loadgen

Overview ¶

Command loadgen sends a steady-rate stream of /v1/messages requests at a Relay endpoint. Designed for stressing the eventlog/CH path: defaults target the loadtest route configured against the fakeanthropic upstream.

Usage:

go run ./bench/loadgen --rps 50 --duration 2m

Flags:

--target    base URL of the relay (default http://localhost:5100)
--api-key   client API key (default test-key)
--model     model name to send in body (default loadtest-model)
--rps       requests/sec (default 50)
--duration  total run length (default 2m)
--timeout   per-request timeout (default 30s — long enough for 20s upstream)