pdf-capture-mcp 0.1.0__tar.gz

pdf_capture_mcp-0.1.0/.github/workflows/ci.yml

@@ -0,0 +1,38 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - run: uv sync --extra dev
+      - run: uv run ruff check src/ tests/
+      - run: uv run ruff format --check src/ tests/
+
+  typecheck:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - run: uv sync --extra dev
+      - run: uv run mypy src/pdf_capture_mcp/ --ignore-missing-imports
+
+  test:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-latest, macos-latest]
+        python-version: ["3.11", "3.12"]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - run: uv python install ${{ matrix.python-version }}
+      - run: uv sync --extra dev --python ${{ matrix.python-version }}
+      - run: uv run pytest tests/ -v --tb=short

pdf_capture_mcp-0.1.0/.github/workflows/publish.yml

@@ -0,0 +1,34 @@
+name: Publish to PyPI
+
+on:
+  release:
+    types: [published]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - name: Build package
+        run: uv build
+      - name: Upload artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+  publish:
+    needs: build
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write  # Required for PyPI trusted publishing
+    steps:
+      - name: Download artifacts
+        uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+      - uses: astral-sh/setup-uv@v5
+      - name: Publish to PyPI
+        run: uv publish --trusted-publishing always

pdf_capture_mcp-0.1.0/.gitignore

@@ -0,0 +1,35 @@
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+dist/
+build/
+.eggs/
+
+# Virtual environments
+.venv/
+venv/
+uv.lock
+
+# Testing & tooling caches
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+.coverage
+htmlcov/
+
+# IDE
+.idea/
+.vscode/
+.qoder/
+
+# OS
+.DS_Store
+
+# Local artifacts
+*.log
+tmp/
+._*
+.pytest_cache/
+.ruff_cache/
+.venv/

pdf_capture_mcp-0.1.0/LICENSE

@@ -0,0 +1,33 @@
+MIT License
+
+Copyright (c) 2026 Luis
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+
+---
+
+Third-party license notices:
+
+- marker-pdf (Apache 2.0): https://github.com/datalab-to/marker
+- MinerU (AGPL-3.0): https://github.com/opendatalab/MinerU
+  MinerU is invoked as a separate subprocess and is not linked into this package.
+- pdfplumber (MIT): https://github.com/jsvine/pdfplumber
+- Table Transformer (MIT): https://github.com/microsoft/table-transformer
+- DePlot (Apache 2.0): https://github.com/google-research/google-research/tree/master/deplot
+- pymupdf4llm (Apache 2.0): https://github.com/pymupdf/pymupdf4llm

pdf_capture_mcp-0.1.0/PKG-INFO

@@ -0,0 +1,39 @@
+Metadata-Version: 2.4
+Name: pdf-capture-mcp
+Version: 0.1.0
+Summary: Multi-phase PDF capture pipeline as an MCP server. Convert PDF to high-quality structured Markdown.
+Project-URL: Repository, https://github.com/ChenHongYu2026/pdf-capture-mcp
+Project-URL: Issues, https://github.com/ChenHongYu2026/pdf-capture-mcp/issues
+Author: ChenHongYu2026
+License-Expression: MIT
+License-File: LICENSE
+Keywords: extraction,markdown,mcp,ocr,pdf
+Classifier: Development Status :: 3 - Alpha
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Text Processing :: Markup :: Markdown
+Requires-Python: >=3.11
+Requires-Dist: fastmcp<4,>=3.2.0
+Requires-Dist: httpx>=0.27
+Requires-Dist: numpy>=1.26
+Requires-Dist: pdfplumber>=0.11.0
+Requires-Dist: pillow>=10.0
+Requires-Dist: pymupdf4llm>=0.0.17
+Requires-Dist: pypdf>=5.0
+Provides-Extra: all
+Requires-Dist: marker-pdf>=2.0; extra == 'all'
+Requires-Dist: torch>=2.0; extra == 'all'
+Requires-Dist: transformers>=4.40; extra == 'all'
+Provides-Extra: dev
+Requires-Dist: mypy>=1.15; extra == 'dev'
+Requires-Dist: pytest-asyncio>=0.24; extra == 'dev'
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: ruff>=0.15; extra == 'dev'
+Provides-Extra: marker
+Requires-Dist: marker-pdf>=2.0; extra == 'marker'
+Provides-Extra: ml
+Requires-Dist: torch>=2.0; extra == 'ml'
+Requires-Dist: transformers>=4.40; extra == 'ml'
+Provides-Extra: vlm

