Spaces:

TypeGPT
/

q

Running

App Files Files Community

yangjianchuan commited on Jan 26

Commit

96601d6

0 Parent(s):

first commit

Browse files

Files changed (6) hide show

Dockerfile +21 -0
README.md +104 -0
index.html +96 -0
qwen.py +299 -0
requirements.txt +3 -0
static/favicon.png +0 -0

Dockerfile ADDED Viewed

	@@ -0,0 +1,21 @@

+FROM python:3.9-slim
+# 设置工作目录
+WORKDIR /app
+# 复制依赖文件
+COPY requirements.txt .
+# 安装依赖
+RUN pip install -r requirements.txt
+# 复制应用代码
+COPY index.html .
+COPY qwen.py .
+COPY static ./static
+# 暴露端口
+EXPOSE 8000
+# 启动命令
+CMD ["python", "qwen.py"]

README.md ADDED Viewed

	@@ -0,0 +1,104 @@

+---
+title: Q
+emoji: 🏢
+colorFrom: indigo
+colorTo: purple
+sdk: docker
+pinned: false
+license: mit
+app_port: 8000
+---
+# 通义千问 API 代理服务器
+这是一个基于 FastAPI 实现的通义千问 API 代理服务器，用于转发和处理与通义千问 API 的通信。
+## 主要功能
+- 模型列表获取 API
+- 聊天完成 API
+- 支持流式响应
+- 内置模型列表缓存机制
+- 自动重试机制
+## 环境要求
+- Python 3.7+
+- FastAPI 0.104.1+
+- Uvicorn 0.24.0+
+- HTTPX 0.25.1+
+## 安装步骤
+1. 克隆项目到本地
+2. 安装依赖：
+```bash
+pip install -r requirements.txt
+```
+3. 使用 Docker（可选）：
+```bash
+docker build -t qwen-api-proxy .
+docker run -p 8000:8000 qwen-api-proxy
+```
+## 运行服务
+```bash
+python qwen.py
+```
+或使用 uvicorn：
+```bash
+uvicorn qwen:app --host 0.0.0.0 --port 8000
+```
+服务将在 http://localhost:8000 上运行。
+## API 接口说明
+### 1. 获取模型列表
+```
+GET /api/models
+Header: Authorization: Bearer <your-token>
+```
+返回可用的模型列表，结果会被缓存1小时。
+### 2. 聊天完成
+```
+POST /api/chat/completions
+Header: Authorization: Bearer <your-token>
+{
+    "model": "string",
+    "messages": [
+        {
+            "role": "user",
+            "content": "string"
+        }
+    ],
+    "stream": boolean,
+    "max_tokens": number (可选)
+}
+```
+支持流式和非流式响应，可以通过 stream 参数控制。
+## 错误处理
+- 服务内置了自动重试机制，最多重试3次
+- 500错误或HTML响应会触发重试
+- 401错误表示未授权，需要检查token
+- 400错误表示请求参数有误
+## 获取 API Key
+1. 访问 https://chat.qwenlm.ai/ 并登录
+2. 打开浏览器开发者工具（通常按 F12）
+3. 切换到"应用程序"选项卡
+4. 在左侧菜单中选择"Cookies" -> "https://chat.qwenlm.ai"
+5. 找到名称为"token"的cookie
+6. 复制其值，这就是你的API Key

