Universal Containers (UC) users are complaining that agent answers are not satisfactory. The agent is using PDF files as a knowledge source. How should UC troubleshoot this issue?

Question

Options

• AAnalyze the data mapping between source fields and Data Cloud object fields.
• BCheck that the agent has the PDF file field permission access for the data library.
• CVerify the retriever's filter criteria and data source connection.

• A
  15% (3)
• B
  10% (2)
• C
  75% (15)

Explanation

When an Agentforce agent uses PDF files as a knowledge source and returns poor answers, the most likely culprits are retrieval-layer issues: the retriever's filter criteria may be too broad or too narrow, causing irrelevant chunks to be returned or relevant ones to be missed, and the data source connection may be misconfigured or pointing to incomplete data. Verifying both ensures the agent is actually grounding on the right content from the PDFs. Option A (data mapping to Data Cloud objects) is relevant to structured data ingestion, not PDF knowledge libraries. Option B (field-level permissions on a PDF file field) is not a meaningful concept for document-based knowledge sources.

No community discussion yet for this question.