pdf_capture_mcp-0.1.0/README.md

@@ -0,0 +1,347 @@
+<a id="english"></a>
+
+# pdf-capture-mcp
+
+**English** | [中文](#chinese)
+
+Multi-phase PDF capture pipeline as an [MCP](https://modelcontextprotocol.io) server.
+Convert PDF documents into high-quality structured Markdown — with formula recognition,
+table extraction, layout cleaning, and a built-in quality gate.
+
+[![CI](https://github.com/ChenHongYu2026/pdf-capture-mcp/actions/workflows/ci.yml/badge.svg)](https://github.com/ChenHongYu2026/pdf-capture-mcp/actions/workflows/ci.yml)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
+[![Python 3.11+](https://img.shields.io/badge/python-3.11+-blue.svg)](https://www.python.org/downloads/)
+
+## Features
+
+- **Triple extraction engines**
+  - [pymupdf4llm](https://github.com/pymupdf/pymupdf4llm) (built-in) — zero setup, fast, always available
+  - [marker](https://github.com/datalab-to/marker) (recommended) — highest quality for complex layouts
+  - [MinerU](https://github.com/opendatalab/MinerU) (optional) — best for multi-column/InDesign PDFs, auto-managed in an isolated venv
+- **7 MCP tools** — `pdf_to_markdown`, `extract_tables`, `classify_document`, `pdf_info`, `setup_vlm`, `check_environment`, `install_engine`
+- **Optional VLM enhancement** — plug in any vision-capable model (Qwen-VL, GLM-4V, MiniMax, Moonshot, OpenAI, local Ollama…) for better table/formula extraction. **No extra dependencies needed** — works out of the box with the base install.
+- **Quality gate** — multi-dimensional QC (text completeness, heading structure, formula integrity, table coverage)
+- **Progressive setup** — works immediately with zero config; enhance with marker/VLM on demand
+- **Privacy-first** — API keys are stored locally with `chmod 600` and never echoed back in responses
+
+## Quick Start
+
+### 1. Install
+
+```bash
+# Base package (pymupdf engine + VLM support, ~80MB) — works immediately
+pip install pdf-capture-mcp
+
+# With marker engine (recommended for complex PDFs, includes PyTorch, ~2.5GB)
+pip install "pdf-capture-mcp[marker]"
+
+# Everything (marker + TATR table detection + DePlot charts, ~3GB)
+pip install "pdf-capture-mcp[all]"
+```
+
+### 2. Add to your MCP client
+
+For Qoder / Claude Desktop / Cursor, add to your `mcp.json`:
+
+```json
+{
+  "mcpServers": {
+    "pdf-capture": {
+      "command": "uvx",
+      "args": ["pdf-capture-mcp"]
+    }
+  }
+}
+```
+
+With marker engine (recommended):
+
+```json
+{
+  "mcpServers": {
+    "pdf-capture": {
+      "command": "uvx",
+      "args": ["pdf-capture-mcp[marker]"]
+    }
+  }
+}
+```
+
+Or if installed via pip:
+
+```json
+{
+  "mcpServers": {
+    "pdf-capture": {
+      "command": "pdf-capture-mcp"
