HN Reader

OntoCast is a Python tool that extracts semantic RDF triples from unstructured documents (PDF, JSON, Markdown, text) using LLMs guided by ontologies.

It co-produces both the fact graph and the ontology, refining both through multi-stage critique and validation. This makes it domain-agnostic and useful even without a predefined schema.

Includes a /process endpoint (MCP-compatible) that returns RDF triples and ontology in Turtle. Supports entity disambiguation, semantic chunking, and KG aggregation. Can run locally or via REST, using OpenAI or local models via Ollama.

Repo: https://github.com/growgraph/ontocast

Docs: https://growgraph.github.io/ontocast