First version of replacements for EMO

Author: SebbeJohansson

README.md

@@ -10,3 +10,12 @@ This part of the knowledge bank is a Proxy server to analyze the traffic that go
 ## Docker setup
 You can find a simplified setup using Docker Compose by SebbeJohansson here: https://github.com/SebbeJohansson/emo-proxy-docker
 It uses nginx, dnsmasq, and mitmproxy to be able to pass through the api requests to the EMO Proxy.
+
+## Experimental
+### ChatGPT Speak server
+One of the most common responses from EMO is to respond using ChatGPT.
+An experimental feature is to use a local Speak server to generate EMO's chatgpt speak responses instead of using the living.ai servers.
+You can find a proof of concept server here: https://github.com/SebbeJohansson/EmoChatGptSpeakPOC
+
+Connect the server by adding the following line to your emoProxy.conf file:
+`"chatGptSpeakServer": "http://localhost:8085"`

emoProxy.conf.example

@@ -7,5 +7,6 @@
     "postFS": "/tmp/",
     "logFileName": "/var/log/emoProxyss.log",
     "enableDatabaseAndAPI": false, # For now, default behavior is still without db and api.
-    "sqliteLocation": "/var/data/emo_logs.db"
+    "sqliteLocation": "/var/data/emo_logs.db",
+    "chatGptSpeakServer": ""
 }

emoProxy.go

@@ -9,6 +9,7 @@ import (
 	"io"
 	"log"
 	"net/http"
+	"net/url"
 	"os"
 	"strconv"
 	"time"
@@ -24,6 +25,53 @@ type emo_code struct {
 	Errmessage string `json:"errmessage"`
 }
 
+type Intent struct {
+	Name       string  `json:"name,omitempty"`
+	Confidence float64 `json:"confidence,omitempty"`
+}
+
+type BehaviorParas struct {
+	UtilityType   string         `json:"utility_type,omitempty"`
+	Time          []string       `json:"time,omitempty"`
+	Txt           string         `json:"txt,omitempty"`
+	Url           string         `json:"url,omitempty"`
+	PreAnimation  string         `json:"pre_animation,omitempty"`
+	PostAnimation string         `json:"post_animation,omitempty"`
+	PostBehavior  string         `json:"post_behavior,omitempty"`
+	RecBehavior   string         `json:"rec_behavior,omitempty"`
+	BehaviorParas *BehaviorParas `json:"behavior_paras,omitempty"`
+	Sentiment     string         `json:"sentiment,omitempty"`
+	Listen        int            `json:"listen,omitempty"`
+	AnimationName string         `json:"animation_name,omitempty"`
+}
+
+type QueryResult struct {
+	ResultCode    string         `json:"resultCode,omitempty"`
+	QueryText     string         `json:"queryText,omitempty"`
+	Intent        *Intent        `json:"intent,omitempty"`
+	RecBehavior   string         `json:"rec_behavior,omitempty"`
+	BehaviorParas *BehaviorParas `json:"behavior_paras,omitempty"`
+}
+
+type QueryResponse struct {
+	QueryId      string       `json:"queryId,omitempty"`
+	QueryResult  *QueryResult `json:"queryResult,omitempty"`
+	LanguageCode string       `json:"languageCode,omitempty"`
+	Index        int          `json:"index,omitempty"`
+}
+
+type TokenResponse struct {
+	AccessToken string `json:"access_token,omitempty"`
+	ExpireIn    int    `json:"expire_in,omitempty"`
+	Type        string `json:"type,omitempty"`
+}
+
+type EmoSpeechResponse struct {
+	Code       int64  `json:"code"`
+	Errmessage string `json:"errmessage"`
+	Url        string `json:"url"`
+}
+
 type Configuration struct {
 	PidFile                 string `json:"pidFile"`
 	Livingio_API_Server     string `json:"livingio_api_server"`
@@ -34,6 +82,7 @@ type Configuration struct {
 	LogFileName             string `json:"logFileName"`
 	EnableDatabaseAndAPI    bool   `json:"enableDatabaseAndAPI"`
 	SqliteLocation          string `json:"sqliteLocation"`
+	ChatGptSpeakServer      string `json:"chatGptSpeakServer"`
 }
 
 var (
@@ -105,6 +154,7 @@ func loadConfig(filename string) error {
 		LogFileName:             "/var/log/emoProxy.log",
 		EnableDatabaseAndAPI:    false,
 		SqliteLocation:          "/var/data/emo_logs.db",
+		ChatGptSpeakServer:      "",
 	}
 
 	bytes, err := os.ReadFile(filename)
@@ -315,6 +365,29 @@ func makeApiRequest(r *http.Request) string {
 	body, _ := io.ReadAll(response.Body)
 	log.Println("Server response: ", string(body))
 
+	typedBody := QueryResponse{}
+	decoder := json.NewDecoder(bytes.NewReader([]byte(body)))
+	decoder.DisallowUnknownFields()
+
+	err = decoder.Decode(&typedBody)
+	if err != nil {
+		log.Println("Error when decoding JSON (", string(body), ") will return unhandled:", err)
+	} else {
+		if typedBody.QueryId != "" {
+			if typedBody.QueryResult.Intent.Name == "chatgpt_speak" && conf.ChatGptSpeakServer != "" {
+				speakResponse := makeChatGptSpeakRequest(typedBody.QueryResult.QueryText, typedBody.LanguageCode, typedBody.QueryResult.BehaviorParas.Txt, r)
+				if speakResponse.Url != "" && speakResponse.Txt != "" {
+					log.Println("Successfully replaced chatgpt_speak response for request.")
+					typedBody.QueryResult.BehaviorParas.Url = speakResponse.Url
+					typedBody.QueryResult.BehaviorParas.Txt = speakResponse.Txt
+				} else {
+					log.Println("Failed to get valid response from ChatGptSpeakServer, keeping original response.")
+				}
+			}
+			body, _ = json.Marshal(typedBody)
+		}
+	}
+
 	logResponse(response)
 
 	if useDatabaseAndAPI {
@@ -346,10 +419,8 @@ func makeTtsRequest(r *http.Request) string {
 	}
 	defer response.Body.Close()
 
