AI Terminal screenshot

AI Terminal Plugin for Revolution EDA

Overview

The aiTerminal plugin provides an AI-powered terminal for editing schematic, symbol, layout, and simulation-bench views using natural-language requests. The current plugin supports Anthropic Claude, Google Gemini, and OpenAI models. AI provider API access and usage are separate from the plugin license.

Features

  • AI-Powered Design Modification: Use natural language to request changes to your designs
  • Multiple AI Providers: Select Claude, Gemini, or OpenAI models in the terminal
  • Validated Design Changes: Candidate edits are checked against design schemas and local library resources before being saved
  • Backup and Undo: Recovery backup and editor undo support for applied modifications
  • Design Commands: Inspect views with read, request changes with ai:, or use ops: and planning commands for structured tasks
  • View Creation: Generate symbol views, testbench schematics, and revbench testbenches with dedicated commands
  • Deterministic Layout Planning: layplan: turns natural-language placement requests into layout records with grid snapping, PCell parameter checks, and via construction
  • Context and References: Pin example schematics (context:) or a cross-view design reference (ref:) to guide AI output
  • Session Memory and Cost Reporting: Per-window conversation history with token and cost tracking; new resets the session
  • Secure API Key Management: Fernet-encrypted storage of provider API keys
  • Customizable AI Instructions: User and project instruction files tailor AI behavior
  • License Management: Commercial license validation at use time

Installation

  1. Check Plugins → Browse Plugins… for a compatible package, or place the plugin in your configured REVEDA_PLUGIN_PATH directory.
  2. Restart Revolution EDA after installing or changing the plugin path.
  3. Activate a valid license and provide your own API key for the selected AI provider. Registry availability may differ by platform.

Usage

Opening the Terminal

Open the AI Terminal via Utilities → AI Terminal or use the keyboard shortcut Ctrl+T. The terminal appears as a panel inside the editor window. It is available in the Schematic Editor, Symbol Editor, and Layout Editor, and in the Simulation Environment via Tools → AI Terminal for testbench requests.

Setting Up API Keys

Before using the AI features, you need to set your API key:

  1. Select your preferred model from the dropdown in the terminal title bar
  2. Type setkey and press Enter
  3. Enter your API key when prompted
  4. The key will be encrypted and stored securely in ~/.reveda/api_keys.enc

Basic Commands

  • help — Display available commands
  • read — Print the current design JSON
  • undo — Revert the last applied AI change
  • new — Start a fresh AI session (clears conversation memory)
  • setkey — Store or update the selected provider’s API key
  • license — Show installed license information
  • ai:<request> — Modify the open design with the selected model
  • ops:<request> — Request targeted edits instead of a full design rewrite
  • plan:<request> — Generate or restructure a schematic with explicit symbol references
  • layplan:<request> — Generate layout records through the deterministic plan pipeline
  • tb:<request> — Generate or modify testbench settings in a revbench view
  • tbnew:<request> — Create a new revbench testbench for the current schematic or config view
  • symnew:<request> — Create a symbol view for the current schematic
  • tbschematic:<request> — Create a testbench schematic in a new <dut>_tb cell
  • context:<library/cell[, ...]> — Select example schematics as read-only style/topology reference (context:show / context:clear to inspect or clear)
  • ref:<library/cell/view> — Select one read-only cross-view design reference (ref:show / ref:clear to inspect or clear)

Example Requests

  • ai:Add a 100fF capacitor between nodes A and B
  • ai:Change the width of NMOS current mirror transistors to 1.0um
  • ai:Connect the output of inverter I1 to the input of NAND gate N1
  • ai:Create a differential pair with matched transistors

Model Selection

Switch between AI models using the dropdown in the terminal title bar:

  • Claude — Anthropic models
  • Gemini — Google models
  • OpenAI — ChatGPT models

Customizing AI Instructions

The AI Terminal supports configurable instruction files that let you tailor AI behavior per model or per project.

Configuration

Create an ai_config.json in one of these locations:

  • ~/.reveda/ai_config.json — user-wide configuration
  • ./ai_config.json — project-specific configuration (overrides user config)

Example configuration:

{
  "instruction_paths": {
    "base_dir": "~/.reveda/ai_instructions",
    "claude": "claude.md",
    "gemini": "gemini.md",
    "openai": "openai.md",
    "general": "general.md"
  },
  "fallback_instructions": {
    "enabled": true,
    "use_builtin": true
  }
}

Instruction Files

Place markdown instruction files in the configured directory (default: ~/.reveda/ai_instructions/). Each model can have its own instruction file, plus a shared general.md that applies to all models. Built-in fallback instructions are used when external files are missing.

Security

  • API keys are encrypted using Fernet symmetric encryption and stored in ~/.reveda/api_keys.enc
  • The encryption key is stored separately in ~/.reveda/key.enc
  • The plugin requires a valid license to operate; license validation runs inside compiled plugin code

License

Commercial — proprietary add-on to Revolution EDA. A valid license key is required.