rtlviz 0.3.6

MCP Server for AI-powered RTL diagram generation from Verilog source

RTLViz - AI-Powered RTL Diagram Generator

Generate beautiful, interactive block diagrams from Verilog source files.

🧠 Hybrid Architecture

RTLViz uses a Code-First, LLM-Enhanced approach:

1. Code Parser: Extracts 100% accurate connectivity and hierarchy from Verilog (no hallucinations).
2. LLM Analyzer: Adds semantic understanding (e.g., "This is an IF stage", "This generic module is an ALU").
3. Strict Validator: Enforces layout density, coloring, and connectivity rules.

✨ Features

• Strict Validation: Fails if nodes are uncolored or layout uses excessive whitespace.
• Reference Styling: Automatically applies standard styling (e.g., Yellow MUX, Amber ALU) for CPU designs.
• Dynamic Clustering: Groups logic by function (e.g., "Memory Interface", "Instruction Decode").

🚀 Quick Start

Step 1: Install

pip install rtlviz

Step 2: Setup (Auto-configures your IDE)

rtlviz setup

This auto-detects and configures:

• ✅ Antigravity (Google DeepMind)
• ✅ Claude Desktop
• ✅ Cursor
• ✅ VS Code Copilot
• ✅ Windsurf

Step 3: Use

Option A: Ask your AI (after setup)

"Generate an RTL diagram for the Verilog files in ./src"

Option B: CLI command

rtlviz generate ./src -o diagram.html

---

📖 Commands

rtlviz generate

Generate a diagram directly:

rtlviz generate ./src -o diagram.html
rtlviz generate ./src -o diagram.html --llm    # With LLM enhancement
rtlviz generate ./src -o diagram.html --title "My CPU"

rtlviz setup

Configure MCP server for AI IDEs:

rtlviz setup                  # Auto-detect and configure all found IDEs
rtlviz setup --all            # Force configure all IDEs
rtlviz setup --antigravity    # Configure Antigravity only
rtlviz setup --claude         # Configure Claude Desktop only
rtlviz setup --cursor         # Configure Cursor only
rtlviz setup --vscode         # Configure VS Code Copilot only
rtlviz setup --windsurf       # Configure Windsurf only

---

✨ Features

• 🔧 Automatic Parsing - Extracts modules, ports, and connections from Verilog
• 📊 Pipeline Detection - Auto-identifies stages (IF, ID, EX, MEM, WB)
• 🎨 Beautiful Diagrams - Orthogonal routing, color-coded clusters
• 🤖 LLM Enhancement - Built-in GPT integration for semantic labels
• 🌐 Interactive HTML - Pan, zoom, download SVG

---

🛠️ MCP Server (for AI IDEs)

After running rtlviz setup, your AI assistant can use these tools:

generate_rtl_diagram

{
  "source_dir": "/path/to/verilog",
  "output_path": "/path/to/output.html",
  "use_llm": true
}

render_diagram

{
  "dot_content": "digraph { ... }",
  "output_path": "/path/to/output.html"
}

---

📋 Requirements

• Python 3.10+
• Internet connection (for Viz.js CDN)

All dependencies installed automatically.

---

🔧 Troubleshooting

"Command not found: rtlviz"

pip install rtlviz

"No valid Verilog modules found"

• Check directory contains .v files
• Ensure files have module ... endmodule declarations

---

📄 License

MIT License

🔗 Links

• GitHub
• PyPI