[juanceresa]

  pip install sift-kg
  sift extract ./docs/
  sift build
  sift view

That's the whole workflow. No code, no database, no Docker.

I built this while working on a forensic document analysis platform for Cuban property restitution cases. Needed a way to extract entities and relations from document dumps and get a browsable knowledge graph without standing up infrastructure.

Uses any LLM provider (OpenAI, Anthropic, Ollama) via LiteLLM. Human-in-the-loop entity resolution — the LLM proposes merges, you approve or reject.

The repo includes a complete FTX case study (9 articles → 373 entities, 1,184 relations). Explore the graph live: https://juanceresa.github.io/sift-kg/graph.html

[digdugdirk]

Looks very useful and very cool! Just a heads up - your graph loads terribly on mobile (android + Firefox), it's just a skinny strip in a container at the top of the page.

[juanceresa]

Thanks! Yeah the pyvis viewer isn't mobile-friendly — it's built for desktop browser exploration. I should add a note about that. Appreciate the heads up.