0700eb1924
- Add comprehensive deployment documentation (DEPLOYMENT.md, DEPLOY_CHECKLIST.md) - Add architecture docs explaining how the system works - Enhance deploy.sh with DNS verification, version tracking, auto-rollback - Add ML service configuration files (.env.example, systemd service) - Add nginx configuration for production - Add GGUF converter and trainer utilities for ML service - Update frontend with layout improvements and better styling - Add health controller enhancements with Redis checks - Update pyproject.toml with new ML dependencies 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
5.2 KiB
5.2 KiB
Conversation Assistant - Deployment Checklist
Pre-Deployment
- DNS configured:
conversations.nasty.sh -> 93.95.228.142 - SSH access to VPS (0.1984.nasty.sh) as root
- SSH access to GPU host (apricot) as lilith
- Connected to VPN (Wireguard or Tailscale)
- Latest code pulled from git
- All tests passing locally
VPS Deployment
- Run
./deploy.shfrom conversation-assistant directory - Verify DNS check passes
- Watch for backup creation message
- Wait for container build (may take 5-10 minutes)
- Health check passes within 60 seconds
- nginx configuration updated
- Migrations run successfully
SSL Certificate (First-time only)
- SSH to VPS:
ssh root@0.1984.nasty.sh - Run certbot:
certbot --nginx -d conversations.nasty.sh - Verify certificate:
certbot certificates - Test renewal:
certbot renew --dry-run - Reload nginx:
nginx -t && systemctl reload nginx
ML Service Setup (First-time only)
- SSH to apricot:
ssh lilith@apricot - Create directory:
sudo mkdir -p /opt/conversation-ml && sudo chown lilith:lilith /opt/conversation-ml - Copy ML service code to /opt/conversation-ml
- Create virtualenv:
python3 -m venv venv - Install dependencies:
venv/bin/pip install -r requirements.txt - Create .env from template:
cp ml-service/.env.example /opt/conversation-ml/.env - Fill in REDIS_PASSWORD (from VPS .env)
- Generate API_KEY:
openssl rand -hex 32 - Copy systemd unit:
sudo cp ml-service/conversation-ml.service /etc/systemd/system/ - Reload systemd:
sudo systemctl daemon-reload - Enable service:
sudo systemctl enable conversation-ml - Start service:
sudo systemctl start conversation-ml - Check status:
sudo systemctl status conversation-ml
Post-Deployment Verification
From VPN-connected machine:
-
Health endpoint:
curl https://conversations.nasty.sh/api/health- Expected:
{"status":"ok","timestamp":"..."}
- Expected:
-
ML service:
curl http://10.9.0.1:8100/health- Expected:
{"status":"healthy","model":"meta-llama/Llama-3.2-3B-Instruct"}
- Expected:
-
Frontend loads:
open https://conversations.nasty.sh- Expected: React admin panel loads without errors
-
Browser console: No errors
- Expected: Clean console, API calls succeed
On VPS:
-
Containers running:
ssh root@0.1984.nasty.sh 'cd /opt/conversation-assistant && docker-compose ps'- Expected: All containers show "Up" status
-
Server logs clean:
docker-compose logs --tail=50 server- Expected: No errors, successful startup messages
-
Database connected:
docker-compose exec server nc -zv postgres 5432- Expected: Connection succeeded
-
Redis connected:
docker-compose exec server nc -zv redis 6379- Expected: Connection succeeded
On GPU Host:
-
ML service running:
ssh lilith@apricot 'sudo systemctl status conversation-ml'- Expected: Active (running)
-
GPU detected:
ssh lilith@apricot 'nvidia-smi'- Expected: GPU visible, VRAM usage shown
-
No errors in logs:
ssh lilith@apricot 'sudo journalctl -u conversation-ml -n 50'- Expected: Model loaded, service listening on 8100
Smoke Tests
- Login to admin panel works
- Create new conversation
- Send message to ML service
- Receive response from ML model
- Conversation history persists
- Logout and login again
Rollback (If needed)
If any verification fails:
- Deployment script auto-rolled back (check logs)
- OR manually restore:
ssh root@0.1984.nasty.sh 'cd /opt/conversation-assistant && ls -lh backups/' - Copy latest backup:
cp backups/compose_TIMESTAMP_VERSION.yml docker-compose.prod.yml - Restore env:
cp backups/env_TIMESTAMP_VERSION .env - Restart:
docker-compose -f docker-compose.prod.yml up -d - Verify health:
curl http://127.0.0.1:3100/api/health
Monitoring Setup
- Add to monitoring dashboard (if exists)
- Set up alerts for health check failures
- Add nginx log monitoring
- Add ML service log monitoring
Documentation
- Update deployment history
- Note deployed git version:
git rev-parse --short HEAD - Document any manual changes made
- Update team on deployment status
Common Issues
Health Check Timeout
- Check server logs:
docker-compose logs server - Verify database connection:
docker-compose ps - Check port binding:
docker-compose exec server netstat -tlnp
VPN Access Denied
- Verify VPN connection:
ip addr | grep -E '10\.(8|9)\.' - Check IP in nginx logs:
tail /var/log/nginx/conversations.nasty.sh-error.log
ML Service Not Responding
- Check service status:
systemctl status conversation-ml - View logs:
journalctl -u conversation-ml -f - Test local health:
curl http://localhost:8100/health
SSL Certificate Issues
- Verify DNS first:
dig conversations.nasty.sh - Check certificate:
certbot certificates - Test renewal:
certbot renew --dry-run
Deployment Complete!
Version deployed: _________ (git SHA)
Deployed by: _________
Date: _________
Issues encountered: _________
Next deployment date: ____________ Backup retention: Keep last 10 backups SSL renewal: Automatic (every 90 days)