first commit

services/mcp-server/Dockerfile

@@ -0,0 +1,38 @@
+FROM python:3.12-slim
+
+WORKDIR /app
+
+# 系统依赖（MinerU 需要 libGL）
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    curl \
+    libgl1-mesa-glx \
+    libglib2.0-0 \
+    libsm6 \
+    libxrender1 \
+    libxext6 \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt .
+RUN pip install --no-cache-dir -r requirements.txt \
+    --index-url https://pypi.tuna.tsinghua.edu.cn/simple \
+    --trusted-host pypi.tuna.tsinghua.edu.cn
+
+# 预下载 MinerU 模型（构建时执行，加速启动）
+RUN python -c "
+import os
+os.environ['HF_ENDPOINT'] = 'https://hf-mirror.com'
+try:
+    from magic_pdf.model.doc_analyze_by_custom_model import ModelSingleton
+    print('MinerU 模型下载完成')
+except Exception as e:
+    print(f'模型下载跳过（将在运行时下载）: {e}')
+" || true
+
+COPY main.py .
+
+HEALTHCHECK --interval=30s --timeout=10s --start-period=120s --retries=3 \
+    CMD curl -f http://localhost:8011/health || exit 1
+
+EXPOSE 8011
+
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "8011", "--workers", "1"]

services/mcp-server/main.py

@@ -0,0 +1,136 @@
+import os
+import tempfile
+import logging
+from pathlib import Path
+
+from fastapi import FastAPI, UploadFile, File, HTTPException
+from pydantic import BaseModel
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+DEVICE = os.getenv("DEVICE", "cpu")
+UPLOAD_DIR = Path(os.getenv("UPLOAD_DIR", "/app/uploads"))
+PARSED_DIR = Path(os.getenv("PARSED_DIR", "/app/parsed"))
+
+UPLOAD_DIR.mkdir(parents=True, exist_ok=True)
+PARSED_DIR.mkdir(parents=True, exist_ok=True)
+
+app = FastAPI(title="MinerU 文档解析服务")
+
+SUPPORTED_TYPES = {
+    "application/pdf": "pdf",
+    "application/vnd.openxmlformats-officedocument.wordprocessingml.document": "docx",
+    "application/msword": "doc",
+    "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet": "xlsx",
+}
+
+
+def parse_pdf_mineru(pdf_path: str) -> str:
+    """使用 MinerU 解析 PDF"""
+    try:
+        from magic_pdf.data.data_reader_writer import FileBasedDataWriter
+        from magic_pdf.pipe.UnicodeFormulaPDFPipe import UnicodeFormulaPDFPipe
+
+        with tempfile.TemporaryDirectory() as tmpdir:
+            writer = FileBasedDataWriter(tmpdir)
+            pipe = UnicodeFormulaPDFPipe(pdf_path, writer)
+            pipe.pipe_classify()
+            pipe.pipe_analyze()
+            pipe.pipe_parse()
+            md_content = pipe.pipe_mk_uni_format(tmpdir, drop_mode="none")
+        return md_content or ""
+    except Exception as e:
+        logger.warning(f"MinerU 解析失败，降级到 PyMuPDF：{e}")
+        return parse_pdf_pymupdf(pdf_path)
+
+
+def parse_pdf_pymupdf(pdf_path: str) -> str:
+    """降级：使用 PyMuPDF 提取文本"""
+    try:
+        import fitz  # PyMuPDF
+        doc = fitz.open(pdf_path)
+        pages = []
+        for i, page in enumerate(doc):
+            text = page.get_text()
+            if text.strip():
+                pages.append(f"## 第 {i+1} 页\n\n{text}")
+        return "\n\n".join(pages)
+    except Exception as e:
+        return f"[解析失败：{e}]"
+
+
+def parse_docx(file_path: str) -> str:
+    """解析 Word 文档"""
+    try:
+        from docx import Document
+        doc = Document(file_path)
+        parts = []
+        for para in doc.paragraphs:
+            if para.text.strip():
+                style = para.style.name if para.style else ""
+                if "Heading" in style:
+                    level = style.replace("Heading ", "").strip()
+                    try:
+                        prefix = "#" * int(level)
+                    except ValueError:
+                        prefix = "##"
+                    parts.append(f"{prefix} {para.text}")
+                else:
+                    parts.append(para.text)
+        for table in doc.tables:
+            rows = []
+            for row in table.rows:
+                rows.append(" | ".join(cell.text.strip() for cell in row.cells))
+            if rows:
+                parts.append("\n".join(rows))
+        return "\n\n".join(parts)
+    except Exception as e:
+        return f"[Word 解析失败：{e}]"
+
+
+class ParseResponse(BaseModel):
+    filename: str
+    markdown: str
+    page_count: int
+    parser: str
+
+
+@app.post("/mineru-parse", response_model=ParseResponse)
+async def mineru_parse(file: UploadFile = File(...)) -> ParseResponse:
+    content = await file.read()
+    suffix = Path(file.filename or "doc.pdf").suffix.lower()
+
+    with tempfile.NamedTemporaryFile(suffix=suffix, delete=False) as tmp:
+        tmp.write(content)
+        tmp_path = tmp.name
+
+    try:
+        if suffix == ".pdf":
+            markdown = parse_pdf_mineru(tmp_path)
+            parser = "mineru"
+        elif suffix in (".docx", ".doc"):
+            markdown = parse_docx(tmp_path)
+            parser = "python-docx"
+        else:
+            raise HTTPException(status_code=415, detail=f"不支持的文件类型：{suffix}")
+
+        page_count = markdown.count("## 第") if suffix == ".pdf" else markdown.count("\n\n")
+        return ParseResponse(
+            filename=file.filename or "unknown",
+            markdown=markdown,
+            page_count=max(page_count, 1),
+            parser=parser,
+        )
+    finally:
+        os.unlink(tmp_path)
+
+
+@app.post("/parse-document", response_model=ParseResponse)
+async def parse_document(file: UploadFile = File(...)) -> ParseResponse:
+    return await mineru_parse(file)
+
+
+@app.get("/health")
+def health():
+    return {"status": "ok", "device": DEVICE}

services/mcp-server/requirements.txt

@@ -0,0 +1,11 @@
+fastapi>=0.115
+uvicorn[standard]>=0.30
+pydantic>=2.7
+python-multipart>=0.0.9
+httpx>=0.27
+# MinerU 文档解析
+mineru[pipeline]>=1.0
+# Word/Excel 降级解析
+python-docx>=1.1
+openpyxl>=3.1
+PyMuPDF>=1.24   # PDF 降级解析