Deepseek-v3 / Dify api接入飞书机器人go程序

准备工作

开通了接收消息权限的飞书机器人，例如我希望用户跟飞书机器人私聊，就需要开通这个权限：读取用户发给机器人的单聊消息 im:message.p2p_msg:readonly
准备好飞书机器人的API key 和Secret
deepseek-v3的api key+secret：https://platform.deepseek.com/api_keys 这里获取，一开始有10元的免费额度，趁能充多充点，经常不让充值。
自己部署一下dify，推荐使用docker-compose方式，这个有很多教程就不赘述了

飞书机器人通过长连接获取用户私聊发的消息

我们使用长连接的方式接收用户消息，需要在飞书开发者后台中配置一下应用，见
配置回调订阅方式
代码如下：

import(larkevent "github.com/larksuite/oapi-sdk-go/v3/event""github.com/larksuite/oapi-sdk-go/v3/event/dispatcher""github.com/larksuite/oapi-sdk-go/v3/service/auth/v3"larkim "github.com/larksuite/oapi-sdk-go/v3/service/im/v1"larkws "github.com/larksuite/oapi-sdk-go/v3/ws"
)
var sent map[string]struct{} // 这里简单去个重 实际使用要自己再写去重部分
// 飞书消息过来Content字段值是{\"text\":\"早上好～\"}这样的，需要再解析一下
type Text struct {Text string `json:"text"`
}
// 处理接收到用户消息的事件
func callback() {sent = make(map[string]struct{})// 注册事件回调，OnP2MessageReceiveV1 为接收消息 v2.0；OnCustomizedEvent 内的 message 为接收消息 v1.0。NewEventDispatcher()里的两个参数都填空字符串eventHandler := dispatcher.NewEventDispatcher("", "").OnP2MessageReceiveV1(func(ctx context.Context, event *larkim.P2MessageReceiveV1) error {// messageid简单去重if _, ok := sent[*event.Event.Message.MessageId]; ok {return nil} else {sent[*event.Event.Message.MessageId] = struct{}{}}fmt.Printf("[ OnP2MessageReceiveV1 access ], data: %s\n", larkcore.Prettify(event))fmt.Println(event.Event.Message.Content) // content中就是用户发过来的消息内容var text Textjson.Unmarshal([]byte(*event.Event.Message.Content), &text)fmt.Println(text.Text)// 这里可以把用户输入发给deepseek或者dify并接收其响应，具体实现后面讲//resp, e := deepseek.CallDeepSeekAPI(text.Text)//if e != nil {//	return e//}resp := dify.ChatMessages(text.Text)fmt.Println(resp)// 这里组织飞书机器人发送消息的content格式，跟接收到消息的一样，也是{\"text\":\"say something\"}var contentStruct struct {Text string `json:"text"`}contentStruct.Text = respcontent, _ := json.Marshal(contentStruct)// 这个messages是机器人发送消息函数，见下方messages(getTernantAccessToken(), *event.Event.Sender.SenderId.OpenId, *event.Event.Message.MessageType, string(content))return nil}).//im:message.p2p_msg:readonly 这个先不用管OnCustomizedEvent("", func(ctx context.Context, event *larkevent.EventReq) error {fmt.Printf("[ OnCustomizedEvent access ], type: message, data: %s\n", string(event.Body))return nil})// 创建Clientcli := larkws.NewClient(AppID, AppSecret,larkws.WithEventHandler(eventHandler),larkws.WithLogLevel(larkcore.LogLevelDebug),)// 启动客户端 保持一个长链接err := cli.Start(context.Background())if err != nil {panic(err)}
}//发送消息
func messages(token, receiveID, msgType, content string) {// 创建 Clientclient := lark.NewClient(AppID, AppSecret)// 创建请求对象receiveIDType := "open_id"if strings.HasPrefix(receiveID, "oc") { // 这里我简单区分了一下群聊和个人receiveIDType = "chat_id"}req := larkim.NewCreateMessageReqBuilder().ReceiveIdType(receiveIDType).Body(larkim.NewCreateMessageReqBodyBuilder().ReceiveId(receiveID).MsgType(msgType).Content(content).Build()).Build()// 发起请求resp, err := client.Im.Message.Create(context.Background(), req, larkcore.WithTenantAccessToken(token))// 处理错误if err != nil {fmt.Println(err)return}// 服务端错误处理if !resp.Success() {fmt.Println(resp.Code, resp.Msg, resp.RequestId())return}// 业务处理//fmt.Println(larkcore.Prettify(resp))fmt.Println(string(resp.RawBody))
}

tips: 获取的用户消息长这样：

{EventV2Base: {Schema: "2.0",Header: {EventID: "xx",EventType: "im.message.receive_v1",AppID: "xx",TenantKey: "xx",CreateTime: "1738892348642",Token: ""}},EventReq: {Body: <binary> len 672,RequestURI: ""},Event: {Sender: {SenderId: {UserId: "xx",OpenId: "xx",UnionId: "xx"},SenderType: "user",TenantKey: "xx"},Message: {MessageId: "xx",CreateTime: "1738892348363",UpdateTime: "1738892348363",ChatId: "xx",ChatType: "p2p",MessageType: "text",Content: "{\"text\":\"早上好～\"}"}}
}

