Verilog-AI-Agent

This is a Python-based AI tool that takes natural language prompts and generates Verilog hardware modules and testbenches using OpenAI GPT-4. It then automatically compiles, simulates, and visualizes the resulting design using Icarus Verilog and waveform generation.

You can either:

• Use the Streamlit web interface, or
• Use the command-line script (CLI)

Features

• Natural Language to Verilog
  Enter prompts like "4-bit full adder" and get clean, syntactically correct Verilog modules.

• Powered by GPT-4 via OpenAI API
  Uses your OpenAI key to generate both the module and a corresponding testbench.

• Testbench Enhancement
  Injects $monitor, $dumpfile, $dumpvars, and $finish if missing — making testbenches simulation-ready.

• Automatic Simulation
  Compiles and simulates the design using Icarus Verilog (iverilog + vvp).

• Waveform Generation
  Dumps simulation data to a VCD file (wave.vcd) viewable in GTKWave or as a PNG (via matplotlib — still improving).

• Modular Design Comes with both a Streamlit GUI and a lightweight CLI script.

How It Works

Option 1: Streamlit Web Interface

1. Run the app locally (streamlit run streamlit_app.py)
2. Enter your OpenAI API key and hardware prompt
3. App:
   • Calls GPT-4 to generate module and testbench
   • Validates and patches testbench if needed
   • Lets you view, download, and simulate your design
4. After simulation:
   • Shows console output
   • Generates wave.vcd
   • Plots waveform as a PNG (basic version)
   • Optionally download waveform file to open in GTKWave

Option 2: CLI Script

1. Add your OpenAI API key in the script (openai.api_key = "sk-...")
2. Run the script (python verilog_cli_generator.py)
3. Flow:
   • Asks you for a hardware design prompt
   • GPT returns module + testbench
   • Automatically saves .v files in current directory
   • Asks if you want to simulate using iverilog
   • If yes, runs simulation and creates wave.vcd
   • Asks if you want to open waveform in GTKWave

Getting Started

Requirements

Tool	Required	Description
Python 3.8+	Yes	For both CLI and Streamlit
OpenAI API key	Yes	To access GPT-4 via OpenAI API
Icarus Verilog	Yes	Icarus Verilog compiler and simulator
gtkwave	Optional	For viewing wave.vcd (CLI + Streamlit)
matplotlib	Optional	For waveform plotting in Streamlit app
vcdvcd	Optional	VCD parsing library for PNG generation

OpenAI API Key Setup

To use either version, you'll need an OpenAI API key.

1. Go to https://platform.openai.com/account/api-keys
2. Click Create new secret key
3. Copy the key

• In Streamlit: Paste the key into the app when prompted
• In CLI: Replace "sk-..." in the script:

  openai.api_key = "sk-..."

Note: GPT-4 access requires a paid account

Installation

Streamlit Version

1. Clone the repository:

   git clone https://github.com/YOUR_USERNAME/verilog-ai-agent.git
   cd verilog-ai-agent/streamlit_app
   

2. Install Python dependencies:

   pip install -r requirements.txt
   

3. Run the app:

   streamlit run streamlit_app.py
   

Once running:

• Enter your OpenAI API key
• Enter a prompt like "3-bit up counter"
• Click Generate
• Click Run Simulation
• View results, waveform, and download files

CLI Version

1. Clone the repository:

   git clone https://github.com/YOUR_USERNAME/verilog-ai-agent.git
   cd verilog-ai-agent/cli_version
   

2.Add your API key to the script

3. Install Python dependencies:

   pip install -r requirements.txt
   

4. Run the script:

   python verilog_cli_generator.py
   

Once running, you will be prompted to:

• Hardware prompt
• Simulation option
• Whether to view waveform in GTKWave

Project Stucture

├── streamlit_app/
│   ├── streamlit_app.py       # Main Streamlit code
│   ├── requirements.txt       # Python dependencies
│   ├── waveform.png           # Rendered waveform (optional)
│   ├── wave.vcd               # Simulation output
│   └── *.v                    # Generated Verilog + Testbench
│
├── cli_version/
│   └── verilog_cli_generator.py # Simple CLI version

Notes

• Waveform visualization is a work in progress — matplotlib rendering is still being improved.
• Use GTKWave (gtkwave wave.vcd) for detailed signal visualization
• The tool assumes a valid module/testbench is generated; please check logs if simulation fails

Why I Built This

As a student passionate about both AI and hardware design, I built this tool to explore hardware-software co-design. It reflects my interest in bridging LLMs with digital design and using Python to streamline RTL workflows.
It lets me:

• Apply generative AI in an EDA context
• Automate repetitive design + simulation tasks
• Build practical AI tooling for Verilog users