+    }
+  }
+}
+```
+
+### 3. Use it
+
+Ask your AI agent things like:
+
+> "Convert ~/Downloads/paper.pdf to Markdown"
+> "Extract all tables from this report"
+> "Is this PDF a scanned document?"
+
+The server works immediately with the built-in pymupdf engine. For higher quality on
+complex layouts, the agent can install marker on demand (or you can pre-install it).
+
+## Tools
+
+| Tool | Description |
+|------|-------------|
+| `pdf_to_markdown` | Full pipeline: extract → clean → QC → structured Markdown |
+| `extract_tables` | Table extraction (pdfplumber rules + optional TATR deep learning) |
+| `classify_document` | Document type detection (academic paper, consulting report, …) |
+| `pdf_info` | Fast metadata: page count, text layer, scanned detection |
+| `setup_vlm` | Configure optional VLM enhancement (any vision-capable provider) |
+| `check_environment` | Verify engines, dependencies, and filesystem compatibility |
+| `install_engine` | Install marker/ml engines on behalf of the user |
+
+## VLM Enhancement (Optional)
+
+VLM re-extracts complex tables and broken formulas from page images.
+Works with **any provider whose model supports image input**:
+
+| Provider | Example model | API base |
+|----------|--------------|----------|
+| Alibaba Qwen | `qwen-vl-max` | `https://dashscope.aliyuncs.com/compatible-mode/v1` |
+| Zhipu AI | `glm-4v` | `https://open.bigmodel.cn/api/paas/v4` |
+| MiniMax | `minimax-m3` | `https://api.minimaxi.com/v1` |
+| Moonshot | `moonshot-v1-vision` | `https://api.moonshot.cn/v1` |
+| OpenAI | `gpt-4o` | `https://api.openai.com/v1` |
+| Ollama (local) | `llama3.2-vision` | `http://localhost:11434/v1` |
+
+Set your key via environment variable (recommended — never typed into chat):
+
+```bash
+export PDF_CAPTURE_VLM_API_KEY=your_key_here
+```
+
+Then tell your agent: *"Enable VLM with qwen-vl-max"* — it validates vision capability
+before saving. **Note: using VLM consumes your API tokens.**
+
+## MinerU Engine (Optional)
+
+For the highest extraction quality on complex layouts (multi-column, InDesign PDFs):
+
+```bash
+pdf-capture-mcp setup-mineru   # requires Python 3.11 on PATH
+```
+
+Models (~2GB) auto-download from ModelScope on first extraction.
+
+## Environment Variables
+
+| Variable | Default | Description |
+|----------|---------|-------------|
+| `PDF_CAPTURE_ENGINE` | `auto` | Default engine: `marker` / `mineru` / `pymupdf` / `auto` |
+| `PDF_CAPTURE_VLM_API_KEY` | — | VLM API key (preferred over passing in chat) |
+| `PDF_CAPTURE_CACHE_DIR` | `~/.cache/pdf-capture-mcp` | Model & config cache |
+| `PDF_CAPTURE_MINERU_VENV` | `<cache>/venv-mineru` | MinerU venv location |
+| `PDF_CAPTURE_LOG_LEVEL` | `INFO` | Logging level |
