Returns 200 once both models have finished loading and the warmup pass has completed; 503 before that. Since main.cpp loads and warms up models synchronously before Drogon starts accepting connections (see Architecture → main.cpp), in practice this endpoint returns 200 from the moment the server accepts any connections at all — there’s no window where the server is reachable but this reports false.
​
boolean
true once InferenceEngine::isReady() is set, which happens at the end of a successful initialize() call.
Gate your orchestrator’s readinessProbe on this endpoint, not /v1/health — see that page for why the distinction matters. In practice, given the synchronous startup described above, if the process is failing to load models it will have already exited with status 1 (see Configuration → model paths) rather than serving 503 — so this endpoint mostly matters for orchestrators that want to hold traffic back during the brief warmup window on a slow start.