Medical Paper Assistant 醫學論文寫作助手

🔬 A Guided & Exploratory Medical Paper Writing Tool
Powered by MCP + GitHub Copilot

🔬 引導式、探索式醫學論文寫作工具
由 MCP + GitHub Copilot 驅動

🇬🇧 English

🎯 Why This Tool?

Traditional paper writing tools require you to know exactly what you want before you start. But research is rarely that linear.

Medical Paper Assistant is different:

🔍 Explore First, Decide Later - Browse literature freely, save interesting papers, then decide your research direction
💬 Conversational Workflow - Chat naturally with AI to refine your ideas, not fight with forms
🧭 Guided Process - Step-by-step prompts guide you from concept to publication-ready manuscript
🔗 Native MCP + Copilot Integration - Works directly inside VS Code, no context switching

💡 What Makes Us Unique

Traditional Tools	Medical Paper Assistant
Fixed templates, rigid workflow	Flexible, exploratory approach
Separate apps for search/write/cite	All-in-one integrated experience
Manual reference management	Auto-save with PDF & citation formats
Export then format	Direct Word export with journal styles
Learn complex UI	Natural language conversation

Overview

Medical Paper Assistant is an AI-powered writing tool designed specifically for medical researchers. Built on the Model Context Protocol (MCP), it integrates seamlessly with VS Code through GitHub Copilot, providing a comprehensive workflow from literature search and data analysis to manuscript drafting and formatted export.

✨ Key Features

Feature	Description
Literature Search & Management	Connect to PubMed API, search articles, download PDFs from PMC Open Access, and build a local reference library
Data Analysis	Automatically read CSV data, perform statistical tests (t-test, correlation, etc.), and generate publication-ready figures
Smart Draft Generation	Generate manuscript drafts based on your research concept and analysis results
Automatic Citations	Insert citation markers `[1]` and generate formatted reference lists (Vancouver, APA, Nature styles)
Interactive Refinement	Fine-tune specific sections through conversational dialogue
Word Export	Export Markdown drafts to `.docx` files conforming to journal templates

🚀 Installation

Prerequisites

Before you begin, ensure you have the following installed:

Requirement	Version	How to Check	Installation Guide
Python	3.10 or higher	`python3 --version`	python.org
Git	Any recent version	`git --version`	git-scm.com
VS Code	Latest	Open VS Code → Help → About	code.visualstudio.com
GitHub Copilot	Extension	VS Code Extensions panel	Search "GitHub Copilot" in Extensions

Quick Installation (Recommended)

Step 1: Clone the Repository

Open a terminal and run:

```bash

Navigate to where you want to store the project

cd ~/projects # or any directory of your choice

Clone the repository

git clone https://github.com/u9401066/med-paper-assistant.git

Enter the project directory

cd med-paper-assistant ```

Step 2: Run the Setup Script

Linux/macOS: ```bash

Make the script executable

chmod +x scripts/setup.sh

Run the setup script

./scripts/setup.sh ```

Windows (PowerShell - Recommended): ```powershell

Run the PowerShell setup script

.\scripts\setup.ps1 ```

Windows (Command Prompt): ```cmd

Run the batch setup script

scripts\setup.bat ```

This script will:

✅ Create a Python virtual environment (`.venv/`)
✅ Install all required dependencies
✅ Create the `.vscode/mcp.json` configuration file (with correct paths for your OS)
✅ Verify the installation

Step 3: Reload VS Code

After setup completes:

Open VS Code
Press `Ctrl+Shift+P` (Windows/Linux) or `Cmd+Shift+P` (Mac)
Type `Developer: Reload Window` and press Enter

Step 4: Verify Installation

In VS Code's Copilot Chat panel:

Type /mcp and press Enter
You should see mdpaper (50 tools) in the list (or with drawio (15 tools) if you enabled Draw.io integration)
If you see this, the installation was successful! 🎉

Optional: Setup Draw.io Integration

For diagram generation (CONSORT/PRISMA flowcharts), run:

# Initialize submodules and setup Draw.io
./scripts/setup-integrations.sh

This adds the drawio MCP server with tools for creating diagrams.

To use Draw.io:

Start the web app: ./scripts/start-drawio.sh
Reload VS Code
Use drawio tools in Copilot Chat

Optional: Recommended VS Code Extensions

These extensions enhance your research workflow:

Extension	Purpose	Install
Project Manager	Quick project switching, sidebar, tags	`ext install alefragnani.project-manager`
Foam	Wikilinks, backlinks, hover preview	`ext install foam.foam-vscode`

Quick Install:

code --install-extension alefragnani.project-manager
code --install-extension foam.foam-vscode

📚 Foam Integration for Reference Management

Medical Paper Assistant integrates with Foam to provide a powerful reference linking and preview system.

What is Foam?

Foam is a personal knowledge management system inspired by Roam Research. It turns your Markdown files into a connected knowledge graph with:

Wikilinks [[link]] - Link between notes
Hover Preview - See content without opening files
Backlinks - See which notes reference the current note
Graph Visualization - Visualize connections

Setup Foam for This Project

Step 1: Install Foam Extension

In VS Code:

Press Ctrl+Shift+X (Extensions panel)
Search for foam.foam-vscode
Click Install
Reload VS Code (Ctrl+Shift+P → Developer: Reload Window)

Step 2: Save References with Citation Keys

When you save a reference, the system automatically creates Foam-compatible files:

# Using MCP tool
save_reference(pmid="38049909")

# Creates:
references/
├── 38049909/
│   ├── content.md       ← Main content with all citation formats
│   └── metadata.json    ← Structured data
└── tang2023_38049909.md ← Foam alias for easy linking

Step 3: Use Wikilinks in Your Draft

# My Research Draft

According to [[tang2023_38049909]], remimazolam shows promising results
for ICU sedation compared to propofol.

The study by [[smith2024_41285088]] also supports this finding...

Step 4: Hover to Preview

Move your mouse over any [[link]] to instantly see:

📄 Paper title and authors
📋 All citation formats (Vancouver, APA, Nature)
📝 Full abstract
🔗 Clickable DOI link

Citation Key Format

References are saved with a human-friendly + verifiable format:

[[author_year_pmid]]
   │      │     │
   │      │     └── PMID for verification
   │      └── Publication year
   └── First author's last name

Example: [[tang2023_38049909]]
         → Tang et al., 2023, PMID: 38049909
         → Verify: https://pubmed.ncbi.nlm.nih.gov/38049909/

Citation Formats in Preview

When you hover over a reference, you'll see all citation formats. The project's preferred format is marked with ⭐:

## 📋 Citation Formats

**⭐ APA**: Tang, Y., Gao, X., ... (2023). Remimazolam besylate versus...

**Vancouver**: Tang Y, Gao X, et al. Remimazolam besylate versus...

**Nature**: Tang Y, et al. Remimazolam besylate versus... *Crit Care* **27**...

**In-text**: (Tang et al., 2023)

To set your preferred citation style:

set_citation_style(style="apa")  # Options: vancouver, apa, nature, harvard, ama

Foam Features for Research

Feature	How to Use	Benefit
Hover Preview	Mouse over `[[link]]`	See abstract without opening file
Backlinks Panel	Open any reference file	See which drafts cite this paper
Graph View	`Ctrl+Shift+P` → `Foam: Show Graph`	Visualize paper connections
Orphan References	`Ctrl+Shift+P` → `Foam: Show Orphans`	Find saved but uncited papers

code --install-extension foam.foam-vscode


#### Manual Installation (If Setup Script Fails)

<details>
<summary>Click to expand manual installation steps</summary>

**Step 1: Clone the Repository**

\`\`\`bash
git clone https://github.com/u9401066/med-paper-assistant.git
cd med-paper-assistant
\`\`\`

**Step 2: Create Virtual Environment**

\`\`\`bash
# Create a virtual environment named .venv
python3 -m venv .venv

# Activate the virtual environment
# On Linux/Mac:
source .venv/bin/activate

# On Windows (Command Prompt):
.venv\Scripts\activate.bat

# On Windows (PowerShell):
.venv\Scripts\Activate.ps1
\`\`\`

> 💡 **What is a virtual environment?**
> A virtual environment is an isolated Python environment that keeps this project's dependencies separate from your system Python. This prevents conflicts with other projects.

**Step 3: Install Dependencies**

\`\`\`bash
# Upgrade pip first (recommended)
pip install --upgrade pip

# Install the project in editable mode
pip install -e .
\`\`\`

**Step 4: Configure VS Code MCP**

Create the file \`.vscode/mcp.json\` in the project root:

\`\`\`bash
# Create .vscode directory if it doesn't exist
mkdir -p .vscode

# Create the mcp.json file
cat > .vscode/mcp.json << 'EOF'
{
  "inputs": [],
  "servers": {
    "mdpaper": {
      "type": "stdio",
      "command": "${workspaceFolder}/.venv/bin/python",
      "args": ["-m", "med_paper_assistant.interfaces.mcp"],
      "env": {
        "PYTHONPATH": "${workspaceFolder}/src"
      }
    }
  }
}
EOF
\`\`\`

> ⚠️ **Windows Users**: Replace \`.venv/bin/python\` with \`.venv\\Scripts\\python.exe\`

**Step 5: Reload VS Code**

Press \`Ctrl+Shift+P\` → Type \`Developer: Reload Window\` → Press Enter

</details>

#### Troubleshooting

<details>
<summary>❌ "Command not found: python3"</summary>

**Problem**: Python is not installed or not in your PATH.

**Solution**:
1. Install Python from [python.org](https://www.python.org/downloads/)
2. During installation, check "Add Python to PATH"
3. Restart your terminal and try again

</details>

<details>
<summary>❌ MCP server not showing in Copilot</summary>

**Problem**: The MCP server is not being detected by VS Code.

**Solutions**:
1. Ensure \`.vscode/mcp.json\` exists and has correct paths
2. Check that the virtual environment was created: \`ls .venv/\`
3. Try reloading VS Code: \`Ctrl+Shift+P\` → \`Developer: Reload Window\`
4. Check VS Code Output panel: \`View\` → \`Output\` → Select "MCP" from dropdown

</details>

<details>
<summary>❌ "ModuleNotFoundError: med_paper_assistant"</summary>

**Problem**: The package is not installed or PYTHONPATH is incorrect.

**Solution**:
\`\`\`bash
# Activate virtual environment first
source .venv/bin/activate  # Linux/Mac
# .venv\Scripts\activate   # Windows

# Reinstall the package
pip install -e .
\`\`\`

</details>

---

### 📖 Usage Guide

This assistant integrates with GitHub Copilot through MCP. Use \`/mcp.mdpaper.*\` commands in Copilot Chat.

#### Workflow Overview

