# Air Container
Source: https://aircloud.mintlify.app/docs/air-container/index

Deploy your AI container to Air Cloud

Air Container enables you to deploy containerized AI services quickly and efficiently.
Here's how it works:

## Deployment Flow

### Step 1. Select a project

<Frame>
  <img />
</Frame>

First, go to the Project Dashboard and select the organization and project you
created during onboarding.

<Frame>
  <img />
</Frame>

From the Service tab on the project page, select **"Container"** and click
**\[+ Create]** button to create a new container.

### Step 2. Base Settings

<Frame>
  <img />
</Frame>

<br />

<Frame>
  <img />
</Frame>

During container creation, the following fields are required:

<ResponseField name="Name" type="string">
  A user-friendly name for the container.
</ResponseField>

<ResponseField name="Category" type="string">
  Type of service the container provides (used for future Playground
  integration).
</ResponseField>

<ResponseField name="Container Image" type="url">
  Docker image URL of the container.
</ResponseField>

<ResponseField name="Registry Provider" type="string">
  The registry where the image is hosted (e.g., GitHub Container Registry,
  Docker Hub).
</ResponseField>

<ResponseField name="Registry Username & Password" type="string">
  Required if pulling from a private registry.
</ResponseField>

### Step 3. Configure Resources

<Frame>
  <img />
</Frame>

* **General Mode**: Define instance type (e.g., RTX 4070, 4090).
* **Autoscaling Mode**: Define minimum and maximum replicas (1 to 30). Higher
  values require manual approval.

### Step 4. Advanced Settings

<Frame>
  <img />
</Frame>

In the "Advanced" section, additional options are available:

<ResponseField name="Start Command" type="string">
  Overrides the default startup command in the container image.
</ResponseField>

<ResponseField name="Port" type="number">
  Overrides the default exposed port.
</ResponseField>

<ResponseField name="Health Check URL" type="url">
  Path to check container health status (e.g., /api/health).
</ResponseField>

<ResponseField name="Environment Variables" type="object">
  Set key-value pairs required for your app (e.g., DB credentials, API keys).
</ResponseField>

### Step 5. Review and Deploy

<Frame>
  <img />
</Frame>

### Step 6. Deployment Completion

<Frame>
  <img />
</Frame>

Once deployed, the container will start immediately, and an API endpoint or service URL will be provided.

* The container will appear in the left-hand list.
* Selecting an item will show its detailed information on the right.
* You can edit settings from the right panel.
* Clicking the "Dashboard" button will redirect you to a detailed management page with container status and activity logs.

## API Request

Once the container status is `RUNNING`, you can access the AI inference API using the exposed endpoint.
Replace the host part with your container's Endpoint URL.

```bash cURL theme={null}
curl --request POST \
     --url ${ENDPOINT_URL}/api/v1/chat/completions \
     --header "Accept: application/json" \
     --header "Authorization: Bearer ${YOUR_API_KEY}" \
     --header "Content-Type: application/json" \
     --data '{
  "messages": [
    {
      "role": "system",
      "content": "You are a helpful assistant."
    },
    {
      "role": "user",
      "content": "Write a haiku about recursion in programming."
    }
  ]
}'
```


# Monitoring & API Testing
Source: https://aircloud.mintlify.app/docs/air-container/monitoring-api-testing



* **Single Request Test**: Use Postman or similar tools to test your vLLM container endpoint.
* **Autoscaling Test**: Use Locust to simulate load and validate replica scaling.
* **Usage Metrics**: Monitor replica count, request rate, and system response time.
* **Runtime Logs**:
  * Logs are displayed per container instance and replica.
  * You can access real-time and historical logs from the Air Cloud dashboard.
  * Logs are shown in reverse chronological order.
  * Use filters to view logs by time range, container, or instance ID.
  * Logs include startup command output, health check status, error messages, and stdout/stderr of model servers.
  * If your container fails, logs will be preserved for a limited retention window for debugging.
* **Settings**: You can update endpoint settings only when the container is stopped.


