Dockerfile

FROM python:3.11-slim

WORKDIR /app

# 1. 安装必要的 Python 库
RUN pip install fastapi uvicorn sentence-transformers torch --index-url https://pypi.tuna.tsinghua.edu.cn/simple

# 2. 创建一个稳健的后端服务 (app.py)
RUN echo "from fastapi import FastAPI, Request, HTTPException" > app.py && \
    echo "from fastapi.responses import JSONResponse" >> app.py && \
    echo "from sentence_transformers import SentenceTransformer" >> app.py && \
    echo "import os, uvicorn" >> app.py && \
    echo "MODEL_ID = os.getenv('MODEL_ID', 'microsoft/harrier-oss-v1-0.6b')" >> app.py && \
    echo "print(f'正在加载模型: {MODEL_ID}')" >> app.py && \
    echo "model = SentenceTransformer(MODEL_ID, device='cpu')" >> app.py && \
    echo "print('模型加载完毕！准备接收请求...')" >> app.py && \
    echo "app = FastAPI()" >> app.py && \
    echo "@app.post('/v1/embeddings')" >> app.py && \
    echo "async def create_embeddings(request: Request):" >> app.py && \
    echo "    try:" >> app.py && \
    echo "        body = await request.json()" >> app.py && \
    echo "        input_text = body.get('input', '')" >> app.py && \
    echo "        if not input_text:" >> app.py && \
    echo "            raise HTTPException(422, '缺少 input 字段')" >> app.py && \
    echo "        embeddings = model.encode(input_text).tolist()" >> app.py && \
    echo "        return JSONResponse({" >> app.py && \
    echo "            'object': 'list'," >> app.py && \
    echo "            'data': [{'object': 'embedding', 'embedding': embeddings, 'index': 0}]," >> app.py && \
    echo "            'model': MODEL_ID" >> app.py && \
    echo "        })" >> app.py && \
    echo "    except HTTPException:" >> app.py && \
    echo "        raise" >> app.py && \
    echo "    except Exception as e:" >> app.py && \
    echo "        return JSONResponse({'error': str(e)}, status_code=500)" >> app.py && \
    echo "if __name__ == '__main__':" >> app.py && \
    echo "    uvicorn.run(app, host='0.0.0.0', port=7860)" >> app.py

# 3. 启动服务
CMD ["python", "app.py"]