AI挺好用的。但是想用时就得打开网页或者其他工具插件才行?太不方便了。记得有次要给媳妇演示下它的强大,竟没带电脑竟啥事干不成。索性把AI接入微信公众号,这样在公众号聊天窗口里发消息,AI自动给我回复内容,且可以分享给好友邀好友一块儿体验,这太好玩儿了。
本文主要来聊聊如何快速使用个人微信接入 AI,欢迎体验,仅供学习参考。
好玩多了,AI接入微信公众号后可以随时随地使用它了。这里总结下AI接入微信的方法。效果展示:
环境准备
1.微信公众号。测试的就行,也建议用测试的,很容易申请,无门槛且不限制次数。
2.AI的访问api-key,这个需要科学上网申请个账号,获取api-key并保存好。
3.一台云服务器。各大云商有提供,包年很便宜,部署下Golang的后台服务。
具备以上条件就ok啦,接下来介绍下微信公众号接入方法。
我的首页里关注公众号可体验功能。测试号支持最多一百个用户,名额有限,目前还有剩余可以体验,欢迎交流。源码下载地址:https://download.csdn.net/download/qq8864/87448657
实现思路
总体方案介绍
通过搭建后台云服务,实现微信公众号与AI的连接。使用微信公众号开发平台提供的接口,主要是客服消息接口,用户在公众号聊天窗口发送问题内容,后台服务收到后,调用ai接口问答,并将AI的结果发送到微信公众号窗口中进行展示。
接入微信公众平台
如何接入微信平台?这部分内容网上有很多,比较简单这里就不详细介绍了。
介绍几个我封装的接入微信公众号的golang接口:
// HandleWxLogin首次接入,成为开发者
func HandleWxLogin(c *gin.Context) {fmt.Printf("==>HandleWxLogin\n")echostr := c.DefaultQuery("echostr", "")if echostr != "" {fmt.Printf("==>echostr:%s\n", echostr)c.String(200, "%s", echostr)return}}// WxGetAccessToken 获取微信accesstoken
func WxGetAccessToken() string {url := fmt.Sprintf("https://api.weixin.qq.com/cgi-bin/token?grant_type=client_credential&appid=%v&secret=%v", APPID, APPSECRET)resp, err := http.Get(url)if err != nil {fmt.Println("获取微信token失败", err)return ""}defer resp.Body.Close()body, err := ioutil.ReadAll(resp.Body)if err != nil {fmt.Println("微信token读取失败", err)return ""}token := token{}err = json.Unmarshal(body, &token)if err != nil {fmt.Println("微信token解析json失败", err)return ""}return token.AccessToken
}
客服回复消息接口
调用此封装的微信接口,向用户回复消息。比如用户在公众号聊天窗口发送问题后,通过向AI的api接口发送请求,把AI回复的内容,通过客服回复消息接口向用户应答。
微信客服回复接口封装
// 客服回复接口
func WxPostCustomTextMsg(accessToken string, touser string, content string) {url := "https://api.weixin.qq.com/cgi-bin/message/custom/send?access_token=" + accessTokenreq:=&WxCustomTextMsg{ToUser:touser,MsgType:"text",Text:WxCustomText{Content:content}}jsonStr := req.ToJson()//fmt.Printf("WxPostCustomTextMsg:%#v\n", jsonStr)request ,_:= http.NewRequest("POST", url, bytes.NewBuffer(jsonStr))request.Header.Set("Content-Type", "application/json")client := &http.Client{}resp, err := client.Do(request)if err != nil {fmt.Println(err)return}defer resp.Body.Close()body, err := ioutil.ReadAll(resp.Body)if err != nil {fmt.Println(err)return}fmt.Println(string(body))
}
需要注意的是,需要使用golang的协程异步处理,因为用户发送信息后若5秒内没响应,会报服务异常的。异步示例如下:
var wxReceiveFunc = func(msg wxapi.WxReceiveCommonMsg) error {fmt.Println("weixin msg received")fmt.Printf("%#v\n", msg)touser := msg.FromUserNamecontent := msg.ContentaccessToken := wxapi.WxGetAccessToken()//异步请求chatAI,成功后调用客服接口回复go func(){resp:=chatapi.AskChatAI(content)if(resp !=""){wxapi.WxPostCustomTextMsg(accessToken,touser,resp)}else{wxapi.WxPostCustomTextMsg(accessToken,touser,"chatGPT服务异常")}}()return nil
}
text-davinci-003介绍
使用text-davinci-003的模型的ai的接口,详细的介绍参照地址:
https://beta.openai.com/docs/api-reference/completions/create#completions/create-model
关于text-davinci-003和官方AI的区别,参照下面文章,其实用起来差别不大。甚至text-davinci-003在写作创造方面更优秀。简单说AI是个产品,text-davinci-003更面向开发者开发产品使用。
curl https://api.openai.com/v1/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"model": "text-davinci-003", "prompt": "Say this is a test", "temperature": 0, "max_tokens": 7}'
{"id": "cmpl-uqkvlQyYK7bGYrRHQ0eXlWi7","object": "text_completion","created": 1589478378,"model": "text-davinci-003","choices": [{"text": "\n\nThis is indeed a test","index": 0,"logprobs": null,"finish_reason": "length"}],"usage": {"prompt_tokens": 5,"completion_tokens": 7,"total_tokens": 12}
