LLM Wiki Pattern

Source Summary

The source describes a pattern for building personal knowledge bases where an LLM does not merely retrieve from raw documents at query time, but incrementally maintains a persistent wiki between the human and the raw sources.

Key Claims

• Traditional RAG repeatedly rediscovers relevant knowledge from scratch at query time.
• A maintained wiki compounds understanding because synthesis is stored persistently.
• The wiki should be LLM-authored and LLM-maintained, while the human focuses on curation and interpretation.
• Three layers matter:
  • raw sources
  • wiki
  • schema/instruction file
• Three recurring operations matter:
  • ingest
  • query
  • lint
• index.md and log.md help the agent navigate and maintain the system without requiring full RAG infrastructure at small scale.

Concrete Design Ideas Mentioned

• use Obsidian as the browsing and editing environment
• keep raw sources immutable
• file good answers back into the wiki
• optionally add local search tools as the wiki scales
• optionally use local image downloads so the agent can inspect visuals separately

Relevance To This Repository

This source directly motivates the structure of vipin wiki.

It establishes:

• why the wiki exists
• why AGENTS.md is important
• why index.md and log.md should be maintained carefully

Related Pages

• llm-wiki
• vipin
• 2026-04-21-vipin-wiki-bootstrap