Session Context

How Quox maintains context across conversations and sessions.

Use this document to maintain continuity between development sessions.

Current State (Last Updated: 2025-01-14)

Completed Features

• HAL-9000 style homepage with audio-reactive visualizer
• Voice mode selector (JARVIS/Cyberpunk/HAL)
• Text input fallback with subtitles
• n8n workflow engine running in Docker
• quox-chat webhook for Claude integration
• quox-speak webhook for TTS (ElevenLabs)
• quox-transcribe webhook for STT (Whisper)
• Context injection - Claude knows all 50+ hosts
• Session continuity via session_id
• Quick actions panel with visual feedback
• React Router navigation structure
• Activity logging in n8n
• Comprehensive documentation suite

In Progress

• Voice testing with actual API keys
• Multi-agent development team architecture
• Settings page UI implementation

Not Started

• Self-modification with approval gates
• RAG integration with Qdrant
• Grafana dashboard synchronization
• Prometheus exporter automation
• System update automation agents

API Keys Status

To configure:

sudo nano /opt/n8n/.env
# Add keys, then:
cd /opt/n8n && docker compose restart n8n

Next Session Priorities

Priority 1: Voice Testing

1. Add OpenAI API key for Whisper
2. Add ElevenLabs API key for TTS
3. Test full voice conversation flow
4. Fix any issues discovered

Priority 2: Automation Agents

Create n8n workflows for:

• apt-update - Check for updates on hosts
• apt-upgrade - Full system upgrade
• grafana-sync - Update dashboards with current hosts

Priority 3: Bastion Tool Exploration

SSH to nw-edge-gw-01 and document bastion tool features:

• SSH key management
• MOTD configuration
• Prometheus exporter installation
• Firewall port management

Priority 4: RAG Setup

• Deploy Qdrant container
• Create embedding pipeline
• Index n8n workflow templates
• Test context-aware queries

Key Files Reference

Dashboard

Infrastructure

Workflows (JSON)

Architecture Decisions

Voice Flow

User speaks → Browser MediaRecorder → WebM blob
    → POST /webhook/transcribe → Whisper API → text
    → POST /webhook/chat → Claude API → response
    → POST /webhook/speak → ElevenLabs API → audio
    → Browser Audio playback

State Machine (HAL Eye)

IDLE (red) → LISTENING (blue) → THINKING (orange) → SPEAKING (green) → IDLE

SSH Access Pattern

Dashboard → n8n → Bastion (nw-edge-gw-01) → Target hosts

Known Issues

1. SSH intermittent - nw-edge-gw-01 sometimes rejects connections
2. Browser audio - Some browsers block autoplay
3. Large responses - May need truncation handling

Useful Commands

# Restart services
cd /opt/n8n && docker compose restart

# View n8n logs
docker logs n8n -f

# Rebuild dashboard
cd /home/control/quox-dashboard && npm run build

# Check workflow status
docker exec n8n n8n list:workflow

Session Notes

2025-01-14 Session

• Built complete HAL interface with voice support
• Created comprehensive documentation
• Designed automation agents (not yet implemented)
• Planned RAG integration with Qdrant
• Ready for voice testing once API keys configured

Next Session Goals

• Configure and test voice features
• Implement at least one automation agent
• Explore bastion tool capabilities
• Begin RAG implementation