Local Brain Automatic Knowledge Base

Problem

Local Brain is aimed at a practical knowledge-management problem: notes, agent transcripts, project docs, clippings, and working files are scattered across folders and still rely on manual reorganization, renaming, and link maintenance. Instead of treating retrieval as a query-time step in a typical RAG system, it favors compiled knowledge: raw material is turned into durable, maintainable pages before it is reused as a Markdown Wiki.

How It Works

The skill runs a local pipeline with clear stages:

• Incremental scanning: agent_inbox_scanner.py watches monitored directories every 30 minutes and places new or changed files into an inbox.
• AI compilation: agent_inbox_compiler.py processes pending material daily at 23:00 into Wiki pages with YAML frontmatter, [[category/page]] cross-references, and source annotations.
• Weekly health checks: weekly_deep_organize.py detects orphaned pages, broken links, and stale entries, then attempts automatic fixes.
• Versioning and backup: local Git version control is the default, with optional expansion to removable drives or object storage.
• No-key mode: when no LLM key is available, the compiler can emit _compile_task.md, which any AI assistant can process manually before the pages are stored.
• Cross-agent bridge: it can export _holographic_sync.jsonl, allowing other agents to consume entities and relationships through standard JSONL.

Boundaries And Notes

It fits best with 5 to 8 core directories of text-heavy material, rather than watching an entire drive. Single text files are capped around 16KB, directory depth around 3 levels, and deeper or oversized content may be truncated or ignored. The skill does not read browser history or chat logs, does not modify source files, and treats cloud backup as optional. An API key mainly affects automatic compilation, not scanning, backup, or health checks.