```mermaid
flowchart LR
    A["📝 Concept<br/>/concept"] --> B["🔍 Search<br/>/strategy"]
    B --> C["📊 Analysis<br/>/analysis"]
    C --> D["✍️ Draft<br/>/draft"]
    D --> E["🔄 Clarify<br/>/clarify"]
    E --> F["📄 Export<br/>/format"]

🆕 Literature Exploration Mode (Recommended for Beginners!)

Don't know what to research? Start by exploring literature first!

Command: /mcp.mdpaper.search

flowchart LR
    A["🔍 Start Exploration"] --> B["📚 Search & Save Papers"]
    B --> C{"Found Direction?"}
    C -->|Not yet| B
    C -->|Yes!| D["📁 Convert to Project"]
    D --> E["✍️ Start Writing"]

How it works:

Start exploring - Automatically creates a temporary workspace
```
start_exploration()
```

Search and save interesting papers - No project commitment required

search_literature(query="your interest topic")
save_reference(pmid="12345678")  # Save interesting ones

When you find a direction - Convert to formal project with all references

convert_exploration_to_project(
    name="Your Research Title",
    paper_type="original-research"
)

💡 Benefit: All references saved during exploration are automatically transferred to your new project!

Step 1: Prepare Your Data

Before starting, organize your files:

Directory	Purpose
`data/`	Place your CSV research data here
`templates/`	(Optional) Journal Word templates

Step 2: Develop Your Research Concept

Command: `/mcp.mdpaper.concept`

This guided workflow helps you develop a well-structured research concept with novelty protection:

flowchart LR
    A["📚 Literature Search"] --> B["🔍 Gap Identification"]
    B --> C{"❓ User Confirms Gap"}
    C -->|Yes| D["✍️ Concept Writing"]
    C -->|No| B
    D --> E["✅ Validate Concept"]

Step-by-step process:

📚 Literature Search (Mandatory)
- Agent searches PubMed for 5-10 relevant studies
- Key papers are saved to your reference library
🔍 Gap Identification (Mandatory)
- Agent presents literature summary
- Identifies research gaps and limitations
- 🛑 Asks you to confirm which gap your research addresses
✍️ Concept Writing (After your confirmation)
- Creates structured concept with protected sections:
  - 🔒 NOVELTY STATEMENT - What's new (protected)
  - 🔒 KEY SELLING POINTS - Your 3-5 key differentiators (protected)
  - 📝 Background, Methods, Expected Outcomes (editable)
✅ Validation
- Use validate_concept tool to verify completeness

The output is saved to `drafts/concept_*.md`.

💡 Why Protected Sections? The 🔒 markers ensure your key innovations are preserved when writing the full paper. The Agent must ask before modifying these sections.

