cancel
Showing results forย 
Search instead forย 
Did you mean:ย 
Databricks Free Edition Help
Engage in discussions about the Databricks Free Edition within the Databricks Community. Share insights, tips, and best practices for getting started, troubleshooting issues, and maximizing the value of your trial experience to explore Databricks' capabilities effectively.
cancel
Showing results forย 
Search instead forย 
Did you mean:ย 

Questions about the Knowledge Assistant during agentBrick

Remy
New Contributor II

hello.
I'm testing agentBricks right now. As I'm testing, I have a few questions that I'd like to ask.

1) Is it possible to input the schema in a format with input and output?

2) In the free trial, when I create a vector search index and utilize it as a knowledge source, I get the following error.

Maximum number of vector search endpoints per workspace exceeded quota of 1

Is it impossible to create a vector search index in the free trial and utilize it as a Knowledge source to build the Knowledge Assistant?

3) Why do I get the following error? I built Knowledge Assistant by setting UC Files as the Type of Knowledge source.

Vector index type DIRECT_ACCESS is not supported for this workspace.

#agentbricks

4 REPLIES 4

Martser
New Contributor II

Did you ever figure this out? Also running into

Vector index type DIRECT_ACCESS is not supported for this workspace.

jamesl
Databricks Employee
Databricks Employee

Hi @Martser and @Remy ,

For the "UC Files" type of knowledge source you want to point to a volume with these file types: TXT, PDF, MD, PPT/PPTX, DOC/DOCX. I assume you were pointing to a location with other files or assets?

In general the free trial only supports DELTA_SYNC for vector search (automatically syncs with a source Delta Table, automatically and incrementally updating the index as the underlying data in the Delta Table changes). I don't believe that is mentioned specifically in documentation, but you can find other limitations here: https://docs.databricks.com/aws/en/getting-started/express-setup

If you need full functionality I would suggest contacting a team member if you don't already have a contact.

As for the other questions:

1) Is it possible to input the schema in a format with input and output?

What type of use case do you have? The information extraction and custom agent have some flexibility around this, but the knowledge assistant uses a standard chat interface. The product is still in Beta though so there may be changes in the future.

2) In the free trial, when I create a vector search index and utilize it as a knowledge source, I get the following error.

Maximum number of vector search endpoints per workspace exceeded quota of 1

Using Agent Bricks with vector search currently includes creating a new vector search endpoint, even if you are pointing to an existing vector search index (i.e., you cannot reuse the existing endpoint with your new agent). This is causing you to exceed the quota of 1 endpoint. You may find other relevant details around functionality and pricing here: https://www.databricks.com/product/pricing/beta-products

I would try deleting your existing vector search index and let Agent Bricks create one for you with the "UC Files" type  for knowledge source. 

I hope that helps, but let me know if you have further questions. If this solves the issue, please use the "Accept as Solution" button to let us know!

- James

Martser
New Contributor II

I am seemingly still getting the 

Vector index type DIRECT_ACCESS is not supported for this workspace.

issue. the folder only contains .txt files. I am a premium user. Seems as if even though I have premium the workspace still 'thinks' its free trial

jamesl
Databricks Employee
Databricks Employee

Hi @Martser,

Try checking the docs on other limitations and requirements that may be preventing the agent brick from working. Here are some things you could check...

Verify prerequisites and region
Ensure your workspace has Agent Bricks preview enabled, serverless compute, Model Serving, partner-powered AI, and Unity Catalog, and that youโ€™re in a supported region for Knowledge Assistant.

Use a supported Vector Search index type
Knowledge Assistant supports Vector Search indexes that use the Databricks-managed embeddings with the databricks-gte-large-en model. It does not support self-managed or direct-access indexes. If youโ€™re selecting an existing index, make sure itโ€™s a Delta Sync index configured with gte-large-en; avoid Direct Access indexes.

Using UC Files
Place your files in a Unity Catalog volume (txt, pdf, md, ppt/pptx, doc/docx). Files larger than 50 MB are skipped. When you create the agent with UC Files, Knowledge Assistant will handle parsing and build a supported index under the hood.

Check workspace limitations
Knowledge Assistant is not supported in workspaces with Enhanced Security and Compliance enabled.

After checking the above, if you made any changes, try re-creating the agent. I hope that helps, but please let me and the team know if it doesn't resolve the issue.

-James

Join Us as a Local Community Builder!

Passionate about hosting events and connecting people? Help us grow a vibrant local communityโ€”sign up today to get started!

Sign Up Now