}
官方貌似只提供了python的demo和接口实现,其实模拟http调用使用其他语言都可以实现。
main.go代码实现
package mainimport ("fmt""net/http""weixin2/apis""weixin2/wxapi"_"weixin2/convert""weixin2/chatapi""github.com/gin-gonic/gin""github.com/jasonlvhit/gocron"
)var wxReceiveFunc = func(msg wxapi.WxReceiveCommonMsg) error {fmt.Println("weixin msg received")fmt.Printf("%#v\n", msg)touser := msg.FromUserNamecontent := msg.ContentaccessToken := wxapi.WxGetAccessToken()go func(){resp:=chatapi.AskChatAI(content)if(resp !=""){wxapi.WxPostCustomTextMsg(accessToken,touser,resp)}else{wxapi.WxPostCustomTextMsg(accessToken,touser,"chatGPT服务异常")}}()return nil
}func main() {// 定时每天早上7点 公众号发送天气预报gocron.Every(1).Day().At("07:00").Do(apis.SendWeather)// 定时每天早上九点 公众号发送每日一句gocron.Every(1).Day().At("10:00").Do(apis.SendExponent)fmt.Println("开启定时触发任务")gocron.Start()wxapi.WxReceiveFunc = wxReceiveFuncrouter := gin.Default()router.GET("/", wxapi.HandleWxLogin)router.POST("/", wxapi.HandleWxPostRecv)//运行的端口router.Run(":8000")
}
至于里面的每日发送是啥?
哈哈,为了好玩。我让每日7点给我发送一份天气预报。每日12点发个当前大盘点位,关注下股市大盘。至于大盘咨询的获取的实现,调用的第三方的接口。这里也给出下。
大盘信息获取接口实现:
package apisimport ("fmt""io/ioutil""net/http""strconv""strings""time""weixin2/util""weixin2/wxapi"
)var (LastPoint string //上次点数LastState string //上次状态LastTime string //上次时间State string //当前状态
)type ExponentInfo struct {CurrentPoints string // 当前点数Gszzl string //涨跌幅度Turnover string //成交额(万元)Name string //名称
}// 查询指数
func QueryBroadMarket(code string) (exponentInfo ExponentInfo, err error) {res, err := http.Get("http://hq.sinajs.cn/list=" + code)if res != nil {defer res.Body.Close()}if err != nil {return}if res.StatusCode != http.StatusOK {return exponentInfo, fmt.Errorf("http statusCode:%d", res.StatusCode)}result, err := ioutil.ReadAll(res.Body)if err != nil {return exponentInfo, err}base := util.ParseStr(string(result))//exponentInfo.Name = BroadMarketName[code]exponentInfo.Name = util.ConvertToString(base[0], "gbk", "utf-8")currentPoints := strings.Split(base[1], ".")exponentInfo.CurrentPoints = currentPoints[0]exponentInfo.Gszzl = base[3]if len(base[5]) > 5 {base[5] = base[5][0 : len(base[5])-4]base[5] += "亿"}exponentInfo.Turnover = base[5]return
}// ExponentTplID 模板ID
// {{title.DATA}} {{date.DATA}} {{curpoint.DATA}} {{turnover.DATA}} {{gszzl.DATA}} {{state.DATA}} {{lastpoint.DATA}} {{laststate.DATA}} {{tips.DATA}}
var ExponentTplID = "RIzqpuRxwlV1IWucjE186Uvu949N0Uly9H3CoTMGskc"// ExponentTpl 结构
type ExponentTpl struct {Title string `json:"title"`Content string `json:"content"`Note string `json:"note"`
}// AfterTime 判断一个时间在现在之前还是之后
// layout 解析的时间格式,
// judeTime 需要判断的时间
// bool =true 在当前时候之后,false在当前时间之前
func AfterTime(tpl, judeTime string) (bool, error) {stringToTime, err := time.Parse(tpl, judeTime)if err != nil {return false, err}beforeOrAfter := stringToTime.After(time.Now())if beforeOrAfter == true {return true, nil} else {return false, nil}
}//SendExponent 发送每日大盘
func SendExponent() {info, err := QueryBroadMarket("s_sh000001")if err != nil {fmt.Printf("QueryBroadMarket error,err=%s\n", err.Error())return}accessToken := wxapi.WxGetAccessToken()if accessToken == "" {return}flist := wxapi.WxGetUserList(accessToken)if flist == nil {return}title := "上证指数"timeStr := time.Now().Format("2006-01-02 15:04:05")tips := "tips:市场有风险,入市需谨慎"lastflt, err := strconv.ParseFloat(LastPoint, 32/64)curflt, err := strconv.ParseFloat(info.CurrentPoints, 32/64)if curflt > lastflt {State = "上升 " + strconv.FormatFloat(curflt-lastflt, 'f', 2, 32)} else {State = "下降 " + strconv.FormatFloat(curflt-lastflt, 'f', 2, 32)}reqdata := "{\"title\":{\"value\":\"" + title + "\", \"color\":\"#0000CD\"},\"date\":{\"value\":\"" + timeStr + "\"},\"curpoint\":{\"value\":\"" + "当前点数:" + info.CurrentPoints + "\"}, \"turnover\":{\"value\":\"" + "成交额:" + info.Turnover + "\"},\"gszzl\":{\"value\":\"" + "估值:" + info.Gszzl + "\"},\"state\":{\"value\":\"" + "今日状态:" + State + "\"},\"lastpoint\":{\"value\":\"" + "昨日点数:" + LastPoint + "\"},\"laststate\":{\"value\":\"" + "昨日状态:" + LastState + "\"},\"tips\":{\"value\":\"" + tips + "\"}}"for _, v := range flist {wxapi.WxPostTemplate(accessToken, reqdata, "blog.csdn.net/qq8864", ExponentTplID, v.Str)}now := time.Now()if now.Hour() > 13 {LastTime = timeStrLastPoint = info.CurrentPointsLastState = State}}
每日天气接口实现:
/**
每日天气api接口
*/
package apisimport ("fmt""github.com/tidwall/gjson""io/ioutil""net/http""weixin2/wxapi"
)var (WeatTemplateID = "AXCDf6fetE2-ebF7Di8izW9avSN4cFWvnfcAQDSW7-k" //天气模板IDWeatherApiVersion = "v6"WeatherAPPID = "45849795"WeatherAppSECRET = "T5xyVwLB"
)// GetWeather 获取天气
func GetWeather(city string) (string, string, string, string) {url := fmt.Sprintf("https://tianqiapi.com/api?version=%s&appid=%s&appsecret=%s&city=%s", WeatherApiVersion, WeatherAPPID, WeatherAppSECRET, city)resp, err := http.Get(url)if err != nil {fmt.Println("获取天气失败", err)return "", "", "", ""}defer resp.Body.Close()body, err := ioutil.ReadAll(resp.Body)if err != nil {fmt.Println("读取内容失败", err)return "", "", "", ""}fmt.Println(string(body))data := string(body)day := gjson.Get(data, "date").Strwea := gjson.Get(data, "wea").Strtem := gjson.Get(data, "tem").Str//tem2 := gjson.Get(thisday, "tem2").StrairTips := gjson.Get(data, "air_tips").Strreturn day, wea, tem, airTips
}// PostWeather 发送天气
func PostWeather(accessToken, city, openid string) {day, wea, tem, airTips := GetWeather(city)if day == "" || wea == "" || tem == "" || airTips == "" {return}reqdata := "{\"city\":{\"value\":\"城市:" + city + "\", \"color\":\"#0000CD\"}, \"day\":{\"value\":\"" + day + "\"}, \"wea\":{\"value\":\"天气:" + wea + "\"}, \"tem1\":{\"value\":\"平均温度:" + tem + "\"}, \"air_tips\":{\"value\":\"tips:" + airTips + "\"}}"//fmt.Println(reqdata)wxapi.WxPostTemplate(accessToken, reqdata, "http://yangqq.xyz", WeatTemplateID, openid)
}//SendWeather 向公众号所有用户发送天气预报
func SendWeather() {accessToken := wxapi.WxGetAccessToken()if accessToken == "" {return}flist := wxapi.WxGetUserList(accessToken)if flist == nil {return}city := "郑州"for _, v := range flist {go PostWeather(accessToken, city, v.Str)}
}
其他资源
ChatGPT API版介绍及使用讲解。 - 哔哩哔哩
ChatGPT又添劲敌?OpenAI核心员工创业,新模型获一片叫好
百度安全验证
【NLP】万字拆解!追溯ChatGPT各项能力的起源_qq62985c01d4e12的技术博客_51CTO博客
GitHub - transitive-bullshit/chatgpt-api: Node.js client for the unofficial ChatGPT API. 🔥
GitHub - malaohu/wechat-chatGPT: 实现微信公众号被动返回接口的ChatGPT
https://github.com/869413421/wechatbot
https://github.com/eatmoreapple/openwechat
GitHub - hktalent/ChatGPT-API: ChatGPT-API for go
GitHub - otiai10/openaigo: OpenAI GPT-3 API Client for Go
回复文本消息 | 微信开放文档
微信公众平台开发概述 | 微信开放文档
text-davinci-003和ChatGPT之间的不同点 - 知乎
OpenAI | GPT-3新模型Davinci,将AI写作提升到新水平!网友惊呼:GPT-4要来了?
OpenAI-Davinci,一个新的GPT-3模型,让AI写作更上一层楼!网友惊呼:GPT-4要_内容_能力_时间
https://www.cnblogs.com/taoshihan/p/17096902.html
百度安全验证
https://beta.openai.com/docs/api-reference/completions/create#completions/create-model