Testing the Hugging Face Inference API (Free Tier)

Learn how to use Hugging Face's free inference API for AI model testing.

Prerequisites

• Python 3.8+
• requests library: pip install requests
• Hugging Face account (free at https://huggingface.co)

Quick Start

import requests
import os

# Get your free token from Hugging Face settings
HF_TOKEN = os.getenv("HF_TOKEN")

# Use a free model (no auth required for some)
MODEL = "google/flan-t5-base"

def query_hf(prompt: str, model: str = MODEL) -> str:
    """Query Hugging Face Inference API."""
    url = f"https://api-inference.huggingface.co/models/{model}"
    headers = {"Authorization": f"Bearer {HF_TOKEN}"} if HF_TOKEN else {}

    response = requests.post(
        url,
        headers=headers,
        json={"inputs": prompt},
        timeout=30
    )

    if response.status_code == 200:
        return response.json()[0]["generated_text"]
    else:
        raise Exception(f"Error: {response.status_code} - {response.text}")

# Usage
if __name__ == "__main__":
    prompt = "Translate to French: Hello, how are you?"
    try:
        result = query_hf(prompt)
        print(f"Prompt: {prompt}")
        print(f"Result: {result}")
    except Exception as e:
        print(f"Error: {e}")

Testing Different Free Models

# Text generation
models = [
    "gpt2",
    "facebook/opt-350m",
    "google/flan-t5-base",
    "microsoft/DialoGPT-small"
]

for model in models:
    try:
        result = query_hf("The future of AI is", model)
        print(f"{model}: {result[:100]}...")
    except Exception as e:
        print(f"{model}: Error - {e}")

Error Handling & Rate Limits

import time

def robust_query(prompt: str, model: str = MODEL, retries: int = 3) -> str:
    """Query with exponential backoff for rate limits."""
    for attempt in range(retries):
        try:
            return query_hf(prompt, model)
        except Exception as e:
            if "429" in str(e) or "503" in str(e):
                wait = 2 ** attempt
                print(f"Rate limited. Waiting {wait}s...")
                time.sleep(wait)
            else:
                raise
    raise Exception("Max retries exceeded")

Conclusion

Hugging Face's free inference API provides access to thousands of models without any cost. Perfect for experimentation, prototyping, and learning AI integration.