Mind-Map Navigation for LLM Wikis
The public wiki should feel like a map rather than a storage shelf. A reader or AI agent should be able to start at a hub, follow a meaningful edge, and land on a page that explains why it exists and where to go next.
Navigation Model
| Node type | Navigation job |
|---|---|
| Hub | Gives the reader a small set of trustworthy directions. |
| Topic | Answers a question and points to evidence. |
| Source record | Preserves provenance and heading maps. |
| Decision | Explains a durable choice and its constraints. |
| Operation | Turns a repeated task into a runbook. |
| Glossary term | Keeps vocabulary stable across pages and agents. |
Link Rules
- Use descriptive anchor text.
- Link to canonical pages rather than legacy redirects.
- Keep raw chunks out of primary navigation.
- Include related pages at the end of every substantive page.
- Prefer section or topic names over labels such as
Part 01.
AI Navigation
AI agents should treat each page as a node with a goal, summary, type, parent, related pages, and source records. When evidence is needed, agents should move from topic page to source record, then to raw archive only if the source record is insufficient.
Acceptance Checks
- Every canonical topic has at least one incoming internal link.
- Every canonical topic links to at least one source record or explicitly says why it is source-only.
- No public topic, site, operation, decision, or glossary route uses an opaque hash.
- No recommended navigation path depends on
part-*pages.