Graphs when the path is known on the AI Agents track. If the happy path is a fixed sequence, use a state machine or graph. The model fills slots and chooses branches you defined. It does not invent new states. This is safer for refunds, bookings, and approvals.
This lesson assumes you already worked through Tasks that take many steps.
The idea in practice
Draw states, allowed edges, and which tool each edge may call. The model output is an edge name plus slot values. Illegal edges are rejected.
A concrete check
goal = {
'track': 'AI Agents',
'lesson': 'Graphs when the path is known',
}
checks = [
'input available at decision time',
'score matches the real decision',
'failure case written down',
]
print(goal['lesson'])
for item in checks:
print('-', item)
Run the sketch locally if you have Python. The printout is a reminder of the checks, not a trained model. Replace the strings with the real inputs from your own example before you treat it as a design.
What usually goes wrong
A free loop on a payment flow can skip the confirmation state because the model 'felt done'. When this happens, stop adding parameters or tools. Fix the check, the data, or the permission, then run the same example again.
What to write down
- The input you are allowed to use at decision time.
- The output and the score or pass rule.
- One failure you will test on purpose.
- What you will not claim the system can do.
Practice
Draw five states for 'change a delivery address' and mark which edge charges money. That edge needs a person.
Self-check
- Say Graphs when the path is known in one sentence that mentions an input and an output.
- Name the failure mode in this lesson and the check that would catch it.
Done when: you can explain this lesson without the page open, and you have a written failure case.