Guidance Required on Image-Heavy PDF Processing in AI Agent Studio

Summary:

Use Case

Our requirement is to analyze PDFs containing product/component images, derive component composition percentages through visual inspection, and drive downstream business actions based on the extracted information.

Content (please ensure you mask any confidential information):

Based on Oracle documentation, we evaluated:

• RAG Document Tool
• Document Processor Node
• Chat Attachment Reader

Since our use case is transactional rather than knowledge retrieval focused, we concentrated on Chat Attachment Reader and Document Processor Node.

1. Chat Attachment Reader

Observations:

• Successfully extracts textual content from PDFs, but does not consistently identify composition percentages that are available only through images.
• Results vary across executions of the same document.