MediumNLPnlp-267

Tool Error Retry Policy

Problem

Simulate a retry policy over attempt outcomes. After a failure that is in retryable, retry only while retries_used < max_retries; stop on 'success', on a non-retryable outcome, or when retries are exhausted.

Return [attempts, status] where status is the last outcome, 'success', or 'exhausted' when outcomes run out mid-retry.

Hint

max_retries counts retries after the first attempt, not total attempts.

Test cases

case 1

in: [["timeout","timeout","success"],["timeout"],2]

expected: [3,"success"]

case 2

in: [["timeout","timeout","success"],["timeout"],1]

expected: [2,"timeout"]

case 3

in: [["auth"],["timeout"],3]

expected: [1,"auth"]

case 4

in: [[],["timeout"],1]

expected: [0,"exhausted"]

case 5

in: [["timeout"],["timeout"],3]

expected: [1,"exhausted"]

Starter codePython

def tool_retry_policy(outcomes, retryable, max_retries):
    # Your code here
    pass

5 test cases · run your code against them in the editor