Get ZML running on your cluster
Deploy the ZML dispatcher, register your accelerator nodes, and serve your first inference request in under 30 minutes. Complete API reference and configuration guides included.
Where do you want to start?
Quickstart
Install ZML, set your API key, write a routing config, and run your first inference call. Takes under 30 minutes for a working local setup.
Get startedAPI Reference
Complete reference for zml.Client, the infer() method, routing policy schema, accelerator IDs, error codes, and rate limits.
Accelerator Compatibility
Supported accelerator IDs, required driver versions, and known limitations for each hardware target. Updated with each ZML release.
View matrixRouting Policies
Configure latency-first, cost-first, or custom weighted routing. Chain multiple policies and set per-model fallback rules. Full schema reference included.
Read policy docsQuestions not covered in the docs?
The ZML team responds to infrastructure questions directly. No support ticket queue.