Calls /v1/chat/completions on a relay (OpenAI-compatible). Connect images for vision models or prompt extraction. — ComfyUI node from ComfyUI Relay API. Run it…

Use text, image and video models through third-party OpenAI-compatible API relays such as new-api and one-api, with several relays at once. A Relay Manager window adds and tests relays, lists their models and places configured nodes on the canvas. You need your own relay account, and the relay bills your calls. No extra dependencies.

ComfyUI Relay API

relay

model

prompt

system

temperature

max_tokens

seed

extra_json

timeout

images

Calls /v1/chat/completions on a relay (OpenAI-compatible). Connect images for vision models or prompt extraction.

RelayAPI_Chat

text

Relay Chat

Name	Type	Default	Description
relay	COMBO		Add relays in the Relay Manager window (tools/start_manager in the plugin folder). After editing relays.json by hand, press R to refresh.
model	STRING		Model name on the relay. The Relay Models node lists them
prompt	STRING		—
system	STRING		—
temperature	FLOAT	-1.00-1–2	-1 means not sent, using the model's default (some reasoning models only accept that)
max_tokens	INT	00–1000000	0 means not sent
seed	INT	00–4294967295	Not sent to the API. It only decides whether to call again: a new value makes a new (billed) call, the same value reuses the last result
extra_json	STRING		JSON object merged into the request body, such as {"quality": "high"}. Use null to remove a field
timeout	INT	30010–7200	—
imagesopt	IMAGE		Optional. Images are sent to the model too (it must support vision)

Relay Chat

Inputs (10)

Outputs (1)