Fix punctuation (#477)

* add punctuate command

* check punctuation before alignment

enjoy/src/commands/index.ts

@@ -3,3 +3,4 @@ export * from "./lookup.command";
 export * from "./translate.command";
 export * from "./ipa.command";
 export * from "./analyze.command";
+export * from "./punctuate.command";

enjoy/src/commands/punctuate.command.ts

@@ -0,0 +1,38 @@
+import { ChatOpenAI } from "@langchain/openai";
+import { ChatPromptTemplate } from "@langchain/core/prompts";
+
+export const punctuateCommand = async (
+  text: string,
+  options: {
+    key: string;
+    modelName?: string;
+    temperature?: number;
+    baseUrl?: string;
+  }
+): Promise<string> => {
+  const { key, temperature = 0, baseUrl } = options;
+  let { modelName = "gpt-4-turbo-preview" } = options;
+
+  const chatModel = new ChatOpenAI({
+    openAIApiKey: key,
+    modelName,
+    temperature,
+    configuration: {
+      baseURL: baseUrl,
+    },
+    cache: false,
+    verbose: true,
+    maxRetries: 2,
+  });
+
+  const prompt = ChatPromptTemplate.fromMessages([
+    ["system", SYSTEM_PROMPT],
+    ["human", text],
+  ]);
+
+  const response = await prompt.pipe(chatModel).invoke({});
+
+  return response.text;
+};
+
+const SYSTEM_PROMPT = `Please add proper punctuation to the text I provide you. Return the corrected text only.`;

enjoy/src/renderer/hooks/use-ai-command.tsx

@@ -8,6 +8,7 @@ import {
   extractStoryCommand,
   translateCommand,
   analyzeCommand,
+  punctuateCommand,
 } from "@commands";
 
 export const useAiCommand = () => {
@@ -102,10 +103,19 @@ export const useAiCommand = () => {
     });
   };
 
+  const punctuateText = async (text: string) => {
+    return punctuateCommand(text, {
+      key: currentEngine.key,
+      modelName: currentEngine.model,
+      baseUrl: currentEngine.baseUrl,
+    })
+  }
+
   return {
     lookupWord,
     extractStory,
     translate,
     analyzeText,
+    punctuateText
   };
 };

enjoy/src/renderer/hooks/use-transcribe.tsx

@@ -9,10 +9,12 @@ import { AI_WORKER_ENDPOINT } from "@/constants";
 import * as sdk from "microsoft-cognitiveservices-speech-sdk";
 import axios from "axios";
 import { AlignmentResult } from "echogarden/dist/api/API.d.js";
+import { useAiCommand } from "./use-ai-command";
 
 export const useTranscribe = () => {
   const { EnjoyApp, user, webApi } = useContext(AppSettingsProviderContext);
   const { whisperConfig, openai } = useContext(AISettingsProviderContext);
+  const { punctuateText } = useAiCommand();
 
   const transcode = async (src: string | Blob): Promise<string> => {
     if (src instanceof Blob) {
@@ -61,9 +63,19 @@ export const useTranscribe = () => {
       throw new Error(t("whisperServiceNotSupported"));
     }
 
+    let transcript = originalText || result.text;
+    // if the transcript does not contain any punctuation, use AI command to add punctuation
+    if (!transcript.match(/\w[.,!?](\s|$)/)) {
+      try {
+        transcript = await punctuateText(transcript);
+      } catch (err) {
+        console.warn(err.message);
+      }
+    }
+
     const alignmentResult = await EnjoyApp.echogarden.align(
       new Uint8Array(await blob.arrayBuffer()),
-      originalText || result.text
+      transcript
     );
 
     return {