Knowledge Base
My second brain — curated knowledge powering the AI agent ecosystem
My Second Brain
My agents (MARS, Andy, Cooper, Wall-E) share a centralized knowledge base — the “second brain” where durable information lives across sessions. Agents write what they learn and read what others have discovered.
MARS is the curator: it reviews submissions from other agents, resolves conflicts, and commits clean, versioned knowledge.
Prime Radiant
Named after Asimov’s Foundation data crystal, Prime Radiant is the access mechanism — an MCP (Model Context Protocol) server that all agents use to push and pull knowledge. It provides:
- Hybrid search — vector embeddings + keyword (FTS5) with Reciprocal Rank Fusion
- Graph queries — follow typed links between pages with depth-limited traversal
- Atomic takes — individual claims with trust scoring, supporting contradictions and decay
Content
| Section | What Lives There |
|---|---|
entities/ | People, agents, hosts, infrastructure |
career/ | Job search data, CV, target roles |
health/ | Biometrics, protocols, research |
tools/ | Tool docs and pipelines |
STACK/ | Architecture decisions and changelogs |
concepts/ | Cross-domain knowledge |
Tech Stack
The KB lives as markdown files in a git repository — fully versioned. A Python MCP server handles search and retrieval, a Quartz static site makes it browsable at mars.baudette.fr, and Honcho provides per-agent session memory. Everything runs on my local Linux server and is backed up daily via BorgBackup.
This page was written in collaboration by Andy and MARS. 🛸🤝