index.html ADDED Viewed

	@@ -0,0 +1,96 @@

+<!DOCTYPE html>
+<html>
+<head>
+    <title>API 接口说明</title>
+    <link rel="icon" type="image/png" href="/static/favicon.png">
+    <style>
+        body { font-family: Arial, sans-serif; margin: 20px; }
+        h1 { color: #333; }
+        .endpoint { margin: 20px 0; padding: 15px; background: #f5f5f5; border-radius: 5px; }
+        .method { font-weight: bold; color: #007bff; }
+        .url { color: #28a745; }
+        .description { margin-top: 10px; }
+    </style>
+</head>
+<body>
+    <h1>API 接口说明</h1>
+    <h2>项目概述</h2>
+    <p>这是一个基于 FastAPI 实现的通义千问 API 代理服务器，用于转发和处理与通义千问 API 的通信。</p>
+    <h2>主要功能</h2>
+    <ul>
+        <li>模型列表获取 API</li>
+        <li>聊天完成 API</li>
+        <li>支持流式响应</li>
+        <li>内置模型列表缓存机制</li>
+        <li>自动重试机制</li>
+    </ul>
+    <h2>环境要求</h2>
+    <ul>
+        <li>Python 3.7+</li>
+        <li>FastAPI 0.104.1+</li>
+        <li>Uvicorn 0.24.0+</li>
+        <li>HTTPX 0.25.1+</li>
+    </ul>
+    <h2>安装步骤</h2>
+    <ol>
+        <li>克隆项目到本地</li>
+        <li>安装依赖：
+            <pre><code>pip install -r requirements.txt</code></pre>
+        </li>
+        <li>使用 Docker（可选）：
+            <pre><code>docker build -t qwen-api-proxy .
+docker run -p 8000:8000 qwen-api-proxy</code></pre>
+        </li>
+    </ol>
+    <h2>运行服务</h2>
+    <pre><code>python qwen.py</code></pre>
+    <p>或使用 uvicorn：</p>
+    <pre><code>uvicorn qwen:app --host 0.0.0.0 --port 8000</code></pre>
+    <p>服务将在 <a href="http://localhost:8000">http://localhost:8000</a> 上运行。</p>
+    <h2>错误处理</h2>
+    <ul>
+        <li>服务内置了自动重试机制，最多重试3次</li>
+        <li>500错误或HTML响应会触发重试</li>
+        <li>401错误表示未授权，需要检查token</li>
+        <li>400错误表示请求参数有误</li>
+    </ul>
+    <h2>获取 API Key</h2>
+    <ol>
+        <li>访问 <a href="https://chat.qwenlm.ai/">https://chat.qwenlm.ai/</a> 并登录</li>
+        <li>打开浏览器开发者工具（通常按 F12）</li>
+        <li>切换到"应用程序"选项卡</li>
+        <li>在左侧菜单中选择"Cookies" -> "https://chat.qwenlm.ai"</li>
+        <li>找到名称为"token"的cookie</li>
+        <li>复制其值，这就是你的API Key</li>
+    </ol>
+    <h2>许可证</h2>
+    <p>本项目采用 MIT License 开源许可证。</p>
+    <div class="endpoint">
+        <div class="method">GET</div>
+        <div class="url">/api/models</div>
+        <div class="description">
+            获取可用模型列表<br>
+            请求头需要包含 Authorization: Bearer {api_key}
+        </div>
+    </div>
+    <div class="endpoint">
+        <div class="method">POST</div>
+        <div class="url">/api/chat/completions</div>
+        <div class="description">
+            与模型进行对话<br>
+            请求头需要包含 Authorization: Bearer {api_key}<br>
+            支持流式响应（stream: true）
+        </div>
+    </div>
+</body>
+</html>

