@@ -1688,12 +1706,16 @@ function _Chat() {
})}
)}
-
}
- text={Locale.Chat.Send}
- type="primary"
- onClick={() => doSubmit(userInput)}
- />
+
+
+
}
text={Locale.Chat.Send}
diff --git a/app/components/voice-input.tsx b/app/components/voice-input.tsx
index ab82dfdcf..11f05908e 100644
--- a/app/components/voice-input.tsx
+++ b/app/components/voice-input.tsx
@@ -1,82 +1,172 @@
-"use client";
+// "use client";
import { Button, Input, Space } from "antd";
-import { useEffect, useMemo, useRef, useState } from "react";
+import {
+ Dispatch,
+ SetStateAction,
+ useEffect,
+ useMemo,
+ useRef,
+ useState,
+} from "react";
import { AudioOutlined, LoadingOutlined } from "@ant-design/icons";
+import * as ms_audio_sdk from "microsoft-cognitiveservices-speech-sdk";
+import {
+ Recognizer,
+ SpeechRecognitionCanceledEventArgs,
+ SpeechRecognitionEventArgs,
+ SpeechRecognitionResult,
+} from "microsoft-cognitiveservices-speech-sdk/distrib/lib/src/sdk/Exports";
+import { useAccessStore } from "@/app/store";
-export default function VoiceInput() {
- const [userInput, setUserInput] = useState("");
- const [loading, setLoading] = useState(false);
- const recognition = useRef(null);
+interface VoiceInputInterface {
+ userInput: string;
+ setUserInput: Dispatch
>;
+}
- const lastLength = useRef(0);
+// @ts-ignore
+export default function VoiceInput({
+ userInput,
+ setUserInput,
+}: VoiceInputInterface) {
+ const [voiceInputText, setVoiceInputText] = useState("");
+ const [voiceInputLoading, setVoiceInputLoading] = useState(false);
+ // const recognition = useRef(null);
+ const recognizer = useRef();
+ const [tempUserInput, setTempUserInput] = useState("");
+ const accessStore = useAccessStore();
+ // const lastLength = useRef(0);
- useEffect(() => {
- if ("webkitSpeechRecognition" in window) {
- if (recognition.current === null) {
- recognition.current = new window.webkitSpeechRecognition();
- }
- } else {
- console.error("此浏览器不支持webkitSpeechRecognition。");
- return;
- }
- if (!recognition.current) return;
- // 设置语言
- recognition.current.lang = "zh";
- // 开启连续识别
- recognition.current.continuous = true;
- // 开启实时识别
- recognition.current.interimResults = true;
+ // useEffect(() => {
+ //
+ // function onresult(event: any) {
+ // // 这个事件会把前面识别的结果都返回回来，所以需要取最后一个识别结果
+ // const length = event.results.length;
+ // // 没有新的识别结果的时候，事件也会触发，所以这里判断一下如果没有新的识别结果，就不取最后一个识别结果了。
+ // if (lastLength.current === length) {
+ // return;
+ // }
+ //
+ // lastLength.current = length;
+ //
+ // console.log(event.results);
+ //
+ // // 获取最后一个识别结果
+ // const transcript = event.results[length - 1]?.[0]?.transcript;
+ //
+ // // 将最后一个识别结果添加到文本
+ // if (transcript) {
+ // setVoiceInputText((voiceInputText) => voiceInputText + transcript);
+ // }
+ // }
+ //
+ // }, []);
- function onresult(event: any) {
- // 这个事件会把前面识别的结果都返回回来，所以需要取最后一个识别结果
- const length = event.results.length;
- // 没有新的识别结果的时候，事件也会触发，所以这里判断一下如果没有新的识别结果，就不取最后一个识别结果了。
- if (lastLength.current === length) {
- return;
- }
+ function onRecognizedResult(result: SpeechRecognitionResult) {
+ // setVoiceInputText("");
+ setVoiceInputText(`${result.text}`);
- lastLength.current = length;
-
- console.log(event.results);
-
- // 获取最后一个识别结果
- const transcript = event.results[length - 1]?.[0]?.transcript;
-
- // 将最后一个识别结果添加到文本
- if (transcript) {
- setUserInput((userInput) => userInput + transcript);
- }
+ let intentJson = result.properties.getProperty(
+ ms_audio_sdk.PropertyId.LanguageUnderstandingServiceResponse_JsonResult,
+ );
+ if (intentJson) {
