mirror of
https://github.com/voson-wang/toon.git
synced 2026-01-29 23:34:10 +08:00
docs: add dedicated docs website
This commit is contained in:
Johann Schopplich
579
docs/guide/benchmarks.md
Normal file
579
docs/guide/benchmarks.md
Normal file
@@ -0,0 +1,579 @@
|
||||
# Benchmarks
|
||||
|
||||
The benchmarks on this page measure TOON's performance across two key dimensions:
|
||||
|
||||
- **Retrieval Accuracy**: How well LLMs understand and extract information from different input formats.
|
||||
- **Token Efficiency**: How many tokens each format requires to represent the same data.
|
||||
|
||||
Benchmarks are organized into two tracks to ensure fair comparisons:
|
||||
|
||||
- **Mixed-Structure Track**: Datasets with nested or semi-uniform structures (TOON vs JSON, YAML, XML). CSV excluded as it cannot properly represent these structures.
|
||||
- **Flat-Only Track**: Datasets with flat tabular structures where CSV is applicable (CSV vs TOON vs JSON, YAML, XML).
|
||||
|
||||
## Retrieval Accuracy
|
||||
|
||||
<!-- automd:file src="../../benchmarks/results/retrieval-accuracy.md" -->
|
||||
|
||||
Benchmarks test LLM comprehension across different input formats using 209 data retrieval questions on 4 models.
|
||||
|
||||
<details>
|
||||
<summary><strong>Show Dataset Catalog</strong></summary>
|
||||
|
||||
#### Dataset Catalog
|
||||
|
||||
| Dataset | Rows | Structure | CSV Support | Eligibility |
|
||||
| ------- | ---- | --------- | ----------- | ----------- |
|
||||
| Uniform employee records | 100 | uniform | ✓ | 100% |
|
||||
| E-commerce orders with nested structures | 50 | nested | ✗ | 33% |
|
||||
| Time-series analytics data | 60 | uniform | ✓ | 100% |
|
||||
| Top 100 GitHub repositories | 100 | uniform | ✓ | 100% |
|
||||
| Semi-uniform event logs | 75 | semi-uniform | ✗ | 50% |
|
||||
| Deeply nested configuration | 11 | deep | ✗ | 0% |
|
||||
| Valid complete dataset (control) | 20 | uniform | ✓ | 100% |
|
||||
| Array truncated: 3 rows removed from end | 17 | uniform | ✓ | 100% |
|
||||
| Extra rows added beyond declared length | 23 | uniform | ✓ | 100% |
|
||||
| Inconsistent field count (missing salary in row 10) | 20 | uniform | ✓ | 100% |
|
||||
| Missing required fields (no email in multiple rows) | 20 | uniform | ✓ | 100% |
|
||||
|
||||
**Structure classes:**
|
||||
- **uniform**: All objects have identical fields with primitive values
|
||||
- **semi-uniform**: Mix of uniform and non-uniform structures
|
||||
- **nested**: Objects with nested structures (nested objects or arrays)
|
||||
- **deep**: Highly nested with minimal tabular eligibility
|
||||
|
||||
**CSV Support:** ✓ (supported), ✗ (not supported – would require lossy flattening)
|
||||
|
||||
**Eligibility:** Percentage of arrays that qualify for TOON's tabular format (uniform objects with primitive values)
|
||||
|
||||
</details>
|
||||
|
||||
#### Efficiency Ranking (Accuracy per 1K Tokens)
|
||||
|
||||
Each format's overall performance, balancing accuracy against token cost:
|
||||
|
||||
```
|
||||
TOON ████████████████████ 26.9 │ 73.9% acc │ 2,744 tokens
|
||||
JSON compact █████████████████░░░ 22.9 │ 70.7% acc │ 3,081 tokens
|
||||
YAML ██████████████░░░░░░ 18.6 │ 69.0% acc │ 3,719 tokens
|
||||
JSON ███████████░░░░░░░░░ 15.3 │ 69.7% acc │ 4,545 tokens
|
||||
XML ██████████░░░░░░░░░░ 13.0 │ 67.1% acc │ 5,167 tokens
|
||||
```
|
||||
|
||||
TOON achieves **73.9%** accuracy (vs JSON's 69.7%) while using **39.6% fewer tokens**.
|
||||
|
||||
**Note on CSV:** Excluded from ranking as it only supports 109 of 209 questions (flat tabular data only). While CSV is highly token-efficient for simple tabular data, it cannot represent nested structures that other formats handle.
