Operational intelligence platform that converts Sentry production errors into a knowledge graph, enables reproduction in isolated sandboxes, and provides evidence-based debugging.

This implementation uses REAL integrations - no stubs:

• ✅ Real Sentry REST API calls
• ✅ Real Daytona CLI execution (with local fallback)
• ✅ Real Git cloning and command execution
• ✅ Real SQLite database storage

cd backend
pip install -r requirements.txt

Create backend/.env:

# Database
DATABASE_URL=sqlite:///./opsgraph.db

# Sentry (REQUIRED for /events/sentry/pull)
# Get token from: https://sentry.io/settings/account/api/auth-tokens/
# Required scopes: event:read, project:read
SENTRY_AUTH_TOKEN=your_sentry_auth_token_here
SENTRY_ORG_SLUG=your_org_slug
SENTRY_PROJECT_SLUG=your_project_slug
SENTRY_BASE_URL=https://sentry.io/api/0

# Daytona (optional - falls back to local runner if not set)
DAYTONA_ENABLED=false
DAYTONA_CLI=daytona
DAYTONA_STRATEGY=A

# CORS
CORS_ORIGINS=http://localhost:3000

Run backend:

cd backend
uvicorn app.main:app --reload

✅ Backend runs on http://localhost:8000

cd frontend
npm install

Create frontend/.env.local:

NEXT_PUBLIC_API_BASE=http://localhost:8000

Run frontend:

npm run dev

✅ Frontend runs on http://localhost:3000

1. Check backend health:

   curl http://localhost:8000/health/details

   Should return integration status.

2. Open frontend: http://localhost:3000

3. Test Sentry sync:

   • Click "Sync Sentry" (requires valid tokens)
   • OR click "Seed Demo Incident" (works without Sentry)

4. View graph: Errors appear as nodes

5. Reproduce error:

   • Click error node → Click "Reproduce"
   • Enter repo URL and command
   • Watch logs stream in real-time

1. Go to https://sentry.io/settings/account/api/auth-tokens/
2. Click "Create New Token"
3. Scopes needed: event:read, project:read
4. Copy token → paste into SENTRY_AUTH_TOKEN in backend/.env

• GET /health - Basic health check
• GET /health/details - Integration status (Sentry, Daytona, GitHub)
• GET /events/sentry/pull?minutes=60&limit=50 - Pulls REAL Sentry issues
• POST /events/sentry - Ingest single Sentry event
• GET /graph - Get knowledge graph (nodes + edges from SQLite)
• GET /graph/summary - Top errors and services
• POST /simulate/reproduce - Runs REAL command in sandbox
• GET /simulate/{run_id} - Get run status and logs
• GET /simulate/{run_id}/logs?tail=4000 - Stream logs
• GET /audit/recent?limit=50 - Audit trail
• POST /daytona/verify - Test Daytona CLI

Backend:

• FastAPI + SQLModel + SQLite
• Real Sentry REST API client (httpx)
• Real Daytona CLI execution (subprocess)
• Real local Git execution (fallback)

Frontend:

• Next.js (App Router) + TypeScript
• React Flow for graph visualization
• Tailwind CSS (monochrome dark theme)
• Real-time log streaming

Data Flow:

1. Sentry API → Normalize → SQLite
2. SQLite → Graph API → Frontend
3. Frontend → Reproduce → Runner → Logs → Frontend

• Python 3.10+
• Node.js 18+
• Git (for local runner)
• Daytona CLI (optional, install from https://www.daytona.io)

Backend won't start:

• Check Python version: python --version
• Install dependencies: pip install -r requirements.txt
• Check port 8000 is free

Frontend won't start:

• Check Node version: node --version
• Install dependencies: npm install
• Check port 3000 is free

Sentry sync fails:

• Verify SENTRY_AUTH_TOKEN is set
• Verify SENTRY_ORG_SLUG and SENTRY_PROJECT_SLUG are correct
• Check token has event:read and project:read scopes
• Error message will show exact fix steps

Reproduce fails:

• If Daytona enabled: Check daytona --version works
• Falls back to local runner automatically
• Check repo URL is accessible (public or authenticated)
• Check command is valid for the repo

✅ REAL:

• Sentry API calls (httpx → sentry.io)
• Daytona CLI execution (subprocess)
• Local Git execution (subprocess)
• SQLite database (persistent storage)
• Graph visualization (React Flow)
• Log streaming (real-time polling)

❌ NO STUBS:

• All integrations are functional
• All endpoints return real data
• All commands execute in real environments

Set NEXT_PUBLIC_DEMO_MODE=true in frontend/.env.local:

• Pre-fills reproduce modal
• Shows "Seed Demo Incident" prominently
• Works without real integrations

See DEMO.md for full demo playbook.