added lib for transcription and audio redaction

lib/audioRedaction.js

@@ -0,0 +1,44 @@
+const ffmpeg = require('fluent-ffmpeg');
+const fs = require('fs').promises;
+
+
+function redactSensitiveInfo(transcriptionData, audioPath, audioOutputPath, delta = 0.15) {
+    const command = ffmpeg(audioPath)
+        .outputFormat('wav'); // Ensure output format is WAV
+
+    // Iterate over transcription data to apply audio filters
+    for (let i = 0; i < transcriptionData.length; i++) {
+        const {word, start} = transcriptionData[i];
+        let end = transcriptionData[i].end;  // Default end time
+
+        // Check if the word needs redaction
+        if (word.startsWith('[') && word.endsWith(']')) {
+            // Find the start of the next non-redacted word
+            for (let j = i + 1; j < transcriptionData.length; j++) {
+                if (!(transcriptionData[j].word.startsWith('[') && transcriptionData[j].word.endsWith(']'))) {
+                    end = transcriptionData[j].start;
+                    break;
+                }
+            }
+
+            // Apply the volume filter to silence from the start of the current word to the start of the next non-redacted word
+            command.audioFilters({
+                filter: 'volume',
+                options: `volume=0:enable='between(t,${start - delta},${end})'`  // Applying silence
+            });
+
+            // Log the redacted segments
+            console.log(`Redacting from ${start}s to ${end}s: "${word}"`);
+        }
+    }
+    // Handlers for command execution
+    command.on('end', () => {
+        console.log(`Redacted audio saved at ${audioOutputPath}`);
+    }).on('error', (err, stdout, stderr) => {
+        console.error('Error processing audio file:', err.message);
+        console.error('ffmpeg stdout:', stdout);
+        console.error('ffmpeg stderr:', stderr);
+    }).saveToFile(audioOutputPath);
+}
+
+module.exports = {redactSensitiveInfo};

lib/transcription.js

@@ -0,0 +1,82 @@
+const {createClient} = require("@deepgram/sdk");
+const fs = require("fs");
+
+class audioProcess {
+    constructor(apiKey, audioFilePath) {
+        this.deepgram = createClient(apiKey);
+        this.audioDir = audioFilePath
+    }
+
+    async transcribeFile(filePath) {
+        const {result, error} = await this.deepgram.listen.prerecorded.transcribeFile(
+            fs.readFileSync(filePath),
+            {
+                model: "nova-2",
+                smart_format: true,
+                detect_entities: true
+            }
+        );
+
+        return result;
+    }
+
+    async redactFile(filePath) {
+        const {result, error} = await this.deepgram.listen.prerecorded.transcribeFile(
+            fs.readFileSync(filePath),
+            {
+                model: "nova-2",
+                smart_format: true,
+                redact: "pii"
+            }
+        );
+
+        return result;
+    }
+
+    async analyzeText(text) {
+        const {result, error} = await this.deepgram.read.analyzeText(
+            {
+                text,
+            },
+            {
+                language: "en",
+                sentiment: true,
+                intents: true,
+                summarize: true
+            }
+        );
+
+        return result;
+    }
+
+    async processAudio(filePath = this.audioDir) {
+        try {
+            const transcription = await this.transcribeFile(filePath);
+            const redaction = await this.redactFile(filePath);
+            const transcript = transcription.results.channels[0].alternatives[0].transcript;
+            const timestamps = transcription.results.channels[0].alternatives[0].words;
+            const redactionTimestamps = redaction.results.channels[0].alternatives[0].words;
+            const redacted = redaction.results.channels[0].alternatives[0].transcript;
+            const entities = transcription.results.channels[0].alternatives[0].entities;
+
+            const analysisResult = await this.analyzeText(transcript);
+            const sentimentSegment = analysisResult.results.sentiments.segments[0];
+            const sentiment = sentimentSegment.sentiment;
+            const sentimentScore = sentimentSegment.sentiment_score;
+
+            return {
+                transcript,
+                timestamps,
+                redactionTimestamps,
+                redacted,
+                sentiment,
+                sentimentScore,
+                entities
+            };
+        } catch (error) {
+            throw error;
+        }
+    }
+}
+
+module.exports = {audioProcess};