SharkSkin/api-error-retry-classification

import api_error_retryability_classifier as classifier

# To make this example runnable, create a data.jsonl file
# in the same directory as api_error_retryability_classifier.py
# with content like:
# {"input": "HTTP 500 Internal Server Error", "label": "retryable", "note": "..."}
# {"input": "HTTP 404 Not Found", "label": "permanent", "note": "..."}

# Load all classification data
all_errors = classifier.load()
print(f"Total error classifications loaded: {len(all_errors)}")

# Get classification for a specific error
classification_500 = classifier.get_classification("HTTP 500 Internal Server Error")
print(f"HTTP 500 is: {classification_500}") # Expected: retryable

# Check if an error is retryable
is_429_retryable = classifier.is_retryable("HTTP 429 Too Many Requests")
print(f"Is HTTP 429 retryable? {is_429_retryable}") # Expected: True

is_404_retryable = classifier.is_retryable("HTTP 404 Not Found")
print(f"Is HTTP 404 retryable? {is_404_retryable}") # Expected: False

{"input": "HTTP 500 Internal Server Error", "label": "retryable", "note": "Generic server-side issue, often transient. Retrying might resolve it."}
{"input": "HTTP 503 Service Unavailable", "label": "retryable", "note": "Server is temporarily unable to handle the request due to overloading or maintenance. Retry after a delay."}
{"input": "HTTP 429 Too Many Requests", "label": "retryable", "note": "Rate limit exceeded. Client should wait and retry with exponential backoff."}