deepgram-local-dev-loop

Deepgram Local Dev Loop

Overview

Set up an efficient local development workflow for Deepgram integration with fast feedback cycles.

Prerequisites

• Completed deepgram-install-auth setup
• Node.js 18+ with npm/pnpm or Python 3.10+
• Sample audio files for testing
• Environment variables configured

Instructions

Step 1: Create Project Structure

mkdir -p src tests fixtures
touch src/transcribe.ts tests/transcribe.test.ts

Step 2: Set Up Environment Files

# .env.development
DEEPGRAM_API_KEY=your-dev-api-key
DEEPGRAM_MODEL=nova-2

# .env.test
DEEPGRAM_API_KEY=your-test-api-key
DEEPGRAM_MODEL=nova-2

Step 3: Create Test Fixtures

set -euo pipefail
# Download sample audio for testing
curl -o fixtures/sample.wav https://static.deepgram.com/examples/nasa-podcast.wav

Step 4: Set Up Watch Mode

{
  "scripts": {
    "dev": "tsx watch src/transcribe.ts",
    "test": "vitest",
    "test:watch": "vitest --watch"
  }
}

Output

• Project structure with src, tests, fixtures directories
• Environment files for development and testing
• Watch mode scripts for rapid iteration
• Sample audio fixtures for testing

Error Handling

Error	Cause	Solution
Fixture Not Found	Missing audio file	Run fixture download script
Env Not Loaded	dotenv not configured	Install and configure dotenv
Watch Mode Fails	Missing tsx	Install tsx: npm i -D tsx
API Rate Limited	Too many dev requests	Use cached responses in tests

Examples

TypeScript Dev Setup

// src/transcribe.ts
import { createClient } from '@deepgram/sdk';
import { config } from 'dotenv';

config(); // Load .env

const deepgram = createClient(process.env.DEEPGRAM_API_KEY!);

export async function transcribeAudio(audioPath: string) {
  const audio = await Bun.file(audioPath).arrayBuffer();

  const { result, error } = await deepgram.listen.prerecorded.transcribeFile(
    Buffer.from(audio),
    { model: process.env.DEEPGRAM_MODEL || 'nova-2', smart_format: true }
  );

  if (error) throw error;
  return result.results.channels[0].alternatives[0].transcript;
}

// Dev mode: run with sample
if (import.meta.main) {
  transcribeAudio('./fixtures/sample.wav').then(console.log);
}

Test Setup with Vitest

// tests/transcribe.test.ts
import { describe, it, expect, beforeAll } from 'vitest';
import { transcribeAudio } from '../src/transcribe';

describe('Deepgram Transcription', () => {
  it('should transcribe audio file', async () => {
    const transcript = await transcribeAudio('./fixtures/sample.wav');
    expect(transcript).toBeDefined();
    expect(transcript.length).toBeGreaterThan(0);
  });

  it('should handle empty audio gracefully', async () => {
    await expect(transcribeAudio('./fixtures/empty.wav'))
      .rejects.toThrow();
  });
});

Mock Responses for Testing

// tests/mocks/deepgram.ts
export const mockTranscriptResponse = {
  results: {
    channels: [{
      alternatives: [{
        transcript: 'This is a test transcript.',
        confidence: 0.99,
        words: [
          { word: 'This', start: 0.0, end: 0.2, confidence: 0.99 },
          { word: 'is', start: 0.2, end: 0.3, confidence: 0.99 },
        ]
      }]
    }]
  }
};

Resources

• Deepgram SDK Reference
• Vitest Documentation
• dotenv Configuration

Next Steps

Proceed to deepgram-sdk-patterns for production-ready code patterns.