[FEAT]: Introduce TOON Encoding for Token Optimization in AI Assistant

[Vatsal-Verma]

Problem

The AI Assistant in Template Playground currently injects raw JSON agreement data into prompts when assisting users with TemplateMark, Concerto models, or explanations.
For medium-to-large agreements, this leads to:

• High token consumption
• Increased latency and cost

Key Features:

• 📊 Token-Efficient & Accurate: TOON reaches 74% accuracy (vs JSON's 70%) while using ~40% fewer tokens.

TOON           ████████████████████   26.9 acc%/1K tok  │  73.9% acc  │  2,744 tokens
JSON compact   █████████████████░░░   22.9 acc%/1K tok  │  70.7% acc  │  3,081 tokens
YAML           ██████████████░░░░░░   18.6 acc%/1K tok  │  69.0% acc  │  3,719 tokens
JSON           ███████████░░░░░░░░░   15.3 acc%/1K tok  │  69.7% acc  │  4,545 tokens
XML            ██████████░░░░░░░░░░   13.0 acc%/1K tok  │  67.1% acc  │  5,167 tokens

how this works

1 → User Input
TemplateMark, Concerto, and JSON data from the Playground editors

2 → TOON Encoding
JSON agreement data is encoded into TOON for token optimization

3 → API Request
TOON-encoded context is sent to the AI via a standard JSON API call

4 → AI Processing
The model uses TOON only as internal context

5 → AI Response
Normal JSON response with generated code or suggestions

6 → User Output
Results shown in standard editor formats (no TOON exposure)

Sample format

JSON

{
  "user": {
    "id": 123,
    "name": "Alice Johnson",
    "email": "alice@example.com",
    "profile": {
      "age": 28,
      "city": "San Francisco",
      "interests": [
        "coding",
        "music",
        "travel"
      ]
    },
    "orders": [
      {
        "id": "ORD-001",
        "item": "Laptop",
        "price": 1299.99,
        "status": "delivered"
      },
      {
        "id": "ORD-002",
        "item": "Headphones",
        "price": 199.99,
        "status": "shipped"
      },
      {
        "id": "ORD-003",
        "item": "Mouse",
        "price": 49.99,
        "status": "processing"
      }
    ]
  }
}

TOON Format:

user:
  id: 123
  name: Alice Johnson
  email: alice@example.com
  profile:
    age: 28
    city: San Francisco
    interests[3]: coding,music,travel
  orders[3]{id,item,price,status}:
    ORD-001,Laptop,1299.99,delivered
    ORD-002,Headphones,199.99,shipped
    ORD-003,Mouse,49.99,processing

Reference

https://github.com/toon-format/toon

Result

• Using TOON internally optimizes AI Assistant performance while keeping the user experience and data formats unchanged.

Note: here we are not altering any data, we are just encoding the json --> TOON before feeding to API, and response will be in JSON. 

handling pr #446

[netlify]

✅ Deploy Preview for ap-template-playground ready!

Name	Link
🔨 Latest commit	a74ec77
🔍 Latest deploy log	https://app.netlify.com/projects/ap-template-playground/deploys/69d17829adab8d0008b19f05
😎 Deploy Preview	https://deploy-preview-456--ap-template-playground.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[Vatsal-Verma]

I tested this PR with and without TOON enabled. The results show a consistent reduction in token usage while producing equivalent responses. This confirms that TOON provides a positive impact on token efficiency for the current prompt sizes, with larger gains expected for more structured and repetitive inputs.

[github-actions]

This PR is stale because it has been open 15 days with no activity. Remove stale label or comment or this will be closed in 10 days.

[github-actions]

This PR is stale because it has been open 15 days with no activity. Remove stale label or comment or this will be closed in 10 days.