# Introduction
Source: https://aircloud.mintlify.app/docs/index



**Air Cloud** is a cloud computing platform built to run AI inference at scale. Whether you're deploying for real-time AI services or cost-sensitive workloads, Air Cloud gives you two options:

* **Air Cloud**: A distributed, crowd-sourced GPU infrastructure offering the lowest possible pricing through our trusted GPU network.
* **Air Cloud+**: Proprietary GPU devices operated by AIEEV, offering guaranteed uptime, performance, and availability.

Forget complex configurations and overpriced legacy cloud platforms. Traditional bare-metal GPU clouds often fall short in dynamic, production-grade environments — Air Cloud is purpose-built to solve that.

[Get started instantly by creating an account.](https://integrated-dev.aieev.cloud:3002/ko)

## Choose Your Deployment Method

* **Air Container**: Seamlessly deploy containers from public or private repositories (AWS, GitHub, Docker Hub, Google). Just enter your image address and secure token, and you're good to go.
* **Air Model** (Coming Q3 2025): Designed for faster AI inference with serverless, pay-per-second billing. Automatically pulls your models, sets up REST endpoints, and enables autoscaling for your production workloads.
* **Air API** (Coming Q3 2025): Offers OpenAI-compatible REST API endpoints using open-source models like Whisper, LLaMA 3.3 8B, Solar Mini, and D.N.A. Features serverless.

## Choose Your Infrastructure

* **Air Cloud**: A distributed network of over 10,000 GPUs in Korea, sourced from our secure peer-to-peer partner network. We’ve pre-secured 1,000+ GPUs that can be dynamically assigned to your job at ultra-competitive pricing.
* **Air Cloud+**: AIEEV-owned GPU servers designed specifically for AI inference, offering high availability, service stability, and enterprise-grade security.

## Our Mission

We make cloud computing accessible and affordable for everyone — without compromising performance, reliability, or usability. Our goal is to empower both individuals and enterprises to unlock the full potential of AI and scalable infrastructure.

<Info>
  Need help? Check our documentation, reach out via [Discord](https://discord.com/channels/912829806415085598/1187492973148115076/threads/1349672733168107541), chat support, or
  [email](mailto:contact@aieev.com). Details are on our [Contact page](https://www.aieev.com/contact).
</Info>

## What’s Next?

* Run Air Cloud Tutorials
* Add funds to your account


# Quickstart
Source: https://aircloud.mintlify.app/docs/quickstart

Deploy your AI model to Air Cloud

Welcome to Air Cloud. This step-by-step guide will get you up and running with your first deployment in under 5 minutes.

To launch your resources, you'll need both a billing account and at least one project. The onboarding flow in the console guides you through this.

## Creating an Account

<Frame>
  <img />
</Frame>

Start by signing up for an Air Cloud account to manage and access your deployments.
You can sign up using GitHub, Google, or Email ID. Once registered, log in to begin
using the platform.

## Adding Funds

New users receive \$20 in credit upon registration. To deploy resources, you must fund your account. Currently, bank transfers and credit card payments are supported.

To load your balance, email us at [sales@aieev.com](mailto:sales@aieev.com) — our team will respond immediately. Bulk credits and reserved-resource discounts are available.

## Creating a Project

<Frame>
  <img />
</Frame>

Once your account is funded, create a project to deploy and organize your workloads. You can deploy multiple containers, models, or APIs within a single project to match your AI service architecture.

## Managing Your Team

You can either create a personal account or get invited by a team member.

<Frame>
  <img />
</Frame>

<br />

<Frame>
  <img />
</Frame>

* **Sign Up**: Use GitHub, Google, or Email ID.
* **Invite Team Members**: Admin users can invite teammates via email through the Member Management page. Team members must register using the exact email address they were invited with. After registration, they must accept the invite to join the organization and project.
* **Role Types**: Admins can manage team settings, invite/delete members, and remove projects. Users have limited access to deployment and monitoring tools only.

