Engine health & verification

Confirm that the API, dependencies, and bundled compute runtime are healthy through supported HTTP probes.

Algenta starts and supervises its bundled runtime as part of the deployment. Health checks verify that integrated service; they are not instructions for managing a second process.

Health endpoints

Endpoint	Authentication	Tells you
GET /v1/health	none	API process liveness.
GET /v1/health/ready	none	Readiness of critical dependencies. A database failure returns 503.
GET /v1/infrastructure	API key	Executes a compute probe and reports node status and runtime mode.
GET /v1/runtime/manifest	API key	Lists the runtime modules included in this deployment.
GET /v1/libraries/health	API key	Runtime-library health when that optional HTTP surface is enabled.

Compute probe

curl -sS "$ALGENTA_API_URL/v1/infrastructure" \
  -H "Authorization: Bearer $ALGENTA_API_KEY"

The current contract reports:

Field	Healthy value	Failure value
primary node status	live	error
engine_mode	binary	binary when the probe failed, or missing when the artifact is absent
engine_version	the executing runtime version	unknown when the probe fails
probe_latency_ms	measured latency of this probe	null when the probe fails
uptime_30d_pct, avg_latency_ms, p95_latency_ms, throughput_rpm	null; historical metrics require an external metrics store	null

engine_mode: "binary" means the bundled compiled runtime is present; status: "live" proves it executed the probe successfully. Treat both fields together as the readiness signal.

End-to-end smoke test

curl -sS -X POST "$ALGENTA_API_URL/v1/simulate" \
  -H "Authorization: Bearer $ALGENTA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"mode":"auto","scenario":{"variables":{"demand":{"low":80,"high":120}},"objective":"maximize_margin"},"runs":100,"seed":7}'

A successful response contains real summary statistics. Use repeated requests to evaluate warm performance; do not mix image download, service boot, or the first connection into a hot-path latency measurement.

Failure handling

The production deployment fails closed when its required compute runtime cannot execute. It must not silently route to a Python implementation. If the infrastructure probe reports missing or error:

1. Confirm the deployment uses a current signed Algenta image or package.

2. Inspect the API startup logs for the first runtime verification or compute error.

3. Redeploy the same integrated service after correcting the artifact or platform mismatch.

4. Rerun both the infrastructure probe and deterministic simulation.

Do not install or operate an additional runtime service beside Algenta.

Related

Verify the bundled engineMonitoring & observability