Deepgram Webhooks Events

Deepgram Webhooks Events

Overview

Implement callback URL handling for asynchronous Deepgram transcription workflows.

Prerequisites

• Publicly accessible HTTPS endpoint
• Deepgram API key with transcription permissions
• Request validation capabilities
• Secure storage for transcription results

Deepgram Callback Flow

1. Client sends transcription request with callback URL
2. Deepgram processes audio asynchronously
3. Deepgram POSTs results to callback URL
4. Your server processes and stores results

Instructions

Step 1: Create Callback Endpoint

Set up an HTTPS endpoint to receive results.

Step 2: Implement Request Validation

Verify callbacks are from Deepgram.

Step 3: Process Results

Handle the transcription response.

Step 4: Store and Notify

Save results and notify clients.

Examples

TypeScript Callback Server (Express)

// server/callback.ts
import express from 'express';
import crypto from 'crypto';
import { logger } from './logger';
import { storeTranscription, notifyClient } from './services';

const app = express();

// Raw body for signature verification
app.use('/webhooks/deepgram', express.raw({ type: 'application/json' }));
app.use(express.json());

interface DeepgramCallback {
  request_id: string;
  metadata: {
    request_id: string;
    transaction_key: string;
    sha256: string;
    created: string;
    duration: number;
    channels: number;
    models: string[];
  };
  results: {
    channels: Array<{
      alternatives: Array<{
        transcript: string;
        confidence: number;
        words: Array<{
          word: string;
          start: number;
          end: number;
          confidence: number;
        }>;
      }>;
    }>;
  };
}

// Verify callback is from Deepgram
function verifyDeepgramSignature(
  payload: Buffer,
  signature: string | undefined,
  secret: string
): boolean {
  if (!signature) return false;

  const expectedSignature = crypto
    .createHmac('sha256', secret)
    .update(payload)
    .digest('hex');

  return crypto.timingSafeEqual(
    Buffer.from(signature),
    Buffer.from(expectedSignature)
  );
}

app.post('/webhooks/deepgram', async (req, res) => {
  const requestId = req.headers['x-request-id'] as string;

  logger.info('Received Deepgram callback', { requestId });

  try {
    // Verify signature if using webhook secret
    const signature = req.headers['x-deepgram-signature'] as string;
    const webhookSecret = process.env.DEEPGRAM_WEBHOOK_SECRET;

    if (webhookSecret && !verifyDeepgramSignature(req.body, signature, webhookSecret)) {
      logger.warn('Invalid signature', { requestId });
      return res.status(401).json({ error: 'Invalid signature' });
    }

    const callback: DeepgramCallback = JSON.parse(req.body.toString());

    // Extract transcript
    const transcript = callback.results.channels[0]?.alternatives[0]?.transcript;
    const confidence = callback.results.channels[0]?.alternatives[0]?.confidence;

    logger.info('Processing transcription', {
      requestId: callback.request_id,
      duration: callback.metadata.duration,
      confidence,
    });

    // Store result
    await storeTranscription({
      requestId: callback.request_id,
      transcript,
      confidence,
      metadata: callback.metadata,
      words: callback.results.channels[0]?.alternatives[0]?.words,
    });

    // Notify client (WebSocket, email, etc.)
    await notifyClient(callback.request_id, {
      status: 'completed',
      transcript,
    });

    res.status(200).json({ received: true });
  } catch (error) {
    logger.error('Callback processing failed', {
      requestId,
      error: error instanceof Error ? error.message : 'Unknown error',
    });

    res.status(500).json({ error: 'Processing failed' });
  }
});

// Health check
app.get('/health', (_, res) => {
  res.json({ status: 'ok' });
});

export default app;

