Agent test runs return 402 on exhausted balance; model_override tier check enforced
Agent test runs return 402 on exhausted balance; model_override tier check enforced
Every agent test-run endpoint admits the run against the workspace’s remaining
credit and its spending limits before the run is created, and documents 402
for a depleted balance or an exhausted spend limit:
POST /v1/agents/tests/{test_id}/runsPOST /v1/agents/{agent_id}/tests/runsPOST /v1/agents/tests/runs/batchPOST /v1/agents/tests/suite-runs/{suite_run_id}/resubmit
The behavioural change is on the single-run path,
POST /v1/agents/tests/{test_id}/runs, which previously skipped the admission
check its three sibling paths already applied - so a run started on a zero
balance, past a tripped per-key spend cap, or past a workspace spend budget.
The other three already refused; their 402 is now documented.
Model access is enforced on every run path. An explicit override - the
per-test model_override or the run-level config_override.model - must name
a catalogued model the workspace’s plan includes, and returns 402 before any
run is queued when it does not. Previously an explicit override was passed
through with no catalog check and no plan-tier check, so a test could run on a
model above the plan. An over-tier model stored on the agent itself is clamped
to the platform default rather than refused, matching what a live call does, so
a test still predicts production. A resubmit re-checks the parent suite’s
stored override against today’s plan instead of replaying it.