chatpulse

ChatPulse

A real-time chat sentiment tracking overlay for Twitch streamers. Monitor your chat's mood during polls, debates, or Q&A sessions with a glassmorphism overlay for OBS.

A dedicated bot account reads chat on behalf of all streamers, making this a multi-tenant SaaS that supports horizontal scaling via Redis.

Features

• Real-time sentiment tracking from Twitch chat messages via EventSub webhooks
• Two display modes: combined tug-of-war bar or split positive/negative bars
• Customizable triggers and labels for "for" and "against" votes
• Sliding-window memory (5-120s) controls how long votes count
• Multi-instance scaling with Redis for horizontal deployment
• Bot account architecture — streamers only grant channel:bot scope; a single bot reads all channels
• Token encryption at rest with AES-256-GCM
• Overlay URL rotation to invalidate old URLs
• Per-user debouncing (1s) to prevent spam
• Zero-cost idle — skips processing when no overlay viewers are connected
• Prometheus metrics for HTTP, vote processing, cache, and WebSocket
• Correlation IDs for end-to-end request tracing
• Audit logging for sensitive operations (login, config changes, UUID rotation)
• Per-IP rate limiting on all route groups
• Security headers (HSTS, CSP, X-Frame-Options, etc.)

Prerequisites

1. Twitch Application: Register at https://dev.twitch.tv/console/apps

   • Get your Client ID and Client Secret
   • Set the OAuth Redirect URL to http://localhost:8080/auth/callback (or your production domain)

2. Twitch Bot Account: A dedicated Twitch account that will read chat

   • Authorize the bot with user:read:chat and user:bot scopes (one-time setup)
   • Note the bot's Twitch user ID for BOT_USER_ID

3. PostgreSQL: Version 18 or higher (required for uuidv7())

4. Redis: Version 7+

5. Public HTTPS URL: Required for EventSub webhook delivery (use ngrok for local development)

6. Go: Version 1.26+ (for local development only)

Quick Start with Docker

1. Clone the repository:

git clone https://github.com/pscheid92/chatpulse.git
cd chatpulse

1. Create a .env file from the example:

cp .env.example .env

1. Edit .env with your credentials:

TWITCH_CLIENT_ID=your_client_id
TWITCH_CLIENT_SECRET=your_client_secret
TWITCH_REDIRECT_URI=http://localhost:8080/auth/callback
SESSION_SECRET=$(openssl rand -hex 32)
TOKEN_ENCRYPTION_KEY=$(openssl rand -hex 32)
WEBHOOK_CALLBACK_URL=https://your-subdomain.ngrok-free.app/webhooks/eventsub
WEBHOOK_SECRET=$(openssl rand -hex 16)
BOT_USER_ID=your_bot_twitch_user_id

1. Start the application:

make docker-up

1. Open http://localhost:8080 in your browser.

Local Development

1. Install dependencies:

make deps

1. Start PostgreSQL and Redis (or use Docker):

docker run -d \
  --name chatpulse-postgres \
  -e POSTGRES_USER=twitchuser \
  -e POSTGRES_PASSWORD=twitchpass \
  -e POSTGRES_DB=twitchdb \
  -p 5432:5432 \
  postgres:18-alpine

docker run -d \
  --name chatpulse-redis \
  -p 6379:6379 \
  redis:8-alpine

1. Expose your local server for webhooks:

ngrok http 8080

1. Set up your .env file as described above (use the ngrok URL for WEBHOOK_CALLBACK_URL).

2. Run the server:

make run

Make Targets

make build             # Build binary -> ./server
make run               # Build and run locally
make test              # Run all tests (unit + integration, ~15s)
make test-short        # Run unit tests only (fast, <2s, no Docker)
make test-unit         # Alias for test-short
make test-integration  # Run integration tests only (~12s, requires Docker)
make test-race         # Run tests with race detector
make test-coverage     # Generate coverage report
make fmt               # Format code
make lint              # Run golangci-lint
make deps              # Download and tidy dependencies
make sqlc              # Regenerate sqlc code
make docker-build      # Build Docker image
make docker-up         # Start with Docker Compose (app + PostgreSQL 18 + Redis 8)
make docker-down       # Stop Docker Compose
make clean             # Remove build artifacts

Testing

make test           # Run all tests (unit + integration, ~15s)
make test-short     # Run unit tests only (skip integration, <2s)
make test-race      # Run with race detector
make test-coverage  # Generate coverage report

TDD workflow:

• Use make test-short for rapid feedback during development (<2s, no Docker)
• Run make test before committing (full suite with testcontainers)
• CI runs full suite on every push

Usage

1. Configure Your Overlay

1. Visit http://localhost:8080 and log in with your Twitch account
2. Configure your sentiment triggers:
   • For Trigger: Word/phrase viewers type to vote "for" (e.g., "yes", "agree")
   • Against Trigger: Word/phrase to vote "against" (e.g., "no", "disagree")
   • Labels: Display labels for each side
   • Memory: How long votes count in the sliding window (5-120s, or infinite)
   • Display Mode: Combined (tug-of-war) or Split (two bars)
3. Click "Save Configuration"

2. Add to OBS

1. Copy your unique overlay URL from the dashboard
2. In OBS, add a new Browser Source
3. Paste your overlay URL
4. Set dimensions: 800x100 (adjust to preference)

3. During Your Stream

• Use the Reset to Center button on the dashboard to reset the sentiment bar
• Use Rotate Overlay URL to generate a new URL and invalidate the old one

Environment Variables

See .env.example for all variables with comments.

