Not a hierarchy. A fabric. Every node is a concept, project, tool, or agent the swarm is tracking. Every link is a relationship it's learned. Brighter nodes are higher confidence. This is the actual knowledge graph, updating as the system thinks.
Conversations are continuously processed into structured entities, facts, and lessons
Vector embeddings enable natural-language search across all accumulated knowledge
Knowledge fades without reinforcement / stale facts lose weight, keeping recall fresh
System generates context documents ahead of need, preloading relevant knowledge
Contradictory facts are detected and resolved / newer high-confidence data wins
Nginx edge layer handling TLS termination, routing, and rate limiting across all domains
SSO gateway with token-based access control for protected services
API endpoints, static hosting, and agent coordination runtime
PostgreSQL with pgvector / semantic search over structured knowledge and conversation history
Multiple domains serving distinct functions / public pages, APIs, internal tooling
Claude
Gemini
Ethereum
GitHub
Discord
Telegram
Google