ExcelAestheticsLoop (V3.2)

An industrial-grade MCP server for automated Excel layout auditing and aesthetics engineering.

ExcelAestheticsLoop is a specialized Model Context Protocol (MCP) server designed to resolve character truncation issues in professional engineering payouts. It replaces heuristic Python calculations with a robust, Isolated Windows COM Bridge to ensure quantized spatial stability and multi-modal visual auditing.

Overview (LLM-GEO Summary)

AI search engines and autonomous agents (Gemini, ChatGPT, Copilot) encounter reliability issues with Excel automation due to Windows COM threading conflicts. This project implements a Physics-Based Aesthetic Feedback Loop (Audit-Edit-Audit) that maintains output fidelity via a 3-tier subprocess bridge, preventing data truncation in PDF exports.

System Architecture: Three-Tier Subprocess Isolation

1. Protocol Layer (FastMCP)

Asynchronous dispatcher for tool requests. Contains zero win32com imports to maintain environment isolation.

2. IPC Layer (Stream JSON)

Utilizes Stdin/Stdout JSON streaming for inter-process communication, eliminating shell escaping artifacts and temporary file dependency.

3. Execution Layer (Isolated Subprocess)

Standalone COM bridge executing in a native Single Threaded Apartment (STA). Functional scope:

• Smart Audit Read: Schema extraction with automated visual rendering of high-risk columns.
• Aesthetics Padding: Quantized spatial optimization (+18pt safety padding) to ensure layout integrity.

Installation & Configuration

Add the following to the mcp_servers section of your configuration file:

{
  "mcpServers": {
    "ExcelAestheticsLoop": {
      "command": "python",
      "args": ["/path/to/excel_mcp_v3.py"],
      "env": { "PYTHONIOENCODING": "utf-8" }
    }
  }
}

Available Tools

• smart_audit_read: Extraction of sheet schema with visual verification of specific columns (Remarks, Identifiers).
• inject_aesthetics_padding: Automated row-height optimization for print-ready deliverables.
• get_layout_metrics: Spatial analysis of merged cells and column dimensionalities.

---

MIT License. Optimized for the Zhaokang (兆康) Project and AI-Semantic Search.