From ChatGPT User to RAG Implementer: A Developer's Journey
summary
A working RAG pipeline that indexes HTML docs into embeddings, stores them in PostgreSQL with pgvector, and answers queries via an LLM is implemented as a small Kotlin Spring app. It uses Kotlin + Spring AI with unstructured.io for parsing, OpenAI text embeddings (examples mention 1,586 dimensions and usual 100–300 sizes), and pgvector + OpenAI embeddings for similarity search (8,191 token chunking cautioned). Demo shows one HTML page becomes two DB rows with vectors and the system reliably answers document questions while prompting “I do not have such information” to avoid hallucinations.
AI-written summary · about
chapters (7)
- 0:00Introduction and my background
- 7:09What are embeddings?
- 17:16From embeddings to LLMs
- 22:15What is RAG?
- 28:00Building the RAG pipeline
- 34:15Demo and code walkthrough
- 41:57Summary and key takeaways
related talks
53mBuild RAG from Scratch
27mBuild a ChatGPT RAG Data Pipeline
41mFrom RAGs to riches: Build an AI document interrogation app in 30 mins
48mGoing Beyond Prompt Engineering
50mModular RAG Architectures with Java and Spring AI
50mBuild a ChatGPT RAG Data Pipeline
58mDemystifying GenAI: Build a ChatGPT App with Vector Store
51mEvent Driven AI: Supercharging ChatGPT with RAG & LangStream