Jason

README excerpt

# Disclaimer

This was made in a total of about 6 hours during the Shack15 GPT-5 hackathon.

Maybe it will serve as useful to folks on a draft version of an interpreter or as inspiration for other projects.

----

# lang-gen

End-to-end DSL generator with grammar-constrained output from OpenAI.

## Features

- Generates strict-subset Lark grammars from natural language descriptions
- Compiles grammars to JavaScript parsers using lark-js
- Auto-derives AST schema from grammar structure
- Generates interpreters based on semantic descriptions
- **VSCode extension generation** with syntax highlighting
- No hardcoded language semantics - everything is generated at runtime

## Installation

```bash
# Install dependencies (this will also set up Python environment)
npm install

# The postinstall script automatically:
# 1. Creates a Python virtual environment
# 2. Installs lark-js in the venv
# 3. Verifies the installation

# If you need to manually run setup:
./setup.sh
```

### Requirements

- Node.js 18+
- Python 3 (for lark-js compiler)
- OpenAI API key

## Usage

```bash
# Basic usage
node index.js --spec "tiny calc with ints + - * / and parens" --semantics "evaluate to a number"

# With sample code to test
node index.js --spec "boolean logic with AND OR NOT" --semantics "evaluate to true/false" --sample "true AND (false OR true)"

# Read semantics from stdin
echo "compile to JavaScript code" | node index.js --spec "simple expression language"

# Specify output directory
node ind