Nova AI API Integration Guide

This guide walks through how Nova sends prompts to Hugging Face using an Azure Function backend, returning generated text for dynamic frontend use.

• API endpoint: https://ambientpixels-meme-api-fn.azurewebsites.net/api/generate-text
• Azure Function: generate-text
• Frontend tester: /lab/nova-ai-test.html
• JS fetch logic: /js/nova-ai.js
• Server function: /generate-text/index.js

1. User enters a prompt on the Nova AI test page or via site UI
2. nova-ai.js sends a POST request with the prompt to Azure Function
3. Azure Function (Node.js) receives the prompt and attaches the Hugging Face API key
4. It then relays the prompt to Hugging Face model inference endpoint (e.g., gpt2)
5. Response is parsed, formatted, and returned back to the frontend

1. func init . --javascript to initialize project
2. func new --name generate-text --template "HTTP trigger"
3. npm install node-fetch (required for API requests)
4. Replace /generate-text/index.js with the custom function logic
5. In Azure > Configuration > Application Settings, set: HUGGINGFACE_API_KEY
6. Under CORS, allow: http://localhost:5500, http://127.0.0.1:5500, and https://ambientpixels.ai
7. Deploy with: func azure functionapp publish ambientpixels-meme-api-fn --force

const fetch = require('node-fetch');

module.exports = async function (context, req) {
  const prompt = req.body?.prompt;
  context.log('📩 Received prompt:', prompt);

  if (!prompt) {
    context.res = {
      status: 400,
      headers: { 'Access-Control-Allow-Origin': '*', 'Access-Control-Allow-Headers': 'Content-Type' },
      body: { error: 'Missing prompt in request body.' }
    };
    return;
  }

  const HUGGINGFACE_API_KEY = process.env.HUGGINGFACE_API_KEY;
  const HF_MODEL = 'gpt2';

  try {
    const hfResponse = await fetch(`https://api-inference.huggingface.co/models/${HF_MODEL}`, {
      method: 'POST',
      headers: {
        'Authorization': `Bearer ${HUGGINGFACE_API_KEY}`,
        'Content-Type': 'application/json'
      },
      body: JSON.stringify({ inputs: prompt })
    });

    const text = await hfResponse.text();
    context.log('📦 Hugging Face raw response:', text);

    let data;
    try {
      data = JSON.parse(text);
    } catch (err) {
      context.log('❌ Failed to parse JSON:', err);
      throw new Error('Hugging Face returned invalid JSON or HTML');
    }

    context.res = {
      status: hfResponse.ok ? 200 : 500,
      headers: {
        'Access-Control-Allow-Origin': '*',
        'Access-Control-Allow-Headers': 'Content-Type'
      },
      body: hfResponse.ok
        ? { text: data[0]?.generated_text || '⚠️ No generated text.' }
        : { error: 'Hugging Face error', details: data }
    };
  } catch (err) {
    context.log.error('🔥 Nova AI Function Error:', err);
    context.res = {
      status: 500,
      headers: {
        'Access-Control-Allow-Origin': '*',
        'Access-Control-Allow-Headers': 'Content-Type'
      },
      body: { error: 'Internal server error', message: err.message }
    };
  }
};

func start

# In another terminal:
Invoke-RestMethod -Uri http://localhost:7071/api/generate-text `
  -Method POST `
  -Headers @{ "Content-Type" = "application/json" } `
  -Body '{ "prompt": "Nova, describe static electricity as a dream." }'

Successful response includes: { text: "Nova, describe static electricity as a dream..." }

• 500 error + HTML response: HF model is unavailable or invalid API key
• CORS error: Make sure all dev/test domains are added in Azure CORS settings
• fetch is not a function: Ensure node-fetch is installed locally
• Deployment not connecting: Check Azure CLI auth and use az login