# Welcome to Your LLM Wiki 🧠

> A **persistent, compounding knowledge base** maintained by LLM agents.
>
> *"The wiki is a persistent, compounding artifact. The cross-references are already there. The contradictions have already been flagged. The synthesis already reflects everything you've read."* — [Andrej Karpathy](https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f)

---

## Quick Start

### 1. Add a source
Drop an article, paper, or note into **`raw/`**, then tell your LLM agent:

> "I added a new source to raw/. Please ingest it into the wiki."

### 2. Ask a question
Your LLM will search the wiki and synthesize an answer with citations.

> "What do we know about [topic]?"

### 3. Save insights back
Good answers can be filed into the wiki:

> "Write that comparison as a new wiki page and update the index."

### 4. Lint periodically
> "Health-check the wiki. Look for contradictions, stale claims, orphan pages, and missing cross-references."

---

## Architecture

| Layer | Directory | Purpose |
|-------|-----------|---------|
| **Raw Sources** | `raw/` | Immutable source documents (articles, papers, notes) |
| **The Wiki** | `wiki/` | LLM-generated markdown pages — entities, concepts, comparisons |
| **The Schema** | `schema/AGENTS.md` | Configuration that tells the LLM how to maintain the wiki |

### Special files
- **`index.md`** — content catalog (LLM-maintained)
- **`log.md`** — chronological changelog (LLM-maintained)
- **`tags.md`** — auto-generated tag index (for `mkdocs-material`)

---

## Tips

- Use **Obsidian Web Clipper** to save articles as markdown into `raw/`
- Run `docker compose --profile setup run init` to initialize the structure
- The wiki viewer runs at **`http://<host>:8787`**
- Your LLM agent (Hermes, Codex, Claude Code) does all the maintenance — you just curate and explore