qwen.py ADDED Viewed

	@@ -0,0 +1,299 @@

+from fastapi import FastAPI, Request, Response, UploadFile, File
+from fastapi.responses import StreamingResponse, FileResponse
+from fastapi.staticfiles import StaticFiles
+import httpx
+import json
+import asyncio
+import time
+import base64
+from typing import Optional, Dict, Any, List
+from io import BytesIO
+# 配置常量
+QWEN_API_URL = "https://chat.qwenlm.ai/api/chat/completions"
+QWEN_MODELS_URL = "https://chat.qwenlm.ai/api/models"
+QWEN_FILES_URL = "https://chat.qwenlm.ai/api/v1/files/"
+MAX_RETRIES = 3
+RETRY_DELAY = 1  # 1秒
+# 缓存设置
+cached_models = None
+cached_models_timestamp = 0
+CACHE_TTL = 60 * 60  # 缓存1小时
+app = FastAPI()
+app.mount("/static", StaticFiles(directory="static"), name="static")
+client = httpx.AsyncClient()
+@app.get("/")
+async def root():
+    return FileResponse("index.html")
+async def sleep(seconds: float):
+    await asyncio.sleep(seconds)
+# 添加 base64 转换为文件的函数
+async def base64_to_file(base64_str: str) -> BytesIO:
+    try:
+        # 去除 data:image/jpeg;base64, 这样的前缀
+        if ',' in base64_str:
+            base64_str = base64_str.split(',', 1)[1]
+        # 解码 base64 数据
+        image_data = base64.b64decode(base64_str)
+        return BytesIO(image_data)
+    except Exception as e:
+        raise Exception(f"Failed to convert base64 to file: {str(e)}")
+# 添加图片上传函数
+async def upload_image_to_qwen(auth_header: str, image_data: BytesIO) -> str:
+    try:
+        files = {'file': ('image.jpg', image_data, 'image/jpeg')}
+        headers = {
+            "Authorization": auth_header,
+            "accept": "application/json"
+        }
+        async with httpx.AsyncClient() as client:
+            response = await client.post(
+                QWEN_FILES_URL,
+                headers=headers,
+                files=files
+            )
+            if response.is_success:
+                data = response.json()
+                if not data.get('id'):
+                    raise Exception("File upload failed: No valid file ID returned")
+                return data['id']
+            else:
+                raise Exception(f"File upload failed with status {response.status_code}")
+    except Exception as e:
+        raise Exception(f"Failed to upload image: {str(e)}")
+# 添加消息处理函数
+async def process_messages(messages: List[Dict], auth_header: str) -> List[Dict]:
+    processed_messages = []
+    for message in messages:
+        if isinstance(message.get('content'), list):
+            new_content = []
+            for content in message['content']:
+                if (content.get('type') == 'image_url' and
+                    content.get('image_url', {}).get('url', '').startswith('data:')):
+                    # 处理 base64 图片
+                    image_data = await base64_to_file(content['image_url']['url'])
+                    image_id = await upload_image_to_qwen(auth_header, image_data)
+                    new_content.append({
+                        'type': 'image',
+                        'image': image_id
+                    })
+                else:
+                    new_content.append(content)
+            message['content'] = new_content
+        processed_messages.append(message)
+    return processed_messages
+async def fetch_with_retry(url: str, options: Dict, retries: int = MAX_RETRIES):
+    last_error = None
+    for i in range(retries):
+        try:
+            response = await client.request(
+                method=options.get("method", "GET"),
+                url=url,
+                headers=options.get("headers", {}),
+                json=options.get("json"),
+            )
+            if response.is_success:
+                return response
+            content_type = response.headers.get("content-type", "")
+            if response.status_code >= 500 or "text/html" in content_type:
+                last_error = {
+                    "status": response.status_code,
+                    "content_type": content_type,
+                    "response_text": response.text[:1000],
+                    "headers": dict(response.headers)
+                }
+                if i < retries - 1:
+                    await sleep(RETRY_DELAY * (i + 1))
+                    continue
+            else:
+                last_error = {
+                    "status": response.status_code,
+                    "headers": dict(response.headers)
+                }
+                break
+        except Exception as error:
+            last_error = error
+            if i < retries - 1:
+                await sleep(RETRY_DELAY * (i + 1))
+                continue
+    raise Exception(json.dumps({
+        "error": True,
+        "message": "All retry attempts failed",
+        "last_error": str(last_error),
+        "retries": retries
+    }))
