🧵 Vector stores for RAG & Agent Context: This thread will be for discussing and sharing about building with Vector stores for RAG and agent context

This event is about building with any vector storage system, so feel free to take a look at any you've been interested in, Weaviate, pineconde, chroma, qdrant, Milvus etc!

Try to build something by May 1st if you want to share it in the stream!

After last week’s “Vector Stores for RAG and Agent Context” meeting I revisited an all-local RAG/Agent example I have based on CrewAI, LangChain and LanceDB. Thank you @powerful-horse-58724. This is currently running on an M1 Macbook Pro with 16GB of memory, which limits the size of the LLM models. • Crew AI (simple researcher/report RAG example) • LangChain Community LanceDB vector store integration (CrewAI tool) • Ollama models: qwen2.5-coder:7b-instruct and granite-embedding:30m (huh, coder?) • Script to chunk and load a handful of Project Gutenberg musical instrument texts. • Run via command line to generate a 1-pager about i.e. “What are the kinds of musical instruments in an orchestra?” I needed to spend some time with configuration and prompts to make sure the tools were called and that the reports were primarily based on the research. I enjoyed the exercise and added a SPARQL graph query for the next discussion.

Thats awesome!