Create knowledge bases

In Azure Logic Apps Automation, create a knowledge base as an external context source that an agent can search during reasoning. Rather than rely only on a model’s training, the agent gets the relevant information from your files, docs, code, and data to ground its answers.

The following items are under development:

Document upload size and format limits.
Per-source token budget controls.
Granular permissions on which environment members can attach a source.

If you encounter problems, report a bug so your feedback can help shape future releases.

Requirements

A Microsoft work or school account in the same Microsoft Entra tenant as the environment creator-owner.

Your account must exist in the same tenant so the environment creator-owner can add you to the environment. You don’t need an Azure subscription to create apps and workflows in an automation environment.
Access to the Azure Logic Apps Automation portal.
Contributor or Author role on the environment resource to create knowledge bases.

The environment Reader role doesn’t have enough permissions to create knowledge bases.

If you don’t have environment access, contact the environment creator-owner so they can add you with the required permissions.

1. Create a knowledge base

Before you can attach a knowledge source to your agent, other than files to upload, create the knowledge base or add an existing knowledge source to your environment.

In the Azure Logic Apps Automation portal, open your environment.
On the environment sidebar, select Knowledge.
Select Add file or Add knowledge source > Add file.
1. Enter a name and description for your knowledge base.
2. Drag or browse and select the files you want. Select Add file.
Continue to Attach a knowledge source to your agent.

Add another knowledge base

To add another knowledge base when you have existing ones, follow these steps:

In the Azure Logic Apps Automation portal, open your environment.
On the environment sidebar, select Knowledge.

The portal opens your existing knowledge base.
In the Knowledge section, select Add file.
In the Add file box, for Knowledge Hub, select Create new.
1. Enter a name and description for your knowledge base.
2. Drag or browse and select the files you want. Select Add file.

2. Attach a knowledge source to your agent

In your workflow, on the designer, select the agent action.
In the agent information pane, select the Knowledge tab.
Select Add files or Add knowledge source.
- Add files: Enter a name and description for your knowledge base. Drag or browse and select the files you want. Select Add file.
- Add knowledge source: Select the knowledge sources you want.
When you’re done, close the pane.

At the next run time, the agent retrieves related information from the source before calling the model. The agent queries each attached source by using agent’s instructions and inputs. The agent adds the top results to the model’s context. The model chooses which results to cite, if any.

Troubleshoot information retrieval problems

To debug problems related to agent responses based on stale data or not finding the correct information, open the workflow run history so you can examine the query and retrieved information. Each time that the agent retrieves information is a step in the agent’s iterations. These steps appear as entries under the agent action’s inputs and outputs:

Inputs: Shows the query used by the runtime. The model often rewrites the agent’s instructions.
Outputs: Shows the retrieved information with a relevance score per result.

Usually, these problems arise due to chunking or queries with rewritten instructions.