核心知识点：
gin 加载html
chat gpt Stream
gin stream
js EventSource
关于gpt3中的上下文:
只要将之前的问题和回答用换行符号拼接 最后再加新的的问题即可
关于参数MaxTokens：
返回最大文本数，openai 也是基于这个参数收费
源码：
用html简单写个简易前端
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>Title</title>
</head>
<body>
<div id="result">
</div>
<div>
<input type="text" id="msg">
<button type="button" onclick="submit()">提交!</button>
</div>
</body>
<script>
class Chat {
constructor(q, a) {
this.q = q;
this.a = a;
}
}
const chats = [];
function submit() {
let question = document.getElementById("msg").value
if (question === undefined || question === "") {
alert("please input question")
}
const chat = new Chat();
chat.q = question;
chat.a = "";
const stream = new EventSource("/stream?msg=" + chat.q + "&chats=" + JSON.stringify(chats))
stream.addEventListener("message", function (e) {
let parse = JSON.parse(e.data);
chat.a += parse["text"]
document.getElementById("result").innerHTML = chat.a
});
stream.addEventListener("stop", function (e) {
stream.close()
chat.a += "\n"
chats.push(chat)
document.getElementById("result").innerHTML = chat.a
console.log(e)
});
stream.onerror = function (event) {
stream.close()
console.log(event)
}
}
</script>
</html>
go
gin提供服务
go-gp3 调用openai接口
package main
import (
"context"
"encoding/json"
"errors"
"fmt"
"github.com/gin-gonic/gin"
gogpt "github.com/sashabaranov/go-gpt3"
"io"
"log"
"net/http"
"os"
"os/signal"
"time"
)
type SteamRequest struct {
Msg string `form:"msg" binding:"required"`
Chats string `form:"chats" binding:"required"`
}
type Chat struct {
A string `form:"a"`
Q string `form:"q"`
}
func setupRouter() *gin.Engine {
r := gin.Default()
r.LoadHTMLGlob("templates/*")
r.GET("/", func(c *gin.Context) {
c.HTML(http.StatusOK, "index.html", gin.H{
"title": "Main website",
})
})
r.GET("/stream", func(c *gin.Context) {
var request SteamRequest
err := c.ShouldBindQuery(&request)
if err != nil {
log.Fatal(err.Error())
}
log.Println("request" + request.Msg)
var chats []Chat
errs := json.Unmarshal([]byte(request.Chats), &chats)
if errs != nil {
fmt.Println("json unmarshal error:", errs)
}
var question = ""
if len(chats) != 0 {
for i := 0; i < len(chats); i++ {
chat := chats[i]
question += chat.Q + "\n" + chat.A
}
}
question += request.Msg
client := gogpt.NewClient("you token")
ctx := context.Background()
req := gogpt.CompletionRequest{
Model: gogpt.GPT3TextDavinci003,
MaxTokens: 5,
Prompt: question,
Stream: true,
}
stream, err := client.CreateCompletionStream(ctx, req)
if err != nil {
return
}
chanStream := make(chan string, 100)
go func() {
defer stream.Close()
defer close(chanStream)
for {
response, err := stream.Recv()
if errors.Is(err, io.EOF) {
fmt.Println("Stream finished")
chanStream <- "<!finish>"
return
}
if err != nil {
fmt.Printf("Stream error: %v\n", err)
chanStream <- "<!error>"
return
}
if len(response.Choices) == 0 {
fmt.Println("Stream finished")
chanStream <- "<!finish>"
return
}
data, err := json.Marshal(response.Choices[0])
chanStream <- string(data)
fmt.Printf("Stream response: %v\n", response.Choices[0])
}
}()
c.Stream(func(w io.Writer) bool {
if msg, ok := <-chanStream; ok {
if msg == "<!finish>" {
c.SSEvent("stop", "finish")
}
if msg == "<!error>" {
c.SSEvent("stop", "error")
}
c.SSEvent("message", msg)
fmt.Printf("message: %v\n", msg)
return true
}
return false
})
})
return r
}
func main() {
gin.SetMode(gin.ReleaseMode)
router := setupRouter()
srv := &http.Server{
Addr: ":8080",
Handler: router,
}
go func() {
// 服务连接
if err := srv.ListenAndServe(); err != nil && err != http.ErrServerClosed {
log.Fatalf("listen: %s\n", err)
}
}()
// 等待中断信号以优雅地关闭服务器（设置 5 秒的超时时间）
quit := make(chan os.Signal)
signal.Notify(quit, os.Interrupt)
<-quit
log.Println("Shutdown Server ...")
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
defer cancel()
if err := srv.Shutdown(ctx); err != nil {
log.Fatal("Server Shutdown:", err)
}
log.Println("Server exiting")
}
网友评论