This is an end-to-end AI RAG and sentiment system that processes crypto news.

1. Text data is extracted from website articles.
2. Extract metadata: publish_date, entities mentioned (coins, tokens, blockchains, bridges)
3. Split text into chunks and calculate embeddings.
4. Compute overall daily sentiment per entity
5. Store data in Tortoise ORM models in Postgresql, also use pgvector for storing vectors from embeddings

1. Setup FastAPI to create endpoints for inference and search
2. Implement Leucene or other search system for BM25 search
3. Implement hybrid search with RAG and BM25
4. Create search endpoint
5. Create QA endpoint with RAG and Reranking