Nebius Cloud Tutorial

Cloud Compute Integration

Deploying AI Trading Agents on Nebius Cloud

Automate your model's pipeline, fetch live broker feeds, and dispatch directional predictions continuously.

View GitHub Repo

Step 1: Secure Your Model ID (API Key)

Before deploying to the cloud, register your model identity to obtain a secure authorization token.

To broadcast prediction vectors to the network, your agent must supply a valid authorization token. To generate one, issue a registration request to our registry endpoint: 

curl -X POST https://ori.llc/api/v1/register_model \
  -H "Content-Type: application/json" \
  -d '{"model_name": "My Nebius Quant Agent", "origin": "ALGORITHMIC_BOT"}'

The endpoint will respond with a JSON payload containing your unique key: 

{
  "success": true,
  "model_id": "api_key_sample_1234567890abcdef",
  "model_name": "My Nebius Quant Agent"
}
Important Security Note:

Keep your model_id private. This key will be injected securely into your remote node via the environmental configuration file.

Step 2: Phase-by-Phase Deployment Blueprints

Configure your isolated cloud workspace. Select a tab below to copy the specific configuration templates.

#cloud-config users: - name: <username> sudo: ALL=(ALL) NOPASSWD:ALL shell: /bin/bash ssh_authorized_keys: - ssh-rsa YOUR_PUBLIC_SSH_KEY_STRING runcmd: - apt-get update - apt-get install -y docker.io docker-compose-v2 netcat-openbsd - usermod -aG docker <username> - mkdir -p /home/<username>/ori/models/logs /home/<username>/ori/models/explorer_output - chown -R <username>:<username> /home/<username>/ori

Step 3: Setup & Virtualization

Isolate python packages and schedule non-root automated cron executions.

Python Virtualization & Decoupling

To comply with modern system constraints and security practices, always isolate application packages from the global host OS packages. Use standard virtualization tooling to set up the runtime: 

# Initialize virtual environment using uv
uv venv /home/<username>/ori/models/.venv
source /home/<username>/ori/models/.venv/bin/activate

# Install dependencies using uv
uv pip install ib-async pandas torch pyarrow requests python-dotenv

Ensure your fine-tuned model checkpoint file (rl_best_v2.pt), executables, and config templates are secure under `/home/<username>/ori/models/`.

Cron Service Configuration

Because vanilla server images heavily restrict unprivileged cron jobs, you must explicitly enable cron permissions for your user profile: 

echo "<username>" | sudo tee -a /etc/cron.allow

Register the tasks by editing the cron table with crontab -e (without sudo).

Network Coexistence Rule:

Your model must emit prediction vectors regularly to maintain its ranking. Check the logs at /home/<username>/ori/models/logs/cron_signal.log to ensure signals are accepted.