Deploying hugging face's free api on cloudflare to support interface forwarding

Since the hugging face is inaccessible for domestic deployments, it's in the Big BrotherDeployment programthat were adapted to be able to deploy to cloudflare workers.

preliminary

1. Registration cloudflare
2. Register hugging face and apply for api key.Request api key address
3. Copy the following code and deploy it to cloudflare workers.
4、Support in oneapi/newapi click "get model list" a key to add available models

The cloudflare workers code is as follows:

//对接one-api/new-api使用
const API_KEY = "sk-1234567890";

//你的hugging face api key去hugging face申请
const HUGGINGFACE_API_KEY = "hf_xxxxxxxxxxx";

//目前发现的可用模型，请求时如模型不在该列表内，则使用你请求的模型
const CUSTOMER_MODEL_MAP = {
"qwen2.5-72b-instruct": "Qwen/Qwen2.5-72B-Instruct",
"gemma2-2b-it": "google/gemma-2-2b-it",
"gemma2-27b-it": "google/gemma-2-27b-it",
"llama-3-8b-instruct": "meta-llama/Meta-Llama-3-8B-Instruct",
"llama-3.2-1b-instruct": "meta-llama/Llama-3.2-1B-Instruct",
"llama-3.2-3b-instruct": "meta-llama/Llama-3.2-3B-Instruct",
"phi-3.5": "microsoft/Phi-3.5-mini-instruct"
};

async function handleRequest(request) {
try {
if (request.method === "OPTIONS") {
return getResponse("", 204);
}

const authHeader = request.headers.get("Authorization");
if (!authHeader || !authHeader.startsWith("Bearer ") || authHeader.split(" ")[1] !== API_KEY) {
return getResponse("Unauthorized", 401);
}

if (request.url.endsWith("/v1/models")) {
const arrs = [];
Object.keys(CUSTOMER_MODEL_MAP).map(element => arrs.push({ id: element, object: "model" }))
const response = {
data: arrs,
success: true
};

return getResponse(JSON.stringify(response), 200);
}

if (request.method !== "POST") {
return getResponse("Only POST requests are allowed", 405);
}

if (!request.url.endsWith("/v1/chat/completions")) {
return getResponse("Not Found", 404);
}

const data = await request.json();
const messages = data.messages || [];
const model = CUSTOMER_MODEL_MAP[data.model] || data.model;
const temperature = data.temperature || 0.7;
const max_tokens = data.max_tokens || 8196;
const top_p = Math.min(Math.max(data.top_p || 0.9, 0.0001), 0.9999);
const stream = data.stream || false;

const requestBody = {
model: model,
stream: stream,
temperature: temperature,
max_tokens: max_tokens,
top_p: top_p,
messages: messages
};

const apiUrl = `https://api-inference.huggingface.co/models/${model}/v1/chat/completions`;
const response = await fetch(apiUrl, {
method: 'POST',
headers: {
'Authorization': `Bearer ${HUGGINGFACE_API_KEY}`,
'Content-Type': 'application/json'
},
body: JSON.stringify(requestBody)
});

if (!response.ok) {
const errorText = await response.text();
return getResponse(`Error from API: ${response.statusText} - ${errorText}`, response.status);
}

const newResponse = new Response(response.body, {
status: response.status,
headers: {
...Object.fromEntries(response.headers),
'Access-Control-Allow-Origin': '*',
'Access-Control-Allow-Methods': '*',
'Access-Control-Allow-Headers': '*'
}
});

return newResponse;
} catch (error) {
return getResponse(JSON.stringify({
error: `处理请求失败: ${error.message}`
}), 500);
}
}

function getResponse(resp, status) {
return new Response(resp, {
status: status,
headers: {
"Content-Type": "application/json",
"Access-Control-Allow-Origin": "*",
"Access-Control-Allow-Methods": "*",
"Access-Control-Allow-Headers": "*"
}
});
}

addEventListener('fetch', event => {
event.respondWith(handleRequest(event.request))
})

The Thousand Questions 2.5 model is recommended, all other good models require a subscription.