Step 3: Configure Search Strategy

Command: `/mcp.mdpaper.strategy`

Set up your literature search parameters:

Keywords and MeSH terms
Exclusion criteria
Article types (Review, Clinical Trial, Meta-Analysis)
Date range
Minimum sample size

Step 4: Analyze Your Data

Command: `/mcp.mdpaper.analysis`

The assistant will:

List available CSV files in `data/`
Ask you to specify grouping and outcome variables
Perform appropriate statistical tests
Generate figures (saved to `results/figures/`)
Create Table 1 (baseline characteristics)

Step 5: Write Your Draft

Command: `/mcp.mdpaper.draft`

The assistant will:

Read your `concept.md` and analysis results
Ask which template to use
Generate a Markdown draft with:
- Embedded figures
- Automatic citations
- Formatted tables

⚠️ Important: This command requires a concept file. Create one using `/mcp.mdpaper.concept` first.

Step 6: Refine Your Content

Command: `/mcp.mdpaper.clarify`

Interactively improve specific sections:

"Make the Introduction more conservative"
"Expand the limitations section"
"Simplify the statistical methods description"

Step 7: Export to Word

Command: /mcp.mdpaper.format

This follows a 7-step workflow:

Read template structure
Read draft content
Map sections
Insert content
Verify document
Check word limits
Save final document

Output: results/your_paper.docx

📂 Project Structure

Directory	Purpose
`projects/`	Research projects (isolated workspaces)
`projects/{slug}/concept.md`	Research concept with protected sections
`projects/{slug}/drafts/`	Markdown drafts
`projects/{slug}/references/{PMID}/`	Local reference library
`projects/{slug}/data/`	Raw data files (CSV)
`projects/{slug}/results/`	Generated outputs (figures, tables, docx)
└ `metadata.json`	Citation info, formatted references
└ `content.md`	Abstract and notes
└ `fulltext.pdf`	PDF (if available from PMC)
`templates/`	Journal Word templates (`.docx` for export)
`src/med_paper_assistant/core/`	Core logic modules
`src/med_paper_assistant/mcp_server/`	MCP server implementation
└ `templates/`	Internal templates for Agent guidance (`.md`)
`tests/`	Test suite
`.memory/`	Project context files
└ `.agent_constitution.md`	Agent behavior rules (dynamically loaded)

⚖️ Agent Constitution

The MCP server dynamically loads the Agent Constitution from .memory/.agent_constitution.md. This file defines the core behavior rules that Copilot must follow.

Constitution Structure:

Part	Content
Part 1: System Development	Language rules, Memory Bank maintenance, Code quality, Privacy
Part 2: Research Operations	Concept workflow, Protected content rules, Draft writing rules

Key Rules for Protected Content (🔒):

## 2.2 🔒 Protected Content Rules

| Section | Description | Rule |
|---------|-------------|------|
| 🔒 NOVELTY STATEMENT | Innovation declaration | Must ask before modifying |
| 🔒 KEY SELLING POINTS | Core differentiators (3-5) | Must ask before modifying |
| 🔒 Author Notes | Private notes | Do not modify |

Modification Rules:
✅ Can refine wording
✅ Can improve academic expression
⛔ MUST ask user before substantial changes
⛔ NEVER delete protected content
⛔ NEVER weaken selling points

To customize: Edit .memory/.agent_constitution.md and reload VS Code. The new rules will be automatically applied.

🎯 Novelty Validation System

The system includes an intelligent Novelty Validation feature that ensures research concepts truly describe novel contributions before proceeding to draft writing.

How It Works

flowchart LR
    A["📝 Concept File"] --> B["🔍 Structure Check"]
    B --> C{"Required sections?"}
    C -->|No| D["❌ FAIL"]
    C -->|Yes| E["🎯 Novelty Scoring"]
    E --> F["Round 1"]
    E --> G["Round 2"]
    E --> H["Round 3"]
    F & G & H --> I{"All ≥ 75?"}
    I -->|No| J["❌ FAIL + Suggestions"]
    I -->|Yes| K["✅ PASS"]

Scoring Mechanism

Setting	Value	Description
Rounds	3	Number of independent evaluations
Threshold	75/100	Minimum score per round
Pass Criteria	All 3 rounds ≥ 75	Must pass all rounds

Evaluation Dimensions

Dimension	Weight	What It Checks
Uniqueness	25%	Is the approach/method unique?
Significance	25%	Is the contribution meaningful?
Gap Alignment	20%	Does novelty address the research gap?
Specificity	15%	Is the claim specific and concrete?
Verifiability	15%	Can the claim be verified/tested?

Example Validation Report

📋 Concept Validation Report
📄 File: concept.md

## 🔒 Protected Sections (Required)
| Section | Found | Has Content | Status |
|---------|-------|-------------|--------|
| 🔒 NOVELTY STATEMENT | ✅ | ✅ | ✅ PASS |
| 🔒 KEY SELLING POINTS (4 points) | ✅ | ✅ | ✅ PASS |

## 🎯 Novelty Evaluation
**Status:** ✅ PASSED
**Average Score:** 82.3/100
**Threshold:** 75

| Round | Score | Status |
|-------|-------|--------|
| 1 | 84 | ✅ |
| 2 | 79 | ✅ |
| 3 | 84 | ✅ |

## ✅ VALIDATION PASSED

Key Design Decisions

Mandatory Before Draft Writing: The system enforces validation before any concept file can be used for draft generation.
Multi-Round Evaluation: Using 3 rounds increases reliability and reduces false positives.
High Threshold: Requiring 75+ ensures only well-articulated novelty claims pass.
Actionable Feedback: Failed validations provide specific suggestions for improvement.

🧠 Skills System

Skills are complete workflow definitions that guide the AI through complex multi-step tasks. Unlike individual tools that perform single actions, Skills define the entire process from start to finish.

📁 See full documentation: .skills/README.md

When to Use Skills

User Request	Skill to Load
"Help me do a literature review"	`literature_review`
"Develop my research concept"	`concept_development`
"Search for papers on [topic]"	`parallel_search`
"Write the Introduction section"	`draft_introduction`
"Write the Discussion"	`draft_discussion`

How It Works

flowchart LR
    A["🗣️ User Request"] --> B["🔍 Identify Intent"]
    B --> C["📖 Load Skill"]
    C --> D["🔧 Follow Workflow"]
    D --> E["✅ Deliverables"]

Example: When you say "help me do a literature review", the AI:

Loads .skills/research/literature_review.md
Follows the defined phases (search → screen → synthesize)
Asks at decision points
Produces the expected deliverables

Skill Tools

Tool	Purpose
`list_skills`	Show all available skills
`load_skill`	Load a specific skill's workflow
`suggest_skill`	Get skill recommendation based on task

⚡ Parallel Search

Parallel Search dramatically speeds up literature discovery by running multiple search strategies simultaneously.