+| `MINERU_MODEL_SOURCE` | `modelscope` | MinerU model source: `modelscope` / `huggingface` / `local` |
+
+## Development
+
+```bash
+git clone https://github.com/ChenHongYu2026/pdf-capture-mcp.git
+cd pdf-capture-mcp
+uv sync --extra dev
+uv run pytest tests/ -v
+uv run ruff check src/ tests/
+uv run mypy src/pdf_capture_mcp/
+```
+
+## Troubleshooting
+
+| Problem | Solution |
+|---------|----------|
+| `Operation not supported` during install | External/exFAT drive: `export UV_LINK_MODE=copy` then retry |
+| `uvx: command not found` | Install uv: `curl -LsSf https://astral.sh/uv/install.sh \| sh` |
+| MCP server not appearing in tools | Restart your MCP client; check `mcp.json` syntax |
+| marker engine slow on first run | Downloads ~1GB models on first use; cached afterwards |
+| Python version too low | Requires 3.11+: `uv python install 3.11` |
+
+## License
+
+MIT — see [LICENSE](LICENSE).
+
+Third-party notices: MinerU (AGPL-3.0, invoked as a separate subprocess, not linked),
+marker (Apache-2.0), pdfplumber (MIT), Table Transformer (MIT), pymupdf4llm (Apache-2.0).
+
+---
+
+<a id="chinese"></a>
+
+# pdf-capture-mcp（中文）
+
+[English](#english) | **中文**
+
+多阶段 PDF 捕获管线，以 [MCP](https://modelcontextprotocol.io) 服务器形式提供。
+将 PDF 文档转换为高质量结构化 Markdown —— 支持公式识别、表格提取、版面清洁和内置质量门控。
+
+## 功能特性
+
+- **三提取引擎**
+  - [pymupdf4llm](https://github.com/pymupdf/pymupdf4llm)（内置）—— 零配置、快速、始终可用
+  - [marker](https://github.com/datalab-to/marker)（推荐）—— 复杂版面提取质量最高
+  - [MinerU](https://github.com/opendatalab/MinerU)（可选）—— 多栏/InDesign 排版最佳，自动管理独立虚拟环境
+- **7 个 MCP 工具** —— `pdf_to_markdown`、`extract_tables`、`classify_document`、`pdf_info`、`setup_vlm`、`check_environment`、`install_engine`
+- **可选 VLM 增强** —— 接入任何具备视觉能力的模型（通义千问 Qwen-VL、智谱 GLM-4V、MiniMax、月之暗面 Moonshot、OpenAI、本地 Ollama 等），提升表格/公式提取质量。**无需额外依赖**，基础安装即可使用。
+- **质量门控** —— 多维度 QC 评估（文本完整度、标题结构、公式完好率、表格覆盖率）
+- **渐进式配置** —— 零配置即可工作；按需增强 marker/VLM
+- **隐私优先** —— API Key 以 `chmod 600` 权限本地存储，绝不在响应中回显
+
+## 快速开始
+
+### 1. 安装
+
+```bash
+# 基础包（pymupdf 引擎 + VLM 支持，约 80MB）—— 安装即可用
+pip install pdf-capture-mcp
+
+# 含 marker 引擎（推荐复杂 PDF，包含 PyTorch，约 2.5GB）
+pip install "pdf-capture-mcp[marker]"
+
+# 完整安装（marker + TATR 表格检测 + DePlot 图表提取，约 3GB）
+pip install "pdf-capture-mcp[all]"
+```
+
+### 2. 添加到 MCP 客户端
+
+Qoder / Claude Desktop / Cursor 用户，在 `mcp.json` 中添加：
+
+```json
+{
+  "mcpServers": {
+    "pdf-capture": {
+      "command": "uvx",
+      "args": ["pdf-capture-mcp"]
+    }
+  }
+}
+```
+
+含 marker 引擎（推荐）：
+
+```json
+{
+  "mcpServers": {
+    "pdf-capture": {
+      "command": "uvx",
+      "args": ["pdf-capture-mcp[marker]"]
+    }
+  }
+}
+```
+
+如果已通过 pip 安装：
+
+```json
+{
+  "mcpServers": {
+    "pdf-capture": {
