gushwork-relay

The Python SDK for the Relay gateway. It is the official OpenAI SDK with Relay's headers and stored prompts added, so chat completions, the Responses API, embeddings, streaming, retries and typed errors work as they do in the OpenAI SDK.

Install

pip install gushwork-relay

Python 3.10 or later.

Quickstart

Set your Relay API key, your gateway URL and the id of a config:

export RELAY_API_KEY=...
export RELAY_BASE_URL=https://<your-gateway-host>/v1
export RELAY_CONFIG=...

from relay_ai import Relay

client = Relay()  # reads the three variables above

reply = client.chat.completions.create(
    model="gpt-4.1-mini",
    messages=[{"role": "user", "content": "Say hello in five words."}],
)
print(reply.choices[0].message.content)

# A prompt stored in Relay: its template, model and variables.
reply = client.prompts.completions.create(prompt="support-reply@staging", variables={"customer": "Ana"})
print(reply.choices[0].message.content)

AsyncRelay is the asynchronous client. Both take api_key, base_url, config, metadata and trace_id as arguments.

Documentation

The full guide is in the SDKs section of the Relay docs: https://docs.gushwork.org/sdks/.