接下来就是实现调用deepseek或dify的api的逻辑了

Deepseek-v3 API调用代码

package deepseekimport ("bytes""encoding/json""fmt""io""net/http""net/http/httputil"
)
var (// DeepSeek-R1 API 的配置DeepSeekAPIURL = "https://api.deepseek.com/chat/completions" // 直接用这个就行DeepSeekAPIKey = "你的key"
)
// DeepSeek-R1 API 请求数据结构
type DeepSeekRequest struct {Model    string        `json:"model"`Messages []RoleContent `json:"messages"`Stream   bool          `json:"stream"`
}
type RoleContent struct {Role    string `json:"role"`Content string `json:"content"`
}// DeepSeek-R1 API 响应数据结构
type DeepSeekResponse struct {Choices []struct {Message struct {Content string `json:"content"`} `json:"message"`} `json:"choices"`
}// 调用 DeepSeek-R1 API
func CallDeepSeekAPI(msg string) (string, error) {requestBody := DeepSeekRequest{Model: "deepseek-chat",Messages: []RoleContent{{Role: "system", Content: "You are a helpful assistant."}, // 这里可以自行修改{Role: "user", Content: msg}, // msg就是用户发的消息},Stream: false, // 这里先不用流式输出}requestBytes, err := json.Marshal(requestBody)if err != nil {return "", err}req, err := http.NewRequest("POST", DeepSeekAPIURL, bytes.NewBuffer(requestBytes))if err != nil {return "", err}req.Header.Set("Authorization", "Bearer "+DeepSeekAPIKey)req.Header.Set("Content-Type", "application/json")// 这里我dump了一下请求看发的是否正确 可以删掉dump, _ := httputil.DumpRequest(req, true)fmt.Println(string(dump))// 发请求client := &http.Client{}resp, err := client.Do(req)if err != nil {return "", err}defer resp.Body.Close()body, err := io.ReadAll(resp.Body)if err != nil {return "", err}// 解析响应var deepSeekResponse DeepSeekResponseif err := json.Unmarshal(body, &deepSeekResponse); err != nil {return "", err}// 拿content返回if len(deepSeekResponse.Choices) > 0 {return deepSeekResponse.Choices[0].Message.Content, nil}return "", fmt.Errorf("no response from DeepSeek API")
}

Dify api调用方法

如何在dify中接入大模型并制作一个问答机器人参考：https://docs.dify.ai/zh-hans/guides/application-orchestrate/conversation-application
点击【发布】之后，去【访问api】页面，右上角有一个在这里插入图片描述
点击这个API密钥保存下来

调用代码如下：

package difyimport ("bytes""encoding/json""fmt""io""log""net/http""net/http/httputil""strconv""strings"
)type ChatMessageRequest struct {Inputs         map[string]interface{} `json:"inputs"`Query          string                 `json:"query"`ResponseMode   string                 `json:"response_mode"`ConversationID string                 `json:"conversation_id,omitempty"`User           string                 `json:"user"`
}type ChatMessageResponse struct {ID             string `json:"id"`Answer         string `json:"answer"`ConversationID string `json:"conversation_id"`CreatedAt      int    `json:"created_at"`
}const (DifyBaseURL = "http://192.168.xx.xx:12345/v1" // 这里是你的dify服务地址DifyApiKey  = "app-xxxx" // dify提供的api密钥ChatMsgPath = "/chat-messages"
)func ChatMessages(msg string) string {requestData := ChatMessageRequest{Query:        msg,ResponseMode: "blocking", // 我们先选择阻塞模式，就是等回答全部生成后发回来，而不是sse那种模拟打字输出的形式（streaming)User:         "abc123",}// 将请求数据序列化为 JSONrequestBody, err := json.Marshal(requestData)if err != nil {fmt.Errorf("failed to marshal request data: %v", err)}// 创建 HTTP 请求req, err := http.NewRequest("POST", DifyBaseURL+ChatMsgPath, bytes.NewBuffer(requestBody))if err != nil {log.Fatalf("Failed to create request: %v", err)}// 设置请求头req.Header.Set("Authorization", "Bearer "+DifyApiKey)req.Header.Set("Content-Type", "application/json")// 发送请求client := &http.Client{}// 这里dump了一下看发送请求是否正确，可以删掉dump, _ := httputil.DumpRequest(req, true)fmt.Println(string(dump))resp, err := client.Do(req)if err != nil {log.Fatalf("Failed to send request: %v", err)}defer resp.Body.Close()// 读取响应body, err := io.ReadAll(resp.Body)if err != nil {log.Fatalf("Failed to read response body: %v", err)}// 输出响应fmt.Println("Response Status:", resp.Status)fmt.Println("Response Body:", string(body))var res ChatMessageResponseif err := json.Unmarshal(body, &res); err != nil {fmt.Errorf("Failed to unmarshal response body: %v", err)return ""}fmt.Println("Answer:", res.Answer)return res.Answer // 这个就是dify调大模型获得的返回内容
}