1-Click Model GPU Droplets: 7 Powerful Assistant Steps

Once the Droplet is running, the inference pattern is straightforward. When you are connected to the GPU Droplet itself, requests can go to http://localhost:8080. When another machine calls the endpoint, replace localhost with the Droplet public IPv4 address and include the bearer token.

The source tutorial describes two practical routes: cURL and Python. A corrected cURL example looks like this:

curl http://localhost:8080/v1/chat/completions \
  -X POST \
  -H 'Content-Type: application/json' \
  -H "Authorization: Bearer $BEARER_TOKEN" \
  -d '{"messages":[{"role":"user","content":"What is deep learning?"}],"temperature":0.7,"top_p":0.95,"max_tokens":128}'

That request sends a user message to the model and returns a chat-completion response with the assistant message, finish reason, model name, and token usage. It is enough to prove that the endpoint is live before building a full interface.

The API shape is one reason 1-Click Model GPU Droplets are approachable: the application can use a familiar chat-completions pattern instead of custom inference plumbing.

Python is usually cleaner once the assistant needs application logic. The tutorial uses the Hugging Face Hub InferenceClient with the local endpoint as the base URL:

import os
from huggingface_hub import InferenceClient

client = InferenceClient(
    base_url="http://localhost:8080",
    api_key=os.getenv("BEARER_TOKEN"),
)

chat_completion = client.chat.completions.create(
    messages=[{"role": "user", "content": "What is deep learning?"}],
    temperature=0.7,
    top_p=0.95,
    max_tokens=128,
)

print(chat_completion.choices[0].message.content)

For a local Droplet demo, this is enough. For anything exposed outside the Droplet, add network controls, secret handling, request logging, and a clear rule for who can call the endpoint.