docint

docint ingests heterogeneous collections — text, markdown, PDF, DOCX, HTML — then chunks and embeds them with a local ONNX sentence-transformer model. It enriches each chunk with summaries, entities, keywords, and taxonomy, and answers queries using hybrid dense-vector plus BM25 retrieval.

No cloud services. No LLM API calls. Everything runs on the machine in front of you.

Keeps a growing pile of documents findable.