Async Transcription Request

// services/async-transcription.ts
import { createClient } from '@deepgram/sdk';
import { v4 as uuidv4 } from 'uuid';
import { redis } from './redis';

interface AsyncTranscriptionOptions {
  language?: string;
  model?: string;
  diarize?: boolean;
  punctuate?: boolean;
}

export class AsyncTranscriptionService {
  private client;
  private callbackBaseUrl: string;

  constructor(apiKey: string, callbackBaseUrl: string) {
    this.client = createClient(apiKey);
    this.callbackBaseUrl = callbackBaseUrl;
  }

  async submitTranscription(
    audioUrl: string,
    options: AsyncTranscriptionOptions = {}
  ): Promise<{ jobId: string; requestId: string }> {
    const jobId = uuidv4();
    const callbackUrl = `${this.callbackBaseUrl}/webhooks/deepgram?job=${jobId}`;

    const { result, error } = await this.client.listen.prerecorded.transcribeUrl(
      { url: audioUrl },
      {
        model: options.model || 'nova-2',
        language: options.language || 'en',
        diarize: options.diarize ?? false,
        punctuate: options.punctuate ?? true,
        smart_format: true,
        callback: callbackUrl,
      }
    );

    if (error) {
      throw new Error(`Transcription submission failed: ${error.message}`);
    }

    // Store job tracking info
    await redis.hset(`transcription:${jobId}`, {
      status: 'processing',
      requestId: result.request_id,
      submittedAt: new Date().toISOString(),
      audioUrl,
    });

    // Set expiration (24 hours)
    await redis.expire(`transcription:${jobId}`, 86400);

    return {
      jobId,
      requestId: result.request_id,
    };
  }

  async getStatus(jobId: string): Promise<{
    status: string;
    result?: unknown;
  }> {
    const data = await redis.hgetall(`transcription:${jobId}`);

    if (!data || Object.keys(data).length === 0) {
      throw new Error('Job not found');
    }

    return {
      status: data.status,
      result: data.result ? JSON.parse(data.result) : undefined,
    };
  }
}

Store and Notify Services

// services/store.ts
import { redis } from './redis';
import { db } from './database';

interface TranscriptionResult {
  requestId: string;
  transcript: string;
  confidence: number;
  metadata: Record<string, unknown>;
  words?: Array<{
    word: string;
    start: number;
    end: number;
    confidence: number;
  }>;
}

export async function storeTranscription(result: TranscriptionResult): Promise<void> {
  // Store in database
  await db.transcriptions.insert({
    request_id: result.requestId,
    transcript: result.transcript,
    confidence: result.confidence,
    metadata: result.metadata,
    words: result.words,
    created_at: new Date(),
  });

  // Update Redis for quick access
  const jobId = await redis.get(`request:${result.requestId}:job`);
  if (jobId) {
    await redis.hset(`transcription:${jobId}`, {
      status: 'completed',
      result: JSON.stringify(result),
      completedAt: new Date().toISOString(),
    });
  }
}

// services/notify.ts
import { WebSocketServer } from './websocket';
import { emailService } from './email';

export async function notifyClient(
  requestId: string,
  data: { status: string; transcript?: string }
): Promise<void> {
  // Get client info for this request
  const clientId = await redis.get(`request:${requestId}:client`);

  if (clientId) {
    // WebSocket notification
    WebSocketServer.sendToClient(clientId, {
      type: 'transcription_complete',
      requestId,
      ...data,
    });
  }

  // Email notification (optional)
  const email = await redis.get(`request:${requestId}:email`);
  if (email) {
    await emailService.send({
      to: email,
      subject: 'Your transcription is ready',
      body: `Transcription for request ${requestId} is complete.`,
    });
  }
}

