Farzapedia and personal AI wikis
Farzapedia is a pioneering, private personal knowledge wiki developed by Farza (founder of buildspace). It serves as the definitive real-world implementation of the LLM Wiki design pattern popularized by llm-wiki.
The architecture of Farzapedia
Rather than relying on complex, black-boxed, and expensive vector database infrastructure (Retrieval-Augmented Generation / RAG), Farzapedia uses a simple, compiled approach:
- Unstructured Source Materials: Approximately 2,500 raw records comprising Apple Notes, personal diaries, and raw iMessage chat histories.
- The LLM Compiler: An LLM agent (e.g., Claude) is instructed to act as a disciplined editor and gardener of the mind.
- The Interlinked Wiki: The LLM parses, deduplicates, and synthesizes the unorganized inputs, generating ~400 structured, clean, and heavily interlinked Markdown files tied together with standard WikiLinks.
The four principles of owned knowledge
Farzapedia champions four crucial properties of modern knowledge engineering:
- Explicit: The memory artifact is human-readable, navigable, and fully inspectable. There are no opaque vector embeddings or proprietary database formats.
- Yours (Local First): All data lives securely on your local hard drive under your control. No vendor lock-in or subscription APIs.
- File Over App: Universal, standardized Markdown files mean your second brain is completely independent of any single app or operating system.
- BYOAI (Bring Your Own AI): A flat directory of clean files allows you to plug in any LLM agent (Claude, OpenAI, Codex) to query, write, or refactor your knowledge base without altering the underlying data.
Dynamic querying in this vault
The llm-wiki-cli is a fully realized software pipeline built to make the Farzapedia pattern deterministic. By treating Markdown frontmatter as a queryable RDF knowledge graph, you can run powerful SPARQL queries and use command-line pipelines to synthesize indexes on the fly—fulfilling the dream of a queryable, self-updating mind map.