open-notebook/README.md

176 lines
3.9 KiB
Markdown

# Open Notebook - Netcup RS 8000 Deployment
Self-hosted NotebookLM alternative integrated with the AI orchestrator stack.
## Architecture
```
Open Notebook
├── Frontend (Next.js) → port 8502 → Traefik → notebook.jeffemmett.com
├── API (FastAPI) → port 5055
├── Database (SurrealDB) → embedded
└── AI Providers:
├── LLM → Ollama (local, FREE)
├── Embeddings → Ollama (local, FREE)
├── STT → Groq/OpenAI (cloud)
└── TTS → ElevenLabs/OpenAI (cloud, for podcasts)
```
## Quick Deploy
```bash
# 1. SSH to Netcup
ssh netcup
# 2. Clone/copy the deployment files
cd /opt/websites
git clone https://gitea.jeffemmett.com/jeff/open-notebook.git
# OR copy files manually
mkdir -p /opt/websites/open-notebook
cd /opt/websites/open-notebook
# 3. Pull required Ollama models
docker exec ollama ollama pull llama3.2:3b # Fast LLM
docker exec ollama ollama pull llama3.1:8b # Better LLM
docker exec ollama ollama pull nomic-embed-text # Embeddings
# 4. Edit docker.env with your API keys (optional)
nano docker.env
# 5. Deploy
docker compose up -d
# 6. Verify
docker logs -f open-notebook
```
## Configure DNS (Cloudflare Tunnel)
### Option A: Add to existing tunnel config
```bash
ssh netcup
nano /root/cloudflared/config.yml
```
Add:
```yaml
- hostname: notebook.jeffemmett.com
service: http://localhost:80
```
Restart cloudflared:
```bash
docker restart cloudflared
```
### Option B: Cloudflare Dashboard
1. Go to Cloudflare Zero Trust → Access → Tunnels
2. Select your tunnel → Public Hostnames
3. Add `notebook.jeffemmett.com``http://localhost:80`
### DNS Records (if not using wildcard)
In Cloudflare DNS, add CNAME:
- Type: CNAME
- Name: notebook
- Target: a838e9dc-0af5-4212-8af2-6864eb15e1b5.cfargotunnel.com
- Proxy: Enabled
## AI Provider Configuration
### Local (FREE) - Already configured
| Feature | Provider | Model | Cost |
|---------|----------|-------|------|
| LLM | Ollama | llama3.2:3b, llama3.1:8b | FREE |
| Embeddings | Ollama | nomic-embed-text | FREE |
### Cloud (for premium features)
| Feature | Recommended Provider | Notes |
|---------|---------------------|-------|
| STT | **Groq** (free tier) | Fast Whisper, 100 hrs/month free |
| TTS | ElevenLabs | Best voice quality for podcasts |
| TTS (alt) | OpenAI | Cheaper, good quality |
### Adding API Keys
Edit `docker.env`:
```bash
# For Speech-to-Text (transcription)
GROQ_API_KEY=gsk_your_key_here
# For Text-to-Speech (podcasts)
ELEVENLABS_API_KEY=your_key_here
# OR
OPENAI_API_KEY=sk-your_key_here
```
Then restart:
```bash
docker compose restart
```
## Useful Commands
```bash
# View logs
docker logs -f open-notebook
# Restart
docker compose restart
# Update to latest version
docker compose pull
docker compose up -d
# Check Ollama models
docker exec ollama ollama list
# Pull new Ollama model
docker exec ollama ollama pull mistral:7b
# Backup data
tar -czvf notebook-backup.tar.gz notebook_data surreal_data
```
## Accessing Open Notebook
- **Web UI**: https://notebook.jeffemmett.com
- **API Docs**: https://notebook.jeffemmett.com/api/docs (if exposed)
- **Local**: http://159.195.32.209:8502
## Integration with AI Orchestrator
The Open Notebook instance connects to the same Ollama service used by the AI orchestrator, sharing:
- Model cache (no duplicate downloads)
- Compute resources
- Network (ai-orchestrator_ai-internal)
For advanced routing (e.g., GPU-accelerated inference via RunPod), configure the AI orchestrator to expose OpenAI-compatible endpoints.
## Troubleshooting
**Container won't start:**
```bash
docker logs open-notebook
# Check if ports are in use
netstat -tlnp | grep -E '8502|5055'
```
**Can't connect to Ollama:**
```bash
# Verify network connectivity
docker exec open-notebook curl http://ollama:11434/api/tags
```
**Database issues:**
```bash
# Reset database (CAUTION: loses data)
docker compose down
rm -rf surreal_data
docker compose up -d
```