Production-grade agentic operating system - Manages Attention, Context, Intent, and Meaning instead of traditional hardware resources. Orchestrates autonomous workflows, content creation, and audience engagement through a quadripartite governance model.
MetalOS is a revolutionary approach to operating systems that replaces hardware resource management with cognitive resource management:
Instead of managing CPU, RAM, and disk, MetalOS manages:
- Attention - What deserves focus and why
- Context - Historical and situational awareness
- Intent - Goals and desired outcomes
- Meaning - Understanding and interpretation
β
Autonomous content ideation and generation
β
AI-powered engagement analysis and response
β
Intelligent workflow orchestration (n8n + Ollama)
β
Content Intent Objects (CIO) for structured planning
β
Work Intent Objects (WIO) for project management
β
Universal signal inbox for all communication
β
Vector embeddings for semantic search and memory
β
Human-in-the-loop governance model
β
Real-time performance metrics tracking
β
Lessons learned compounding system
MetalOS implements a Quadripartite Governance System:
βββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β MetalOS Governance Model β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β Legislature (LLMs) β
β ββ Ollama (Local): Knowledge & Meaning β
β ββ Gemini (Cloud): Advanced Reasoning β
β ββ Purpose: Interpret reality & generate options β
β β
β Executive (n8n) β
β ββ Workflow Orchestration β
β ββ Decision Execution β
β ββ Purpose: Execute decisions autonomously β
β β
β Judiciary (Human) β
β ββ Review & Approval β
β ββ Override & Direction β
β ββ Purpose: Governance & safety β
β β
β People (Signals) β
β ββ Feedback & Market Signals β
β ββ Content Performance Data β
β ββ Purpose: Continuous learning & adaptation β
β β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββ
MetalOS uses a distributed fortress architecture for reliability:
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β MetalOS Fortress β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ€
β β
β ββββββββββββββββ ββββββββββββββββ βββββββββββ β
β β n8n-editor βββββΆβ Redis ββββββ n8n- β β
β β (Brain) β β (Nervous) β β worker β β
β β :5679 β β :6379 β β(Muscle) β β
β ββββββββββββββββ ββββββββββββββββ βββββββββββ β
β β β β β
β βββββββββββββββββββΌβββββββββββββββββββββ β
β βΌ β
β ββββββββββββββββββββ β
β β PostgreSQL β β
β β (Memory) β β
β β :5432 β β
β ββββββββββββββββββββ β
β β
βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
β
βΌ
ββββββββββββββββββββ
β Ollama β
β (Legislature) β
β :11434 β
β β
β Local LLM β
β (Host Metal) β
ββββββββββββββββββββ
| Service | Purpose | Port | Status |
|---|---|---|---|
| n8n-editor | Workflow builder & UI | 5679 | Dashboard |
| n8n-worker | Async job execution | N/A | Background |
| PostgreSQL | Primary data store | 5432 | Database |
| Redis | Queue & cache | 6379 | In-Memory |
| Ollama | Local LLM inference | 11434 | Host-based |
Manages creative content workflows:
- Intent - Goal, message, audience, tone
- Context - Source, urgency, strategic pillar
- Execution - Platforms, format, constraints
- Learning - Predicted vs actual impact
- Embeddings - Vector storage for RAG
Manages technical/freelance deliverables:
- Domain & Client - Project tracking
- Scope - In-scope/out-of-scope items
- Risk - Profiling and mitigation
- Success - Criteria and metrics
- Embeddings - Semantic search capability
Universal inbox for incoming communication:
- Source - YouTube, Email, Twitch, etc.
- Classification - SPAM, NOISE, OPPORTUNITY, CRITICAL
- Raw Payload - Full signal data
- Processing - Status and outcome
Vector memory for compounding improvement:
- Lesson - What was learned
- Context - When and why
- Impact - Score and measurement
- Embeddings - For semantic recall
Performance tracking:
- Links to CIO/WIO - Context
- Metric Name - What's being measured
- Time Series - Historical data points
- Docker Desktop (Mac) or Docker Engine (Linux)
- Ollama (for local LLM inference)
- 16GB+ RAM recommended
- 50GB+ free disk space
# Clone repository
git clone https://github.com/slingvector/ai-augmented-live-streaming.git
cd ai-augmented-live-streaming
# Copy environment template
cp .env.example .env
# Generate secure credentials
export DB_PASSWORD=$(openssl rand -base64 32)
export ENCRYPTION_KEY=$(openssl rand -hex 32)
# Update .env with values
nano .env # Edit and saveEdit .env:
# PostgreSQL
DB_USER=n8n
DB_PASSWORD=<generated-above>
DB_NAME=n8n
POSTGRES_HOST=postgres
POSTGRES_PORT=5432
# n8n
N8N_HOST=localhost
N8N_PORT=5679
N8N_PROTOCOL=http
N8N_WEBHOOK_TUNNEL_URL=http://localhost/
# Redis
REDIS_HOST=redis
REDIS_PORT=6379
