Show HN: I built a Privacy First local AI RAG GUI for your own documents
github.comByte-Vision is a privacy-first document intelligence platform that transforms static documents into an interactive, searchable knowledge base. Built on Elasticsearch with RAG (Retrieval-Augmented Generation) capabilities, it offers document parsing, OCR processing, and conversational AI interfaces.
Error at step 4: cd frontend && npm install && cd .. doesn't grok on my mac.
How are you dealing with documents that exceed the context window? Chunking an db vectorization? One issue with some approaches is the Paragraph 40 that references Paragraph 6 in a document but the LLM doesn’t really have direct capability to “remember” that reference.
I chunk the documents and use Elasticsearch to store the vectors. On a laptop with 8GB of GPU I can have a pretty large context window and not hallucinate.
Nice! Will have to give it a go, looks interesting. Thanks for sharing.