Back to projects

llmwiki

LLM-managed personal knowledge base

Add notes in plain text. Ask questions. Get answers from your own knowledge.

View on GitHub

Demo

terminal
$ wiki add "Transformers use self-attention"
> embedding... done
> stored in knowledge base
>
$ wiki ask "how does attention work?"
> searching knowledge base...
>
Transformers use self-attention to...

Problem

Note-taking tools optimise for writing, not retrieval. Months later, the note you need is buried — and keyword search returns 40 results for a term that appears everywhere.

What it is

llmwiki is a Rust CLI and TUI that turns your plain-text notes into a queryable knowledge base. You write notes in natural language; the LLM synthesises answers by retrieving and combining the relevant ones.

  • add notes from the CLI or TUI
  • ask questions in plain English
  • semantic retrieval over your entire note history
  • local-first — nothing leaves your machine

Features

CLI + TUI

Fast terminal interface for adding, searching, and browsing notes.

LLM Synthesis

Ask questions; the model synthesises answers from your own notes.

Semantic Search

Retrieval over embeddings — finds relevant notes by meaning.

Local-first

Everything stays on your machine. No cloud, no sync service.

Query flow

Note / Query
Embed
Vector Store
LLM Synthesis
Answer

Example

Question

wiki ask "what did I learn about attention mechanisms?"

Answer

Based on your notes:

Transformers use self-attention to weight the relevance of each
token in a sequence relative to every other token. The attention
score is computed as softmax(QKᵀ / √d_k) · V.

Sources: [note #12, note #34]

Quick start

cargo install llmwiki
wiki add "your note here"
wiki ask "your question"

Why I Built This

I write notes constantly but rarely find them when I need them. Keyword search breaks down the moment you forget the exact phrase you used.

llmwiki lets me ask questions in the way I think, not in the way I typed.

Repository

View on GitHub

Links