Open Source Document Chat
ParsePal turns PDF exploration into a small, inspectable learning system.
Upload a document, generate embeddings locally, get an AI summary, and keep chatting against the retrieved context. The project is intentionally open so the product, prompts, and tradeoffs are easy to study.
What It Does
ParsePal is a focused interface for chatting with a single PDF at a time. The current flow ingests a document, stores embeddings in ChromaDB, generates a concise AI summary, and keeps conversation history in the app so each upload becomes an inspectable thread.
Why It Exists
This project is for learning in public. That includes RAG plumbing, release discipline, layout decisions, and the less glamorous work of making a small tool feel coherent across empty, active, and public-facing states.
Learning Principles
Build with enough structure to learn from every release.
Learn in public through a working product, not isolated experiments
Keep the stack understandable enough to modify without ceremony
Ship UI changes that feel intentional, not generic demo polish
Follow The Work
The easiest way to track progress is through the repository and the release notes. The changelogs page in this app mirrors that release history so the product can explain itself without leaving the interface.