-	// read response
 	body, _ := io.ReadAll(response.Body)
 
-	// write post request body to fs
 	logBody(response.Header.Get("Content-Type"), body, "tts_")
 	logResponse(response)
 
@@ -382,10 +453,8 @@ func makeApiTtsRequest(r *http.Request) string {
 	}
 	defer response.Body.Close()
 
-	// read response
 	body, _ := io.ReadAll(response.Body)
 
-	// write post request body to fs
 	logBody(response.Header.Get("Content-Type"), body, "apitts_")
 	logResponse(response)
 
@@ -418,7 +487,6 @@ func makeResRequest(r *http.Request, w http.ResponseWriter) string {
 	}
 	defer response.Body.Close()
 
-	// read response
 	body, _ := io.ReadAll(response.Body)
 
 	logBody(response.Header.Get("Content-Type"), body, "res_")
@@ -434,3 +502,89 @@ func makeResRequest(r *http.Request, w http.ResponseWriter) string {
 	}
 	return string(body)
 }
+
+func makeEmoSpeechRequest(text string, languageCode string, r *http.Request) EmoSpeechResponse {
+	request, _ := http.NewRequest("GET", "https://"+conf.Livingio_API_Server+"/emo/speech/tts?q="+url.QueryEscape(text)+"&l="+url.QueryEscape(languageCode), nil)
+
+	val, exists := r.Header["Authorization"]
+	if exists {
+		request.Header.Add("Authorization", val[0])
+	}
+
+	val, exists = r.Header["Secret"]
+	if exists {
+		request.Header.Add("Secret", val[0])
+	}
+
+	request.Header.Del("User-Agent")
+
+	httpclient := &http.Client{}
+	response, err := httpclient.Do(request)
+
+	if err != nil {
+		log.Fatalf("An Error Occured %v", err)
+	}
+	defer response.Body.Close()
+
+	body, _ := io.ReadAll(response.Body)
+
+	var emoSpeechResponse EmoSpeechResponse
+	if err := json.Unmarshal([]byte(body), &emoSpeechResponse); err != nil {
+		log.Printf("Error unmarshaling ChatGptSpeakServer response: %v\n", err)
+		return EmoSpeechResponse{}
+	}
+
+	return emoSpeechResponse
+}
+
+func makeChatGptSpeakRequest(queryText string, languageCode string, fallbackResponse string, r *http.Request) BehaviorParas {
+	type ChatGptSpeakRequest struct {
+		QueryText        string `json:"queryText"`
+		LanguageCode     string `json:"languageCode"`
+		FallbackResponse string `json:"fallbackResponse,omitempty"`
+	}
+	type ChatGptSpeakResponse struct {
+		ResponseText string `json:"responseText"`
+	}
+
+	chatGptRequestData := ChatGptSpeakRequest{
+		QueryText:        queryText,
+		LanguageCode:     languageCode,
+		FallbackResponse: fallbackResponse,
+	}
+	chatGptRequestBody, _ := json.Marshal(chatGptRequestData)
+	chatGptRequest, _ := http.NewRequest("POST", conf.ChatGptSpeakServer+"/speak", bytes.NewBuffer(chatGptRequestBody))
+	chatGptRequest.Header.Add("Content-Type", "application/json")
+
+	chatGptClient := &http.Client{}
+	chatGptResponse, err := chatGptClient.Do(chatGptRequest)
+	if err != nil {
+		log.Fatalf("An Error Occured while calling ChatGptSpeakServer %v", err)
+	}
+	defer chatGptResponse.Body.Close()
+
+	chatGptResponseBody, _ := io.ReadAll(chatGptResponse.Body)
+
+	var chatGptTypedResponse ChatGptSpeakResponse
+	if err := json.Unmarshal([]byte(chatGptResponseBody), &chatGptTypedResponse); err != nil {
+		log.Printf("Error unmarshaling ChatGptSpeakServer response: %v\n", err)
+		return BehaviorParas{}
+	}
+
+	if chatGptTypedResponse.ResponseText == "" {
+		log.Println("ChatGptSpeakServer returned empty response text")
+		return BehaviorParas{}
+	}
+
+	emoSpeechResponse := makeEmoSpeechRequest(chatGptTypedResponse.ResponseText, languageCode, r)
+	if emoSpeechResponse.Code != 200 || emoSpeechResponse.Url == "" {
+		log.Printf("Error in EmoSpeechResponse: Code %d, Errmessage: %s\n", emoSpeechResponse.Code, emoSpeechResponse.Errmessage)
+		return BehaviorParas{}
+	}
+	behaviorParasResponse := BehaviorParas{
+		Txt: chatGptTypedResponse.ResponseText,
+		Url: emoSpeechResponse.Url,
+	}
+
+	return behaviorParasResponse
+}