Ignoring a request while extracting structured output

API
1

Thanks in advance for your help.

I am using an LLM to parse into a JSON schema one or more shipments out of a plaintext request such as:

“Can I get a quote for 20 pallets of oranges, 10000 lbs, and 30 pallets of lemons, 10000 lbs, going from Chicago to Houston, pickup tomorrow?”

I want to capture the shipment data but do not want the model to waste time calculating a quote. How do I use prompts/JSON elements to direct the LLM to ignore the quote request and just provide the shipment data? Is it enough to simply leave the quote elements out of the schema? Or should I leave them in the schema so the model does not get confused?

I will be using program logic to generate the quote, so I do not need the LLM to do any quote estimation. I do not want to throw off logprobs or burn CPU by including the quote attempt in the LLM reasoning.

Here is the schema I am filling:

[ {
“pickupStreetAddress” : null,
“pickupStateCode” : “IL”,
“pickupPostalCode” : “60601”,
“pickupDate” : “2024-10-05”,
“pickupCountryCode” : “US”,
“pickupCity” : “Chicago”,
“deliveryStreetAddress” : null,
“deliveryStateCode” : “TX”,
“deliveryPostalCode” : “77001”,
“deliveryDate” : null,
“deliveryCountryCode” : “US”,
“deliveryCity” : “Houston”,
“confidence” : 98,
“cargoList” : [ {
“cargoWeightInPounds” : 10000,
“cargoNMFCClass” : “65”,
“cargoName” : “Oranges”,
“cargoContainerName” : “Pallets”,
“cargoContainerCount” : 20
}, {
“cargoWeightInPounds” : 10000,
“cargoNMFCClass” : “65”,
“cargoName” : “Lemons”,
“cargoContainerName” : “Pallets”,
“cargoContainerCount” : 30
} ]
} ]