Failure Cause Classifier

Apply transparent keyword rules to a redacted failure record and route it to a conservative next action.

Input and output

Input is the normalizer output or an object with error_message, failed_node, and status. Output is a failure class, matched rules, and a suggested path; it never claims root-cause certainty.

Execution and boundaries

Run python scripts/classify_failure.py --input examples/input.json --output examples/output.json. Rules cover timeout, rate-limit, validation, missing-field, unauthorized, mapping, and unknown patterns. No API or log source is contacted.

Acceptance

The sample returns transient with a retry candidate; an unrecognized message returns unknown and human review.