watsonx.ai

 View Only

Q&A with RAG Accelerator on watsonx.ai

By Thomas Schaeck posted Wed September 04, 2024 10:02 PM

  

We released our new watsonx Q&A with RAG Accelerator 1.3 update for watsonx.ai with watsonx Discovery / ElasticSearch Enterprise or watsonx.data Milvus for vector-indexing and retrieval.

New watsonx.data Milvus Integration
- Convert, split, vectorize docs into Milvus collection with E5 embeddings
- Q&A-with-RAG Python Function can retrieve from Milvus
- Q&A + user feedback logging to Milvus with vectors for questions
- User feedback analytics notebook can analyse log data from Milvus

New Prompt Templates for Llama 3.1 8b and 70b models

Simplified Parameter Sets to streamline configuration steps and included required libs in Software Spec assets

You can try it here https://lnkd.in/eS4N4CwA with watsonx.ai on IBM Cloud or get the project template ZIP for use with watsonx.ai software.

You can start with the accelerator as it is, and configure parameters and customise code as needed for your solution.

Kudos to our watsonx Accelerators, watsonx Development Solutions Architecture CoE, and watsonx.ai teams for this great new update!

No alt text provided for this image


#watsonx.ai


#GenerativeAI
1 comment
32 views

Permalink

Comments

Kenji Matsuo
Wed September 11, 2024 05:27 AM

Hello,

Thank you for providing the helpful project samples.
I am currently trying out the "Q&A with RAG Accelerator" in the TechZone environment.
However, I encountered an error while running the notebook "Process and Ingest Data into Vector DB" specifically in the "Inserting Documents using Langchains Vector Store" section when executing "vector_store.add_documents()"
Could you advise on how I can receive support for this issue? I am an employee of an IBM business partner.

(Update)
With the November 14 update, the error was resolved, and I was finally able to complete the hands-on session. Thank you!