Skip to main content
Aria Compute PIN is the end-side post-training service for fine-tuning, LoRA adaptation, reinforcement learning, evaluation, and model export. The PIN API exposes these capabilities over HTTP at base URL http://<pin-host>:8001, under the /v1 path prefix. All /v1/* routes require a Bearer API key in the Authorization header, except public routes such as /health, /version, /v1/auth/login, and /v1/auth/register. Node-agent callback routes under /v1/agent/* also require header X-Agent-Secret, equal to the server AGENT_REGISTER_SECRET. Separate admin routes (/v1/admin/*) exist for operators and are not documented on this public site. Error responses use this shape:
Successful responses use { "code": 0, "data": …, "message": "" }. Common status codes include 400 for validation errors, 401 for unauthorized requests, 404 for missing resources, and 500 for internal errors.

Authentication

Bearer API keys and node-agent X-Agent-Secret.

Create a Job

Submit fine-tuning, LoRA, or RL jobs through the PIN API.

List Models

Browse available base models and exported checkpoints.

Billing Summary

Retrieve usage and billing summaries for your organization.