refactor & optimize property.json (#242)

* 1. prevent changing property.json anywhere
2. every worker start will have a new property.json, strip property.json with non-use graphs
3. except for channel name, token and asr_language, move all other
   properties settings to frontend

* 1. moved properties to frontend
2. frontend can now overwrite property values, while channelName,
   remoteStreamUid still takes highest priority
3. use process group to better kill worker

* fix: remove legacy graph mappings

* fix: use native strings.Join

* fix: add page description

.env.example

@@ -11,9 +11,6 @@ LOG_PATH=/tmp/astra
 LOG_STDOUT=true
 # Graph designer server port
 GRAPH_DESIGNER_SERVER_PORT=49483
-# The corresponding graph name based on the language
-GRAPH_NAME_ZH=va.openai.azure
-GRAPH_NAME_EN=va.openai.azure
 # Server port
 SERVER_PORT=8080
 # Maximum number of workers

agents/.gitignore

@@ -27,4 +27,3 @@ xdump_config
 .vscode
 *.pyc
 *.pyc.*
-/property.json

playground/src/app/layout.tsx

@@ -7,7 +7,7 @@ import './global.css'
 
 export const metadata: Metadata = {
   title: "Astra.ai",
-  description: "Generated by create next app",
+  description: "A multimodal agent powered by TEN",
   appleWebApp: {
     capable: true,
     statusBarStyle: "black",

playground/src/common/constant.ts

@@ -69,3 +69,45 @@ export const COLOR_LIST: ColorItem[] = [{
   default: "#481C3F"
 }]
 
+export type VoiceTypeMap = {
+  [voiceType: string]: string;
+};
+
+export type VendorNameMap = {
+  [vendorName: string]: VoiceTypeMap;
+};
+
+export type LanguageMap = {
+  [language: string]: VendorNameMap;
+};
+
+export const voiceNameMap: LanguageMap = {
+  "zh-CN": {
+    azure: {
+      male: "zh-CN-YunxiNeural",
+      female: "zh-CN-XiaoxiaoNeural",
+    },
+    elevenlabs: {
+      male: "pNInz6obpgDQGcFmaJgB", // Adam
+      female: "Xb7hH8MSUJpSbSDYk0k2", // Alice
+    },
+    polly: {
+      male: "Zhiyu",
+      female: "Zhiyu",
+    },
+  },
+  "en-US": {
+    azure: {
+      male: "en-US-BrianNeural",
+      female: "en-US-JaneNeural",
+    },
+    elevenlabs: {
+      male: "pNInz6obpgDQGcFmaJgB", // Adam
+      female: "Xb7hH8MSUJpSbSDYk0k2", // Alice
+    },
+    polly: {
+      male: "Matthew",
+      female: "Ruth",
+    },
+  },
+};

playground/src/common/graph.ts

@@ -0,0 +1,46 @@
+import { voiceNameMap } from "./constant"
+
+export const getGraphWithLanguage = (graphName: string, language: string) => {
+    if (graphName == "camera.va.openai.azure") {
+        if (language == "zh-CN") {
+            return `${graphName}.cn`
+        }
+        return `${graphName}.en`
+    }
+    return graphName
+}
+
+export const getGraphProperties = (graphName: string, language: string, voiceType: string) => {
+    if (graphName == "camera.va.openai.azure") {
+        return {
+            "agora_rtc": {
+                "agora_asr_language": language,
+            },
+            "openai_chatgpt": {
+                "model": "gpt-4o"
+            },
+            "azure_tts": {
+                "azure_synthesis_voice_name": voiceNameMap[language]["azure"][voiceType]
+            }
+        }
+    } else if (graphName == "va.openai.azure") {
+        return {
+            "agora_rtc": {
+                "agora_asr_language": language,
+            },
+            "azure_tts": {
+                "azure_synthesis_voice_name": voiceNameMap[language]["azure"][voiceType]
+            }
+        }
+    } else if (graphName == "va.qwen.rag") {
+        return {
+            "agora_rtc": {
+                "agora_asr_language": language,
+            },
+            "azure_tts": {
+                "azure_synthesis_voice_name": voiceNameMap[language]["azure"][voiceType]
+            }
+        }
+    }
+    return {}
+}

playground/src/common/index.ts

@@ -4,3 +4,4 @@ export * from "./utils"
 export * from "./storage"
 export * from "./request"
 export * from "./mock"
+export * from "./graph"

playground/src/common/request.ts

@@ -1,12 +1,12 @@
+import { AnyObject } from "antd/es/_util/type"
 import { REQUEST_URL } from "./constant"
 import { genUUID } from "./utils"
 
 interface StartRequestConfig {
   channel: string
   userId: number,
-  language: string
-  voiceType: string
   graphName: string
+  properties: AnyObject
 }
 
 interface GenAgoraDataConfig {
@@ -35,15 +35,14 @@ export const apiGenAgoraData = async (config: GenAgoraDataConfig) => {
 
 export const apiStartService = async (config: StartRequestConfig): Promise<any> => {
   const url = `${REQUEST_URL}/start`
-  const { language, channel, userId, voiceType, graphName } = config
+  const { channel, userId, graphName, properties } = config
   const data = {
     request_id: genUUID(),
-    agora_asr_language: language,
     channel_name: channel,
     openai_proxy_url: "",
     remote_stream_id: userId,
-    voice_type: voiceType,
     graph_name: graphName,
+    properties,
   }
   let resp: any = await fetch(url, {
     method: "POST",

