Havoc412
1ca75f6567
- 将 ES TopK 查询返回的文档数量从 1 增加到 2 - 注释中提到未来可能采用重排策略，从 10 篇文档中筛选出 2 篇 - 此修改旨在改善聊天机器人的回复质量
268 lines
8.1 KiB
Go
268 lines
8.1 KiB
Go
package web
|
||
|
||
import (
|
||
"catface/app/global/consts"
|
||
"catface/app/global/errcode"
|
||
"catface/app/global/variable"
|
||
"catface/app/model_res"
|
||
"catface/app/service/nlp"
|
||
"catface/app/service/rag/curd"
|
||
"catface/app/utils/llm_factory"
|
||
"catface/app/utils/micro_service"
|
||
"catface/app/utils/response"
|
||
"io"
|
||
"net/http"
|
||
"strconv"
|
||
|
||
"github.com/gin-gonic/gin"
|
||
"github.com/gorilla/websocket"
|
||
"go.uber.org/zap"
|
||
)
|
||
|
||
type Rag struct {
|
||
}
|
||
|
||
func (r *Rag) Release(context *gin.Context) {
|
||
token := context.GetString(consts.ValidatorPrefix + "token")
|
||
if ok := variable.GlmClientHub.ReleaseOneGlmClient(token); ok {
|
||
variable.ZapLog.Info("释放一个 GLM Client",
|
||
zap.String("token", token),
|
||
zap.String("当前空闲连接数", strconv.Itoa(variable.GlmClientHub.Idle)))
|
||
} else {
|
||
variable.ZapLog.Warn("尝试释放一个 GLM Client，但是 token 无效",
|
||
zap.String("当前空闲连接数", strconv.Itoa(variable.GlmClientHub.Idle)))
|
||
}
|
||
|
||
response.Success(context, consts.CurdStatusOkMsg, "")
|
||
}
|
||
|
||
// v1 Http-POST 版本; chat 需要不使用 ch 的版本。
|
||
// func (r *Rag) Chat(context *gin.Context) {
|
||
// // 1. query embedding
|
||
// query := context.GetString(consts.ValidatorPrefix + "query")
|
||
// embedding, ok := nlp.GetEmbedding(query)
|
||
// if !ok {
|
||
// code := errcode.ErrPythonService
|
||
// response.Fail(context, code, errcode.ErrMsg[code], "")
|
||
// return
|
||
// }
|
||
|
||
// // 2. ES TopK
|
||
// docs, err := model_es.CreateDocESFactory().TopK(embedding, 1)
|
||
// if err != nil || len(docs) == 0 {
|
||
// variable.ZapLog.Error("ES TopK error", zap.Error(err))
|
||
|
||
// code := errcode.ErrNoDocFound
|
||
// response.Fail(context, code, errcode.ErrMsg[code], errcode.ErrMsgForUser[code])
|
||
// }
|
||
|
||
// // 3. LLM answer
|
||
// if answer, err := nlp.ChatKnoledgeRAG(docs[0].Content, query); err == nil {
|
||
// response.Success(context, consts.CurdStatusOkMsg, gin.H{
|
||
// "answer": answer,
|
||
// })
|
||
// } else {
|
||
// response.Fail(context, consts.CurdStatusOkCode, consts.CurdStatusOkMsg, "")
|
||
// }
|
||
// }
|
||
|
||
func (r *Rag) ChatSSE(context *gin.Context) {
|
||
query := context.Query("query")
|
||
token := context.Query("token")
|
||
|
||
mode := context.Query("mode")
|
||
if mode == "" {
|
||
mode = consts.RagChatModeKnowledge
|
||
}
|
||
|
||
// 0-1. 测试 python
|
||
if !micro_service.TestLinkPythonService() {
|
||
code := errcode.ErrPythonService
|
||
response.Fail(context, code, errcode.ErrMsg[code], "")
|
||
return
|
||
}
|
||
|
||
// 0-2. 获取一个 GLM Client
|
||
if token == "" {
|
||
token = variable.SnowFlake.GetIdAsString()
|
||
}
|
||
client, ercode := variable.GlmClientHub.GetOneGlmClient(token, llm_factory.GlmModeKnowledgeHub)
|
||
