PKParth krishan Goswamiinparthkg.hashnode.dev·Sep 21 · 8 min readDay 19: app.py - From Notebook to src/, and the Index That Went StaleMy RAG pipeline told me it couldn't find anything about retrieval-augmented generation. I had just added the original RAG paper to it. Getting to that bug meant first turning four days of notebook cel11M
PKParth krishan Goswamiinparthkg.hashnode.dev·Sep 20 · 7 min readDay 18: pdf_loader.ipynb — Adding Groq, and Teaching RAG to Say NoMy first call to an LLM failed before a single chunk reached it. The model the course uses has been retired. Once that was fixed, I found a bigger problem: my RAG pipeline confidently answered a quest12AM
PKParth krishan Goswamiinparthkg.hashnode.dev·Sep 19 · 5 min readDay 17: pdf_loader.ipynb — ChromaDB, a Retriever, and a Lying ScoreMy retriever told me nothing in my four PDFs was about hard negative mining. Page 4 of the embeddings report is about exactly that. On Day 16 I turned 359 chunks into 384-dimension vectors. Today they12AM
PKParth krishan Goswamiinparthkg.hashnode.dev·Sep 17 · 5 min readDay 16: pdf_loader.ipynb — 64 Pages, 359 Chunks, 384 Dimensions64 pages went in. 359 chunks came out, and every chunk is now a list of 384 numbers. On Day 15, loaders turned my PDFs into Document objects. Today they get cut into pieces and turned into vectors, th00
PKParth krishan Goswamiinparthkg.hashnode.dev·Sep 16 · 5 min readDay 15: document.ipynb — What RAG Fixes and LangChain's DocumentI thought RAG was a kind of model. It's a pipeline, and its first half never touches an LLM. Today I started Krish Naik's RAG crash course. Before any embeddings or vector databases, I had to learn ho00
