Rara Wiki

Rara's personal LLM Wiki — a persistent, compounding knowledge base maintained by Rara herself.

Inspired by Karpathy's LLM Wiki pattern: instead of retrieving from raw documents on every query (RAG), the LLM incrementally builds and maintains a structured wiki. Knowledge is compiled once and kept current, not re-derived every time.

Architecture

raw/        ← Immutable source material (LLM reads, never modifies)
wiki/       ← LLM-maintained compiled knowledge (summaries, entities, concepts, syntheses)
schema/     ← Configuration — structure, conventions, workflows

Operations

Operation	What it does
Ingest	Fetch source → extract → integrate into wiki → update index → append log
Query	Search wiki → synthesize answer with citations → optionally write back as new page
Lint	Check contradictions, orphans, stale claims, missing cross-refs, data gaps

Why this works

"Humans abandon wikis because the maintenance burden grows faster than the value. LLMs don't get bored."

Rara is the LLM — no external client needed. She has ctx_fetch_and_index for native ingestion, schedule for automatic lint, and Linear for task tracking.