Graphify

Graphify is an MIT-licensed knowledge graph engine that ingests heterogeneous inputs — source code (with AST awareness), markdown docs, PDFs and papers, meeting transcripts, browser history, even images and diagrams — and decodes them into a single traversable graph. The pitch is 'any input, one graph, complete recall': instead of throwing everything into a vector store and praying for retrieval, it builds explicit nodes and edges that you can walk to surface relationships like 'this RFC ↔ this commit ↔ this meeting decision'. It runs on-device by default with an optional cloud mode.

The differentiator versus generic RAG stacks is the incremental graph maintenance and the local-first posture: when a file changes, only affected nodes and edges update, so the corpus stays coherent at millions of files without re-embedding everything. That makes it interesting for engineers and researchers who want long-horizon memory over a private corpus rather than a chatbot wrapper. As of this writing the project is in waitlist / early-access; the marketing copy leans heavily on MIT licensing and 'free forever' framing, so expect a community-driven OSS release with cloud add-ons later rather than a polished SaaS today.