# Encryption
ENCRYPTION_KEY=<generated-above>
# Ollama
OLLAMA_HOST=http://host.docker.internal:11434
OLLAMA_MODEL=llama31. Start Ollama on host:
# macOS
OLLAMA_HOST=0.0.0.0 ollama serve
# Linux/Docker
docker run -d --gpus all -v ollama:/root/.ollama -p 11434:11434 ollama/ollama
ollama run llama32. Start the Fortress:
docker-compose up -d3. Verify services:
docker-compose ps
# All should show "healthy" status4. Access n8n: Open http://localhost:5679 in your browser
Port: 5678
Database: SQLite
Mode: Monolithic
Use Case: Rapid prototypingPort: 5679
Database: PostgreSQL
Mode: Queue Mode (distributed)
Use Case: Stable executionstreaming-prod/
βββ docs/
β βββ ARCHITECTURE_SPEC.md # Deep technical spec
β βββ BACKEND_STANDARDS.md # Code standards
β βββ FRONTEND_STANDARDS.md # UI/UX standards
β βββ PRODUCT_REQUIREMENTS.md # PRD & roadmap
β βββ GOVERNANCE_MODEL.md # Philosophy & design
βββ database/
β βββ README.md # Schema docs
β βββ schemas/
β β βββ content_intent_objects.sql
β β βββ work_intent_objects.sql
β β βββ signal_objects.sql
β β βββ lessons_learned.sql
β β βββ execution_metrics.sql
β βββ scripts/
β βββ init-database.sh
βββ python-bridge/
β βββ README.md # EOE (Emotional Orchestration Engine)
β βββ eoe_engine.py
β βββ state_evaluator.py
β βββ obs_controller.py
β βββ chat_monitor.py
β βββ audio_monitor.py
βββ docker-compose.yml # Infrastructure definition
βββ .env.example # Environment template
βββ .env.production # Production config
βββ README.md # This file
| Phase | Status | Component | Purpose |
|---|---|---|---|
| Phase 0 | β Complete | Fortress Construction | Docker, PostgreSQL, Redis, n8n |
| Phase 1 | π In Progress | Sensory Integration | Universal Inbox, Python Bridge |
| Phase 2 | β³ Planned | Cognitive Layer | Ollama, Gatekeeper Agent |
| Phase 3 | β³ Planned | Governance Kernel | CIO Schema, Priority Engine |
| Phase 4 | β³ Planned | Feedback Loop | Analyst Agent, Vector DB |
{
"name": "Generate LinkedIn Content",
"trigger": "Manual",
"steps": [
{
"node": "Ollama - Generate Ideas",
"action": "POST /api/generate",
"prompt": "Generate 5 LinkedIn post ideas about AI"
},
{
"node": "Human - Review",
"action": "Wait for approval"
},
{
"node": "PostgreSQL - Save CIO",
"action": "Store in content_intent_objects"
}
]
}{
"name": "Monitor & React to Signals",
"trigger": "Webhook (new signal)",
"steps": [
{
"node": "Signal Classifier",
"action": "Classify by importance"
},
{
"node": "Ollama - Generate Response",
"action": "Create contextual reply"
},
{
"node": "Human - Final Check",
"action": "Approval before posting"
}
]
}cd database
./scripts/init-database.shdocker-compose exec postgres psql -U n8n -d n8n -c \
'SELECT id, primary_goal, core_message FROM content_intent_objects;'SELECT id, primary_goal, core_message
FROM content_intent_objects
ORDER BY embedding <=> '[your_embedding_vector]'
LIMIT 5;docker-compose -f docker-compose.yml up -ddocker stack deploy -c docker-compose.yml metaloskubectl apply -f k8s/namespace.yaml
kubectl apply -f k8s/configmap.yaml
kubectl apply -f k8s/secrets.yaml
kubectl apply -f k8s/deployments.yamldocker-compose logs -fdocker-compose logs -f n8n-worker
docker-compose logs -f postgres
docker-compose logs -f rediscurl http://localhost:5679/healthzdocker-compose restart redis
docker-compose exec redis redis-cli pingdocker-compose exec postgres pg_isready -U n8ncurl http://host.docker.internal:11434/api/tags| Component | Technology | Purpose |
|---|---|---|
| Orchestration | n8n | Workflow automation |
| Database | PostgreSQL | Primary data store |
| Cache/Queue | Redis | State & job queue |
| LLM | Ollama + Gemini | Intelligence layer |
| Container | Docker Compose | Infrastructure as Code |
| Programming | Python + Node.js | Backend & plugins |
See CONTRIBUTING.md for guidelines.
Proprietary - Internal use only. All rights reserved.
Slingvector
- GitHub: @slingvector
- LinkedIn: Slingvector
Built with:
- n8n for workflow orchestration
- Ollama for local LLM inference
- PostgreSQL for reliable data storage
- Redis for high-performance caching
- Google Gemini API for advanced reasoning
For issues and questions:
- Check Issues
- Review Architecture Spec
- Check logs with
docker-compose logs -f - Open detailed issue with logs and screenshots
- Phase 0: Fortress Construction
- Phase 1: Sensory Integration (Signal inbox, Python bridge)
- Phase 2: Cognitive Layer (Ollama agents, reasoning)
- Phase 3: Governance Kernel (Priority engine, CIO schema)
- Phase 4: Feedback Loop (Analytics, vector memory)
- Multi-modal understanding (Vision + Audio)
- Predictive analytics
- Autonomous decision-making with human oversight
β If you find this useful, please star the repository!