|
||||
|
||||
#### Per-Model Accuracy
|
||||
|
||||
Accuracy across 4 LLMs on 209 data retrieval questions:
|
||||
|
||||
```
|
||||
claude-haiku-4-5-20251001
|
||||
→ TOON ████████████░░░░░░░░ 59.8% (125/209)
|
||||
JSON ███████████░░░░░░░░░ 57.4% (120/209)
|
||||
YAML ███████████░░░░░░░░░ 56.0% (117/209)
|
||||
XML ███████████░░░░░░░░░ 55.5% (116/209)
|
||||
JSON compact ███████████░░░░░░░░░ 55.0% (115/209)
|
||||
CSV ██████████░░░░░░░░░░ 50.5% (55/109)
|
||||
|
||||
gemini-2.5-flash
|
||||
→ TOON ██████████████████░░ 87.6% (183/209)
|
||||
CSV █████████████████░░░ 86.2% (94/109)
|
||||
JSON compact ████████████████░░░░ 82.3% (172/209)
|
||||
YAML ████████████████░░░░ 79.4% (166/209)
|
||||
XML ████████████████░░░░ 79.4% (166/209)
|
||||
JSON ███████████████░░░░░ 77.0% (161/209)
|
||||
|
||||
gpt-5-nano
|
||||
→ TOON ██████████████████░░ 90.9% (190/209)
|
||||
JSON compact ██████████████████░░ 90.9% (190/209)
|
||||
JSON ██████████████████░░ 89.0% (186/209)
|
||||
CSV ██████████████████░░ 89.0% (97/109)
|
||||
YAML █████████████████░░░ 87.1% (182/209)
|
||||
XML ████████████████░░░░ 80.9% (169/209)
|
||||
|
||||
grok-4-fast-non-reasoning
|
||||
→ TOON ███████████░░░░░░░░░ 57.4% (120/209)
|
||||
JSON ███████████░░░░░░░░░ 55.5% (116/209)
|
||||
JSON compact ███████████░░░░░░░░░ 54.5% (114/209)
|
||||
YAML ███████████░░░░░░░░░ 53.6% (112/209)
|
||||
XML ███████████░░░░░░░░░ 52.6% (110/209)
|
||||
CSV ██████████░░░░░░░░░░ 52.3% (57/109)
|
||||
```
|
||||
|
||||
**Key tradeoff:** TOON achieves **73.9% accuracy** (vs JSON's 69.7%) while using **39.6% fewer tokens** on these datasets.
|
||||
|
||||
<details>
|
||||
<summary><strong>Performance by dataset, model, and question type</strong></summary>
|
||||
|
||||
#### Performance by Question Type
|
||||
|
||||
| Question Type | TOON | JSON compact | JSON | CSV | YAML | XML |
|
||||
| ------------- | ---- | ---- | ---- | ---- | ---- | ---- |
|
||||
| Field Retrieval | 99.6% | 99.3% | 99.3% | 100.0% | 98.2% | 98.9% |
|
||||
| Aggregation | 54.4% | 47.2% | 48.8% | 44.0% | 47.6% | 41.3% |
|
||||
| Filtering | 56.3% | 57.3% | 50.5% | 49.1% | 51.0% | 47.9% |
|
||||
| Structure Awareness | 88.0% | 83.0% | 83.0% | 85.9% | 80.0% | 80.0% |
|
||||
| Structural Validation | 70.0% | 45.0% | 50.0% | 80.0% | 60.0% | 80.0% |
|
||||
|
||||
#### Performance by Dataset
|
||||
|
||||
##### Uniform employee records
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `csv` | 72.0% | 2,352 | 118/164 |
|
||||
| `toon` | 73.8% | 2,518 | 121/164 |
|
||||
| `json-compact` | 69.5% | 3,953 | 114/164 |
|
||||
| `yaml` | 68.3% | 4,982 | 112/164 |
|
||||
| `json-pretty` | 68.3% | 6,360 | 112/164 |
|
||||
| `xml` | 69.5% | 7,324 | 114/164 |
|
||||
|
||||
##### E-commerce orders with nested structures
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `toon` | 81.1% | 7,232 | 133/164 |
|
||||
| `json-compact` | 76.8% | 6,794 | 126/164 |
|
||||
| `yaml` | 75.6% | 8,347 | 124/164 |
|
||||
| `json-pretty` | 76.2% | 10,713 | 125/164 |
|
||||
| `xml` | 74.4% | 12,023 | 122/164 |
|
||||
|
||||
##### Time-series analytics data
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `csv` | 73.3% | 1,406 | 88/120 |
|
||||
| `toon` | 72.5% | 1,548 | 87/120 |
|
||||
| `json-compact` | 71.7% | 2,349 | 86/120 |
|
||||
| `yaml` | 71.7% | 2,949 | 86/120 |
|
||||
| `json-pretty` | 68.3% | 3,676 | 82/120 |
|
||||
| `xml` | 68.3% | 4,384 | 82/120 |
|
||||
|
||||
##### Top 100 GitHub repositories
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `toon` | 62.9% | 8,779 | 83/132 |
|
||||
| `csv` | 61.4% | 8,527 | 81/132 |
|
||||
| `yaml` | 59.8% | 13,141 | 79/132 |
|
||||
| `json-compact` | 55.3% | 11,464 | 73/132 |
