Submitting AI Tasks

Nodia transforms your compute hardware into part of a global AI mesh. Whether you're operating a single Nodia Core at home or managing a fleet of Atlas devices in a data center, submitting workloads is seamless, secure, and efficient.

Submitting via Nodia Dashboard

Nodia uses a simple visual interface to make workload submission accessible to everyone — no coding required.

1. Log In

• Visit dashboard.nodia.io

• Connect your Phantom Wallet to authenticate

2. Submit a Task

• Go to Workloads → Submit New Task

• Choose a task type:

  • Inference

  • Training

  • Video Rendering

  • Data Analytics

  • Custom Plugin (coming soon)

3. Define Your Input

• Upload Files: Drag and drop datasets (max 1 GB per upload)

• Decentralized Storage: Enter an IPFS or Arweave link

4. Choose Model

• Select from Nodia’s preloaded models (e.g., YOLOv4, ResNet50)

• Or upload your own in ONNX, TensorFlow, or PyTorch format under Models → Manage

5. Configure Parameters

• Priority: Standard (default) Accelerated (faster dispatch, higher budget use) Enterprise (reserved for advanced task types)

• Resources: Set optional limits for CPU cores, GPU memory, or time per task

• Parallelism: Automatically optimized per network conditions (can be overridden)

6. Allocate Budget

• Enter how many NODIA tokens you want to allocate

• Estimated costs appear in real-time based on task complexity and network demand

• Any unused tokens are refunded after task completion

7. Launch & Monitor

• Click Deploy Task

• Track your Task ID and live status in the Live Task Monitor:

  • Node assignments

  • Completion rate

  • Latency and performance graphs

8. Retrieve Results

• Outputs are aggregated automatically

• Access final results under Workloads → Completed

• Download or stream from IPFS (if applicable)

---

Coming Soon: Programmatic Submission

REST API and SDK integration are under development to support CI/CD pipelines, automation, and dApp integrations. These will enable:

• Secure token-authenticated task submission

• Webhooks for real-time callbacks

• Batch job dispatch and monitoring via CLI tools

Stay tuned for updates via [e-mail] or the Nodia community channels.

Via REST API & SDKs

For automation and integration into CI/CD pipelines, Nodia will soon provide a fully featured REST API and client SDKs.

Example: Submitting an Inference Task (cURL)

bashCopyEditcurl -X POST https://api.nodia.io/v1/task/submit \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "taskType": "Inference",
    "model": "resnet50-v2",
    "dataUrl": "ipfs://QmYourDatasetHash",
    "priority": "Standard",
    "resources": {
      "maxCpu": 4,
      "maxGpuMemory": 2048
    },
    "budgetNODIA": 100
  }'

• Response: Returns { "taskId": "TASK12345", "estimatedCost": 23.5 }.

Example: Submitting via JavaScript SDK

javascriptCopyEditimport { NodiaClient } from "@nodia/sdk";

async function submitInference() {
  const client = new NodiaClient({ apiKey: process.env.NODIA_API_KEY });
  const { taskId, estimatedCost } = await client.submitTask({
    taskType: "Inference",
    model: "yolov4",
    dataUrl: "ipfs://QmYourDatasetHash",
    priority: "Accelerated",
    budget: 50,
  });
  console.log(`Task ${taskId} submitted; estimated cost = ${estimatedCost} NODIA`);
}

submitInference();

6.1.3 Advanced Submission Options

• Batch Mode: Group multiple small data items into a single submit call for improved throughput. Add "batch": true in API payload.

• Callback Notifications: Include a webhookUrl to receive HTTP POST callbacks on task status changes (e.g., completed, failed, timeout).

• Presigned Uploads: For large datasets, request a presigned URL via GET /v1/task/upload-url, then upload to that endpoint before submission.

---

By providing both intuitive UI workflows and robust programmatic interfaces, Nodia ensures every user—from non-technical hobbyists to DevOps teams—can deploy AI workloads quickly, securely, and at scale. Next, we’ll dive into Real-Time Inference, showcasing how Nodia achieves sub-50 ms latency for mission-critical applications.