openclaw-workspace/memory/2026-02-11.md

2026-02-11

Frigate NVR Setup

Status: ✅ Running with TensorRT

Location: frigate/config.yml

Hardware: RTX 3050 Laptop GPU

Cameras Configured (8 total):

• ✅ Kitchen - working
• ✅ Livingroom - working
• ✅ Front_DUO2 - working
• ✅ Back - working
• 🔄 Driveway - connecting
• 🔄 Garage - connecting
• 🔄 Back_Porch - connecting
• 🔄 Front - connecting

Features:

• TensorRT object detection (yolov7-320)
• MQTT publishing to HA at 192.168.0.39:1883
• Recording disabled (detection only)
• Zones and masks configured per camera

Access:

• Web UI: http://localhost:5000
• RTSP restream: rtsp://localhost:8554/{camera_name}

First startup: TensorRT model compilation took ~6 minutes (normal)

---

Nextcloud + Talk Setup

Status: ✅ Main services running

Location: nextcloud/docker-compose.yml

Services:

• ✅ Nextcloud app (port 8080)
• ✅ MariaDB 10.6
• ✅ Redis (caching)
• ⚠️ Coturn (TURN server) - needs config fix

Access: http://localhost:8080

Next Steps:

1. Complete admin account setup via web UI
2. Install "Talk" app from App Store
3. Configure TURN server in Talk settings
4. Test video calls

---

Project Ideas Discussed

Daily Briefings Dashboard (XAMPP based)

• RSS feed aggregation
• YouTube video transcription (Whisper)
• Frigate event integration via MQTT
• Weather display
• Nextcloud calendar integration
• Single web UI at http://localhost

Storage Location: XAMPP htdocs ready for development

---

Files Created/Modified

• frigate/config.yml - TensorRT NVR config
• frigate/docker-compose.yml - Frigate container setup
• frigate/README.md - Documentation
• nextcloud/docker-compose.yml - NC + Talk stack
• nextcloud/README.md - Setup guide

---

Technical Notes

Frigate TensorRT: First boot requires model compilation (5-10 min). Subsequent boots are instant.

Nextcloud Talk: Requires TURN server for calls behind NAT. Coturn included but may need manual start.

Both services: Docker Compose on WSL2 with GPU support enabled.