012025 · Open Source

graphify-rs

RoleMaintainer
TagsRust · Knowledge Graphs · CLI · LLM Tooling

What it is

Inspired by Andrej Karpathy's /raw folder workflow: drop anything into a folder — papers, tweets, screenshots, code, notes — and get a structured knowledge graph that surfaces what you didn't know was connected.

Three things graphify-rs does that an LLM alone can't:

# Feature Why it matters
1 Persistent graph Relationships survive across sessions. Query weeks later without re-reading.
2 Honest audit trail Every edge tagged EXTRACTED, INFERRED, or AMBIGUOUS. Facts vs. guesses, always clear.
3 Cross-document surprise Community detection finds connections you'd never think to ask about.

Why a rewrite

The original was a fine prototype. Two things pushed me to a Rust rewrite: (1) the graph build needed to be fast and reproducible enough to run on every commit of a research repo, and (2) the audit trail needed to be a first-class data type, not a comment in an extraction script. Rust gave me the right constraints for both.

Using it

cargo install graphify-rs
 
# Build a knowledge graph (free, fast, no API key needed)
graphify-rs build --no-llm
 
# Explore interactively
open graphify-out/graph.html
 
# Query the graph
graphify-rs query "how does auth work?"
 
# (Optional) Add semantic extraction via Claude API
graphify-rs build --llm anthropic

The interactive viewer is a single HTML file with a force-directed layout; the CLI query path is great for piping into other tools.

← All workTalk about this →