Hugging Face Inference API 开发指南

import requests
import os

# 最佳实践：将您的令牌存储在环境变量中
# 在本示例中，我们直接定义它，但在生产环境请使用 os.getenv("HF_API_TOKEN")。
API_TOKEN = "YOUR_API_TOKEN"
API_URL = "https://api-inference.huggingface.co/models/distilbert/distilbert-base-uncased-finetuned-sst-2-english"

def query_model(payload):
    headers = {"Authorization": f"Bearer {API_TOKEN}"}
    response = requests.post(API_URL, headers=headers, json=payload)
    response.raise_for_status()  # 为错误的响应代码引发异常
    return response.json()

# 让我们对一个句子进行分类
data_payload = {
    "inputs": "I love the new features in this software, it's amazing!"
}

try:
    output = query_model(data_payload)
    print(output)
    # 预期输出可能如下所示： [[{'label': 'POSITIVE', 'score': 0.9998...}]]
except requests.exceptions.RequestException as e:
    print(f"发生错误： {e}")

import requests

API_URL = "https://api-inference.huggingface.co/models/gpt2"
headers = {"Authorization": "Bearer YOUR_API_TOKEN"}

def query_text_generation(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

output = query_text_generation({
    "inputs": "The future of AI in software development will be",
    "parameters": {"max_length": 50, "temperature": 0.7}
})
print(output)
# 预期输出： [{'generated_text': 'The future of AI in software development will be...'}]

import requests

API_URL = "https://api-inference.huggingface.co/models/google/vit-base-patch16-224"
headers = {"Authorization": "Bearer YOUR_API_TOKEN"}

def query_image_classification(filename):
    with open(filename, "rb") as f:
        data = f.read()
    response = requests.post(API_URL, headers=headers, data=data)
    return response.json()

# 请确保在同一目录下有一个图像文件（例如 'cat.jpg'）
try:
    output = query_image_classification("cat.jpg")
    print(output)
    # 预期输出： [{'score': 0.99..., 'label': 'Egyptian cat'}, {'score': 0.00..., 'label': 'tabby, tabby cat'}, ...]
except FileNotFoundError:
    print("错误： 'cat.jpg' 未找到。请提供有效的图像文件路径。")

// 使用 fetch 的 JavaScript 示例
async function queryZeroShot(data) {
    const response = await fetch(
        "https://api-inference.huggingface.co/models/facebook/bart-large-mnli",
        {
            headers: { Authorization: "Bearer YOUR_API_TOKEN" },
            method: "POST",
            body: JSON.stringify(data),
        }
    );
    const result = await response.json();
    return result;
}

queryZeroShot({
    "inputs": "Our new feature launch was a massive success!",
    "parameters": {"candidate_labels": ["marketing", "customer feedback", "technical issue"]}
}).then((response) => {
    console.log(JSON.stringify(response));
    // 预期输出： {"sequence": "...", "labels": ["customer feedback", ...], "scores": [0.98..., ...]}
});

# 之前： 直接 API 调用 Hugging Face
# 这会创建单点故障。
import requests

HF_API_URL = "https://api-inference.huggingface.co/models/gpt2"
HF_TOKEN = "YOUR_HF_TOKEN"

def direct_hf_call(payload):
    headers = {"Authorization": f"Bearer {HF_TOKEN}"}
    response = requests.post(HF_API_URL, headers=headers, json=payload)
    return response.json()

# 之后： 调用统一的 EvoLink API (兼容 OpenAI)
# 您的应用程序现在通过自动故障转移和负载均衡具备了弹性。
import requests

# EvoLink 统一 API 端点 (兼容 OpenAI)
EVOLINK_API_URL = "https://api.evolink.ai/v1"
EVOLINK_TOKEN = "YOUR_EVOLINK_TOKEN"

def evolink_image_generation(prompt):
    """
    使用 EvoLink 的智能路由生成图像。
    EvoLink 自动路由到为您选择的模型提供服务的成本最低的提供商。
    """
    headers = {"Authorization": f"Bearer {EVOLINK_TOKEN}"}

    # 典型示例： 使用 Seedream 4.0 生成故事驱动的 4K 图像
    payload = {
        'model': 'doubao-seedream-4.0',  # 或 'gpt-4o-image', 'nano-banana'
        'prompt': prompt,
        'size': '1024x1024'
    }

    response = requests.post(f"{EVOLINK_API_URL}/images/generations",
                            headers=headers, json=payload)
    return response.json()

def evolink_video_generation(prompt):
    """
    使用 EvoLink 的视频模型生成视频。
    """
    headers = {"Authorization": f"Bearer {EVOLINK_TOKEN}"}

    # 典型示例： 使用 Sora 2 生成带音频的 10 秒视频
    payload = {
        'model': 'sora-2',  # 或 'veo3-fast' 生成 8 秒视频
        'prompt': prompt,
        'duration': 10
    }

    response = requests.post(f"{EVOLINK_API_URL}/videos/generations",
                            headers=headers, json=payload)
    return response.json()