A Question-Answering API on a specific topic using RAG.
##1. Usage
There are three services in total, each in separate containers.
- weaviate : Local setup of vector store
- embedding_generator : To add any topic to knowledge base in its chunked + embedded format.
- qa_service : Main end-user facing service that takes user question and answers it based on context retrieved from vector store.
To test, run the following commands from app directory
docker compose up -d After four minutes of waiting ( poor , but conscious design choices..still poor overall ), try the following:
POST request to localhost:8081/generate_and_save_embeddings
-- Parses StreamingLLM.pdf, chunks it and stores text + embeddings in weaviateGET request to localhost:8082/answer
-- Default question "What is a KV cache" is used
Clean up after testing as follows:
docker compose down- *Note : Make sure the bloated images are removed after testing.