How It Works

flowchart LR
    A["🎯 Topic"] --> B["generate_search_queries"]
    B --> C["5 Query Strategies"]
    C --> D["⚡ Parallel Execution"]
    D --> E["merge_search_results"]
    E --> F["📊 Deduplicated Results"]

Usage Example

User: "Find papers about remimazolam ICU sedation"

AI executes:
1. generate_search_queries(topic="remimazolam ICU sedation")
   → Returns 5 different query strategies

2. Parallel execution of search_literature × 5
   → All queries run simultaneously

3. merge_search_results(...)
   → Returns: 42 unique papers (56 total, 14 duplicates removed)
   → Papers found by multiple strategies are flagged as "high relevance"

Strategy Integration

Configure your search strategy once, and it automatically applies to all parallel queries:

configure_search_strategy({
    "date_range": "2020-2025",
    "exclusions": ["animal", "review"],
    "article_types": ["Clinical Trial", "RCT"]
})

generate_search_queries(topic="...", use_saved_strategy=True)
→ All 5 queries include date filters, exclusions, and article type restrictions

🛠️ Available Tools (52 mdpaper + 15 drawio)

📚 mdpaper Tools (52 total)

Category	Tools	Description
Search (10)	`search_literature`, `find_related_articles`, `find_citing_articles`, `fetch_article_details`, `configure_search_strategy`, `get_search_strategy`, `generate_search_queries`, `merge_search_results`, `expand_search_queries`, `search_local_references`	Literature discovery via pubmed-search-mcp submodule
Reference (8)	`save_reference`, `list_saved_references`, `get_reference_details`, `read_reference_fulltext`, `retry_pdf_download`, `format_references`, `set_citation_style`, `insert_citation`	Reference management
Writing (16)	`write_draft`, `read_draft`, `list_drafts`, `draft_section`, `get_section_template`, `count_words`, `validate_concept`, `validate_concept_quick`, `list_templates`, `read_template`, `start_document_session`, `insert_section`, `verify_document`, `check_word_limits`, `save_document`, `export_word`	Manuscript preparation and export
Project (15)	`create_project`, `list_projects`, `switch_project`, `get_current_project`, `update_project_status`, `get_project_paths`, `get_paper_types`, `update_project_settings`, `setup_project_interactive`, `start_exploration`, `get_exploration_status`, `convert_exploration_to_project`, `save_diagram`, `save_diagram_standalone`, `list_diagrams`	Project & diagram management
Skills (3)	`list_skills`, `load_skill`, `suggest_skill`	Workflow automation via Skills System

Note: Analysis tools (analyze_dataset, run_statistical_test, create_plot, generate_table_one) have been moved to a separate data-analysis-mcp project.

🎨 drawio Tools (15 total) - Optional Integration

Category	Tools	Description
Diagram	`create_diagram`, `edit_diagram`, `get_diagram_info`, `apply_diagram_changes`	Create and edit diagrams
Templates	`list_templates`, `create_from_template`	Architecture templates (AWS, GCP, Azure)
Guidelines	`get_drawing_guidelines`, `get_style`, `list_styles`	Drawing best practices
File	`load_file`, `save_tab`	File management

🧪 Development & Testing

```bash

Activate virtual environment

source .venv/bin/activate

Run all tests

pytest tests/

Run specific test file

pytest tests/test_search.py

Run with verbose output

pytest tests/ -v ```

🗺️ Roadmap

We're actively developing new features. Here's what's coming:

Status	Feature	Description
✅	Draw.io Integration	Generate CONSORT/PRISMA flowcharts with real-time WebSocket sync via next-ai-draw-io
✅	Skills System	Workflow definitions guiding AI through complex tasks - see documentation
✅	Parallel Search	Generate multiple search queries and execute in parallel - how it works
✅	Search Strategy Integration	Configure once, apply everywhere - strategy auto-integrates into parallel queries
✅	Table 1 Generator	Auto-generate baseline characteristics table from CSV via `generate_table_one`
✅	Foam Integration	Wikilinks `[[citation_key]]`, hover preview, backlinks panel for reference management
✅	PubMed MCP	Independent MCP server for literature search - can be used standalone
🔜	Multi-language Support	Full UI localization beyond English/Chinese
🔜	Journal Style Library	Pre-configured styles for major medical journals
📋	REST API Mode	Expose MCP tools as REST API for external integrations
📋	Dashboard File Browser	Integrate Chonky React file browser into Dashboard for visual project file management
📋	Collaboration Mode	Multi-author workflow with version control
📋	AI-Assisted Review	Use LLM to review and improve manuscript sections

Legend: ✅ Complete | 🔜 In Progress | 📋 Planned

Want to help? Check out CONTRIBUTING.md or open an issue to discuss!

🤝 Contributing

We welcome contributions from the community! Whether you're a medical researcher with workflow ideas, a developer who wants to add features, or someone who found a bug - we'd love to hear from you.

Ways to contribute:

🐛 Report bugs - Open an issue describing the problem
💡 Suggest features - Share your ideas for new tools or workflows
📝 Improve documentation - Help make the README clearer
🔧 Submit code - Fork, code, and send a PR
🌐 Translations - Help translate to other languages

Getting started:

Fork this repository
Create a feature branch (git checkout -b feature/amazing-feature)
Commit your changes (git commit -m 'Add amazing feature')
Push to the branch (git push origin feature/amazing-feature)
Open a Pull Request

See CONTRIBUTING.md for detailed guidelines.

📄 License

This project is licensed under the MIT License. See LICENSE for details.

⭐ Star History

If you find this project helpful, please consider giving it a star! It helps others discover the project.

🇹🇼 繁體中文

🎯 為什麼選擇這個工具？

傳統論文寫作工具要求你在開始前就知道確切的方向。但研究很少是這麼線性的。

Medical Paper Assistant 不一樣：

🔍 先探索，後決定 - 自由瀏覽文獻、儲存有興趣的論文，再決定研究方向
💬 對話式工作流程 - 用自然語言與 AI 對話來精煉想法，不用填表單
🧭 引導式流程 - 一步步的提示引導你從構思到可投稿的論文
🔗 原生 MCP + Copilot 整合 - 直接在 VS Code 中運作，不用切換應用程式

💡 我們的獨特之處

傳統工具	Medical Paper Assistant
固定模板、僵化流程	彈性、探索式方法
搜尋/寫作/引用分開多個 App	一站式整合體驗
手動管理參考文獻	自動儲存 + PDF + 引用格式
匯出後再排版	直接匯出符合期刊格式的 Word
學習複雜介面	自然語言對話

概述

Medical Paper Assistant（醫學論文寫作助手）是一款專為醫學研究人員設計的 AI 輔助寫作工具。基於 Model Context Protocol (MCP) 架構，可與 VS Code 的 GitHub Copilot 無縫整合，提供從文獻檢索、數據分析、草稿生成到格式化輸出的完整工作流程。

✨ 主要功能

功能	說明
文獻檢索與管理	連接 PubMed API 搜尋文獻，從 PMC Open Access 下載 PDF，建立本地文獻庫
數據分析	自動讀取 CSV 數據，執行統計檢定（t-test、相關性分析等），生成出版品質的圖表
智慧草稿生成	根據您的研究構想與分析結果，自動生成論文草稿
自動引用	插入引用標記 `[1]` 並生成格式化的參考文獻列表（支援 Vancouver、APA、Nature 格式）
互動式修正	透過對話方式微調特定段落
Word 匯出	將 Markdown 草稿匯出為符合期刊格式的 `.docx` 檔案

🚀 安裝指南

系統需求

開始之前，請確認您已安裝以下軟體：

需求	版本	檢查方式	安裝指南
Python	3.10 或更高	終端機輸入 `python3 --version`	python.org
Git	任何近期版本	終端機輸入 `git --version`	git-scm.com
VS Code	最新版	開啟 VS Code → 說明 → 關於	code.visualstudio.com
GitHub Copilot	擴充功能	VS Code 擴充功能面板	在擴充功能中搜尋 "GitHub Copilot"

快速安裝（推薦）

步驟 1：複製專案

開啟終端機並執行：

