Add speech to text sample

samples/whisper-speech-to-text/.gitignore

@@ -0,0 +1,4 @@
+node_modules/
+*.wasm
+build/
+dist/

samples/whisper-speech-to-text/captions/.gitignore

@@ -0,0 +1,7 @@
+node_modules
+dist
+target
+.spin/
+build/
+.spin
+.spin-aka

samples/whisper-speech-to-text/captions/.npmrc

@@ -0,0 +1 @@
+KNITWIT_SOURCE=./config/knitwit.json

samples/whisper-speech-to-text/captions/config/knitwit.json

@@ -0,0 +1,14 @@
+{
+  "version": 1,
+  "project": {
+    "worlds": [
+      "spin-http"
+    ]
+  },
+  "packages": {
+    "@fermyon/spin-sdk": {
+      "witPath": "../../bin/wit",
+      "world": "spin-imports"
+    }
+  }
+}

samples/whisper-speech-to-text/captions/package.json

@@ -0,0 +1,25 @@
+{
+    "name": "captions",
+    "version": "1.0.0",
+    "description": "",
+    "main": "index.js",
+    "scripts": {
+        "build": "knitwit --out-dir build/wit/knitwit --out-world combined && npx webpack --mode=production && npx mkdirp dist && npx j2w -i build/bundle.js -d build/wit/knitwit -n combined -o dist/captions.wasm",
+        "test": "echo \"Error: no test specified\" && exit 1"
+    },
+    "keywords": [],
+    "author": "",
+    "license": "ISC",
+    "devDependencies": {
+        "mkdirp": "^3.0.1",
+        "ts-loader": "^9.4.1",
+        "typescript": "^4.8.4",
+        "webpack": "^5.74.0",
+        "webpack-cli": "^4.10.0",
+        "@fermyon/knitwit": "0.3.0"
+    },
+    "dependencies": {
+        "@fermyon/spin-sdk": "3.2.0",
+        "itty-router": "^5.0.18"
+    }
+}

samples/whisper-speech-to-text/captions/src/index.ts

@@ -0,0 +1,59 @@
+import { Variables } from '@fermyon/spin-sdk';
+import { AutoRouter as Router } from 'itty-router';
+
+let router = Router();
+
+
+// Helper function to create form data with audio blob in the 
+// format required by the transcription API.
+//
+// This is compatible with the OpenAI Whisper API.
+function createFormData(audioBlob: ArrayBuffer): FormData {
+    const formData = new FormData();
+    formData.append('model', 'whisper-1');
+    formData.append('file', new Blob([audioBlob], { type: 'audio/wav' }), 'audio.wav');
+    formData.append('stream', 'false');
+    formData.append('response_format', 'json');
+
+    return formData;
+};
+
+// Main transcription endpoint
+router.post('/api/transcribe', async (request: Request) => {
+    try {
+        console.log('Received request:', request);
+        // Get raw audio data from request body
+        const audioData = await request.arrayBuffer();
+        console.log(`Received audio data of size ${audioData.byteLength}`);
+
+        if (!audioData || audioData.byteLength === 0) {
+            return new Response(JSON.stringify({ error: 'No audio data provided' }), { status: 400 });
+        }
+
+        // Create form data for the transcription API
+        const formData = createFormData(audioData);
+
+        // Send request to transcription API
+        const transcriptionResponse = await fetch(Variables.get('transcription_api')!, {
+            method: 'POST',
+            body: formData
+        });
+
+        if (!transcriptionResponse.ok) {
+            throw new Error(`Transcription API error: ${transcriptionResponse.status}`);
+        }
+
+        const transcription = await transcriptionResponse.json();
+
+        return new Response(JSON.stringify({
+            text: transcription.text,
+            segments: transcription.segments
+        }));
+    } catch (error) {
+        console.error('Transcription error:', error);
+        return new Response(JSON.stringify({ error: 'Internal server error' }), { status: 500 });
+    }
+});
+
+//@ts-ignore
+addEventListener('fetch', async (event: FetchEvent) => { event.respondWith(router.fetch(event.request, { event })) });

samples/whisper-speech-to-text/captions/tsconfig.json

@@ -0,0 +1,18 @@
+{
+    "compilerOptions": {
+        "outDir": "./dist/",
+        "noImplicitAny": true,
+        "module": "es6",
+        "target": "es2020",
+        "jsx": "react",
+        "skipLibCheck": true,
+        "lib": [
+            "ES2020",
+            "WebWorker"
+        ],
+        "allowJs": true,
+        "strict": true,
+        "noImplicitReturns": true,
+        "moduleResolution": "node"
+    }
+}

samples/whisper-speech-to-text/captions/webpack.config.js

@@ -0,0 +1,38 @@
+const path = require('path');
+const SpinSdkPlugin = require("@fermyon/spin-sdk/plugins/webpack")
+
+module.exports = {
+    entry: './src/index.ts',
+    experiments: {
+        outputModule: true,
+    },
+    module: {
+        rules: [
+            {
+                test: /\.tsx?$/,
+                use: 'ts-loader',
+                exclude: /node_modules/,
+            },
+        ],
+    },
+    resolve: {
+        extensions: ['.tsx', '.ts', '.js'],
+    },
+    output: {
+        path: path.resolve(__dirname, './build'),
+        filename: 'bundle.js',
+        module: true,
+        library: {
+            type: "module",
+        }
+    },
+    plugins: [
+        new SpinSdkPlugin()
+    ],
+    optimization: {
+        minimize: false
+    },
+    performance: {
+        hints: false,
+    }
+};