How I created Bookshelf : An open source AI-powered personal knowledge base
summary
Bookshelf is an open source RAG app that actually exists and runs locally using Streamlit, LlamaIndex, Chroma, LangChain and pandas, with optional Ollama-hosted inference models. It stores embedding model in Chroma metadata so queries use the correct embeddings, supports token/sentence/semantic and hierarchical node parser chunking, implements an automerging retriever plus cross‑encoder and BGE re‑rankers, and shows GPU/CUDA setup commands and performance tradeoffs for local evaluation.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 3:05Embedding models
- 4:13Matching and metadata
- 5:45Advanced RAG strategies
- 8:29Optimization techniques
- 11:32Local execution and wrap up
related talks
10mBooks reimagined: AI to create new experiences for things you know
31mSponsored Session: Building a Production Open Source RAG Application using LlamaIndex and Cassandra
41mFrom RAGs to riches: Build an AI document interrogation app in 30 mins
27mAI Demystified Unveiling the Intelligence Behind the Machine
27mFrom RAGs to riches: Build an AI document interrogation app in 30 mins
40mCreating Interactive GenAI applications powered by Streamlit, open source LLMs
40mRolling out Enterprise AI: Tools, Insights, & Team Empowerment
29mUsing LangChain and LlamaIndex to customize LLMs...