```bash

切換到您想存放專案的目錄

cd ~/projects # 或任何您選擇的目錄

複製專案

git clone https://github.com/u9401066/med-paper-assistant.git

進入專案目錄

cd med-paper-assistant ```

💡 什麼是 `git clone`？ `git clone` 會從 GitHub 下載完整的專案副本到您的電腦。這包含所有程式碼、歷史記錄和分支。

步驟 2：執行安裝腳本

Linux/macOS： ```bash

給予腳本執行權限

chmod +x scripts/setup.sh

執行安裝腳本

./scripts/setup.sh ```

Windows（PowerShell - 推薦）： ```powershell

執行 PowerShell 安裝腳本

.\scripts\setup.ps1 ```

Windows（命令提示字元）： ```cmd

執行批次安裝腳本

scripts\setup.bat ```

這個腳本會自動完成：

✅ 建立 Python 虛擬環境（`.venv/`）
✅ 安裝所有必要的套件
✅ 建立 `.vscode/mcp.json` 設定檔（自動偵測作業系統）
✅ 驗證安裝

💡 什麼是虛擬環境？ 虛擬環境是一個隔離的 Python 環境，讓這個專案的套件與系統的 Python 分開。這樣可以避免與其他專案產生衝突。

步驟 3：重新載入 VS Code

安裝完成後：

開啟 VS Code
按下 `Ctrl+Shift+P`（Windows/Linux）或 `Cmd+Shift+P`（Mac）
輸入 `Developer: Reload Window` 並按 Enter

💡 為什麼要重新載入？ VS Code 需要重新讀取 `.vscode/mcp.json` 設定檔才能啟動 MCP 伺服器。

步驟 4：驗證安裝

在 VS Code 的 Copilot Chat 面板中：

輸入 /mcp 並按 Enter
您應該會看到列表中有 mdpaper (50 tools)（如果啟用 Draw.io 整合，還會有 drawio (15 tools)）
如果看到這個，表示安裝成功！🎉

選用：設定 Draw.io 整合

如需圖表生成功能（CONSORT/PRISMA 流程圖），執行：

# 初始化子模組並設定 Draw.io
./scripts/setup-integrations.sh

這會新增 drawio MCP 伺服器，提供圖表生成工具。

使用方式：

啟動網頁應用：./scripts/start-drawio.sh
重新載入 VS Code
在 Copilot Chat 中使用 drawio 工具

選用：推薦的 VS Code 擴展

這些擴展可以增強你的研究工作流程：

擴展	用途	安裝
Project Manager	快速切換專案、側邊欄、標籤	`ext install alefragnani.project-manager`
Foam	Wikilinks、反向連結、浮動預覽	`ext install foam.foam-vscode`

快速安裝：

code --install-extension alefragnani.project-manager
code --install-extension foam.foam-vscode

📚 Foam 整合：參考文獻管理

Medical Paper Assistant 整合了 Foam 來提供強大的參考文獻連結和預覽系統。

什麼是 Foam？

Foam 是一個受 Roam Research 啟發的個人知識管理系統。它將你的 Markdown 文件轉換為互相連結的知識圖譜：

Wikilinks [[連結]] - 在筆記之間建立連結
浮動預覽 - 滑鼠懸停即可看到內容，無需打開文件
反向連結 (Backlinks) - 查看哪些筆記引用了當前文件
圖譜視覺化 - 視覺化你的文獻關聯

設定 Foam

步驟 1：安裝 Foam 擴展

在 VS Code 中：

按 Ctrl+Shift+X 打開擴展面板
搜尋 foam.foam-vscode
點擊安裝
重新載入 VS Code（Ctrl+Shift+P → Developer: Reload Window）

步驟 2：儲存參考文獻

當你儲存參考文獻時，系統會自動建立 Foam 相容的檔案：

# 使用 MCP 工具
save_reference(pmid="38049909")

# 建立的檔案：
references/
├── 38049909/
│   ├── content.md       ← 主內容，包含所有引用格式
│   └── metadata.json    ← 結構化資料
└── tang2023_38049909.md ← Foam 別名檔，方便連結

步驟 3：在草稿中使用 Wikilinks

# 我的研究草稿

根據 [[tang2023_38049909]] 的研究，remimazolam 在 ICU 鎮靜中
顯示出比 propofol 更好的效果。

另一篇 [[smith2024_41285088]] 的研究也支持這個發現...

步驟 4：懸停預覽

將滑鼠移到任何 [[連結]] 上，即可立即看到：

📄 論文標題和作者
📋 所有引用格式（Vancouver、APA、Nature）
📝 完整摘要
🔗 可點擊的 DOI 連結

引用鍵格式

參考文獻以「人性化 + 可驗證」的格式儲存：

[[作者_年份_pmid]]
   │      │     │
   │      │     └── PMID 用於驗證
   │      └── 出版年份
   └── 第一作者姓氏

範例：[[tang2023_38049909]]
      → Tang et al., 2023, PMID: 38049909
      → 驗證：https://pubmed.ncbi.nlm.nih.gov/38049909/

