Platform and health
On this page
These endpoints describe the data plane's health and readiness, and the sample data the Mortar website and console are built around. Responses that hold sample data say so with "illustrative": true, so nothing can mistake them for measurements of your traffic.
Check health
/health
Reports whether the data plane is up and what it serves. It's the one endpoint outside /v1. No key is needed.
curl https://api.mortar.one/health{
"status": "ok",
"version": "1.4.0",
"uptimeSeconds": 86400,
"inFlightGenerations": 2,
"sessions": 14,
"rooms": 31,
"servedModel": "Qwen/Qwen2.5-7B-Instruct",
"servedContextTokens": 4096,
"upstream": {
"mode": "mock",
"target": null,
"label": "mock"
}
}Response
| Field | Type | Description |
|---|---|---|
status |
string | ok whenever the data plane can answer. |
version |
string | The build's version: a release number such as 1.4.0, the module version of an unreleased build, or dev. |
uptimeSeconds |
integer | Seconds since the process started. |
inFlightGenerations |
integer | Chat completions being generated right now. |
sessions |
integer | Cache sessions the router is tracking. |
rooms |
integer | Saved conversations across every owner. |
servedModel |
string | The default model. |
servedContextTokens |
integer | The default model's context window, in tokens. |
upstream.mode |
string | proxy when the data plane forwards requests to a model server, mock when replies come from its built-in demo responder. |
upstream.target |
string or null | The model server's address in proxy mode, otherwise null. |
upstream.label |
string | A short name for where replies come from. |
Check readiness
/readyz
Reports whether this instance should receive new traffic, for a load balancer to poll. It answers 503 as soon as the instance starts shutting down, while it finishes the requests it already has; in production it keeps answering for 5 seconds after that, so the balancer can move traffic away before the instance stops listening. The control plane isn't checked, so a control-plane outage never takes data-plane instances out of rotation. No key is needed.
curl https://api.mortar.one/readyz{
"status": "ready"
}Response
| Field | Type | Description |
|---|---|---|
status |
string | ready when the instance should receive traffic. |
While the instance shuts down, the status is 503 and the body is the error envelope with type set to draining.
Get the platform overview
/v1/platform
Returns the sample organization, its environments, an in-flight rollout and the full model catalog, as the website's product preview shows them. No key is needed.
curl https://api.mortar.one/v1/platformResponse
| Field | Type | Description |
|---|---|---|
org |
object | The sample organization: id, name and initials. |
environments |
array | Where the sample deployments run. Each has id, label, deployment, region and byoc. |
rollout |
object | A deployment being rolled out: deployment, target and its steps. |
privateEndpoint |
object | The address applications call in the sample Bring Your Own Cloud environment: url and note. |
servedModel |
string | The default model. |
servedContextTokens |
integer | The default model's context window, in tokens. |
models |
array | The catalog, with the fields GET /v1/models summarises plus parameters, contextTokens, gpuClass, gpuCount, p50TtftMs and decodeTokensPerSec. |
illustrative |
boolean | Always true. |
Get the fleet
/v1/fleet
Returns the sample fleet the console's deployments view shows: deployments, team GPU quotas and headline numbers for the last 30 days. Needs an API key.
curl https://api.mortar.one/v1/fleet \
-H "Authorization: Bearer $MORTAR_API_KEY"Response
| Field | Type | Description |
|---|---|---|
deployments |
array | The sample deployments. |
deployments[].name |
string | The deployment's name, such as support-assistant. |
deployments[].modelId |
string | The ID of the model it serves. |
deployments[].model |
string | That model's display name. |
deployments[].environmentId |
string | The environment it runs in, such as byoc-acme-prod. |
deployments[].env |
string | That environment's display label, such as BYOC · acme-prod. |
deployments[].byoc |
boolean | Whether the environment is in the customer's own cloud account, with Bring Your Own Cloud. |
deployments[].gpu |
string | The GPU each replica runs on, such as L4. |
deployments[].replicas.ready |
integer | Replicas serving now. |
deployments[].replicas.max |
integer | The most the deployment may scale to. |
deployments[].status |
string | serving, rolling or idle. |
deployments[].spendUsd |
integer | Sample spend, in whole US dollars. |
deployments[].tokensMillions |
integer | Sample tokens served, in millions. |
quotas |
array | Each team's GPU reservation: team, gpu, used and quota. |
kpis |
array | Headline numbers: key, label, value, raw, delta, good and up. |
illustrative |
boolean | Always true. |
List prompt suggestions
/v1/suggestions
Returns the starter prompts the console shows in an empty conversation. No key is needed.
Query parameters
| Parameter | Type | Description |
|---|---|---|
surface |
string | landing for the shorter set the website uses. Leave it out for the console's set. |
curl "https://api.mortar.one/v1/suggestions?surface=landing"{
"data": [
{
"text": "What is Bring Your Own Cloud?",
"value": "In BYOC, what actually leaves my cloud account?"
}
]
}Response
| Field | Type | Description |
|---|---|---|
data |
array | The suggestions, in display order. |
data[].text |
string | The label to show. |
data[].value |
string | The prompt to send when the suggestion is chosen. |