+async def process_line(line: str, previous_content: str) -> tuple[str, Optional[dict]]:
+    try:
+        data = json.loads(line[6:])  # 移除 "data: " 前缀
+        if (data.get("choices") and data["choices"][0].get("delta")):
+            delta = data["choices"][0]["delta"]
+            current_content = delta.get("content", "")
+            # 计算增量内容
+            if previous_content and current_content:
+                if current_content.startswith(previous_content):
+                    new_content = current_content[len(previous_content):]
+                else:
+                    new_content = current_content
+            else:
+                new_content = current_content
+            # 构造新的响应数据
+            new_data = {
+                "choices": [{
+                    "delta": {
+                        "role": delta.get("role", "assistant"),
+                        "content": new_content
+                    }
+                }]
+            }
+            return current_content, new_data
+        return previous_content, data
+    except:
+        return previous_content, None
+async def stream_generator(response: httpx.Response):
+    buffer = ""
+    previous_content = ""
+    async for chunk in response.aiter_bytes():
+        chunk_text = chunk.decode()
+        buffer += chunk_text
+        lines = buffer.split("\n")
+        buffer = lines.pop() if lines else ""
+        for line in lines:
+            line = line.strip()
+            if line.startswith("data: "):
+                previous_content, data = await process_line(line, previous_content)
+                if data:
+                    yield f"data: {json.dumps(data)}\n\n"
+    if buffer:
+        previous_content, data = await process_line(buffer, previous_content)
+        if data:
+            yield f"data: {json.dumps(data)}\n\n"
+    yield "data: [DONE]\n\n"
+@app.get("/healthz")
+async def health_check():
+    return {"status": "ok"}
+@app.get("/api/models")
+async def get_models(request: Request):
+    global cached_models, cached_models_timestamp
+    auth_header = request.headers.get("Authorization")
+    if not auth_header or not auth_header.startswith("Bearer "):
+        return Response(status_code=401, content="Unauthorized")
+    now = time.time()
+    if cached_models and now - cached_models_timestamp < CACHE_TTL:
+        return Response(
+            content=cached_models,
+            media_type="application/json"
+        )
+    try:
+        response = await fetch_with_retry(
+            QWEN_MODELS_URL,
+            {"headers": {"Authorization": auth_header}}
+        )
+        cached_models = response.text
+        cached_models_timestamp = now
+        return Response(
+            content=cached_models,
+            media_type="application/json"
+        )
+    except Exception as error:
+        return Response(
+            content=json.dumps({"error": True, "message": str(error)}),
+            status_code=500
+        )
+@app.post("/api/chat/completions")
+async def chat_completions(request: Request):
+    auth_header = request.headers.get("Authorization")
+    if not auth_header or not auth_header.startswith("Bearer "):
+        return Response(status_code=401, content="Unauthorized")
+    request_data = await request.json()
+    messages = request_data.get("messages")
+    stream = request_data.get("stream", False)
+    model = request_data.get("model")
+    max_tokens = request_data.get("max_tokens")
+    if not model:
+        return Response(
+            content=json.dumps({"error": True, "message": "Model parameter is required"}),
+            status_code=400
+        )
+    try:
+        # 处理消息中的图片
+        processed_messages = await process_messages(messages, auth_header)
+        qwen_request = {
+            "model": model,
+            "messages": processed_messages,
+            "stream": stream
+        }
+        if max_tokens is not None:
+            qwen_request["max_tokens"] = max_tokens
+        response = await client.post(
+            QWEN_API_URL,
+            headers={
+                "Content-Type": "application/json",
+                "Authorization": auth_header
+            },
+            json=qwen_request
+        )
+        if stream:
+            return StreamingResponse(
+                stream_generator(response),
+                media_type="text/event-stream"
+            )
+        return Response(
+            content=response.text,
+            media_type="application/json"
+        )
+    except Exception as error:
+        return Response(
+            content=json.dumps({"error": True, "message": str(error)}),
+            status_code=500
+        )
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=8000)

requirements.txt ADDED Viewed

	@@ -0,0 +1,3 @@

+fastapi==0.104.1
+uvicorn==0.24.0
+httpx==0.25.1

static/favicon.png ADDED Viewed