預覽中的引用格式

懸停參考文獻時，你會看到所有引用格式。專案選擇的格式會用 ⭐ 標記並置頂：

## 📋 Citation Formats

**⭐ APA**: Tang, Y., Gao, X., ... (2023). Remimazolam besylate versus...

**Vancouver**: Tang Y, Gao X, et al. Remimazolam besylate versus...

**Nature**: Tang Y, et al. Remimazolam besylate versus... *Crit Care* **27**...

**In-text**: (Tang et al., 2023)

設定你偏好的引用格式：

set_citation_style(style="apa")  # 選項：vancouver, apa, nature, harvard, ama

Foam 研究功能

功能	使用方式	好處
懸停預覽	滑鼠移到 `[[連結]]`	不用開檔案就能看摘要
反向連結面板	開啟任何參考文獻檔案	查看哪些草稿引用了這篇論文
圖譜視圖	`Ctrl+Shift+P` → `Foam: Show Graph`	視覺化論文之間的關聯
孤立參考文獻	`Ctrl+Shift+P` → `Foam: Show Orphans`	找出已儲存但未被引用的論文

手動安裝（如果腳本失敗）

點擊展開手動安裝步驟

步驟 1：複製專案

```bash git clone https://github.com/u9401066/med-paper-assistant.git cd med-paper-assistant ```

步驟 2：建立虛擬環境

```bash

建立名為 .venv 的虛擬環境

python3 -m venv .venv

啟動虛擬環境

Linux/Mac：

source .venv/bin/activate

Windows（命令提示字元）：

.venv\Scripts\activate.bat

Windows（PowerShell）：

.venv\Scripts\Activate.ps1 ```

💡 如何確認虛擬環境已啟動？ 啟動後，您的終端機提示符前面會出現 `(.venv)`，例如： ``` (.venv) user@computer:~/med-paper-assistant$ ```

步驟 3：安裝套件

```bash

先更新 pip（建議）

pip install --upgrade pip

以可編輯模式安裝專案

pip install -e . ```

💡 什麼是 `pip install -e .`？ `-e` 代表「可編輯模式」（editable mode），這表示當您修改程式碼時，不需要重新安裝就能看到變更。`.` 代表當前目錄。

步驟 4：設定 VS Code MCP

在專案根目錄建立 `.vscode/mcp.json` 檔案：

```bash

建立 .vscode 目錄（如果不存在）

mkdir -p .vscode

建立 mcp.json 檔案

cat > .vscode/mcp.json << 'EOF' { "inputs": [], "servers": { "mdpaper": { "type": "stdio", "command": "${workspaceFolder}/.venv/bin/python", "args": ["-m", "med_paper_assistant.interfaces.mcp"], "env": { "PYTHONPATH": "${workspaceFolder}/src" } } } } EOF ```

⚠️ Windows 使用者注意：請將 `.venv/bin/python` 改為 `.venv\Scripts\python.exe`

步驟 5：重新載入 VS Code

按下 `Ctrl+Shift+P` → 輸入 `Developer: Reload Window` → 按 Enter

疑難排解

❌ 「找不到命令：python3」

問題：Python 未安裝或不在系統 PATH 中。

解決方案：

從 python.org 下載並安裝 Python
安裝時，務必勾選「Add Python to PATH」選項
重新啟動終端機後再試一次

❌ Copilot 中看不到 MCP 伺服器

問題：VS Code 無法偵測到 MCP 伺服器。

解決方案：

確認 `.vscode/mcp.json` 檔案存在且路徑正確
確認虛擬環境已建立：執行 `ls .venv/`
嘗試重新載入 VS Code：`Ctrl+Shift+P` → `Developer: Reload Window`
檢查 VS Code 輸出面板：`檢視` → `輸出` → 從下拉選單選擇「MCP」

❌ 「ModuleNotFoundError: med_paper_assistant」

問題：套件未安裝或 PYTHONPATH 設定錯誤。

解決方案： ```bash

先啟動虛擬環境

source .venv/bin/activate # Linux/Mac

.venv\Scripts\activate # Windows

重新安裝套件

pip install -e . ```

📖 使用指南

本助手透過 MCP 與 GitHub Copilot 整合。在 Copilot Chat 中使用 `/mcp.mdpaper.*` 指令。

工作流程概覽

flowchart LR
    A["📝 研究構想<br/>/concept"] --> B["🔍 文獻搜尋<br/>/strategy"]
    B --> C["📊 數據分析<br/>/analysis"]
    C --> D["✍️ 撰寫草稿<br/>/draft"]
    D --> E["🔄 內容修正<br/>/clarify"]
    E --> F["📄 格式匯出<br/>/format"]

🆕 文獻探索模式（推薦新手使用！）

還不確定研究方向？ 先從探索文獻開始！

指令：/mcp.mdpaper.search

flowchart LR
    A["🔍 開始探索"] --> B["📚 搜尋並儲存論文"]
    B --> C{"找到方向了嗎？"}
    C -->|還沒| B
    C -->|找到了！| D["📁 轉換成正式專案"]
    D --> E["✍️ 開始撰寫"]

使用方式：

開始探索 - 自動建立臨時工作區
```
start_exploration()
```

搜尋並儲存有興趣的論文 - 不需要事先決定專案

search_literature(query="您感興趣的主題")
save_reference(pmid="12345678")  # 儲存有興趣的論文

找到研究方向後 - 將探索內容轉換成正式專案

convert_exploration_to_project(
    name="Your Research Title",  # 英文名稱
    paper_type="original-research"
)

💡 好處：探索期間儲存的所有文獻都會自動轉移到新專案！

步驟 1：準備您的資料

開始之前，請整理您的檔案：

目錄	用途
`data/`	將您的 CSV 研究數據放在這裡
`templates/`	（選用）期刊 Word 範本

步驟 2：發展研究構想

指令：`/mcp.mdpaper.concept`

這個引導式流程會幫助您建構具有創新性保護的結構化研究構想：

flowchart LR
    A["📚 文獻搜尋"] --> B["🔍 缺口識別"]
    B --> C{"❓ 用戶確認缺口"}
    C -->|是| D["✍️ 概念撰寫"]
    C -->|否| B
    D --> E["✅ 驗證概念"]

逐步流程：

📚 文獻搜尋（強制）
- Agent 搜尋 PubMed 找出 5-10 篇相關研究
- 重要文獻自動儲存到參考文獻庫
🔍 缺口識別（強制）
- Agent 呈現文獻摘要
- 識別研究缺口與限制
- 🛑 詢問您確認您的研究要解決哪個缺口
✍️ 概念撰寫（用戶確認後）
- 建立具有保護區塊的結構化概念：
  - 🔒 NOVELTY STATEMENT - 創新性聲明（受保護）
  - 🔒 KEY SELLING POINTS - 您的 3-5 個核心賣點（受保護）
  - 📝 背景、方法、預期結果（可編輯）
✅ 驗證
- 使用 validate_concept 工具驗證完整性

