Getting started / Quickstart

Quickstart

Three commands. You'll be streaming tokens in under thirty seconds. Already using the OpenAI SDK? Point base_url at us. No other change.

1 — Install

$ pip install openai$ export TACHYON_API_KEY=sk-tch-…

2 — Stream a completion

from openai import OpenAIimport os client = OpenAI(    base_url="/v1",    api_key=os.environ["TACHYON_API_KEY"],) stream = client.chat.completions.create(    model="llama3.1-8B",    messages=[{"role": "user", "content": "hello"}],    stream=True,)for chunk in stream:    print(chunk.choices[0].delta.content or "", end="", flush=True)

3 — Measure

First byte under 5ms. Cache the client.

NOTE

Auth. During open beta, any value is accepted as an API key — both "anything" and an empty string will work. Keys are not currently enforced. Auth enforcement will be announced separately. base_url