Retry Mechanism for Callbacks

// services/callback-retry.ts
import { logger } from './logger';

interface RetryConfig {
  maxRetries: number;
  baseDelay: number;
  maxDelay: number;
}

export class CallbackRetryHandler {
  private config: RetryConfig;
  private pendingRetries: Map<string, NodeJS.Timeout> = new Map();

  constructor(config: Partial<RetryConfig> = {}) {
    this.config = {
      maxRetries: config.maxRetries ?? 3,
      baseDelay: config.baseDelay ?? 5000,
      maxDelay: config.maxDelay ?? 60000,
    };
  }

  async processWithRetry(
    requestId: string,
    processor: () => Promise<void>
  ): Promise<void> {
    let attempt = 0;

    while (attempt < this.config.maxRetries) {
      try {
        await processor();
        return;
      } catch (error) {
        attempt++;
        logger.warn('Callback processing failed, will retry', {
          requestId,
          attempt,
          error: error instanceof Error ? error.message : 'Unknown',
        });

        if (attempt >= this.config.maxRetries) {
          throw error;
        }

        const delay = Math.min(
          this.config.baseDelay * Math.pow(2, attempt - 1),
          this.config.maxDelay
        );

        await new Promise(resolve => setTimeout(resolve, delay));
      }
    }
  }

  scheduleRetry(requestId: string, callback: () => Promise<void>, attempt: number): void {
    const delay = Math.min(
      this.config.baseDelay * Math.pow(2, attempt),
      this.config.maxDelay
    );

    const timeout = setTimeout(async () => {
      try {
        await callback();
        this.pendingRetries.delete(requestId);
      } catch (error) {
        if (attempt < this.config.maxRetries) {
          this.scheduleRetry(requestId, callback, attempt + 1);
        } else {
          logger.error('Callback retry exhausted', { requestId });
        }
      }
    }, delay);

    this.pendingRetries.set(requestId, timeout);
  }

  cancel(requestId: string): void {
    const timeout = this.pendingRetries.get(requestId);
    if (timeout) {
      clearTimeout(timeout);
      this.pendingRetries.delete(requestId);
    }
  }
}

Testing Callbacks Locally

# Use ngrok to expose local server
ngrok http 3000

# Test callback endpoint
curl -X POST https://your-ngrok-url.ngrok.io/webhooks/deepgram \
  -H "Content-Type: application/json" \
  -d '{
    "request_id": "test-123",
    "metadata": {
      "request_id": "test-123",
      "duration": 10.5
    },
    "results": {
      "channels": [{
        "alternatives": [{
          "transcript": "This is a test transcript.",
          "confidence": 0.95
        }]
      }]
    }
  }'

Client SDK for Async Transcription

// client/async-client.ts
export class AsyncTranscriptionClient {
  private baseUrl: string;
  private pollInterval: number;

  constructor(baseUrl: string, pollInterval = 2000) {
    this.baseUrl = baseUrl;
    this.pollInterval = pollInterval;
  }

  async submit(audioUrl: string): Promise<string> {
    const response = await fetch(`${this.baseUrl}/transcribe/async`, {
      method: 'POST',
      headers: { 'Content-Type': 'application/json' },
      body: JSON.stringify({ audioUrl }),
    });

    const { jobId } = await response.json();
    return jobId;
  }

  async waitForResult(jobId: string, timeout = 300000): Promise<{
    transcript: string;
    confidence: number;
  }> {
    const startTime = Date.now();

    while (Date.now() - startTime < timeout) {
      const response = await fetch(`${this.baseUrl}/transcribe/status/${jobId}`);
      const data = await response.json();

      if (data.status === 'completed') {
        return data.result;
      }

      if (data.status === 'failed') {
        throw new Error('Transcription failed');
      }

      await new Promise(r => setTimeout(r, this.pollInterval));
    }

    throw new Error('Transcription timeout');
  }

  async transcribe(audioUrl: string): Promise<{
    transcript: string;
    confidence: number;
  }> {
    const jobId = await this.submit(audioUrl);
    return this.waitForResult(jobId);
  }
}

Resources

• Deepgram Callback Documentation
• Webhook Best Practices
• ngrok Documentation

Next Steps

Proceed to deepgram-performance-tuning for optimization.