Get started

Seldon documentation

Route OpenAI-compatible requests through one gateway, and turn repeated LLM calls into measured Functions you activate and call from your code.

Seldon has two parts that share one API key and one base URL.

Gateway. An OpenAI-compatible API. Point an existing SDK at Seldon, route requests across providers, and see every call's tokens, latency and cost in Calls & usage.
Functions. Save a repeated LLM call or a short workflow. Seldon builds alternative pipelines, measures them against your original call, and you activate the one that fits. Your application calls it through the gateway like any other model.

Routing traffic through the gateway does not start optimization on its own. You choose which calls become Functions.

Start here

Section	Use it to
Get started	Follow a short tutorial from zero to a working result.
Functions	Prepare, evaluate, optimize, activate and call a function.
Gateway	Configure SDKs, routing, streaming, caching and observability.
Concepts	Understand how Functions measure alternatives, billing and data handling.
Reference	Look up request fields, headers, errors and limits.

Contact us with the X-Seldon-Trace-Id response header of the request in question. It identifies the call without you sharing its content.