+      "command": "pdf-capture-mcp"
+    }
+  }
+}
+```
+
+### 3. 开始使用
+
+直接对你的 AI 助手说：
+
+> “把 ~/Downloads/论文.pdf 转成 Markdown”
+> “提取这份报告里的所有表格”
+> “这个 PDF 是扫描件吗？”
+
+服务器使用内置 pymupdf 引擎即可立即工作。对于复杂版面，助手可按需安装 marker 引擎。
+
+## 工具列表
+
+| 工具 | 说明 |
+|------|------|
+| `pdf_to_markdown` | 完整管线：提取 → 清洁 → QC → 结构化 Markdown |
+| `extract_tables` | 表格提取（pdfplumber 规则 + 可选 TATR 深度学习） |
+| `classify_document` | 文档类型检测（学术论文、咨询报告等） |
+| `pdf_info` | 快速元数据：页数、文本层、扫描件检测 |
+| `setup_vlm` | 配置可选的 VLM 增强（支持任何具备视觉能力的供应商） |
+| `check_environment` | 校验引擎、依赖和文件系统兼容性 |
+| `install_engine` | 代用户安装 marker/ml 引擎 |
+
+## VLM 增强（可选）
+
+VLM 会从页面图像中重新提取复杂表格和损坏的公式。
+支持**任何模型具备图片输入能力的供应商**：
+
+| 供应商 | 示例模型 | API 地址 |
+|--------|----------|----------|
+| 阿里通义千问 | `qwen-vl-max` | `https://dashscope.aliyuncs.com/compatible-mode/v1` |
+| 智谱 AI | `glm-4v` | `https://open.bigmodel.cn/api/paas/v4` |
+| MiniMax | `minimax-m3` | `https://api.minimaxi.com/v1` |
+| 月之暗面 | `moonshot-v1-vision` | `https://api.moonshot.cn/v1` |
+| OpenAI | `gpt-4o` | `https://api.openai.com/v1` |
+| Ollama（本地） | `llama3.2-vision` | `http://localhost:11434/v1` |
+
+推荐通过环境变量设置 Key（避免在对话中输入）：
+
+```bash
+export PDF_CAPTURE_VLM_API_KEY=你的密钥
+```
+
+然后告诉助手："启用 VLM，用 qwen-vl-max" —— 系统会先验证模型的视觉能力再保存配置。
+**注意：使用 VLM 功能会消耗你的 Token。**
+
+## MinerU 引擎（可选）
+
+针对复杂版面（多栏、InDesign 排版 PDF）获得最高提取质量：
+
+```bash
+pdf-capture-mcp setup-mineru   # 需要 PATH 中有 Python 3.11
+```
+
+首次提取时会从 ModelScope 自动下载模型（约 2GB）。
+
+## 环境变量
+
+| 变量 | 默认值 | 说明 |
+|------|--------|------|
+| `PDF_CAPTURE_ENGINE` | `auto` | 默认引擎：`marker` / `mineru` / `pymupdf` / `auto` |
+| `PDF_CAPTURE_VLM_API_KEY` | — | VLM API Key（推荐方式，避免对话中传递） |
+| `PDF_CAPTURE_CACHE_DIR` | `~/.cache/pdf-capture-mcp` | 模型与配置缓存目录 |
+| `PDF_CAPTURE_MINERU_VENV` | `<cache>/venv-mineru` | MinerU 虚拟环境位置 |
+| `PDF_CAPTURE_LOG_LEVEL` | `INFO` | 日志级别 |
+| `MINERU_MODEL_SOURCE` | `modelscope` | MinerU 模型源：`modelscope` / `huggingface` / `local` |
+
+## 本地开发
+
+```bash
+git clone https://github.com/ChenHongYu2026/pdf-capture-mcp.git
+cd pdf-capture-mcp
+uv sync --extra dev
+uv run pytest tests/ -v
+uv run ruff check src/ tests/
+uv run mypy src/pdf_capture_mcp/
+```
+
+## 故障排查
+
+| 问题 | 解决方案 |
+|------|----------|
+| 安装时报 `Operation not supported` | 外置/exFAT 磁盘：`export UV_LINK_MODE=copy` 后重试 |
+| `uvx: command not found` | 安装 uv：`curl -LsSf https://astral.sh/uv/install.sh \| sh` |
+| MCP 服务器未出现在工具列表 | 重启 MCP 客户端；检查 `mcp.json` 格式 |
+| marker 引擎首次运行慢 | 首次使用需下载约 1GB 模型，后续使用缓存 |
+| Python 版本过低 | 需要 3.11+：`uv python install 3.11` |
+
+## 许可证
+
+MIT —— 详见 [LICENSE](LICENSE)。
+
+第三方声明：MinerU（AGPL-3.0，通过独立子进程调用，未链接）、
+marker（Apache-2.0）、pdfplumber（MIT）、Table Transformer（MIT）、pymupdf4llm（Apache-2.0）。

