通过API和Web界面查询Dify+Ollama+DeepSeek

参考文档:</strong><ul><strong>回答:

Leon_Jinhai_Sun

1124人浏览 · 2025-03-26 08:44:48

Leon_Jinhai_Sun · 2025-03-26 08:44:48 发布

通过API和Web界面查询Dify+Ollama+DeepSeek知识库的具体实现

一、通过Web界面查询

访问Dify Web界面：
- 默认地址：http://localhost:3000 (如果在本地部署)
- 登录后选择你创建的知识库应用
直接问答：
- 在聊天界面输入问题，系统会自动从知识库中检索相关信息并生成回答

二、通过API查询知识库

1. 获取API密钥

在Dify的"设置" > "API密钥"中创建一个新的API密钥

2. Python代码示例

import requests
import json

# 配置参数
API_KEY = "your-dify-api-key"  # 替换为你的Dify API密钥
APP_ID = "your-app-id"        # 替换为你的应用ID
API_URL = "http://localhost/v1/chat-messages"  # 本地部署URL
# 如果是云服务可能是: "https://api.dify.ai/v1/chat-messages"

def query_knowledge_base(question):
    headers = {
        "Authorization": f"Bearer {API_KEY}",
        "Content-Type": "application/json"
    }
    
    payload = {
        "inputs": {},
        "query": question,
        "response_mode": "blocking",  # 也可以是"streaming"流式响应
        "user": "api-user"  # 可自定义用户标识
    }
    
    response = requests.post(
        f"{API_URL}/{APP_ID}",
        headers=headers,
        data=json.dumps(payload)
    )
    
    if response.status_code == 200:
        return response.json()
    else:
        raise Exception(f"API请求失败: {response.status_code}, {response.text}")

# 使用示例
if __name__ == "__main__":
    question = "你们公司的主要产品是什么？"  # 替换为你的问题
    try:
        result = query_knowledge_base(question)
        print("回答内容:")
        print(result['answer'])
        print("\n引用的知识文档:")
        for doc in result.get('retriever_resources', []):
            print(f"- {doc['content'][:100]}...")  # 只打印前100字符
    except Exception as e:
        print(f"查询出错: {str(e)}")

3. 流式响应API调用

如果需要流式响应（适合长回答）：

def stream_knowledge_base(question):
    headers = {
        "Authorization": f"Bearer {API_KEY}",
        "Content-Type": "application/json",
        "Accept": "text/event-stream"  # 重要：请求流式响应
    }
    
    payload = {
        "inputs": {},
        "query": question,
        "response_mode": "streaming",
        "user": "api-user"
    }
    
    with requests.post(
        f"{API_URL}/{APP_ID}",
        headers=headers,
        data=json.dumps(payload),
        stream=True
    ) as response:
        
        if response.status_code != 200:
            raise Exception(f"API请求失败: {response.status_code}, {response.text}")
        
        print("开始接收流式响应:")
        for line in response.iter_lines():
            if line:
                decoded_line = line.decode('utf-8')
                if decoded_line.startswith('data:'):
                    data = json.loads(decoded_line[5:])
                    if 'event' in data and data['event'] == 'message':
                        print(data['answer'], end='', flush=True)
        print("\n回答完成")

# 使用示例
stream_knowledge_base("请详细说明你们的产品特点")

三、使用cURL命令查询

curl -X POST "http://localhost/v1/chat-messages/your-app-id" \
  -H "Authorization: Bearer your-dify-api-key" \
  -H "Content-Type: application/json" \
  -d '{
    "inputs": {},
    "query": "你们公司的主要产品是什么？",
    "response_mode": "blocking",
    "user": "curl-user"
  }'

四、前端集成示例（HTML+JavaScript）

<!DOCTYPE html>
<html>
<head>
    <title>知识库查询</title>
    <style>
        #chat-container { width: 600px; margin: 0 auto; }
        #response { border: 1px solid #ccc; min-height: 200px; padding: 10px; margin: 10px 0; }
    </style>
</head>
<body>
    <div id="chat-container">
        <h2>知识库查询</h2>
        <input type="text" id="question" placeholder="输入你的问题..." style="width: 80%">
        <button onclick="queryKnowledge()">提问</button>
        <div id="response"></div>
    </div>

    <script>
        const API_KEY = "your-dify-api-key";
        const APP_ID = "your-app-id";
        const API_URL = "http://localhost/v1/chat-messages";
        
        function queryKnowledge() {
            const question = document.getElementById('question').value;
            const responseDiv = document.getElementById('response');
            responseDiv.innerHTML = "思考中...";
            
            fetch(`${API_URL}/${APP_ID}`, {
                method: 'POST',
                headers: {
                    'Authorization': `Bearer ${API_KEY}`,
                    'Content-Type': 'application/json'
                },
                body: JSON.stringify({
                    inputs: {},
                    query: question,
                    response_mode: "blocking",
                    user: "web-user"
                })
            })
            .then(response => response.json())
            .then(data => {
                responseDiv.innerHTML = `<strong>回答:</strong><p>${data.answer}</p>`;
                if(data.retriever_resources && data.retriever_resources.length > 0) {
                    responseDiv.innerHTML += `<strong>参考文档:</strong><ul>`;
                    data.retriever_resources.forEach(doc => {
                        responseDiv.innerHTML += `<li>${doc.content.substring(0, 150)}...</li>`;
                    });
                    responseDiv.innerHTML += `</ul>`;
                }
            })
            .catch(error => {
                responseDiv.innerHTML = `查询出错: ${error.message}`;
            });
        }
    </script>
</body>
</html>

五、API响应示例

成功的API响应会返回类似以下JSON结构：

{
  "answer": "我们公司的主要产品是AI开发平台，提供自然语言处理、知识库管理和模型部署等功能...",
  "retriever_resources": [
    {
      "content": "公司产品手册2023版...我们的旗舰产品是Dify AI开发平台...",
      "file_name": "product_manual.pdf",
      "page_number": 5
    }
  ],
  "metadata": {
    "usage": {
      "prompt_tokens": 45,
      "completion_tokens": 128,
      "total_tokens": 173
    },
    "model": "deepseek-llm:7b",
    "timestamp": 1698765432
  }
}

六、高级查询参数

你可以在API请求中添加更多参数来优化查询：

payload = {
    "inputs": {},
    "query": question,
    "response_mode": "blocking",
    "user": "advanced-user",
    "conversation_id": "12345",  # 用于多轮对话保持上下文
    "temperature": 0.7,  # 控制创造性(0-1)
    "top_k": 50,  # 从top_k个最可能的token中选择
    "max_tokens": 500,  # 限制响应长度
    "retrieve_config": {
        "top_k": 3,  # 从知识库返回的最相关文档数
        "score_threshold": 0.7  # 相关性阈值
    }
}