Upload 3 files

Procfile

@@ -0,0 +1 @@
+web: gunicorn app:app

app.py

@@ -1,56 +1,108 @@
+from flask import Flask, request, jsonify
 from sentence_transformers import SentenceTransformer
-import gradio as gr
 import numpy as np
+import os
+import time
 
-# 加载模型（首次运行会自动下载，约 500MB）
-model = SentenceTransformer('BAAI/bge-small-en-v1.5')
-
-def get_embedding(text: str) -> list:
-    """生成文本嵌入向量"""
-    if not text.strip():
-        return "请输入非空文本"
-    # 生成嵌入（返回 numpy 数组）
-    embedding = model.encode(text, normalize_embeddings=True)
-    # 转换为列表返回（方便 API 传输）
-    return embedding.tolist()
-
-def similarity_score(text1: str, text2: str) -> float:
-    """计算两个文本的余弦相似度"""
-    if not text1.strip() or not text2.strip():
-        return 0.0
-    emb1 = model.encode(text1, normalize_embeddings=True)
-    emb2 = model.encode(text2, normalize_embeddings=True)
-    # 余弦相似度 = 向量点积（已归一化）
-    return float(np.dot(emb1, emb2))
-
-# 创建 Gradio 界面
-with gr.Blocks(title="开源文本嵌入 API") as demo:
-    gr.Markdown("# 文本嵌入服务（基于 BAAI/bge-small-en-v1.5）")
+app = Flask(__name__)
+
+# 加载模型
+model_name = "BAAI/bge-small-en-v1.5"
+model = SentenceTransformer(model_name)
+
+# 支持的模型列表
+SUPPORTED_MODELS = {
+    "text-embedding-3-small": model,
+    "bge-small-en-v1.5": model
+}
+
+# 简单的API密钥验证（可选）
+API_KEY = os.getenv("API_KEY", "your-default-api-key")
+
+def verify_api_key(headers):
+    """验证API密钥"""
+    auth_header = headers.get("Authorization")
+    if not auth_header or not auth_header.startswith("Bearer "):
+        return False
+    return auth_header.split("Bearer ")[1] == API_KEY
+
+@app.route('/v1/embeddings', methods=['POST'])
+def create_embedding():
+    """生成嵌入向量，兼容OpenAI API格式"""
+    # 验证API密钥
+    if not verify_api_key(request.headers):
+        return jsonify({
+            "error": {
+                "message": "Invalid API key",
+                "type": "invalid_request_error",
+                "param": None,
+                "code": "invalid_api_key"
+            }
+        }), 401
     
-    with gr.Tab("生成嵌入向量"):
-        input_text = gr.Textbox(label="输入文本", placeholder="请输入需要生成嵌入的文本...")
-        embedding_output = gr.Textbox(label="嵌入向量（前10位）")
-        generate_btn = gr.Button("生成嵌入")
-        generate_btn.click(
-            fn=lambda x: str(get_embedding(x)[:10]) + "...",  # 只显示前10位
-            inputs=input_text,
-            outputs=embedding_output
-        )
+    # 解析请求
+    data = request.json
+    if not data or "input" not in data:
+        return jsonify({
+            "error": {
+                "message": "Missing input",
+                "type": "invalid_request_error",
+                "param": None,
+                "code": "missing_input"
+            }
+        }), 400
     
-    with gr.Tab("计算语义相似度"):
-        text1 = gr.Textbox(label="文本1", placeholder="输入第一个文本...")
-        text2 = gr.Textbox(label="文本2", placeholder="输入第二个文本...")
-        similarity_output = gr.Number(label="余弦相似度（0~1，越高越相似）")
-        similarity_btn = gr.Button("计算相似度")
-        similarity_btn.click(
-            fn=similarity_score,
-            inputs=[text1, text2],
-            outputs=similarity_output
-        )
+    # 获取模型（默认为text-embedding-3-small）
+    model_name = data.get("model", "text-embedding-3-small")
+    if model_name not in SUPPORTED_MODELS:
+        return jsonify({
+            "error": {
+                "message": f"Model {model_name} not found",
+                "type": "invalid_request_error",
+                "param": None,
+                "code": "model_not_found"
+            }
+        }), 404
     
-    # 启用队列，支持并发请求
-    demo.queue()
+    # 处理输入（支持单文本或文本列表）
+    inputs = data["input"]
+    if isinstance(inputs, str):
+        inputs = [inputs]
+    
+    # 计算嵌入向量
+    start_time = time.time()
+    embeddings = model.encode(inputs, normalize_embeddings=True)
+    processing_time = time.time() - start_time
+    
+    # 准备响应数据
+    response_data = {
+        "object": "list",
+        "data": [
+            {
+                "object": "embedding",
+                "embedding": embedding.tolist(),
+                "index": i
+            } for i, embedding in enumerate(embeddings)
+        ],
+        "model": model_name,
+        "usage": {
+            "prompt_tokens": sum(len(text.split()) for text in inputs),  # 简单估算
+            "total_tokens": sum(len(text.split()) for text in inputs)
+        }
+    }
+    
+    return jsonify(response_data)
+
+@app.route('/health', methods=['GET'])
+def health_check():
+    """健康检查接口"""
+    return jsonify({
+        "status": "healthy",
+        "model": model_name,
+        "supported_models": list(SUPPORTED_MODELS.keys())
+    })
 
-if __name__ == "__main__":
-    # 部署到 Hugging Face Spaces 时，不需要指定 server_name 和 server_port
-    demo.launch()
+if __name__ == '__main__':
+    # 生产环境应使用Gunicorn等WSGI服务器
+    app.run(host='0.0.0.0', port=int(os.getenv('PORT', 7860)))
+    

requirements.txt

@@ -1,4 +1,5 @@
-gradio==4.28.3
+flask==2.3.3
 sentence-transformers==2.7.0
 torch==2.2.2
 numpy==1.26.4
+gunicorn==21.2.0  # 用于生产环境部署