I've been following the guide How to connect Watsonx to Watsonx Assistant on how to connect Watsonx to Watsonx Assistant and successfully connected my Watsonx.ai project to my IBM Orchestrate Assistant using an extension.
Now, I'd like to take the next step: I want to build an assistant that can answer questions based on documents I upload — including formats like .txt, .pdf, .docx, .yaml, and others.
.txt
.pdf
.docx
.yaml
My question is: Where can I upload these documents so that my assistant can access them and provide answers based on their content?Are there any best practices or recommended storage options (object storage, embedded document index, etc.) for this use case?
Also, are there any recommended ways to handle this without requiring a cloud account with billing (i.e., without adding credit card details)?
I'm looking for an approach that allows local file ingestion or free-tier options that support basic document Q&A.
I did a test recently and didn't find any free services, including ElasticSearch, Watsonx Discovery, and Vilmus, all requiring at least $500 for basic use. Finally, we used Pinecone as an external vector database.
Hi, we recommend to try it with our new agentic technology where we have ability to upload files directly with an embedded repo from the agent builder: https://www.ibm.com/docs/en/watsonx/watson-orchestrate/base?topic=agents-adding-knowledge#uploading-files-directly-to-the-agent