LLM Wiki

A self-maintaining knowledge base: sources are ingested automatically, entities and concepts extracted, and everything connected in a searchable knowledge graph.

Type: Knowledge base + knowledge graph
Status: In active use
Role: Concept, design & development
Stack: Local AI · Python · Markdown · graph visualisation

RESULT

Sources ingested automatically and connected in a knowledge graph with hundreds of nodes — fully local.

Talk to me about this Read how it works

The knowledge graph: sources, entities, concepts and syntheses connected through hundreds of relations (label-free view)

01 THE PROBLEM

Knowledge gets scattered across notes, documents and conversations. I wanted a system that ingests, summarises and connects sources automatically — so the overview grows along instead of fading away.

02 WHAT IT DOES

Automatic ingest: Drop a source in the inbox and the wiki ingests it: summary, key claims, entities and concepts are extracted automatically.
Knowledge graph: Everything is connected in a searchable graph with hundreds of nodes and relations, with separate types for sources, entities, concepts and syntheses.
Living synthesis: An overview page is rewritten on every ingest, so the common thread across all sources stays up to date.
Local & private: Runs on local AI; the entire knowledge base stays on your own machine.
Kept healthy: Built-in checks for orphan pages, broken links and contradictions keep the knowledge base consistent.

03 WHAT I LEARNED

The value wasn't in even more notes, but in the connections between them. Once ingest, graph and synthesis worked together, a loose collection of documents became a real knowledge system.

INTERESTED?

Let's talk.

Tell me what you think — or what you'd want to do with it.

Contact me Next project: Jarvis