From 96b87641ca5655c4a1f5d1f0821ab21d80b0f709 Mon Sep 17 00:00:00 2001
From: Cohee <18619528+Cohee1207@users.noreply.github.com>
Date: Sun, 19 Nov 2023 20:30:34 +0200
Subject: [PATCH] Add OpenAI Whisper API

---
 server.js     |  2 +-
 src/openai.js | 57 ++++++++++++++++++++++++++++++++++++++++++++++++---
 2 files changed, 55 insertions(+), 4 deletions(-)

diff --git a/server.js b/server.js
index 400045f16..8e7393b7f 100644
--- a/server.js
+++ b/server.js
@@ -3530,7 +3530,7 @@ async function fetchJSON(url, args = {}) {
 // ** END **
 
 // OpenAI API
-require('./src/openai').registerEndpoints(app, jsonParser);
+require('./src/openai').registerEndpoints(app, jsonParser, urlencodedParser);
 
 // Tokenizers
 require('./src/tokenizers').registerEndpoints(app, jsonParser);
diff --git a/src/openai.js b/src/openai.js
index 654545713..6c5d551d3 100644
--- a/src/openai.js
+++ b/src/openai.js
@@ -1,12 +1,15 @@
 const { readSecret, SECRET_KEYS } = require("./secrets");
 const fetch = require('node-fetch').default;
+const FormData = require('form-data');
+const fs = require('fs');
 
 /**
  * Registers the OpenAI endpoints.
- * @param {import("express").Express} app
- * @param {any} jsonParser
+ * @param {import("express").Express} app Express app
+ * @param {any} jsonParser JSON parser
+ * @param {any} urlencodedParser Form data parser
  */
-function registerEndpoints(app, jsonParser) {
+function registerEndpoints(app, jsonParser, urlencodedParser) {
     app.post('/api/openai/caption-image', jsonParser, async (request, response) => {
         try {
             let key = '';
@@ -85,6 +88,54 @@ function registerEndpoints(app, jsonParser) {
         }
     });
 
+    app.post('/api/openai/transcribe-audio', urlencodedParser, async (request, response) => {
+        try {
+            const key = readSecret(SECRET_KEYS.OPENAI);
+
+            if (!key) {
+                console.log('No OpenAI key found');
+                return response.sendStatus(401);
+            }
+
+            if (!request.file) {
+                console.log('No audio file found');
+                return response.sendStatus(400);
+            }
+
+            const formData = new FormData();
+            console.log('Processing audio file', request.file.path);
+            formData.append('file', fs.createReadStream(request.file.path), { filename: 'audio.wav', contentType: 'audio/wav' });
+            formData.append('model', request.body.model);
+
+            if (request.body.language) {
+                formData.append('language', request.body.language);
+            }
+
+            const result = await fetch('https://api.openai.com/v1/audio/transcriptions', {
+                method: 'POST',
+                headers: {
+                    'Authorization': `Bearer ${key}`,
+                    ...formData.getHeaders(),
+                },
+                body: formData,
+            });
+
+            if (!result.ok) {
+                const text = await result.text();
+                console.log('OpenAI request failed', result.statusText, text);
+                return response.status(500).send(text);
+            }
+
+            fs.rmSync(request.file.path);
+            const data = await result.json();
+            console.log('OpenAI transcription response', data);
+            return response.json(data);
+        } catch (error) {
+            console.error('OpenAI transcription failed', error);
+            response.status(500).send('Internal server error');
+        }
+    });
+
     app.post('/api/openai/generate-voice', jsonParser, async (request, response) => {
         try {
             const key = readSecret(SECRET_KEYS.OPENAI);