Appwrite Function & NVIDIA NIM Setup

1. Architecture Flow

2. Function File Structure

The Appwrite Function code lives inside the repository at functions/ask-nvidia/:

functions/
└── ask-nvidia/
    ├── package.json
    └── src/
        └── main.js

Serverless Handler (`functions/ask-nvidia/src/main.js`)

export default async ({ req, res, log, error }) => {
  const headers = {
    'Access-Control-Allow-Origin': '*',
    'Access-Control-Allow-Headers': 'Content-Type, Authorization',
    'Access-Control-Allow-Methods': 'POST, OPTIONS',
  };

  if (req.method === 'OPTIONS') {
    return res.empty(204, headers);
  }

  try {
    // 1. Fetch secret key securely from server environment
    const apiKey = process.env.NVIDIA_API_KEY || req.variables?.NVIDIA_API_KEY;
    if (!apiKey) {
      error('NVIDIA_API_KEY missing on server environment.');
      return res.json({ success: false, error: 'NVIDIA_API_KEY is not configured on server.' }, 500, headers);
    }

    // 2. Parse request payload
    let body = {};
    try {
      body = typeof req.body === 'string' ? JSON.parse(req.body) : req.body;
    } catch (e) {
      body = {};
    }

    const promptText = body.prompt || body.question;
    if (!promptText) {
      return res.json({ success: false, error: 'Prompt is required.' }, 400, headers);
    }

    const model = body.model || "deepseek-ai/deepseek-r1";

    // 3. Proxy request to NVIDIA NIM Serverless Endpoint
    const response = await fetch("https://integrate.api.nvidia.com/v1/chat/completions", {
      method: "POST",
      headers: {
        "Content-Type": "application/json",
        "Authorization": `Bearer ${apiKey}`
      },
      body: JSON.stringify({
        model: model,
        messages: [
          { role: "system", content: "You are an expert Neovim tutor. Answer clearly with markdown code blocks." },
          { role: "user", content: promptText }
        ],
        temperature: 0.2,
        max_tokens: 2048
      })
    });

    if (!response.ok) {
      const errText = await response.text();
      error(`NVIDIA API Error (${response.status}): ${errText}`);
      return res.json({ success: false, error: `Upstream API error (${response.status})` }, response.status, headers);
    }

    const data = await response.json();
    const reply = data.choices?.[0]?.message?.content || "No response generated.";

    return res.json({ success: true, reply }, 200, headers);
  } catch (err) {
    error(`Unexpected Error: ${err.message}`);
    return res.json({ success: false, error: err.message }, 500, headers);
  }
};

3. Deploying Function via Git

Appwrite Functions supports automatic deployment directly from your connected Git repository whenever you push commits.

1. Open Appwrite Console: Go to Functions > Create Function.
2. Select Git Integration: Click Connect Git Repository and select repository RTFM.
3. Configure Production Settings:

   Setting Field	Value	Description
   Function Name	ask-nvidia	Function identifier
   Runtime	Node.js 20.x / 22.x	Execution runtime environment
   Root Directory	functions/ask-nvidia	Directory containing main.js
   Entrypoint	src/main.js	Relative path to entry file
   Execute Access	Any (Public)	Allows website visitors to trigger AI Tutor

4. Save & Trigger Deploy: Click Deploy. Any future push to main will trigger automatic function deployments!

4. Adding Appwrite Secrets (Console Config)

To give your serverless function access to NVIDIA NIM endpoints without putting keys in code:

1. Navigate to Appwrite Console > Functions > ask-nvidia.
2. Select the Settings tab $\rightarrow$ Scroll down to Variables.
3. Click Add Variable:
   • Key: NVIDIA_API_KEY
   • Value: nvapi-YOUR_ACTUAL_NVIDIA_KEY_HERE
4. Click Update to save the environment variable.