System status
Whether the services behind Mortar are answering, checked from this web server at most once every 15 seconds.
Every component passed its check. Checked , just now.
Components
-
Inference API
Operational The OpenAI-compatible endpoint that serves completions.
Ready to take traffic.
GET /readyz, answered in 6 ms -
Control plane
Operational Sign-in, sessions, organisations and API keys.
Ready to take traffic.
GET /readyz, answered in 11 ms -
Website and console
Operational These pages, the sign-in forms and the console.
Ready to take traffic.
GET /readyz, answered in 3 ms
Models
Each offered model follows the live inference API check. Active scheduled maintenance is shown on the affected model.
-
Qwen 2.5 7B Instruct
Operational Alibaba Qwen
Ready to serve requests.
-
Llama 3.1 8B Instruct
Operational Meta Llama
Ready to serve requests.
-
Mistral 7B Instruct
Operational Mistral AI
Ready to serve requests.
-
Qwen 2.5 32B Instruct
Operational Alibaba Qwen
Ready to serve requests.
-
Llama 3.3 70B Instruct
Operational Meta Llama
Ready to serve requests.
-
DeepSeek-V3
Not offered DeepSeek
Listed in the catalog, but not offered by this deployment.
Scheduled maintenance
No maintenance scheduled
Upcoming, active and completed maintenance windows are published here with their affected components and models.
Incident history
Follow every incident update in a feed reader: the incident feed (Atom).
No incidents recorded
When an incident affects a component, it is published here with every update until it is resolved.
How these checks work
- What is asked
- Each component is asked
GET /readyz: should it receive traffic right now? A component too old to have that endpoint answersGET /healthinstead, which only shows that it is running. - Down and degraded
- No answer within 2 seconds, or an answer other than ready, is reported as down. An answer that takes longer than 1 second is reported as degraded.
- How often
- Results are kept for 15 seconds. However many people load this page, each component is checked at most once in that time.
- No uptime figure
- Checks run only when someone loads this page, are held in memory, and start again empty after every deploy. An uptime percentage calculated from them would not be a measurement, so this page does not show one.
- Where this page runs
- This page is served by the deployment it checks, so if the website is down, this page is down with it. A status page that has to stay reachable during an outage needs separate hosting.
- As JSON
/status.jsonreturns the same snapshot as this page.