Architecture

The system has three distinct layers. Each has a clear owner and a clear role.

┌─────────────────────────────────────────────┐
│              Raw Sources                    │
│  articles, papers, images, data files       │
│  (immutable — LLM reads, never writes)      │
└───────────────────┬─────────────────────────┘
                    │ ingest
                    ▼
┌─────────────────────────────────────────────┐
│                The Wiki                     │
│  LLM-generated markdown files               │
│  summaries, entities, concepts, synthesis   │
│  (LLM owns this layer entirely)             │
└───────────────────┬─────────────────────────┘
                    │ governed by
                    ▼
┌─────────────────────────────────────────────┐
│               The Schema                   │
│  CLAUDE.md / AGENTS.md                      │
│  conventions, workflows, page formats       │
│  (you and the LLM co-evolve this)           │
└─────────────────────────────────────────────┘

Layer 1: Raw Sources

Your curated collection of source documents — articles, papers, images, data files. These are immutable: the LLM reads from them but never modifies them. This is your source of truth.

Sources can arrive as:

• Web articles clipped to markdown (Obsidian Web Clipper is ideal)
• PDFs, papers, reports
• Images, screenshots, diagrams
• Exported transcripts (podcasts, meetings, video lectures)
• Data files (CSVs, JSON exports)

Layer 2: The Wiki

A directory of LLM-generated markdown files. The LLM owns this layer entirely — it creates pages, updates them when new sources arrive, maintains cross-references, and keeps everything consistent. You read it; the LLM writes it.

Typical page types:

Type	Purpose
Summary pages	One per source — what it says, key claims, quality notes
Entity pages	People, organizations, products — aggregated across sources
Concept pages	Ideas, techniques, theories — explained and cross-referenced
Comparison pages	Side-by-side analysis of related things
Synthesis pages	High-level essays drawing across many sources
index.md	Catalog of all pages with one-line summaries
log.md	Append-only timeline of all wiki activity

Layer 3: The Schema

A configuration document (e.g. CLAUDE.md for Claude Code, AGENTS.md for Codex) that tells the LLM:

• How the wiki directory is structured
• What the naming and formatting conventions are
• What workflows to follow when ingesting a new source
• How to handle contradictions between sources
• What constitutes a "complete" entity or concept page
• When to create a new page vs. update an existing one

This is the key piece that turns the LLM from a generic chatbot into a disciplined wiki maintainer. You and the LLM co-evolve it over time as you learn what works for your domain.

Example schema excerpt

## Wiki conventions

- All pages live under wiki/
- Source summaries: wiki/sources/<slug>.md
- Entity pages: wiki/entities/<name>.md
- Concept pages: wiki/concepts/<name>.md

## Ingest workflow

When asked to ingest a source:
1. Read the source fully
2. Discuss key takeaways with the user
3. Write wiki/sources/<slug>.md
4. Update wiki/index.md
5. Update or create any entity/concept pages touched
6. Append an entry to wiki/log.md