Draft AI-powered Email Replies Using Hugging Face

Get Your Digital Key (API Token)

First, let's get your unique digital key, known as an API token. An API (Application Programming Interface) is like a restaurant's digital menu and ordering system, allowing different computer programs to talk to each other. Your API token is like your personal, secret VIP pass that tells Hugging Face's "waiter" (the API) that you're authorised to place an order.

Open your web browser and go to the Hugging Face website. Log in to your account. Once logged in, you'll typically see a clean page, perhaps with your activity feed. Look for your profile picture (usually a circle with your initial or chosen image) in the top-right corner of the screen. Click on it. A menu will slide down or drop open, showing options like "Profile," "Settings," and "Sign Out." Select "Settings" from this menu. On the next screen, you'll see a sidebar on the left with various account settings. Click on "Access Tokens". You'll then see a button that says "New token." Click it. A small window will pop up asking for details. Give your token a descriptive name like "Email-Reply-Bot" and set the Role to "read." This "read" role means the token only allows you to receive information (like AI-generated text), not change anything on your account. Click "Generate a token." A very long, random string of characters (your API token) will appear on the screen. Immediately copy this token! You can click the copy icon next to it. Treat this token like a password – you won't see it again once you leave this page. Save it somewhere safe temporarily, like a simple text document, as you'll need it soon.

The "Settings" or "Access Tokens" might be under slightly different names but will generally be found within your account management area.

Next step ↓

Find an AI "Chef" (Model) for Your Email

Now you need to choose an AI model, which is like selecting a specific chef for your meal. Each model is a unique computer program trained for different tasks. For drafting emails, you'll want a "text generation" model – one specifically designed to produce new text based on your instructions.

Open a new browser tab and navigate to the Hugging Face Hub. On this page, you'll see a prominent search bar, usually in the middle or near the top. In this search bar, type text-generation. As you type, a list of models might appear, or you might need to press Enter to see the results. You'll then see many different AI models. For a beginner, a good starting point is gpt2 because it's well-known and generally effective for various text tasks. You might also see distilgpt2 (a smaller, faster version) or microsoft/DialoGPT-medium. Click on the model you've chosen (e.g., click on the gpt2 model card). On the model's specific page, at the very top, you'll see its Model ID. For gpt2, the Model ID is simply gpt2. Make a note of this Model ID – you'll need it for the next step.

Next step ↓

Prepare Your "Order" (API Request)

Now you'll craft the specific instruction you want to send to the AI. This instruction is called a prompt. You'll combine your API token, the model ID, and your prompt into a command using a tool called curl (pronounced "curl"). curl is a common way to send requests from your computer's terminal to web services like Hugging Face's API.

Open your computer's terminal. On Windows, you can search for "Command Prompt" or "PowerShell" in the Start menu. On Mac or Linux, search for "Terminal." Once open, you'll see a window with a blinking cursor, ready for your input. We'll use a curl command, which is a specific line of code that sends your request. Your command will include:

• -X POST: This tells Hugging Face you're sending data (like placing an order).
• -H "Authorization: Bearer YOUR_API_TOKEN": This includes your VIP pass (API token) to confirm you're allowed to make the request.
• -H "Content-Type: application/json": This tells Hugging Face that you're sending information in JSON (JavaScript Object Notation) format, which is a structured way for computer programs to exchange data.
• -d '{ "inputs": "YOUR PROMPT", "parameters": { "max_new_tokens": 50, "return_full_text": false } }': This is the actual data you're sending. "inputs" contains your prompt (the instruction for the AI). "parameters" are extra instructions, like "max_new_tokens": 50 (asking for about 50 new pieces of text, called tokens – a token is roughly four characters) and "return_full_text": false (which ensures the AI only sends back its generated reply, not your original prompt as well).
• https://api-inference.huggingface.co/models/YOUR_MODEL_ID: This is the web address of the AI "kitchen" where your chosen model is waiting.

You need to replace YOUR_API_TOKEN with your actual token from Step 1 and YOUR_MODEL_ID with the model ID from Step 2.

Some corporate networks might block curl commands. If you encounter errors like "command not found," you might need to contact your IT department or try a different computer.

curl -X POST \
 -H "Authorization: Bearer hf_YOURUNIQUEKEY123" \
 -H "Content-Type: application/json" \
 -d '{ "inputs": "The customer is unhappy with the service and wants a refund. Draft a polite and empathetic reply.", "parameters": { "max_new_tokens": 80, "return_full_text": false } }' \
 https://api-inference.huggingface.co/models/gpt2

Next step ↓

Send Your Order to the AI "Kitchen"

With your complete curl command ready, it's time to send it. Carefully copy the entire command, including curl at the start and the model URL at the end. Then, go back to your open terminal window.

Paste the entire curl command into the terminal window at the blinking cursor. After pasting, press the Enter key on your keyboard. What happens next is that your computer sends this request over the internet to Hugging Face. The AI will then process your request – this process is called inference. It might take a few seconds, especially the first time you use a specific model or if Hugging Face needs to "load" the model onto its servers for you. The terminal window will usually remain blank for a short moment while it waits for a response.

A slow response might just mean the model is taking a moment to load.

This text will typically start with a square bracket [ or an opening curly brace {, indicating that data has been received.

Next step ↓