Agent vector databases

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

In Budibase, vector databases are configured in AI Configs and then assigned to each Agent in the Knowledge tab.

Supported provider

Currently we only support pgvector as a provider.

Configure a vector database

1. Open Workspace Settings
2. Go to AI Configs
3. Click on the Embeddings tab
4. In Vector databases, click Add database
5. Enter connection settings:
   • Name
   • Provider (defaulted to pgvector)
   • Host
   • Port
   • Database
   • User
   • Password
6. Save and confirm the database appears in the list

Attach a vector database to an Agent

1. Open your Agent
2. Go to the Knowledge tab
3. 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
• Agent tools and permissions
• Agent troubleshooting