123,
    "total_tokens": 123,
    "total_calls": 123,
    "prompt_tokens_details": {
      "cached_tokens": 123
    },
    "completion_tokens_details": {
      "reasoning_tokens": 123
    }
  }
}

{
  "code": 123,
  "message": "<string>"
}

Authorizations

Authorization · string · header · required

标准的 HTTP Bearer 认证方式，在 API Keys 页面获取密钥。

Headers

X-Session-Id · string

会话 ID，续聊时传入

Body

application/json

messages · object[] · required

当前消息列表，支持多模态内容

Show child attributes

retrieval · object · required

检索预设参数。预设后 LLM 仅决定是否调用检索，无需自行填写参数

Show child attributes

model · string · default: glm-5v-turbo

LLM 模型名称，默认为 glm-5v-turbo

temperature · number · default: 0.7

采样温度，默认为 0.7

max_steps · integer · default: 10

最大推理步数，默认为 10

enable_thinking · boolean · default: false

是否启用思考模式。启用后模型输出推理过程，通过 reasoning 事件流式返回

Response

200

text/event-stream

SSE 流式响应，返回 AgentStreamEvent 事件流

SSE 事件流中的单个事件对象

type · enum<string>

事件类型

Available options: session_created, reasoning, thought, tool_call, tool_result, answer, done, error

sessionId · string

会话 ID

messageId · string

消息 ID（仅 done 事件）

data · Option 1 · string Option 2 · object Option 3 · object Option 4 · object

事件负载，结构取决于 type

usage · object

Token 用量信息（仅 done 事件）

Show child attributes