+ setVoiceInputText(voiceInputText + `${intentJson}`);
}
- // 监听语音识别结果
- recognition.current.addEventListener("result", onresult);
+ // setTempUserInput("");
+ console.log("3333", tempUserInput, "2", voiceInputText);
- return () => {
- if (recognition.current) {
- recognition.current.removeEventListener("result", onresult);
- }
- };
- }, []);
+ // if (result?.translations) {
+ // let resultJson = JSON.parse(result.json);
+ // resultJson['privTranslationPhrase']['Translation']['Translations'].forEach(
+ // function (translation: { Language: any; Text: any; }) {
+ // setVoiceInputText(voiceInputText + ` [${translation.Language}] ${translation.Text}\r\n`);
+ // });
+ // }
+ }
+ function onCanceled(
+ sender: Recognizer,
+ event: SpeechRecognitionCanceledEventArgs,
+ ) {
+ console.log(event);
- function click() {
- if (loading) {
- recognition.current.stop();
- setLoading(false);
- return;
- }
- setLoading(true);
+ // 展示取消事件
+ // statusDiv.innerHTML += "(cancel) Reason: " + ms_audio_sdk.CancellationReason[event.reason];
+ // if (event.reason === ms_audio_sdk.CancellationReason.Error) {
+ // statusDiv.innerHTML += ": " + event.errorDetails;
+ // }
+ // statusDiv.innerHTML += "\r\n";
+ }
+ function onRecognizing(
+ sender: Recognizer,
+ event: SpeechRecognitionEventArgs,
+ ) {
+ let result = event.result;
+ setUserInput(
+ tempUserInput +
+ voiceInputText.replace(/(.*)(^|[\r\n]+).*\[\.\.\.][\r\n]+/, "$1$2") +
+ `${result.text} [...]`,
+ );
- lastLength.current = 0;
- recognition.current.start();
+ setVoiceInputText(
+ voiceInputText.replace(/(.*)(^|[\r\n]+).*\[\.\.\.][\r\n]+/, "$1$2") +
+ `${result.text} [...]`,
+ );
}
+ const startRecognition = () => {
+ if (voiceInputLoading) {
+ recognizer.current?.close();
+ setVoiceInputLoading(false);
+ // setVoiceInputText("");
+ // setUserInput(tempUserInput);
+ return;
+ }
+
+ setVoiceInputLoading(true);
+ setTempUserInput(userInput); // 开始的时候拷贝一份用于复原
+ setVoiceInputText("");
+
+ const speechConfig = ms_audio_sdk.SpeechConfig.fromSubscription(
+ accessStore.azureVoiceKey,
+ "eastasia",
+ );
+ const audioConfig = ms_audio_sdk.AudioConfig.fromDefaultMicrophoneInput();
+ speechConfig.speechRecognitionLanguage = "zh-CN";
+ speechConfig.setProperty(
+ ms_audio_sdk.PropertyId.SpeechServiceConnection_EndSilenceTimeoutMs,
+ "2500",
+ );
+ recognizer.current = new ms_audio_sdk.SpeechRecognizer(
+ speechConfig,
+ audioConfig,
+ );
+ recognizer.current.recognizing = onRecognizing; // 自定义分段显示
+ recognizer.current.canceled = onCanceled; // 自定义中断
+ recognizer.current.recognizeOnceAsync(
+ (result) => {
+ // onRecognizedResult(result);
+ setVoiceInputText(`${result.text}`);
+ console.log("3333", tempUserInput, "2", voiceInputText);
+ setUserInput(tempUserInput + voiceInputText + `${result.text}`);
+ // setVoiceInputText(result.text);
+ console.log("result", result.text);
+ setVoiceInputLoading(false);
+ // recognizer.close();
+ },
+ (err) => {
+ console.error("Recognition error: ", err); // 错误处理
+ setVoiceInputLoading(false);
+ },
+ );
+ };
+
const icon = useMemo(() => {
- if (loading) {
+ if (voiceInputLoading) {
return (
);
@@ -85,17 +175,17 @@ export default function VoiceInput() {
);
- }, [loading]);
+ }, [voiceInputLoading]);
return (
-
-
-
-
+
+
+ {/**/}
+
);
diff --git a/app/config/server.ts b/app/config/server.ts
index eea8a18d6..72998cc17 100644
--- a/app/config/server.ts
+++ b/app/config/server.ts
@@ -27,6 +27,7 @@ declare global {
AZURE_URL?: string; // https://{azure-url}/openai/deployments/{deploy-name}
AZURE_API_KEY?: string;
AZURE_API_VERSION?: string;
