Looks good but seems like a specific implementation. This is what I am looking for:
Vectorizer base class, has all the not model/lib specific stuff, leaving run_model function not implemented
VectorStore base class, has ingest_docs, get_k, put functions.
Then the llm class has a self.rag that uses the above functions to retrieve stuff, and pass it to the llm
Then using the above base classes implement a langchain-faiss rag or any other combo.
Looks good but seems like a specific implementation. This is what I am looking for:
Then using the above base classes implement a langchain-faiss rag or any other combo.