TOON Format Support

TOON (Token-Oriented Object Notation) is an alternative output format for LogicStamp context bundles, providing a compact text-based encoding optimized for AI workflows and efficient storage.

stamp context --format toon

Overview

TOON combines YAML's indentation-based structure for nested objects with a CSV-style tabular layout for uniform arrays. It encodes the same LogicStamp bundle structure as JSON, but in a more compact representation that uses approximately 40% fewer tokens than standard JSON.

Key Features

Token-efficient - Uses ~40% fewer tokens than JSON while maintaining 74% accuracy (vs JSON's 70%) in LLM benchmarks
JSON data model - Lossless encoding of the same objects, arrays, and primitives as JSON
LLM-friendly guardrails - Explicit [N] length declarations and {fields} headers provide clear schema for models
Tabular arrays - Uniform arrays of objects collapse into tables that declare fields once and stream row values line by line
Human-readable - More readable than JSON-compact, though less readable than pretty-printed JSON

Note: The main index file (context_main.json) is always in JSON format, even when using --format toon. Only the folder bundle files use the .toon extension.

Format Example

TOON format uses indentation for nested objects and tabular arrays for uniform data:

graph:
  nodes[2]:
    -
      entryId: App.tsx
      contract:
        type: UIFContract
        kind: react:component
    -
      entryId: Card.tsx
      contract:
        type: UIFContract
        kind: react:component

For uniform arrays of objects, TOON uses CSV-style tabular format:

hikes[3]{id,name,distanceKm,elevationGain}:
  1,Blue Lake Trail,7.5,320
  2,Ridge Overlook,9.2,540
  3,Wildflower Loop,5.1,180

Usage

Generate context bundles in TOON format:

# Generate TOON format bundles
stamp context --format toon

# With style metadata (alternative syntax)
stamp context --format toon --include-style
stamp context style --format toon

# Custom output directory
stamp context --format toon --out ./output

Output Structure

When using --format toon, the output structure is:

output/
├── context_main.json          # Main index (always JSON)
├── context.toon              # Root folder bundles (TOON format)
├── src/
│   └── context.toon          # Bundles from src/ folder (TOON format)
└── src/components/
    └── context.toon          # Bundles from src/components/ (TOON format)

Each folder containing components gets its own context.toon file. The directory structure mirrors your project layout, just like JSON format.

Decoding TOON Files

TOON files can be decoded using the @toon-format/toon package:

import { decode } from '@toon-format/toon';
import { readFile } from 'fs/promises';

// Read and decode a TOON file
const toonContent = await readFile('src/components/context.toon', 'utf-8');
const bundles = decode(toonContent);

// bundles is an array of LogicStampBundle objects
console.log(bundles[0].entryId);
console.log(bundles[0].graph.nodes);

The decoded structure is identical to JSON format bundles - same schema, same contracts, same dependency graphs.

When to Use TOON

Use TOON format when:

You need smaller file sizes (especially for large codebases) - ~40% token savings vs JSON
You're building tools that process context files programmatically
You want efficient storage for CI/CD artifacts
You're working with AI systems that can decode TOON natively
Your data contains uniform arrays of objects (TOON's sweet spot)

Use JSON format when:

You need human-readable output for debugging
You're manually inspecting context files
You want to use stamp context validate, compare, or clean commands
You're using tools that expect JSON
You want to diff context files in git
Your data is deeply nested or non-uniform (JSON-compact may use fewer tokens)

When NOT to use TOON:

Deeply nested structures - For data with minimal tabular eligibility (~0%), JSON-compact often uses fewer tokens
Semi-uniform arrays - When only ~40-60% of arrays are tabular, token savings diminish
Latency-critical applications - Some deployments may process compact JSON faster despite TOON's lower token count (measure TTFT and tokens/sec for your setup)

Format Comparison

All formats contain the same bundle data, just encoded differently:

Format	Extension	Human-readable	Token Efficiency	Use Case
`json`	`.json`	✅ Yes	Baseline	Default, human-readable
`pretty`	`.json`	✅ Yes	Larger	Human inspection, debugging
`ndjson`	`.json`	✅ Yes	Similar to json	Streaming, line-by-line processing
`toon`	`.toon`	✅ Yes (less readable)	~40% fewer tokens	Compact storage, AI workflows

Token Efficiency: TOON achieves approximately 40% token savings compared to standard JSON (2-space indentation) while maintaining similar or better LLM accuracy. Actual savings vary by data structure - uniform arrays of objects see the greatest benefit.

