Getting Started

Quick start

If you already use the OpenAI SDK, migrating takes two lines. Point baseURL at the Celon gateway and swap your API key for an sk-celon-* key. The rest of your code runs unchanged.

A two-line migration

Celon exposes a drop-in OpenAI-compatible API. Request and response formats, streaming, and function calling all stay compatible with your existing code, while the gateway handles routing, caching, and governance in between.

client.ts

 const client = new OpenAI({
-  baseURL: "https://api.openai.com/v1",
-  apiKey: process.env.OPENAI_API_KEY,
+  baseURL: "https://api.celon.ai/v1", // ← Celon gateway
+  apiKey: process.env.CELON_API_KEY,   // ← sk-celon-…
 });

Sending your first request — curl

Pass "celon/auto" as the model and the classifier scores the prompt’s difficulty and routes it to the cheapest model in an appropriate tier. If you want a specific model, just use its catalog id (for example openai/gpt-5.2).

terminal

curl https://api.celon.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-celon-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "celon/auto",
    "messages": [
      { "role": "user", "content": "Explain PostgreSQL indexes in one sentence" }
    ]
  }'

The celon field in the response body and the x-celon-* headers tell you which model handled the request and how much it cost and saved. For the full picture, see the Smart routing guide.

Python — openai SDK

main.py

from openai import OpenAI

client = OpenAI(
    base_url="https://api.celon.ai/v1",  # Celon gateway
    api_key="sk-celon-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx",
)

res = client.chat.completions.create(
    model="celon/auto",  # the classifier picks the tier that matches the difficulty
    messages=[{"role": "user", "content": "Explain PostgreSQL indexes in one sentence"}],
)
print(res.choices[0].message.content)

TypeScript — openai package

main.ts

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.celon.ai/v1", // Celon gateway
  apiKey: process.env.CELON_API_KEY,   // sk-celon-…
});

const res = await client.chat.completions.create({
  // "celon/code" for coding, "celon/quality" for quality, "celon/nitro" for speed, "celon/floor" for the lowest price
  model: "celon/auto",
  messages: [{ role: "user", content: "Explain PostgreSQL indexes in one sentence" }],
});

console.log(res.choices[0].message.content);

Running the gateway locally

Three lines at the monorepo root bring the gateway up. Because CELON_ENABLE_MOCK=1 is the default, you can demo the entire routing and caching flow with the built-in mock model (celon/mock) even without a single upstream provider key.

terminal

cp .env.example .env   # CELON_ENABLE_MOCK=1 by default — runs without upstream keys
pnpm install
pnpm dev:gateway       # → http://localhost:8787

Provider keys (OPENAI_API_KEY, ANTHROPIC_API_KEY, and so on) are all optional — only the adapters for providers you hold a key for get registered.