Default versus additional Data Libraries
Key Takeaway
Each feature can use only one data library at a time. To give an agent more than one library, you first create an ensemble retriever in Data 360 that combines the libraries' retrievers. Each library also has one data source that you can't change after you choose it.
An Agentforce Data Library connects AI features, such as Agentforce agents, to your trusted unstructured or semi-structured data, so their answers are grounded in that data. You can create many data libraries, but two limits shape how you plan them. Each library has one data source. Each feature uses only one library at a time.
What a data library does
A data library turns large sets of information, such as Knowledge articles or uploaded files, into searchable data in Data 360. Data libraries require Data 360 to be set up. Data libraries also consume Data 360 credits.
You can create and configure data libraries from the Agentforce Data Library setup page or from Agentforce Builder.
When you save a library, Salesforce automatically creates these parts of the pipeline, so you don't build them by hand:
- A data stream
- Data lake objects and data model objects
- The mapping between the data lake objects and the data model objects
- A search index
- A retriever
A new library isn't usable right away. Salesforce creates the search index only after the earlier pipeline steps succeed and are active. Chunking time depends on the number, size, and complexity of the articles or files, and on how many Knowledge fields you select. A status card on the library setup page tracks each stage.