@chirag127/keyless-ovh

Keyless OVHcloud AI Endpoints client. OpenAI-compatible chat completions. No API key — anonymous endpoint, no auth header.

Framework-agnostic, zero deps, works in Node ≥18 + browser via global fetch.

Note: IP-rate-limited (429). On 429, chat() throws with err.retryable = true — check this flag to distinguish rate-limit from hard failures.

Install

npm i @chirag127/keyless-ovh

Usage

import { chat, MODELS, DEFAULT_MODEL } from '@chirag127/keyless-ovh';

// string prompt
const reply = await chat('Explain vector databases in one line.');

// messages array + options
const reply2 = await chat(
	[
		{ role: 'system', content: 'You are terse.' },
		{ role: 'user', content: 'Hello' },
	],
	{ model: 'gpt-oss-20b', temperature: 0.2 }
);

// handle 429
try {
  await chat('hi');
} catch (err) {
  if (err.retryable) { /* rate limited — retry later */ }
  else throw err;
}

API

chat(messages, opts?) => Promise<string>

Non-streaming. POSTs {model, messages} to {baseUrl}/chat/completions — no auth header. Returns choices[0].message.content.

On 429 the thrown Error has .retryable = true.

| opt | default | note | |---|---|---| | model | 'gpt-oss-120b' | any id from MODELS | | baseUrl | https://oai.endpoints.kepler.ai.cloud.ovh.net/v1 | OVH anonymous endpoint | | signal | — | AbortSignal | | ...rest | — | passed through (temperature, max_tokens, …) |

Models

Source: OmniRoute freeModelCatalog.data.ts — provider: "ovhcloud", freeType: "keyless", poolKey: "ovhcloud-anon".

| Model | Notes | |---|---| | gpt-oss-120b | Default — GPT-OSS 120B | | Qwen3.6-27B | Qwen 3.6 27B | | Qwen2.5-VL-72B-Instruct | Qwen 2.5 VL 72B (vision) | | Mistral-Small-3.2-24B-Instruct-2506 | Mistral Small 3.2 | | gpt-oss-20b | GPT-OSS 20B (fast) |

Notes

• tos: "ok" — keyless anonymous endpoint, no registration needed.
• IP rate-limited (429) — check err.retryable to detect.
• Throws on non-2xx with the status and response body.

License

MIT