Fully automated audio transcription workflow using n8n, AssemblyAI, and OpenAI GPT-5. Upload audio to Google Drive and get AI-powered transcriptions with speaker diarization in Notion + Google Drive.

• 🎙️ Speaker Diarization - Identifies who said what with timestamps
• 🤖 AI Processing - GPT-5 powered summaries and key insights
• 📊 Dual Output - Saves to both Notion database and Google Drive
• ⚡ Auto-Triggered - Processes new files automatically (checks every minute)
• 🔄 Smart Chunking - Splits long content into Notion-friendly 2000-char blocks
• 🛡️ Error Handling - Continues processing even if OpenAI fails
• ☁️ Cloud Deployment - Runs 24/7 on AWS (no local computer needed)

# 1. Clone repository
git clone <your-repo>
cd iWatcherPartner

# 2. Copy environment template
cp .env.example .env

# 3. Add your API keys to .env
# - ASSEMBLYAI_API_KEY
# - OPENAI_API_KEY
# - NOTION_API_TOKEN
# - NOTION_DATABASE_ID

# 4. Start local n8n (requires Docker)
docker-compose up -d

# 5. Access n8n UI
open http://localhost:5678

# 6. Import workflow
python3 import-workflow.py

# 7. Configure Google Drive OAuth in n8n UI
# Settings → Credentials → Add Google Drive OAuth2 API

📘 Complete Step-by-Step Guide: See docs/COMPLETE-DEPLOYMENT-GUIDE.md

This guide covers everything:

• AWS infrastructure setup
• n8n container configuration
• Workflow import
• Google OAuth setup (with SSM tunnel trick)
• HTTPS setup with custom domain
• Complete testing and verification

Quick deploy (if you know what you're doing):

cd infrastructure
npm install
NOTIFICATION_PHONE_NUMBER="+1234567890" npx cdk deploy SimpleIWatcherStack

For HTTPS setup: docs/HTTPS-DOMAIN-SETUP.md

Google Drive "New" Folder
         ↓
    [n8n Trigger]
         ↓
  Download Audio
         ↓
   AssemblyAI API
  (Speaker Diarization)
         ↓
    OpenAI GPT-5
  (AI Summary) ← [Error Handling]
         ↓
   ┌──────┴──────┐
   ↓             ↓
Notion DB    Google Drive
(Chunked)   (Full .txt)
   ↓
Move to "Completed"

iWatcherPartner/
├── iwatcher-gdrive-trigger.json    # Main n8n workflow
├── import-workflow.py              # Workflow deployment script
├── google_credentials.json         # Google OAuth (keep secure!)
├── .env.example                    # Environment template
├── infrastructure/                 # AWS CDK deployment
│   ├── lib/                        # CDK stack definitions
│   ├── n8n-workflows/              # Workflow backups
│   └── cdk-app.ts                  # Main CDK app
└── docs/                           # Documentation
    ├── AWS-DEPLOYMENT-PLAN.md      # AWS deployment guide
    └── WORKFLOW-SUMMARY.md         # Workflow documentation

• Sign up: https://www.assemblyai.com/
• Get API key from dashboard
• Cost: ~$0.37/hour of audio

• Sign up: https://platform.openai.com/
• Create API key
• Uses: GPT-5 model (gpt-5-2025-08-07)
• Cost: ~$0.05-0.10 per transcript

• Create integration: https://www.notion.so/my-integrations
• Get API token (starts with ntn_)
• Share database with integration
• Cost: Free

• Create project: https://console.cloud.google.com
• Enable Google Drive API
• Create OAuth 2.0 credentials (Desktop app)
• Download google_credentials.json
• Cost: Free

1. Google Drive Trigger

• Monitors "New" folder every minute
• Detects new audio files automatically

2. Speaker Diarization

• Identifies multiple speakers
• Adds timestamps to each utterance
• Format: [MM:SS] Speaker A: text

3. OpenAI Processing

• Model: gpt-5-2025-08-07
• No token limits (uses default 128K)
• Error handling: Falls back if API fails

4. Dual Delivery

• Notion: Chunked into 2000-char blocks
• Google Drive: Full transcript as .txt

5. File Management

• Success: Moves to "Completed" folder
• Failure: Moves to "Failed" folder

Create .env file:

# AssemblyAI
ASSEMBLYAI_API_KEY=your_key_here

# OpenAI
OPENAI_API_KEY=sk-proj-...

# Notion
NOTION_API_TOKEN=ntn_...
NOTION_DATABASE_ID=your_database_id

# Google Drive (folder IDs)
GOOGLE_DRIVE_NEW_FOLDER_ID=...
GOOGLE_DRIVE_COMPLETED_FOLDER_ID=...
GOOGLE_DRIVE_FAILED_FOLDER_ID=...

• Drop audio file into Google Drive "New" folder
• Supported formats: M4A, MP3, WAV, AAC

• Workflow triggers within 1 minute
• Transcription takes ~3-10 minutes
• Check Notion database for results

• Notion: AI summary + chunked transcript
• Google Drive "Completed": Full transcript with metadata
• Original audio: Moved to "Completed" folder

• Cost: $0
• Setup time: 15 minutes
• Use case: Testing, development

• Cost: ~$17/month
• Setup time: 30 minutes
• Use case: Single-user testing

• Cost: ~$88/month
• Setup time: 2-3 hours
• Use case: Production, high availability
• See: docs/AWS-DEPLOYMENT-PLAN.md

# Check n8n logs
docker logs iwatcher-n8n --tail 50

# Verify workflow is active
curl -H "X-N8N-API-KEY: your_key" \
  http://localhost:5678/api/v1/workflows

• Delete credential and recreate
• Ensure redirect URI is http://localhost:5678/rest/oauth2-credential/callback
• Check Google Cloud Console for authorized redirect URIs

• Workflow continues even if OpenAI fails
• Check fallback summary in output
• Verify API key and quotas

Monthly costs for 40 hours of audio:

• AssemblyAI: ~$15 (40 hours × $0.37)
• OpenAI: ~$2 (40 transcripts × $0.05)
• AWS (Production): ~$88/month
• Total: ~$105/month

Free tier options:

• Notion: Free
• Google Drive: Free (15GB)

• Credentials: Stored in .env and google_credentials.json (gitignored)
• AWS: Uses Secrets Manager for API keys
• n8n: Encrypted credential storage
• HTTPS: Recommended for production (see domain setup)

• Complete Deployment Guide ⭐ START HERE - Full step-by-step deployment
• HTTPS Domain Setup - Configure SSL/TLS with Let's Encrypt
• AWS Deployment Plan - AWS architecture options (EC2, ECS, RDS)
• Deployment Learnings - Lessons learned and rabbit holes

• Workflow Summary - Detailed workflow documentation
• README.md (this file) - Project overview

• setup.py - Import workflow and verify environment
• test-complete-pipeline.py - End-to-end testing and monitoring
• import-workflow.py - Basic workflow import utility

1. Fork the repository
2. Create feature branch
3. Test changes locally
4. Submit pull request

MIT License - see LICENSE file for details

For issues and questions:

1. Check Troubleshooting section
2. Review docs/ folder
3. Open GitHub issue

Ready to start? Follow the Quick Start guide above!