System Documentation¶
This section contains the technical system documentation for Sovereign MoE — the fully self-hosted Multi-Model LLM-System.
Contents¶
| Page | Description |
|---|---|
| Overview & System Docs | Complete system documentation: hardware, pipeline, experts, API, deployment |
| Architecture | LangGraph pipeline, service topology, caching architecture, expert routing |
| Kafka Event Streaming | Kafka architecture, topics, message schemas, HowTo admin & dev |
| Docker Services | Service overview, ports, volumes, operational commands |
| HABE 2.0 Operations Manual | Vector Symbolic Architecture (VSA), hierarchical graph compilation, virtual prefix attention modulation |
| Advice-Taker Rule-Engine | John McCarthy's rule-engine, Jaccard 3-gram semantic matching, regex parameter extraction |
| Eurisko Heuristic Breeder | Self-referential template optimizer, roulette-wheel selection, feedback loops, crossover |
| Dynamic System Prompts | Meta-prompter architecture, zero-latency structured fallback pathways |
Quick Overview¶
The system consists of the following core components:
- LangGraph Orchestrator (Port 8002) — FastAPI + LangGraph pipeline, OpenAI API-compatible
- MCP Precision Tools Server (Port 8003) — 16 deterministic calculation tools
- Admin UI (Port 8088) — Web configuration, user management, monitoring
- ChromaDB (Port 8001) — Vector cache for semantic caching
- Valkey (Port 6379) — LangGraph checkpoints, expert performance scores, user cache
- Neo4j (Port 7687) — Knowledge graph (GraphRAG)
- Kafka (Port 9092) — Decoupled event streaming for ingest and audit logging
- Prometheus (Port 9090) + Grafana (Port 3001) — Observability