Integration with Other Commands

TOON format works with all stamp context options:

# TOON with style metadata (both syntaxes work)
stamp context --format toon --include-style
stamp context style --format toon

# TOON with custom depth
stamp context --format toon --depth 2

# TOON with profile
stamp context --format toon --profile llm-safe

# TOON with code inclusion mode
stamp context --format toon --include-code header

File Naming

Folder bundles: context.toon (one per folder)
Main index: context_main.json (always JSON, even with --format toon)
Variants: context_*.toon (if any variants are generated)

.gitignore Patterns

When using stamp init, the following patterns are automatically added to .gitignore:

context.toon - Per-folder TOON bundles
context_*.toon - TOON format variants

These patterns prevent generated TOON files from being committed to version control. See init.md for details.

Working with TOON Files

Note: Currently, TOON format is only supported for generation (stamp context --format toon). The validate, compare, and clean commands do not yet support TOON format files.

To work with TOON files using these commands, you'll need to decode them to JSON first:

import { decode } from '@toon-format/toon';
import { readFile, writeFile } from 'fs/promises';

// Decode TOON to JSON for validation/comparison
const toonContent = await readFile('src/components/context.toon', 'utf-8');
const bundles = decode(toonContent);
const jsonContent = JSON.stringify(bundles, null, 2);
await writeFile('src/components/context.json', jsonContent);

// Now you can validate or compare the JSON file
// stamp context validate src/components/context.json

Limitations

Human readability - TOON files are human-readable but less readable than JSON. While you can inspect TOON files directly, JSON format is better for manual inspection and debugging.
Tool compatibility - Some tools may not support TOON format. JSON is more universally supported.
Index file - The main index (context_main.json) is always JSON, even when using TOON format for bundles.
Command support - Currently, only the stamp context command supports TOON format for generation. The validate, compare, and clean commands do not yet support TOON files and will only work with JSON format.
Deeply nested data - For structures with minimal tabular arrays, JSON-compact may be more token-efficient than TOON.

TOON Format Specification

LogicStamp uses the official @toon-format/toon package (v1.0.0) for encoding and decoding TOON files. This ensures compatibility with the TOON format specification.

For complete TOON format documentation, syntax reference, and benchmarks, see:

Official TOON Format: https://toon-format.dev
TOON Specification: https://github.com/toon-format/toon/blob/main/SPEC.md
TOON Package: https://www.npmjs.com/package/@toon-format/toon

Media Type: TOON files use the .toon extension and the text/toon media type (UTF-8 encoded).

Examples

Generate and inspect TOON bundles

# Generate TOON format
stamp context --format toon --out ./output

# Decode and inspect (Node.js)
node -e "
  const { decode } = require('@toon-format/toon');
  const fs = require('fs');
  const content = fs.readFileSync('./output/src/components/context.toon', 'utf-8');
  const bundles = decode(content);
  console.log('Bundles:', bundles.length);
  console.log('First bundle:', bundles[0].entryId);
"

Programmatic usage

import { decode } from '@toon-format/toon';
import { readFile } from 'fs/promises';
import { join } from 'path';

async function loadContextBundles(projectRoot) {
  const toonPath = join(projectRoot, 'src/components/context.toon');
  const content = await readFile(toonPath, 'utf-8');
  const bundles = decode(content);
  
  return bundles;
}

// Use the bundles
const bundles = await loadContextBundles('./my-project');
bundles.forEach(bundle => {
  console.log(`Component: ${bundle.entryId}`);
  console.log(`Nodes: ${bundle.graph.nodes.length}`);
});

Related Commands

stamp context - Generate context files (supports all formats including TOON)
stamp context validate - Validate context files (JSON format only)
stamp context compare - Compare context files (JSON format only)
stamp context clean - Clean context files (JSON format only)
stamp init - Set up .gitignore patterns for TOON files

Tips

Use JSON format during development for easier debugging and to use validation/comparison tools
Switch to TOON format for production/CI to save storage (when you don't need validation/comparison)
The main index is always JSON, so you can always inspect the project structure
TOON files decode to the same structure as JSON - no data loss
To validate or compare TOON files, decode them to JSON first using the @toon-format/toon package

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

TOON Format Support

Overview

Key Features

Format Example

Usage

Output Structure

Decoding TOON Files

When to Use TOON

Format Comparison

Integration with Other Commands

File Naming

.gitignore Patterns

Working with TOON Files

Limitations

TOON Format Specification

Examples

Generate and inspect TOON bundles

Programmatic usage

Related Commands

Tips

FilesExpand file tree

toon.md

Latest commit

History

toon.md

File metadata and controls

TOON Format Support

Overview

Key Features

Format Example

Usage

Output Structure

Decoding TOON Files

When to Use TOON

Format Comparison

Integration with Other Commands

File Naming

.gitignore Patterns

Working with TOON Files

Limitations

TOON Format Specification

Examples

Generate and inspect TOON bundles

Programmatic usage

Related Commands

Tips