Snorkle

100% private, local document analysis with LLMs.

Features

• Deep Document Search

This is a fork of Patense.local with a modified system prompt for generic document searches. Sometimes ctl+F doesn't cut it and you need a broader or more robust way to find relevant text.

Tech

The core tech and features are the same. Break reference documents into chunks and ask an LLM if the content is relevant to a given query. Return a brief quote and the full text for verification. Uses Next.js on the frontend with a vLLM server.