輸出會儲存到 `drafts/concept_*.md`。

💡 為什麼需要受保護區塊？ 🔒 標記確保您的關鍵創新點在撰寫完整論文時被保留。Agent 必須在修改這些區塊前先詢問您。

步驟 3：設定搜尋策略

指令：`/mcp.mdpaper.strategy`

設定文獻搜尋參數：

關鍵字與 MeSH 術語
排除條件
文章類型（Review、Clinical Trial、Meta-Analysis）
日期範圍
最小樣本數

步驟 4：分析數據

指令：`/mcp.mdpaper.analysis`

助手會：

列出 `data/` 中可用的 CSV 檔案
詢問您指定分組變數與結果變數
執行適當的統計檢定
生成圖表（儲存到 `results/figures/`）
建立 Table 1（基線特徵表）

步驟 5：撰寫草稿

指令：`/mcp.mdpaper.draft`

助手會：

讀取您的 `concept.md` 和分析結果
詢問使用哪個範本
生成 Markdown 草稿，包含：
- 嵌入的圖表
- 自動引用
- 格式化的表格

⚠️ 重要提示：此指令需要 concept 檔案。請先使用 `/mcp.mdpaper.concept` 建立。

步驟 6：修正內容

指令：`/mcp.mdpaper.clarify`

互動式改善特定段落：

「讓 Introduction 寫得更保守一點」
「擴充 limitations 段落」
「簡化統計方法的描述」

步驟 7：匯出到 Word

指令：/mcp.mdpaper.format

依照 7 步驟流程：

讀取範本結構
讀取草稿內容
對應章節
插入內容
驗證文件
檢查字數限制
儲存最終文件

輸出：results/your_paper.docx

📂 專案結構

目錄	用途
`projects/`	研究專案（獨立工作區）
`projects/{slug}/concept.md`	研究概念（含受保護區塊）
`projects/{slug}/drafts/`	Markdown 草稿
`projects/{slug}/references/{PMID}/`	本地文獻庫
`projects/{slug}/data/`	原始資料檔案（CSV）
`projects/{slug}/results/`	生成的輸出（圖表、表格、docx）
└ `metadata.json`	引用資訊、格式化的參考文獻
└ `content.md`	摘要與筆記
└ `fulltext.pdf`	PDF（若可從 PMC 取得）
`templates/`	期刊 Word 範本（`.docx` 用於匯出）
`src/med_paper_assistant/core/`	核心邏輯模組
`src/med_paper_assistant/mcp_server/`	MCP 伺服器實作
└ `templates/`	Agent 引導用內部範本（`.md`）
`tests/`	測試套件
`.memory/`	專案脈絡檔案
└ `.agent_constitution.md`	Agent 行為規則（動態載入）

⚖️ Agent 憲法

MCP 伺服器會動態載入 .memory/.agent_constitution.md 中的 Agent 憲法。此檔案定義 Copilot 必須遵循的核心行為規則。

憲法結構：

部分	內容
第一部分：系統開發	語言規則、Memory Bank 維護、程式碼品質、隱私
第二部分：研究操作	Concept 流程、受保護內容規則、Draft 撰寫規則

受保護內容規則 (🔒)：

## 2.2 🔒 受保護內容規則

| 區塊 | 說明 | 規則 |
|------|------|------|
| 🔒 NOVELTY STATEMENT | 研究創新性聲明 | 必須詢問才能修改 |
| 🔒 KEY SELLING POINTS | 核心賣點（3-5點）| 必須詢問才能修改 |
| 🔒 Author Notes | 作者私人備註 | 不可修改 |

修改規則：
✅ 可以潤飾文字
✅ 可以改善學術表達
⛔ 實質修改前必須詢問用戶
⛔ 絕不可刪除受保護內容
⛔ 絕不可弱化核心賣點

自訂方式： 編輯 .memory/.agent_constitution.md 並重新載入 VS Code，新規則將自動套用。

🎯 新穎性驗證系統

系統內建智慧型新穎性驗證功能，確保研究概念在撰寫草稿前真正描述了創新貢獻。

運作方式

設定	值	說明
評分輪數	3	獨立評估的次數
通過門檻	75/100	每輪最低分數
通過條件	3 輪皆 ≥ 75	必須全部通過

評估維度

維度	權重	檢查項目
獨特性	25%	方法/途徑是否獨特？
重要性	25%	貢獻是否有意義？
缺口對齊	20%	新穎性是否解決研究缺口？
具體性	15%	宣稱是否具體明確？
可驗證性	15%	宣稱是否可被驗證/測試？

設計原則

草稿撰寫前強制驗證：系統強制在使用概念檔案生成草稿前進行驗證
多輪評估：使用 3 輪評估提高可靠性，減少誤判
高門檻：要求 75+ 確保只有表達清晰的新穎性宣稱能通過
可操作的回饋：驗證失敗時提供具體改進建議

🧠 技能系統

技能（Skills）是完整的工作流程定義，引導 AI 完成複雜的多步驟任務。與執行單一動作的工具不同，技能定義了從開始到完成的整個流程。

📁 完整說明文件：.skills/README.md

何時使用技能

使用者請求	要載入的技能
「幫我做文獻回顧」	`literature_review`
「發展研究概念」	`concept_development`
「搜尋 [主題] 的論文」	`parallel_search`
「寫 Introduction」	`draft_introduction`
「寫 Discussion」	`draft_discussion`

技能工具

工具	用途
`list_skills`	顯示所有可用技能
`load_skill`	載入特定技能的工作流程
`suggest_skill`	根據任務描述建議適合的技能

⚡ 並行搜尋

並行搜尋透過同時執行多組搜尋策略，大幅加速文獻探索。

運作方式

1. generate_search_queries(topic="remimazolam ICU sedation")
   → 返回 5 組不同的查詢策略

2. 並行執行 search_literature × 5
   → 所有查詢同時執行

3. merge_search_results(...)
   → 返回：42 篇獨特論文（總共 56 篇，移除 14 篇重複）
   → 被多個策略找到的論文標記為「高相關性」

策略整合

設定一次搜尋策略，自動套用到所有並行查詢：

configure_search_strategy({
    "date_range": "2020-2025",
    "exclusions": ["animal", "review"],
    "article_types": ["Clinical Trial", "RCT"]
})

generate_search_queries(topic="...", use_saved_strategy=True)
→ 所有 5 組查詢都包含日期篩選、排除詞和文章類型限制

🛠️ 可用工具（共 52 + 15 個）

📚 mdpaper 工具（52 個）

類別	工具	說明
搜尋 (10)	`search_literature`, `find_related_articles`, `find_citing_articles`, `fetch_article_details`, `configure_search_strategy`, `get_search_strategy`, `generate_search_queries`, `merge_search_results`, `expand_search_queries`, `search_local_references`	文獻探索，透過 pubmed-search-mcp 子模組
參考文獻 (8)	`save_reference`, `list_saved_references`, `get_reference_details`, `read_reference_fulltext`, `retry_pdf_download`, `format_references`, `set_citation_style`, `insert_citation`	參考文獻管理
寫作 (16)	`write_draft`, `read_draft`, `list_drafts`, `draft_section`, `get_section_template`, `count_words`, `validate_concept`, `validate_concept_quick`, `list_templates`, `read_template`, `start_document_session`, `insert_section`, `verify_document`, `check_word_limits`, `save_document`, `export_word`	草稿準備與匯出
專案 (15)	`create_project`, `list_projects`, `switch_project`, `get_current_project`, `update_project_status`, `get_project_paths`, `get_paper_types`, `update_project_settings`, `setup_project_interactive`, `start_exploration`, `get_exploration_status`, `convert_exploration_to_project`, `save_diagram`, `save_diagram_standalone`, `list_diagrams`	專案與圖表管理
技能 (3)	`list_skills`, `load_skill`, `suggest_skill`	透過技能系統自動化工作流程

