---
updatedAt: 2026-06-08T11:16:47.000Z
agentTools:
  projectIndex: https://docs.budibase.com/llms.txt
---

# Agent vector databases

Vector databases store embeddings and enable semantic retrieval for Agent knowledge.

In Budibase, vector databases are configured in **Workspace Settings > Connections > Knowledge bases** and then assigned from an Agent's **Operations** panel.

## Supported provider

Currently we only support pgvector as a provider.

## Configure a vector database

1. Open **Settings**
2. Go to **Connections > Knowledge bases**
3. In **Vector databases**, click **+ Vector database**
4. Enter connection settings:
   * Name
   * Provider (defaulted to pgvector)
   * Host
   * Port
   * Database
   * User
   * Password
5. Save and confirm the database appears in the list

## Attach a vector database to an Agent

1. Open your Agent
2. Open the **Operations** panel and select the operation you want to configure
3. In the **Knowledge** area, select a **Vector database**
4. Select an **Embedding model**
5. Save

After both are selected, you can tune:

* **Minimum similarity**
* **Chunks to retrieve**

You can then add files/knowledge content in the same area to support retrieval.

## Related guides

* [Agent embedding models](https://docs.budibase.com/docs/agent-embedding-models)
* [Agent tools and permissions](https://docs.budibase.com/docs/agent-tools-and-permissions)
* [Agent troubleshooting](https://docs.budibase.com/docs/agent-troubleshooting)
* [Connections](https://docs.budibase.com/docs/connections)