Hi everyone,
I've created an academic tutor agent designed to help students with their assignments. The main knowledge source for this agent is a large (1000 page) pdf of a social studies textbook. It's important that the agent only refer to the content of the textbook when answering student's questions and that it not use other soruces (including web and "general knowledge").
The trouble is, when I shut off, "Use general knowledge" and "Use information from web" in the settings, the agent frequently gives the following error in response to queries: “no information was found that could help answer this," despite the fact that the relevant information is present in the textbook.
I was troubleshooting with Copilot and it suggested the following:
1. Verify the pdf is actually text (possible OCR issue). I checked and I CAN select, copy, and paste text from the textbook, so I don't think that's the issue.
2. Consider breaking the pdf into smaller chunks (30 to 80 pages) with clear headers, section titles, and a consistent structure -> I haven't tried this, yet.
Is there anything else I should consider doing to fix this problem?