備註： 分析工具（analyze_dataset, run_statistical_test, create_plot, generate_table_one）已移至獨立的 data-analysis-mcp 專案。

🎨 drawio 工具（15 個）- 選用整合

類別	工具	說明
圖表	`create_diagram`, `edit_diagram`, `get_diagram_info`, `apply_diagram_changes`	建立和編輯圖表
範本	`list_templates`, `create_from_template`	架構範本（AWS、GCP、Azure）
指引	`get_drawing_guidelines`, `get_style`, `list_styles`	繪圖最佳實踐
檔案	`load_file`, `save_tab`	檔案管理

🧪 開發與測試

```bash

啟動虛擬環境

source .venv/bin/activate

執行所有測試

pytest tests/

執行特定測試檔案

pytest tests/test_search.py

顯示詳細輸出

pytest tests/ -v ```

🗺️ 開發藍圖

我們正在積極開發新功能，以下是規劃中的項目：

狀態	功能	說明
✅	Draw.io 整合	透過 next-ai-draw-io 生成 CONSORT/PRISMA 流程圖，支援即時 WebSocket 同步
✅	技能系統	工作流程定義，引導 AI 完成複雜任務 - 見說明文件
✅	並行搜尋	生成多組搜尋語法並同時執行 - 運作方式
✅	搜尋策略整合	設定一次，處處套用 - 策略自動整合到並行查詢
✅	Table 1 生成器	透過 `generate_table_one` 從 CSV 自動生成基線特徵表
✅	Foam 整合	Wikilinks `[[citation_key]]`、懸停預覽、反向連結面板 - 參考文獻管理
✅	PubMed MCP	獨立的 MCP 伺服器用於文獻搜尋 - 可單獨使用
🔜	多語言支援	完整 UI 本地化，不只英文/中文
🔜	期刊樣式庫	預設主要醫學期刊的格式設定
📋	REST API 模式	將 MCP 工具以 REST API 形式公開，供外部整合
📋	Dashboard 檔案瀏覽器	整合 Chonky React 檔案瀏覽元件到 Dashboard，視覺化管理專案檔案
📋	協作模式	多作者工作流程與版本控制
📋	AI 輔助審稿	使用 LLM 審閱並改善論文段落

圖例： ✅ 已完成 | 🔜 進行中 | 📋 規劃中

想幫忙？查看 CONTRIBUTING.md 或開 issue 來討論！

🤝 參與貢獻

我們歡迎社群的貢獻！無論你是有工作流程想法的醫學研究人員、想添加功能的開發者，或是發現 bug 的使用者 - 我們都很樂意聽到你的聲音。

貢獻方式：

🐛 回報 Bug - 開 issue 描述問題
💡 建議功能 - 分享你對新工具或工作流程的想法
📝 改善文件 - 幫忙讓 README 更清楚
🔧 提交程式碼 - Fork、開發、送 PR
🌐 翻譯 - 幫忙翻譯成其他語言

開始貢獻：

Fork 這個 repository
建立功能分支 (git checkout -b feature/amazing-feature)
Commit 你的修改 (git commit -m 'Add amazing feature')
Push 到分支 (git push origin feature/amazing-feature)
開一個 Pull Request

詳細指南請參閱 CONTRIBUTING.md。

📄 授權

本專案採用 MIT 授權。詳見 LICENSE 檔案。

⭐ 給顆星星

如果你覺得這個專案有幫助，請考慮給它一顆星！這能幫助其他人發現這個專案。

Name		Name	Last commit message	Last commit date
Latest commit History 131 Commits
.github		.github
.memory		.memory
.skills		.skills
.vscode		.vscode
dashboard		dashboard
integrations		integrations
scripts		scripts
src/med_paper_assistant		src/med_paper_assistant
templates		templates
tests		tests
.foam.json		.foam.json
.gitignore		.gitignore
.gitmodules		.gitmodules
ARCHITECTURE.md		ARCHITECTURE.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
ROADMAP.md		ROADMAP.md
pyproject.toml		pyproject.toml
search_strategy.json		search_strategy.json

License

u9401066/med-paper-assistant

Folders and files

Latest commit

History

Repository files navigation

Medical Paper Assistant 醫學論文寫作助手

🇬🇧 English

🎯 Why This Tool?

💡 What Makes Us Unique

Overview

✨ Key Features

🚀 Installation

Prerequisites

Quick Installation (Recommended)

Navigate to where you want to store the project

Clone the repository

Enter the project directory

Make the script executable

Run the setup script

Run the PowerShell setup script

Run the batch setup script

Optional: Setup Draw.io Integration

Optional: Recommended VS Code Extensions

📚 Foam Integration for Reference Management

What is Foam?

Setup Foam for This Project

Citation Key Format

Citation Formats in Preview

Foam Features for Research

🆕 Literature Exploration Mode (Recommended for Beginners!)

Step 1: Prepare Your Data

Step 2: Develop Your Research Concept

Step 3: Configure Search Strategy

Step 4: Analyze Your Data

Step 5: Write Your Draft

Step 6: Refine Your Content

Step 7: Export to Word

📂 Project Structure

⚖️ Agent Constitution

🎯 Novelty Validation System

How It Works

Scoring Mechanism

Evaluation Dimensions

Example Validation Report

Key Design Decisions

🧠 Skills System

When to Use Skills

How It Works

Skill Tools

⚡ Parallel Search

How It Works

Usage Example

Strategy Integration

🛠️ Available Tools (52 mdpaper + 15 drawio)

📚 mdpaper Tools (52 total)

🎨 drawio Tools (15 total) - Optional Integration

🧪 Development & Testing

Activate virtual environment

Run all tests

Run specific test file

Run with verbose output

🗺️ Roadmap

🤝 Contributing

📄 License

⭐ Star History

🇹🇼 繁體中文

🎯 為什麼選擇這個工具？

💡 我們的獨特之處

概述

✨ 主要功能

🚀 安裝指南

系統需求

快速安裝（推薦）

切換到您想存放專案的目錄

複製專案

進入專案目錄

給予腳本執行權限

執行安裝腳本

執行 PowerShell 安裝腳本