playground/src/common/utils.ts

@@ -56,4 +56,4 @@ export const genUUID = () => {
 
 export const isMobile = () => {
   return /Mobile|iPhone|iPad|Android|Windows Phone/i.test(navigator.userAgent)
-}
+}

playground/src/platform/mobile/description/index.tsx

@@ -1,7 +1,10 @@
 import { setAgentConnected } from "@/store/reducers/global"
 import {
   DESCRIPTION, useAppDispatch, useAppSelector, apiPing, genUUID,
-  apiStartService, apiStopService, REQUEST_URL
+  apiStartService, apiStopService, REQUEST_URL,
+  voiceNameMap,
+  getGraphWithLanguage,
+  getGraphProperties
 } from "@/common"
 import { message } from "antd"
 import { useEffect, useState } from "react"
@@ -48,9 +51,8 @@ const Description = () => {
       const res = await apiStartService({
         channel,
         userId,
-        language,
-        voiceType,
-        graphName,
+        graphName: getGraphWithLanguage(graphName, language),
+        properties: getGraphProperties(graphName, language, voiceType)
       })
       const { code, msg } = res || {}
       if (code != 0) {

playground/src/platform/pc/description/index.tsx

@@ -1,7 +1,10 @@
 import { setAgentConnected } from "@/store/reducers/global"
 import {
   DESCRIPTION, useAppDispatch, useAppSelector, apiPing, genUUID,
-  apiStartService, apiStopService, REQUEST_URL
+  apiStartService, apiStopService, REQUEST_URL,
+  voiceNameMap,
+  getGraphWithLanguage,
+  getGraphProperties
 } from "@/common"
 import { Select, Button, message, Upload } from "antd"
 import { useEffect, useState, MouseEventHandler } from "react"
@@ -48,9 +51,8 @@ const Description = () => {
       const res = await apiStartService({
         channel,
         userId,
-        language,
-        voiceType,
-        graphName
+        graphName: getGraphWithLanguage(graphName, language),
+        properties: getGraphProperties(graphName, language, voiceType)
       })
       const { code, msg } = res || {}
       if (code != 0) {

server/internal/config.go

@@ -2,7 +2,6 @@ package internal
 
 import (
 	"log/slog"
-	"os"
 )
 
 type Prop struct {
@@ -34,9 +33,6 @@ const (
 	PropertyJsonFile = "./agents/property.json"
 	// Token expire time
 	tokenExpirationInSeconds = uint32(86400)
-	// Voice type
-	voiceTypeMale   = "male"
-	voiceTypeFemale = "female"
 )
 
 var (
@@ -138,17 +134,8 @@ var (
 		},
 	}
 
-	// The corresponding graph name based on the language
-	graphNameMap = map[string]string{
-		languageChinese: "va.openai.azure",
-		languageEnglish: "va.openai.azure",
-	}
-
 	// Retrieve parameters from the request and map them to the property.json file
 	startPropMap = map[string][]Prop{
-		"AgoraAsrLanguage": {
-			{ExtensionName: extensionNameAgoraRTC, Property: "agora_asr_language"},
-		},
 		"ChannelName": {
 			{ExtensionName: extensionNameAgoraRTC, Property: "channel"},
 		},
@@ -158,54 +145,8 @@ var (
 		"Token": {
 			{ExtensionName: extensionNameAgoraRTC, Property: "token"},
 		},
-		"VoiceType": {
-			{ExtensionName: extensionNameAzureTTS, Property: "azure_synthesis_voice_name"},
-			{ExtensionName: extensionNameElevenlabsTTS, Property: "voice_id"},
-		},
 		"WorkerHttpServerPort": {
 			{ExtensionName: extensionNameHttpServer, Property: "listen_port"},
 		},
 	}
-
-	// Map the voice name to the voice type
-	voiceNameMap = map[string]map[string]map[string]string{
-		languageChinese: {
-			extensionNameAzureTTS: {
-				voiceTypeMale:   "zh-CN-YunxiNeural",
-				voiceTypeFemale: "zh-CN-XiaoxiaoNeural",
-			},
-			extensionNameElevenlabsTTS: {
-				voiceTypeMale:   "pNInz6obpgDQGcFmaJgB", // Adam
-				voiceTypeFemale: "Xb7hH8MSUJpSbSDYk0k2", // Alice
-			},
-			extensionNamePollyTTS: {
-				voiceTypeMale:   "Zhiyu",
-				voiceTypeFemale: "Zhiyu",
-			},
-		},
-		languageEnglish: {
-			extensionNameAzureTTS: {
-				voiceTypeMale:   "en-US-BrianNeural",
-				voiceTypeFemale: "en-US-JaneNeural",
-			},
-			extensionNameElevenlabsTTS: {
-				voiceTypeMale:   "pNInz6obpgDQGcFmaJgB", // Adam
-				voiceTypeFemale: "Xb7hH8MSUJpSbSDYk0k2", // Alice
-			},
-			extensionNamePollyTTS: {
-				voiceTypeMale:   "Matthew",
-				voiceTypeFemale: "Ruth",
-			},
-		},
-	}
 )
-
-func SetGraphNameMap() {
-	if graphNameZH := os.Getenv("GRAPH_NAME_ZH"); graphNameZH != "" {
-		graphNameMap[languageChinese] = graphNameZH
-	}
-
-	if graphNameEN := os.Getenv("GRAPH_NAME_EN"); graphNameEN != "" {
-		graphNameMap[languageEnglish] = graphNameEN
-	}
-}