TEST BENCH — ENGINE INVITATION
RUN THE AGENT.
CHECK ITS WORK.
Every decision below is the real deterministic engine output (double-run verified). Compare it against the submitted output.csv, or add your own custom test cases and score them with the same invariants the validator uses.
RUN A REQUEST
LOADING ENGINE BASELINE…
LIVE ENGINE CHECKING…
The same real Python pipeline, executed on demand. Change the requested amount, move the request date, or toggle partial payment and the engine decides again — then diff the answer against the submitted baseline. Start it with python3 code/serve_api.py; without it the bench stays on baseline.json.
PROBING THE LIVE ENGINE…
CUSTOM TEST CASES
REQUEST ID
REQUESTED AMOUNT
EXPECTS
FLOOR HOLDS?
NO CUSTOM CASES YET. EXAMPLE: EXPECT request_28 TO BE not_affordable WITH THE FLOOR HOLDING.