Skip to content

API reference

Base URL: https://api.keln.ai/v1 · Auth: Authorization: Bearer sk-keln-…

Keln is OpenAI-wire-compatible, an OpenAI SDK works by changing the base URL. Behind the endpoint, Keln normalizes across its serving pool: whatever capacity serves your request, the parameters you sent behave the same way and the response has the same shape.

Endpoints

Endpoint What it does
POST /v1/chat/completions Chat completions, streaming and non-streaming, tools, structured output, reasoning control. Needs an API key.
GET /v1/models The servable catalog: context windows, live pricing, input modalities (text / image), and current routing performance figures. No key needed, CORS-open.
GET /healthz Liveness check, for uptime monitoring. No key needed.

There is no legacy /v1/completions text endpoint.

Model ids are lowercase (deepseek-ai/deepseek-v4-flash) and matched case-insensitively.

Authentication

Every POST /v1/chat/completions requires an API key, created in the portal (Keys page):

Authorization: Bearer sk-keln-…

Keys are shown once, stored hashed, and can be revoked. A key belongs to a workspace; usage bills to that workspace's balance and can be capped with per-key budgets.

GET /v1/models requires no key and is served CORS-open (Access-Control-Allow-Origin: *).

SDKs

Any OpenAI-compatible client works:

from openai import OpenAI
client = OpenAI(base_url="https://api.keln.ai/v1", api_key="sk-keln-…")

import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://api.keln.ai/v1", apiKey: "sk-keln-…" });

Coming from another gateway? Migration guide →

Reading order