asset-aware-mcp 0.2.0

Medical RAG with Asset-Aware MCP - Precise PDF asset retrieval (tables, figures, sections) for AI Agents

asset-aware-mcp

🏥 Medical RAG with Asset-Aware MCP - 讓 AI Agent 精準存取 PDF 文獻中的表格、章節與知識圖譜

🌐 繁體中文

🎯 Why Asset-Aware MCP?

AI 無法直接讀取你電腦裡的圖片檔案。 這是一個常見的誤解。

方式	AI 能分析圖片內容？	說明
❌ 給 PNG 路徑	否	AI 無法存取本地檔案系統
✅ Asset-Aware MCP	是	透過 MCP 取得 Base64，AI 視覺能力可直接理解

實際效果

# 透過 MCP 取得圖片後，AI 可以直接分析：

User: 這張圖在講什麼？

AI: 這是 Scaled Dot-Product Attention 的架構圖：
    1. 輸入 Q (Query)、K (Key)、V (Value)
    2. Q 和 K 做 MatMul（矩陣乘法）
    3. 經過 Scale（縮放 1/√dₖ）
    4. 可選的 Mask（用於 decoder）
    5. SoftMax 歸一化
    6. 與 V 做最後一次 MatMul 得到輸出

這就是 Asset-Aware MCP 的價值 - 讓 AI Agent 真正「看懂」你的 PDF 文獻中的圖表。

---

✨ Features

• 📄 Asset-Aware ETL - PDF → Markdown, using PyMuPDF to automatically identify tables, sections, and images
• 🔄 Async Job Pipeline - Supports asynchronous task processing, tracking progress for large documents
• 🗺️ Document Manifest - Structured list, allowing Agents to "see the map" before precisely accessing data
• 🧠 LightRAG Integration - Knowledge Graph + Vector Index, supporting cross-document comparison and reasoning
• 📊 A2T (Anything to Table) - Automatically orchestrate information extracted by Agents into professional Excel tables, supporting CRUD, Drafting, and Token-efficient resumption.
• �🔌 MCP Server - Exposes tools and resources to Copilot/Claude via FastMCP
• 🏥 Medical Research Focus - Optimized for medical literature, supporting Base64 image transmission for Vision AI analysis

🏗️ Architecture

┌─────────────────────────────────────────────────────────┐
│                    AI Agent (Copilot)                   │
└─────────────────────┬───────────────────────────────────┘
                      │ MCP Protocol (Tools & Resources)
┌─────────────────────▼───────────────────────────────────┐
│                 MCP Server (server.py)                  │
│  ┌─────────────┐ ┌─────────────┐ ┌─────────────────┐   │
│  │   ingest    │ │  inspect    │ │     fetch       │   │
│  │  documents  │ │  manifest   │ │     asset       │   │
│  └─────────────┘ └─────────────┘ └─────────────────┘   │
│  ┌─────────────────────────────────────────────────┐   │
│  │          A2T (Anything to Table) Workflow       │   │
│  │  [Plan] → [Draft] → [Batch Add] → [Commit]      │   │
│  └─────────────────────────────────────────────────┘   │
└─────────────────────┬───────────────────────────────────┘
                      │
┌─────────────────────▼───────────────────────────────────┐
│                  ETL Pipeline (DDD)                     │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐              │
│  │ PyMuPDF  │  │  Asset   │  │ LightRAG │              │
│  │ Adapter  │→ │  Parser  │→ │  Index   │              │
│  └──────────┘  └──────────┘  └──────────┘              │
└─────────────────────┬───────────────────────────────────┘
                      │
┌─────────────────────▼───────────────────────────────────┐
│                   Local Storage                         │
│  ./data/                                                │
│  ├── doc_{id}/        # Document Assets                 │
│  ├── tables/          # A2T Tables (JSON/MD/XLSX)       │
│  │   └── drafts/      # Table Drafts (Persistence)      │
│  └── lightrag/        # Knowledge Graph                 │
└─────────────────────────────────────────────────────────┘

📁 Project Structure (DDD)

asset-aware-mcp/
├── src/
│   ├── domain/              # 🔵 Domain: Entities, Value Objects, Interfaces
│   ├── application/         # 🟢 Application: Doc Service, Table Service (A2T), Asset Service
│   ├── infrastructure/      # 🟠 Infrastructure: PyMuPDF, LightRAG, Excel Renderer
│   └── presentation/        # 🔴 Presentation: MCP Server (FastMCP)
├── data/                    # Document and Asset Storage
├── docs/
│   └── spec.md              # Technical Specification
├── tests/                   # Unit and Integration Tests
├── vscode-extension/        # VS Code Management Extension
└── pyproject.toml           # uv Project Config

🚀 Quick Start

# Install dependencies (using uv)
uv sync

# Run MCP Server
uv run python -m src.presentation.server

# Or use the VS Code extension for graphical management

🔌 MCP Tools

Tool	Purpose
fetch_document_asset	Precisely retrieve tables (MD) / figures (B64) / sections
consult_knowledge_graph	Knowledge graph query, cross-document comparison
plan_table_schema	AI-driven schema planning & brainstorming (🆕)
create_table_draft	Start a persistent draft session (Token-efficient)
add_rows_to_draft	Batch add data to draft
commit_draft_to_table	Finalize draft into a formal table
resume_draft / resume_table	Resume work with minimal context (Save tokens)
update_cell	Precise cell-level editing
render_table	Render to professional Excel file (with conditional formatting)

🔧 Tech Stack

Category	Technology
Language	Python 3.10+
ETL	PyMuPDF (fitz)
RAG	LightRAG (lightrag-hku)
MCP	FastMCP
Storage	Local filesystem (JSON/Markdown/PNG)

📋 Documentation

• Technical Spec - 詳細技術規格
• Architecture - 系統架構
• Constitution - 專案原則

📄 License

Apache License 2.0