<Warning>
  If a user receives an invite, ensure they log out from all accounts before
  accepting the invitation to avoid errors (e.g., 404 page).
</Warning>

## Billing Information

* **Air Container** and **Air Model** are billed hourly. For full pricing details, see our [Billing FAQ](https://www.aieev.com/pricing).
* **Payment Method**: supports local and international wire transfers as well as credit card payments.


# Support
Source: https://aircloud.mintlify.app/docs/support

We're here to help you get the most out of Air Cloud

<div>
  <Card title="Ask our docs" icon="message-circle-question" href="https://docs.aieev.cloud/docs/overview/">
    Ask our AI assistant that's been trained on our documentation.
  </Card>
</div>

<CardGroup>
  <Card title="Email Us for Support" icon="mail" href="mailto:hk.kim@aieev.com">
    If you have any questions, email us at
    [hk.kim@aieev.com](mailto:hk.kim@aieev.com).
  </Card>

  <Card title="Email Us for Development" icon="mail" href="mailto:bspark@aieev.com">
    If you have any questions, email us at
    [bspark@aieev.com](mailto:bspark@aieev.com).
  </Card>
</CardGroup>

## Frequently Asked Questions

Below are common issues and how to resolve them based on real user deployments and our internal diagnostics:

<AccordionGroup>
  <Accordion title="401 Unauthorized">
    Authentication failed. Please verify that your API key or access credentials are correct.
  </Accordion>

  <Accordion title="404 Not Found">
    The requested container ID or endpoint path is invalid. Double-check the
    container ID or route.
  </Accordion>

  <Accordion title="500 Internal Server Error">
    An internal server error has occurred. Try again later or contact the Air Cloud support team.
  </Accordion>
</AccordionGroup>

Below are common issues and how to resolve them based on real user deployments and our internal diagnostics:

<AccordionGroup>
  <Accordion title="API Key Authentication Failure">
    * Symptom: Fails to authenticate API request
    * Cause: Expired or incorrect API key
    * Fix:
      * Ensure your key is still valid (check expiration)
      * Generate a new API key from the Air Cloud dashboard
  </Accordion>

  <Accordion title="GPU Assignment Errors">
    * Symptom: No GPU detected on container startup
    * Cause: Node instability or driver mismatch
    * Fix:
      * Re-deploy to a new instance
      * Check compatibility of driver with your container runtime
  </Accordion>

  <Accordion title="Network Connectivity Failures">
    * Symptom: Container cannot reach external endpoints
    * Cause: Port configuration, DNS issues, or firewall blocks
    * Fix:
      * Verify container exposes correct port and service is listening
      * Use curl inside container to test outbound connectivity
      * Ensure no security group restrictions block egress traffic
  </Accordion>

  <Accordion title="Health Check Failures / Stuck in Initializing">
    * Symptom: Deployment hangs on health check
    * Cause: Container is not returning 200 OK on /health
    * Fix:
      * Confirm that the correct health check path and port are exposed
      * Check logs for stack traces or model loading delays
  </Accordion>

  <Accordion title="Autoscaling Not Triggering">
    * Symptom: No scaling despite high load
    * Cause: Scaling threshold too high, or unhealthy replicas
    * Fix:
      * Tune autoscaling parameters (e.g., CPU threshold, latency window)
      * Verify all replicas pass health checks
  </Accordion>

  <Accordion title="502/503 Gateway Errors">
    * Symptom: API responds with server error
    * Cause: App not bound to expected port or crashed during startup
    * Fix:
      * Verify service listens on `0.0.0.0:<PORT>`
      * Increase health check timeout if model loading takes long
  </Accordion>
</AccordionGroup>

<Info>
  We do our best to respond to all requests within 24 hours but delays may occur
  during busy times.
</Info>