Correct way to build a database wide RAG chatbot #12712
Replies: 1 comment
-
For building a production-ready chatbot capable of handling a large and diverse document database, consider a multi-faceted approach that enhances both retrieval accuracy and efficiency:
This holistic approach addresses scalability, relevance, and accuracy challenges, making it suitable for a production environment. For implementation details, consider looking into resources like LlamaIndex's documentation on optimizing production RAG, which provides insights into decoupling retrieval and synthesis chunks, structured retrieval, dynamic retrieval strategies, and optimizing context embeddings. Sources
|
Beta Was this translation helpful? Give feedback.
-
We can create simple RAG chatbot powered by a retriever , or Agent with tools. However , what's the ideal way to deal with cases where user's database is large enough say 100s of files and they need to find an answer based off of a certain document , retrievers won't work the best(or would they?) or the better method is to create a query engine for each document(PDF,DOCX,etc) or there's a method that I might be missing out.
The goal is to build a production ready chatbot which can act as a personal assistant.
Thank you for your answer!
Beta Was this translation helpful? Give feedback.
All reactions