Hey, as you can see in the image, I have a spreadsheet as a tool to the OpenAI message model node. However, even though I’ve mentioned in its prompt to fill out the spreadsheet (with damages from a PDF), it still doesn’t do it 50% of the time. 50% of the time it does, and 50% it doesn’t, please help with this.
Thanks!
Improve the request with clear instructions and examples. If you can share it for analysis, the request should be structured like tools: {name: …, parameters: {…}} if it is in explicit tool call mode.
For example, prompt user:
You should use the “Comparison” spreadsheet tool to record the damage data extracted from the PDF. Always use the tool, even if the data is partial or appears redundant. Do not return anything in the response message; simply activate the tool.
Just want to add to @Erick_Torres answer. A bigger (smarter model) usally fixes this issue if you notice the issue still sticks after adding instructions.
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.
