example with transcript generation

This commit is contained in:
surajshivakumar
2024-06-25 04:33:54 -04:00
parent 7985b2a937
commit dfa8d895c4

View File

@@ -0,0 +1,195 @@
const { request } = require('undici');
const { pipeline } = require('stream');
const fs = require('fs');
const path = require('path');
const os = require('os');
const fsP = require('fs').promises;
const createAudioService = require('..');
const audioService = createAudioService();
// Constants
const JAMBONZ_API_BASE_URL = 'https://jambonz.one/api/v1';
// Utility functions
async function readJSONFile(filePath) {
try {
const data = await fsP.readFile(filePath, 'utf8');
return JSON.parse(data);
} catch (err) {
throw new Error(`Error reading JSON file: ${err.message}`);
}
}
async function writeJSONFile(filePath, data) {
try {
await fs.writeFileSync(filePath, JSON.stringify(data, null, 2));
} catch (err) {
throw new Error(`Error writing JSON file: ${err.message}`);
}
}
function extractJambonzTrace(data, audioService) {
const transcriptJson = {
parties: { N: 2 },
conversation: { 'as heard': {} }
};
data.resourceSpans.forEach((resourceSpan) => {
resourceSpan.instrumentationLibrarySpans.forEach((librarySpan) => {
librarySpan.spans.forEach((span) => {
span.attributes.forEach((attribute) => {
switch (attribute.key) {
case 'from':
if (attribute.value) transcriptJson.parties.from = attribute.value.stringValue;
break;
case 'duration':
if (attribute.value) transcriptJson.duration = attribute.value.stringValue;
break;
case 'to':
if (attribute.value) transcriptJson.parties.to = attribute.value.stringValue;
break;
case 'http.body':
if (attribute.value) {
const httpBody = JSON.parse(attribute.value.stringValue);
if (httpBody.speech && httpBody.speech.alternatives && httpBody.speech.alternatives.length > 0) {
const alternative = httpBody.speech.alternatives[0];
transcriptJson.conversation['as heard']['full transcript'] = alternative.transcript;
transcriptJson.conversation['as heard']['confidence'] = alternative.confidence;
transcriptJson.conversation['as heard']['transcription vendor'] = httpBody.speech.vendor.name;
// eslint-disable-next-line max-len
transcriptJson.conversation['as heard'].timestamps = httpBody.speech.vendor.evt.channel.alternatives[0].words;
}
}
break;
}
});
});
});
});
return transcriptJson;
}
async function fetchJambonzTrace() {
const url = `${JAMBONZ_API_BASE_URL}/Accounts/${process.env.ACCOUNT_SID}/RecentCalls/trace/${process.env.TRACE_ID}`;
const headers = {
'Authorization': `Bearer ${process.env.JAMBONZ_API_TOKEN}`,
'Accept': '*/*',
};
try {
const { statusCode, body } = await request(url, { headers });
if (statusCode !== 200) {
throw new Error(`Request failed with status code: ${statusCode}`);
}
const data = await body.json();
await writeJSONFile('response.json', data);
console.log('Data has been received');
const jsonData = await readJSONFile('./response.json');
return extractJambonzTrace(jsonData, audioService);
} catch (error) {
console.error('Error fetching data:', error);
throw error;
}
}
async function downloadFile(url, tempFilePath, finalFilePath, token) {
try {
const { statusCode, body } = await request(url, {
method: 'GET',
headers: {
'Authorization': `Bearer ${token}`,
'Accept': '*/*'
}
});
if (statusCode !== 200) {
throw new Error(`Failed to download file: ${statusCode}`);
}
const tempWriteStream = fs.createWriteStream(tempFilePath);
pipeline(
body,
tempWriteStream,
async(err) => {
if (err) {
console.error('Pipeline failed:', err);
} else {
try {
await processAndRedactFile(tempFilePath, finalFilePath);
} catch (processingError) {
console.error('Error processing file:', processingError);
} finally {
// Clean up the temporary file
fs.unlink(tempFilePath, (err) => {
if (err) {
console.error('Error deleting temp file:', err);
} else {
console.log('Temporary file deleted.');
}
});
}
}
}
);
} catch (error) {
console.error('Error fetching file:', error);
}
}
async function processAndRedactFile(audioFilePath, outputFilePath) {
const credentials = { 'vendor': 'deepgram', 'apiKey': process.env.DEEPGRAM_API_KEY };
try {
const transcriptionResults = await audioService.transcribe(credentials, audioFilePath);
const timestampsToRedact = transcriptionResults.redactionTimestamps;
await audioService.redact({
credentials: credentials,
transcriptionData: timestampsToRedact,
audioPath: audioFilePath,
audioOutputPath: outputFilePath
});
const jambonzTrace = await fetchJambonzTrace();
jambonzTrace.conversation['after the fact'] = {
'full transcript': transcriptionResults.transcript,
'confidence': transcriptionResults.confidence,
'transcription vendor': transcriptionResults.vendor,
'timestamps': timestampsToRedact
};
await writeJSONFile(`${process.env.OUTPUT_PATH}/transcription.json`, jambonzTrace);
console.log('File redacted and saved successfully.');
} catch (error) {
console.error('Failed to redact audio:', error);
throw error;
}
}
// Main execution
async function main() {
try {
const [,, callId, date, format] = process.argv;
if (!callId || !date || !format) {
throw new Error('Missing required command line arguments. Usage: node app.js <callId> <date> <format>');
}
const url = `${JAMBONZ_API_BASE_URL}/Accounts/${process.env.ACCOUNT_SID}/RecentCalls/${callId}/record/${date}/${format}`;
const finalFilePath = path.resolve(__dirname, path.join(process.env.OUTPUT_PATH, 'redacted_audio.wav'));
const tempFolder = process.env.TEMP_FOLDER || os.tmpdir();
const tempFilePath = path.join(tempFolder, 'tempDownloadedFile.mp3');
await downloadFile(url, tempFilePath, finalFilePath, process.env.JAMBONZ_API_TOKEN);
} catch (error) {
console.error('Error in main process:', error);
process.exit(1);
}
}
main();