if ercode != 0 {
|
||
response.Fail(context, ercode, errcode.ErrMsg[ercode], errcode.ErrMsgForUser[ercode])
|
||
return
|
||
}
|
||
defer variable.GlmClientHub.UnavtiveOneGlmClient(token) // INFO ws 结束时，取消 Avtive 的占用。
|
||
|
||
// 1. query embedding
|
||
embedding, ok := nlp.GetEmbedding([]string{query})
|
||
if !ok {
|
||
code := errcode.ErrPythonService
|
||
response.Fail(context, code, errcode.ErrMsg[code], "")
|
||
return
|
||
}
|
||
|
||
// 2. ES TopK
|
||
dochub, err := curd.TopK(mode, embedding, 1)
|
||
if err != nil || dochub.Length() == 0 {
|
||
variable.ZapLog.Error("ES TopK error", zap.Error(err))
|
||
|
||
code := errcode.ErrNoDocFound
|
||
response.Fail(context, code, errcode.ErrMsg[code], errcode.ErrMsgForUser[code])
|
||
}
|
||
|
||
// UPDATE
|
||
closeEventFromVue := context.Request.Context().Done()
|
||
ch := make(chan string) // TIP 建立通道。
|
||
|
||
// 3. LLM answer
|
||
go func() {
|
||
err := nlp.ChatRAG(query, mode, dochub, ch, client)
|
||
if err != nil {
|
||
variable.ZapLog.Error("ChatKnoledgeRAG error", zap.Error(err))
|
||
}
|
||
close(ch)
|
||
}()
|
||
|
||
context.Stream(func(w io.Writer) bool {
|
||
select {
|
||
case c, ok := <-ch:
|
||
if !ok {
|
||
return false
|
||
}
|
||
context.SSEvent("chat", c)
|
||
return true
|
||
case <-closeEventFromVue:
|
||
return false
|
||
}
|
||
})
|
||
}
|
||
|
||
var upgrader = websocket.Upgrader{ // TEST 测试，先写一个裸的 wss
|
||
ReadBufferSize: 1024,
|
||
WriteBufferSize: 1024,
|
||
CheckOrigin: func(r *http.Request) bool {
|
||
return true // info 在生产环境中可能需要更安全的检查
|
||
},
|
||
}
|
||
|
||
func (r *Rag) ChatWebSocket(context *gin.Context) {
|
||
query := context.Query("query")
|
||
token := context.Query("token")
|
||
|
||
// INFO 查询模式
|
||
mode := context.Query("mode")
|
||
if mode == "" {
|
||
mode = consts.RagChatModeKnowledge
|
||
}
|
||
|
||
if token == "" {
|
||
token = variable.SnowFlake.GetIdAsString()
|
||
}
|
||
|
||
// 0-1. 协议升级
|
||
ws, err := upgrader.Upgrade(context.Writer, context.Request, nil)
|
||
if err != nil {
|
||
variable.ZapLog.Error("OnOpen error", zap.Error(err))
|
||
response.Fail(context, errcode.ErrWebsocketUpgradeFail, errcode.ErrMsg[errcode.ErrWebsocketUpgradeFail], "")
|
||
return
|
||
}
|
||
defer ws.Close()
|
||
|
||
// 0-2. 测试 Python 微服务是否启动
|
||
if !micro_service.TestLinkPythonService() {
|
||
code := errcode.ErrPythonServierDown
|
||
err := ws.WriteMessage(websocket.TextMessage, model_res.CreateNlpWebSocketResult("", errcode.ErrMsgForUser[code]).JsonMarshal())
|
||
if err != nil {
|
||
variable.ZapLog.Error("Failed to send error message via WebSocket", zap.Error(err))
|
||
}
|
||
return
|
||
}
|
||
|
||
// 0-3. 从 GLM_HUB 中获取一个可用的 glm client;
|
||
clientInfo, ercode := variable.GlmClientHub.GetOneGlmClientInfo(token, llm_factory.GlmModeKnowledgeHub)
|
||
if ercode != 0 {
|
||
variable.ZapLog.Error("GetOneGlmClient error", zap.Error(err))
|
||
