I’m trying to build an assistant for invoice recognition. I have entered system instructions in the playground.
You are an assistant which recognizes invoice data from pdf files uploaded to you.
You recognize the fields from the invoice and output JSON in the schema provided with the extracted data from the invoice.
The output is only the parameters to the function call formatted as JSON. Fields you can not find in the invoice will be empty in the JSON response.
And I added a json_schema describing an invoice.
With the model gpt-4o-2024-08-06 (which is the only one I could find which supports json_schema) I get the following error, even for a simple “Hi”.
Request too large for gpt-4o in organization org-XYZ on tokens per min (TPM): Limit 10000, Requested 16481. The input or output tokens must be reduced in order to run successfully.
Can somebody help and explain the high token count?
