CoursesBuilding AI Applications
Models and the API
The request an application actually sends: messages and roles, the parameters that change the answer, the token budget that sets the bill, and the errors a production client must survive.
- Lessons
- 7
- Exercises
- 40
- Minutes
- 42
- 1
The Request
After this lesson you can read a chat API request as what it is: a stateless call that carries the whole conversation every time, with three roles that mean different things.
DebugMultiple choiceTrace - 2
Parameters That Change the Answer
After this lesson you can set temperature, max_tokens and the system message for a task on purpose, and read the stop reason to know whether you got a whole answer.
DebugFill blankMultiple choice - 3
Tokens, Budget and Cost
After this lesson you can size a request's token budget, explain why a long conversation costs more every turn, and estimate a feature's bill before it ships.
Code orderMultiple choiceTrace - 4
Errors, Rate Limits and Retries
After this lesson you can tell which API errors to retry, how to back off, and how to make a retry safe when the first attempt may have succeeded.
DebugMultiple choiceTrace - 5
Images and Documents as Input
After this lesson you can send an image or a document to a model, predict what it costs and what it will get wrong, and put the checks between what the model says it sees and anything that acts on it.
DebugCode orderMultiple choice - 6
Checkpoint: The RequestCheckpoint
Roles, parameters, budget and errors in fresh situations.
Fill blankMultiple choice - 7
Boss: The First EndpointBoss
One endpoint from first request to first incident: roles, parameters, budget, truncation, errors and a privacy near-miss.
DebugMultiple choiceTrace