err := ws.WriteMessage(websocket.TextMessage, model_res.CreateNlpWebSocketResult("", errcode.ErrMsgForUser[ercode]).JsonMarshal())
|
||
if err != nil {
|
||
variable.ZapLog.Error("Failed to send error message via WebSocket", zap.Error(err))
|
||
}
|
||
return
|
||
}
|
||
defer variable.GlmClientHub.UnavtiveOneGlmClient(token) // INFO ws 结束时，取消 Avtive 的占用。
|
||
|
||
// 1. query embedding
|
||
clientInfo.AddQuery(query)
|
||
embedding, ok := nlp.GetEmbedding(clientInfo.UserQuerys)
|
||
if !ok {
|
||
code := errcode.ErrPythonServierDown
|
||
err := ws.WriteMessage(websocket.TextMessage, model_res.CreateNlpWebSocketResult("", errcode.ErrMsgForUser[code]).JsonMarshal())
|
||
if err != nil {
|
||
variable.ZapLog.Error("Failed to send error message via WebSocket", zap.Error(err))
|
||
}
|
||
return
|
||
}
|
||
|
||
// 2. ES TopK // INFO 这里需要特化选取不同知识库的文档；目前是依靠显式的路由。
|
||
dochub, err := curd.TopK(mode, embedding, 2) // 更好的做法是【重排】 10 -> 2;
|
||
if err != nil || dochub.Length() == 0 {
|
||
variable.ZapLog.Error("ES TopK error", zap.Error(err))
|
||
|
||
code := errcode.ErrNoDocFound
|
||
err := ws.WriteMessage(websocket.TextMessage, model_res.CreateNlpWebSocketResult("", errcode.ErrMsgForUser[code]).JsonMarshal())
|
||
if err != nil {
|
||
variable.ZapLog.Error("Failed to send error message via WebSocket", zap.Error(err))
|
||
}
|
||
return
|
||
}
|
||
|
||
// STAGE websocket 的 defer 关闭函数，但是需要 ES 拿到的 doc—id
|
||
defer func() { // UPDATE 临时"持久化"方案，之后考虑结合 jwt 维护的 token 处理。
|
||
// 0. 传递参考资料的信息
|
||
docMsg := model_res.CreateNlpWebSocketResult(consts.AiMessageTypeDoc, dochub.Docs)
|
||
err := ws.WriteMessage(websocket.TextMessage, docMsg.JsonMarshal())
|
||
if err != nil {
|
||
variable.ZapLog.Error("Failed to send doc message via WebSocket", zap.Error(err))
|
||
}
|
||
|
||
// 1. 传递 token 信息； // UPDATE 临时方案
|
||
tokenMsg := model_res.CreateNlpWebSocketResult(consts.AiMessageTypeToken, token)
|
||
err = ws.WriteMessage(websocket.TextMessage, tokenMsg.JsonMarshal())
|
||
if err != nil {
|
||
variable.ZapLog.Error("Failed to send token message via WebSocket", zap.Error(err))
|
||
}
|
||
// ws.Close() // 在上面调用了 defer；// TIP defer 的“栈”性质。
|
||
}()
|
||
|
||
// 3.
|
||
closeEventFromVue := context.Request.Context().Done() // 接收前端传来的中断信号。
|
||
ch := make(chan string) // TIP 建立通道。
|
||
|
||
go func() {
|
||
err := nlp.ChatRAG(query, mode, dochub, ch, clientInfo.Client) // TIP 接口
|
||
if err != nil {
|
||
variable.ZapLog.Error("ChatKnoledgeRAG error", zap.Error(err))
|
||
}
|
||
close(ch) // 这里 close，使得下方 for 结束。
|
||
}()
|
||
|
||
for {
|
||
select {
|
||
case c, ok := <-ch:
|
||
if !ok {
|
||
return
|
||
}
|
||
// variable.ZapLog.Info("ChatKnoledgeRAG", zap.String("c", c))
|
||
err := ws.WriteMessage(websocket.TextMessage, model_res.CreateNlpWebSocketResult("", c).JsonMarshal())
|
||
if err != nil {
|
||
return
|
||
}
|
||
case <-closeEventFromVue:
|
||
return
|
||
}
|
||
}
|
||
}
|
||
|
||
func (r *Rag) HelpDetectCat(context *gin.Context) {
|
||
// TODO 也许也可以同样用上面那个接口了。
|
||
}
|