+ AZURE_VOICE_KEY?: string;
// google only
GOOGLE_API_KEY?: string;
@@ -93,6 +94,7 @@ export const getServerSideConfig = () => {
azureUrl: process.env.AZURE_URL ?? "",
azureApiKey: process.env.AZURE_API_KEY ?? "",
azureApiVersion: process.env.AZURE_API_VERSION ?? "",
+ azureVoiceKey: process.env.AZURE_VOICE_KEY ?? "",
isGoogle,
googleApiKey: process.env.GOOGLE_API_KEY,
diff --git a/app/constant.ts b/app/constant.ts
index aa0b6be7e..dce78c587 100644
--- a/app/constant.ts
+++ b/app/constant.ts
@@ -107,7 +107,7 @@ Latex inline: $x^2$
Latex block: $$e=mc^2$$
`;
-export const SUMMARIZE_MODEL = "gpt-3.5-turbo-1106";
+export const SUMMARIZE_MODEL = "gpt-3.5-turbo-0125";
export const GEMINI_SUMMARIZE_MODEL = "gemini-pro";
export const KnowledgeCutOffDate: Record = {
@@ -132,8 +132,8 @@ export const DEFAULT_MODELS = [
// available: false,
// },
{
- name: "gpt-3.5-turbo-1106",
- describe: "GPT-3,最快,笨,最便宜",
+ name: "gpt-3.5-turbo-0125",
+ describe: "GPT-3,最快,效果一般,最便宜",
available: true,
provider: {
id: "openai",
diff --git a/app/masks/cn.ts b/app/masks/cn.ts
index efeecf802..c040ff4ec 100644
--- a/app/masks/cn.ts
+++ b/app/masks/cn.ts
@@ -58,7 +58,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -84,7 +84,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -110,7 +110,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -136,7 +136,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -162,7 +162,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -188,7 +188,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -214,7 +214,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -240,7 +240,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -272,7 +272,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 0.5,
max_tokens: 2000,
presence_penalty: 0,
@@ -298,7 +298,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -331,7 +331,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -364,7 +364,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
@@ -422,7 +422,7 @@ export const CN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 1,
max_tokens: 2000,
presence_penalty: 0,
diff --git a/app/masks/en.ts b/app/masks/en.ts
index 0a3939de5..d315a5b6a 100644
--- a/app/masks/en.ts
+++ b/app/masks/en.ts
@@ -86,7 +86,7 @@ export const EN_MASKS: BuiltinMask[] = [
},
],
modelConfig: {
- model: "gpt-3.5-turbo-1106",
+ model: "gpt-3.5-turbo-0125",
temperature: 0.5,
max_tokens: 2000,
presence_penalty: 0,
diff --git a/app/store/access.ts b/app/store/access.ts
index dd8d45b4a..3e38f695e 100644
--- a/app/store/access.ts
+++ b/app/store/access.ts
@@ -30,6 +30,7 @@ const DEFAULT_ACCESS_STATE = {
azureUrl: "",
azureApiKey: "",
azureApiVersion: "2023-05-15",
+ azureVoiceKey: "",
// google ai studio
googleUrl: "",
diff --git a/app/store/config.ts b/app/store/config.ts
index 4f32c590d..497f0dd0a 100644
--- a/app/store/config.ts
+++ b/app/store/config.ts
@@ -51,7 +51,7 @@ export const DEFAULT_CONFIG = {
dontUseModel: DISABLE_MODELS,
modelConfig: {
- model: "gpt-3.5-turbo-1106" as ModelType,
+ model: "gpt-3.5-turbo-0125" as ModelType,
temperature: 0.8,
top_p: 1,
max_tokens: 2000,
@@ -137,7 +137,7 @@ export const useAppConfig = createPersistStore(
}),
{
name: StoreKey.Config,
- version: 3.8991,
+ version: 3.8992,
migrate(persistedState, version) {
const state = persistedState as ChatConfig;
@@ -168,7 +168,7 @@ export const useAppConfig = createPersistStore(
if (version < 3.8) {
state.lastUpdate = Date.now();
}
- if (version < 3.8991) {
+ if (version < 3.8992) {
state.lastUpdate = Date.now();
return { ...DEFAULT_CONFIG };
}