Skip to main content
Aria Compute builds device-native foundation models (AFM) for edge devices such as mobiles, robots, and wearables, with both cloud and local runtimes. This site covers signing up and signing in, creating API keys, and downloading and integrating AFM models; routing decisions through ROUTER’s OpenAI-compatible gateway; managing cloud agents and sessions with AGENT; running PIN post-training jobs; and storing and recalling long-term memory with MEMO. It also documents the PIN, ROUTER, and AGENT REST APIs, the local runtimes (Ollaya and aria-engine), and account and billing.

Quickstart

Create an account, generate an API key, and make your first call.

Authentication

Session tokens, API keys, and query-token downloads explained.

AFM foundation models

Model download and runtime.

PIN training platform

Fine-tuning, LoRA, RL, and evaluation.

ROUTER decision routing

Chat routing across three decisioners.

AGENT cloud agents

beta Agents API, governed execution and tool families.

MEMO memory storage

Aria Memo on device, AriaMem in the cloud.

API Reference

PIN, ROUTER, and AGENT endpoints.

PIN API

Post-training, jobs, and GPU nodes.

ROUTER API

OpenAI-compatible gateway with three decisioners.

AGENT API

agent.turn.* events, read-only governance endpoints and five cloud-client SDKs.

What you can do

  • Download AFM foundation models.
  • Create API keys.
  • Top up your prepaid wallet with Stripe (USD) on the international site or WeChat Pay / Alipay (CNY) on the China site.
  • Pull the latest aria-engine, aria-router, aria-agent, and aria-memo releases from GitHub (international) or Gitee (China).
  • Manage sessions, enable two-factor authentication, and view metered usage.
The platform runs as two independent deployments: ariacompute.com (international, USD, Stripe) and ariacompute.cn (China, CNY, WeChat Pay + Alipay). Accounts, API keys, and wallets are not shared between the two sites.