pdf_capture_mcp-0.1.0/pyproject.toml

@@ -0,0 +1,69 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "pdf-capture-mcp"
+version = "0.1.0"
+description = "Multi-phase PDF capture pipeline as an MCP server. Convert PDF to high-quality structured Markdown."
+license = "MIT"
+requires-python = ">=3.11"
+authors = [{ name = "ChenHongYu2026" }]
+keywords = ["pdf", "markdown", "mcp", "extraction", "ocr"]
+classifiers = [
+    "Development Status :: 3 - Alpha",
+    "Intended Audience :: Developers",
+    "License :: OSI Approved :: MIT License",
+    "Programming Language :: Python :: 3.11",
+    "Programming Language :: Python :: 3.12",
+    "Topic :: Text Processing :: Markup :: Markdown",
+]
+dependencies = [
+    "fastmcp>=3.2.0,<4",
+    "pymupdf4llm>=0.0.17",
+    "pdfplumber>=0.11.0",
+    "pypdf>=5.0",
+    "Pillow>=10.0",
+    "numpy>=1.26",
+    "httpx>=0.27",
+]
+
+[project.optional-dependencies]
+marker = ["marker-pdf>=2.0"]
+ml = ["torch>=2.0", "transformers>=4.40"]
+vlm = []  # VLM uses OpenAI-compatible HTTP API via httpx (already a base dep)
+all = ["pdf-capture-mcp[marker,ml]"]
+dev = [
+    "pytest>=8.0",
+    "pytest-asyncio>=0.24",
+    "ruff>=0.15",
+    "mypy>=1.15",
+]
+
+[project.scripts]
+pdf-capture-mcp = "pdf_capture_mcp.__main__:main"
+
+[project.urls]
+Repository = "https://github.com/ChenHongYu2026/pdf-capture-mcp"
+Issues = "https://github.com/ChenHongYu2026/pdf-capture-mcp/issues"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/pdf_capture_mcp"]
+
+[tool.ruff]
+line-length = 100
+target-version = "py311"
+
+[tool.ruff.lint]
+select = ["E", "F", "I", "N", "W", "UP"]
+
+[tool.mypy]
+# No python_version pin: numpy's type stubs use PEP 695 syntax (3.12+),
+# which mypy rejects when targeting 3.11. Checks run against the active interpreter.
+warn_return_any = true
+warn_unused_configs = true
+ignore_missing_imports = true
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+asyncio_mode = "auto"

pdf_capture_mcp-0.1.0/src/pdf_capture_mcp/__init__.py

@@ -0,0 +1,3 @@
+"""pdf-capture-mcp: Multi-phase PDF capture pipeline as an MCP server."""
+
+__version__ = "0.1.0"

pdf_capture_mcp-0.1.0/src/pdf_capture_mcp/__main__.py

@@ -0,0 +1,35 @@
+"""Entry point: python -m pdf_capture_mcp"""
+
+from __future__ import annotations
+
+import sys
+
+
+def main() -> None:
+    """Start the PDF Capture MCP server."""
+    from pdf_capture_mcp.config import setup_logging
+    from pdf_capture_mcp.server import mcp
+
+    setup_logging()
+
+    # Handle setup-mineru subcommand
+    if len(sys.argv) > 1 and sys.argv[1] == "setup-mineru":
+        from pdf_capture_mcp.engines.mineru_engine import ensure_mineru_env
+
+        try:
+            python_exe = ensure_mineru_env()
+            print(f"MinerU environment ready: {python_exe}")
+        except RuntimeError as exc:
+            print(f"ERROR: {exc}", file=sys.stderr)
+            sys.exit(1)
+        return
+
+    # Determine transport
+    if "--sse" in sys.argv:
+        mcp.run(transport="sse")
+    else:
+        mcp.run(transport="stdio")
+
+
+if __name__ == "__main__":
+    main()