Required:

• DATABASE_URL — PostgreSQL connection string
• REDIS_URL — Redis connection string (e.g., redis://localhost:6379)
• TWITCH_CLIENT_ID / TWITCH_CLIENT_SECRET — Twitch app credentials
• TWITCH_REDIRECT_URI — OAuth callback URL
• SESSION_SECRET — Secret for session cookies
• TOKEN_ENCRYPTION_KEY — 64 hex chars for AES-256-GCM token encryption (generate: openssl rand -hex 32)
• WEBHOOK_CALLBACK_URL — Public HTTPS URL for EventSub webhook delivery
• WEBHOOK_SECRET — HMAC secret for webhook verification (10-100 chars)
• BOT_USER_ID — Twitch user ID of the bot account

Optional:

• APP_ENV — development (default) or production (controls secure cookies)
• PORT — Server port (default: 8080)
• LOG_LEVEL — debug, info, warn, error (default: info)
• LOG_FORMAT — text (default) or json (recommended for production)
• MAX_WEBSOCKET_CONNECTIONS — File descriptor limit check (default: 10000)
• SESSION_MAX_AGE — Cookie expiry (default: 168h / 7 days)
• SHUTDOWN_TIMEOUT — Graceful shutdown deadline (default: 10s)

How It Works

• Bot Account: A single bot account reads chat in all connected channels via EventSub webhooks
• Webhooks + Conduits: Chat messages arrive via Twitch EventSub webhooks transported through a Conduit, verified with HMAC-SHA256
• Vote Processing: Messages matching trigger words exactly (case-insensitive) are counted as votes
• Debouncing: Each viewer can vote once per second to prevent spam
• Sliding-window Counting: Votes are recorded in Redis Streams; sentiment is computed over a configurable time window (old votes naturally expire)
• Real-time Broadcast: Updates are pushed to overlay clients via Centrifuge WebSocket with Redis broker for cross-instance delivery
• Client-side Lerp: The overlay uses requestAnimationFrame for smooth animation toward server ratios with zero server cost
• Rate Limiting: Per-IP rate limits on auth, dashboard/API, and webhook routes
• Correlation IDs: Every request gets a unique ID propagated through logs for tracing
• Audit Logging: Sensitive operations (login, logout, config save, reset, URL rotation) emit structured audit logs

Architecture

• Backend: Single Go binary (Echo v4) serving HTTP, WebSocket, and webhook endpoints
• Database: PostgreSQL 18+ with auto-migrations (tern) for streamers, configs, and EventSub subscriptions
• Caching: 3-layer read-through cache (in-memory 10s → Redis 1h → PostgreSQL) with pub/sub invalidation
• Scaling: Multi-instance via Redis (Streams for vote counting, pub/sub for broadcasting, Centrifuge Redis broker for WebSocket fan-out)
• Observability: Structured logging (slog) with correlation IDs, audit logs, Prometheus metrics (/metrics endpoint)
• Security: Per-IP rate limiting, security headers (HSTS, CSP), WebSocket origin validation, production SSL enforcement
• Frontend: Minimal HTML/CSS/JS with no external dependencies, embedded via go:embed

Production Deployment

1. Use HTTPS with a reverse proxy (nginx/Caddy) for SSL termination
2. Set TWITCH_REDIRECT_URI to your production domain
3. Generate strong secrets:

   SESSION_SECRET=$(openssl rand -hex 32)
   WEBHOOK_SECRET=$(openssl rand -hex 16)
   TOKEN_ENCRYPTION_KEY=$(openssl rand -hex 32)
   

4. Set APP_ENV=production for secure cookies (also enforces DATABASE_URL SSL — rejects sslmode=disable/allow)
5. Set LOG_FORMAT=json for structured logging
6. Configure PostgreSQL backups
7. Prometheus metrics are available at /metrics for monitoring

Troubleshooting

Overlay not connecting?

• Verify the server is running and the overlay URL is correct
• Check browser console for WebSocket errors

Chat messages not being tracked?

• Ensure webhook delivery is working (check server logs for EventSub notifications)
• Verify the bot account has authorized user:read:chat + user:bot scopes
• Confirm BOT_USER_ID matches the bot's Twitch user ID
• Check that WEBHOOK_CALLBACK_URL is publicly reachable over HTTPS

Webhooks not arriving?

• For local dev, ensure ngrok is running and the URL in .env matches
• Check that WEBHOOK_SECRET is at least 10 characters

License

MIT License - see LICENSE file for details.