I’m trying to build a Claude Desktop/ChatGPT-like PDF analysis workflow where I can ask questions about the entire PDF context (text + images). I tried using an HTTP PUT node which works but isn’t as efficient as the AI Agent node that can connect to memory, MCP, and tools. When I pass PDF binary data to the AI Agent node, I can convert base64 to binary successfully, but including the binary in the prompt gives me NaN errors. Has anyone successfully passed PDF binary data to AI Agent nodes for full document analysis? Any tips on proper binary handling or alternative approaches would be appreciated!
1 Like
Hi @exingous
Have you tried converting the pdf to text, then giving the model that as the prompt/system message? Here’s an example.
If this response helped you, please click the heart to show that it is useful
