AI JSON Optimizer (Reduce Tokens)

Optimize JSON for AI models by reducing token usage with intelligent TOON format conversion.

Features

• Smart Optimization: Automatically chooses JSON or TOON format for maximum estimated token savings
• Recursive Conversion: Handles nested objects and arrays
• Token Estimation: Shows estimated token reduction percentage
• In-Place Editing: Replaces content directly in the editor
• Clipboard Copy: Optimized result copied to clipboard automatically
• Minified Output: Compact format for AI efficiency by default

Example

Before (JSON)

{
  "users": [
    {"name": "Alice", "age": 30},
    {"name": "Bob", "age": 25}
  ]
}

After (TOON)

{"k":["users"],"d":[{"k":["name","age"],"d":[["Alice",30],["Bob",25]]}]}

Result: around 35% estimated token reduction

Benefits

• Reduce AI API token usage by around 20-50%
• Faster AI responses
• Lower API costs
• Maintains data integrity

Usage

1. Open a JSON file in VS Code
2. Select JSON text (or run on full document)
3. Run Optimize JSON for AI from Command Palette (Ctrl+Shift+P)
4. Review the token savings notification
5. Optimized output is written to editor and clipboard

Requirements

• VS Code ^1.80.0

Settings

• aiOptimizer.prettyPrint: Enable formatted output (default: false)

Smart Optimization Logic

• Arrays of objects -> attempt TOON conversion
• Large objects (more than 3 keys or nested content) -> attempt TOON if beneficial
• Small/simple objects -> keep JSON
• Final decision -> choose the output with fewer estimated tokens

TOON Format

Optimized with shorter keys:

• Objects: {"k": ["key1", "key2"], "d": [value1, value2]}
• Arrays: {"k": ["key1", "key2"], "d": [[val1, val2], [val3, val4]]}

Nested structures are supported recursively.