Serverless Inference with the DigitalOcean GenAI Platform

To get started, we can either use the DigitalOcean API or the Cloud Console. Let’s look at each methodology.

Step 1a: Create a DigitalOcean API Key

To get started, you first need to create your own DigitalOcean account and login. Once you’ve done so, navigate to the team space of your choice. From here, we will create a DigitalOcean API key. This will help us create our model access token later on, so we can actually skip this step and go to the section “Step 2b: Create a Model Access key with the Cloud Console”.

Use the navigation bar on the left hand side of the homepage, and scroll down until you see “API”. Click in to open the API homepage. We can then use the button on the top right to create a new key.

In the key creation page, name your key as needed and give appropriate permissions. To do this, either give full permission or scroll down to “genai” in the custom scope selection and select all. Then, create the key. Save this value for later.

Step 2A: Create a Model Access key with the API

Next, we are going to make our model access key for GenAI’s Serverless Inference. To do so, we can either use the console or API. To use the API, use the saved API key from earlier with the following curl request in your terminal. Replace the value for “$DIGITALOCEAN_TOKEN” with your own.

curl -X POST -H 'Authorization: Bearer $DIGITALOCEAN_TOKEN' https://api.digitalocean.com/v2/gen-ai/models/api_keys

This will output our model access key. Save this value for later. We will need it to query the model.

Step 2B: Create a Model Access key with the Cloud Console

If we want to use the console, simply navigate to the GenAI platform tab “Serverless Inference” in the Console. Once there, use the button on the bottom right to “Create model access key”.

Then, all we have to do is name our key. Save the output value for later! We will need it for Serverless Inference with Python.

Step 3: Generating Text with Python and Serverless Inference

With our new model access key, we can begin running DigitalOcean Serverless Inference from any machine with internet access! We recommend working from a Jupyter Notebook. Follow the steps outlined in this tutorial for tips on setting up your environment for this article.

Once your env is set up, open a new .ipynb IPython Notebook file. Use the code snippet below to begin generating text, specifically answering what the capital of France is. Edit the snippet to reflect your changed API key value on line 5.

from openai import OpenAI
import os

client = OpenAI(
    api_key= **“your_model_access_key_here”**,
    base_url="https://inference.do-ai.run/v1"
)

stream = client.chat.completions.create(
    model="llama3-8b-instruct",
    messages=[
        {
            "role": "developer",
            "content": "You are a helpful assistant.",
        },
        {
            "role": "user",
            "content": "What is the capital of France?",
        },
    ],
    stream=True,
    max_completion_tokens=10
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

This should produce the output “The capital of France is Paris”. We recommend changing the values like max_completion_tokens to better suit your prompt as needed. We can also edit the model value on line 10 to be any of the available models listed in the DigitalOcean GenAI Model offerings. This model zoo is constantly being updated to reflect the growing industry, so check back frequently for updates.

Now that we have set up our environment to run Serverless Inference, we have a plethora of possible activities we can use the model for. From everything an LLM is capable of, we can create powerful agentic applications that leverage the strength of the AI model. Some possible use cases for this include:

• Event driven applications: Where a specific occurrence triggers the run of the LLM, this is common in agentic use cases
• Scalable backend services: When the inference on the backend may need to be scaled indefinitely, serverless can guarantee that your users will never be left waiting
• Data processing: batch jobs and data processing can be handled effectively and efficiently without requiring costly setup Source
• And much more!

Serverless Inference is a true answer for companies seeking LLM based solutions without the hassle of hiring or learning the required steps to deploying your own server. With DigitalOcean’s GenAI Platform, accessing Serverless Inference from powerful NVIDIA GPUs just became easier than ever! We encourage everyone to try the new solution!