A personal knowledge base built and maintained by an LLM. You curate the sources and ask the questions — the LLM handles all the writing, cross-referencing, and maintenance.

Based on the LLM Wiki pattern.

Instead of retrieving from raw documents every time you ask a question (like RAG), the LLM incrementally builds a persistent wiki — a structured, interlinked collection of markdown files. When you add a new source, the LLM reads it, extracts key information, and integrates it into the existing wiki: updating entity pages, revising topic summaries, flagging contradictions, and strengthening the evolving synthesis.

The wiki is a compounding artifact. Cross-references are already in place. Contradictions are already flagged. Every source you add and every question you ask makes it richer.

Knowledge-Wiki/
├── CLAUDE.md              # Schema — conventions, workflows, and page formats
├── README.md              # This file
├── raw/                   # Source documents (immutable — LLM reads, never modifies)
│   └── assets/            # Images and other media
└── wiki/                  # LLM-generated and maintained markdown pages
    ├── index.md           # Content catalog for navigating the wiki
    ├── log.md             # Chronological activity log
    ├── sources/           # Source summary pages
    ├── entities/          # People, places, organizations
    ├── concepts/          # Topics and ideas
    └── analyses/          # Comparisons, syntheses, and saved query results

1. Drop a file into raw/.
2. Tell the LLM to process it.
3. The LLM reads the source, discusses key takeaways with you, creates a source summary page, updates entity and concept pages across the wiki, and logs the activity.

A single source might touch 10-15 wiki pages. You can stay involved and guide emphasis, or batch-ingest with less supervision.

1. Ask any question about your collected knowledge.
2. The LLM searches the wiki index, reads relevant pages, and synthesizes an answer with citations.
3. Substantial answers get filed back into the wiki as analysis pages — so your explorations compound just like ingested sources.

Periodically ask the LLM to audit the wiki for:

• Contradictions between pages
• Stale claims superseded by newer sources
• Orphan pages with no inbound links
• Important concepts mentioned but lacking their own page
• Missing cross-references
• Data gaps worth investigating

• An LLM agent with file read/write access (e.g., Claude Code)
• A markdown viewer (e.g., Obsidian) for browsing the wiki

1. Clone the repo and open it with your LLM agent.
2. Add a source — drop an article, paper, or note into raw/.
3. Tell the LLM to ingest it — e.g., "Read and ingest raw/my-article.md."
4. Browse the results — open the wiki in Obsidian or any markdown viewer and explore the generated pages.
5. Ask questions — query the wiki to get synthesized answers from everything you've ingested.

• Obsidian Web Clipper — browser extension that converts web articles to markdown, great for quickly adding sources.
• Download images locally — in Obsidian, set an attachment folder path (e.g., raw/assets/) and bind "Download attachments" to a hotkey so images are preserved locally.
• Graph view — Obsidian's graph view is the best way to visualize wiki structure and spot orphan pages.
• Marp — markdown-based slide decks. Obsidian has a plugin. Useful for generating presentations from wiki content.
• Dataview — Obsidian plugin that queries YAML frontmatter across pages for dynamic tables and lists.
• Version history — the wiki is just a git repo of markdown files, so you get history, branching, and collaboration for free.

• Personal — goals, health, self-improvement. Journal entries, articles, podcast notes building a structured picture over time.
• Research — deep dives over weeks or months. Papers, articles, reports becoming a comprehensive wiki with an evolving thesis.
• Reading a book — chapter-by-chapter companion wiki with characters, themes, and plot threads interlinked.
• Business/team — internal wiki fed by Slack threads, meeting transcripts, project docs, customer calls.
• Anything else — competitive analysis, due diligence, trip planning, course notes, hobby deep-dives.

The bottleneck of a knowledge base isn't reading or thinking — it's the bookkeeping. Updating cross-references, keeping summaries current, noting contradictions, maintaining consistency. Humans abandon wikis because the maintenance burden outpaces the value. LLMs handle all of that at near-zero cost, so the wiki stays maintained and keeps compounding.

This project is for personal use. Adapt it however you like.