A TypeScript-based RAG pipeline for local document querying. It uses nomic-embed-text, llama3, and Qdrant for on-device processing. Users can upload documents and ask questions through a chat interface, with answers sourced from the uploaded files.
From the README
LocalCortex Chat with your documents. Entirely on your machine. Upload PDFs, Word files, Markdown or plain text, then ask questions in a chat UI. Answers come with their sources, and when your documents don't cover a question, it says so. No cloud APIs, no API keys, nothing leaves your computer. Real answers from llama3 on a laptop CPU. The waits while it thinks are cut, and each caption shows the actual time taken. --- Why LocalCortex • Private by design. Embedding, search, generation and chat history all run in containers on your machine. Your files are never sent anywhere. • Answers you can check. Every answer shows the documents it came from as source chips. • Honest when it...
Skills
- TypeScript
- MDX
- CSS
- HTML
- JavaScript
