So I’m new to this and i need to know how to setup a batch api.
Can i run it locally from my laptop for testing? Is it actually cheaper?
Feel free to share anything. Also enlighten me about how this works.
thanks in advance
The “Chat Completions” and “Responses” endpoint are ways to have an AI produce an output for your input (to ask a question).
Batches, as a service provided by OpenAI, allow you to submit a special file containing the plain JSON RESTful request bodies of multiple API calls, as single lines, in a format called JSONL (json-lines).
You upload the file to the storage in “files” endpoint by API, and then produce the job to be run with the “batch” endpoint against that file as an API call also. Then monitor progress.
Then, there is a 24 hour turnaround to try to fit those calls in, at a 50% reduced price. They sometimes don’t complete in the 24 hours and get cancelled, other times can be much quicker. You can download your results file.
OpenAI has a dashboard interface for starting the jobs instead of making API calls, but the format itself means you will need to be versed in constructing http call bodies to the OpenAI API (not using an OpenAI library/module SDK normally).