|
||||
| `json-pretty` | 56.1% | 15,157 | 74/132 |
|
||||
| `xml` | 48.5% | 17,105 | 64/132 |
|
||||
|
||||
##### Semi-uniform event logs
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `json-compact` | 63.3% | 4,819 | 76/120 |
|
||||
| `toon` | 57.5% | 5,799 | 69/120 |
|
||||
| `json-pretty` | 59.2% | 6,797 | 71/120 |
|
||||
| `yaml` | 48.3% | 5,827 | 58/120 |
|
||||
| `xml` | 46.7% | 7,709 | 56/120 |
|
||||
|
||||
##### Deeply nested configuration
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `json-compact` | 92.2% | 574 | 107/116 |
|
||||
| `toon` | 95.7% | 666 | 111/116 |
|
||||
| `yaml` | 91.4% | 686 | 106/116 |
|
||||
| `json-pretty` | 94.0% | 932 | 109/116 |
|
||||
| `xml` | 92.2% | 1,018 | 107/116 |
|
||||
|
||||
##### Valid complete dataset (control)
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `toon` | 100.0% | 544 | 4/4 |
|
||||
| `json-compact` | 100.0% | 795 | 4/4 |
|
||||
| `yaml` | 100.0% | 1,003 | 4/4 |
|
||||
| `json-pretty` | 100.0% | 1,282 | 4/4 |
|
||||
| `csv` | 25.0% | 492 | 1/4 |
|
||||
| `xml` | 0.0% | 1,467 | 0/4 |
|
||||
|
||||
##### Array truncated: 3 rows removed from end
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `csv` | 100.0% | 425 | 4/4 |
|
||||
| `xml` | 100.0% | 1,251 | 4/4 |
|
||||
| `toon` | 0.0% | 474 | 0/4 |
|
||||
| `json-compact` | 0.0% | 681 | 0/4 |
|
||||
| `json-pretty` | 0.0% | 1,096 | 0/4 |
|
||||
| `yaml` | 0.0% | 859 | 0/4 |
|
||||
|
||||
##### Extra rows added beyond declared length
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `csv` | 100.0% | 566 | 4/4 |
|
||||
| `toon` | 75.0% | 621 | 3/4 |
|
||||
| `xml` | 100.0% | 1,692 | 4/4 |
|
||||
| `yaml` | 75.0% | 1,157 | 3/4 |
|
||||
| `json-compact` | 50.0% | 917 | 2/4 |
|
||||
| `json-pretty` | 50.0% | 1,476 | 2/4 |
|
||||
|
||||
##### Inconsistent field count (missing salary in row 10)
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `csv` | 75.0% | 489 | 3/4 |
|
||||
| `yaml` | 100.0% | 996 | 4/4 |
|
||||
| `toon` | 100.0% | 1,019 | 4/4 |
|
||||
| `json-compact` | 75.0% | 790 | 3/4 |
|
||||
| `xml` | 100.0% | 1,458 | 4/4 |
|
||||
| `json-pretty` | 75.0% | 1,274 | 3/4 |
|
||||
|
||||
##### Missing required fields (no email in multiple rows)
|
||||
|
||||
| Format | Accuracy | Tokens | Correct/Total |
|
||||
| ------ | -------- | ------ | ------------- |
|
||||
| `csv` | 100.0% | 329 | 4/4 |
|
||||
| `xml` | 100.0% | 1,411 | 4/4 |
|
||||
| `toon` | 75.0% | 983 | 3/4 |
|
||||
| `yaml` | 25.0% | 960 | 1/4 |
|
||||
| `json-pretty` | 25.0% | 1,230 | 1/4 |
|
||||
| `json-compact` | 0.0% | 755 | 0/4 |
|
||||
|
||||
#### Performance by Model
|
||||
|
||||
##### claude-haiku-4-5-20251001
|
||||
|
||||
| Format | Accuracy | Correct/Total |
|
||||
| ------ | -------- | ------------- |
|
||||
| `toon` | 59.8% | 125/209 |
|
||||
| `json-pretty` | 57.4% | 120/209 |
|
||||
| `yaml` | 56.0% | 117/209 |
|
||||
| `xml` | 55.5% | 116/209 |
|
||||
| `json-compact` | 55.0% | 115/209 |
|
||||
| `csv` | 50.5% | 55/109 |
|
||||
|
||||
##### gemini-2.5-flash
|
||||
|
||||
| Format | Accuracy | Correct/Total |
|
||||
| ------ | -------- | ------------- |
|
||||
| `toon` | 87.6% | 183/209 |
|
||||
| `csv` | 86.2% | 94/109 |
|
||||
| `json-compact` | 82.3% | 172/209 |
|
||||
| `yaml` | 79.4% | 166/209 |
|
||||
| `xml` | 79.4% | 166/209 |
|
||||
| `json-pretty` | 77.0% | 161/209 |
|
||||
|
||||
##### gpt-5-nano
|
||||
|
||||
| Format | Accuracy | Correct/Total |
|
||||
| ------ | -------- | ------------- |
|
||||
| `toon` | 90.9% | 190/209 |
|
||||
| `json-compact` | 90.9% | 190/209 |
|
||||
| `json-pretty` | 89.0% | 186/209 |
|
||||
| `csv` | 89.0% | 97/109 |
|
||||
| `yaml` | 87.1% | 182/209 |
|
||||
| `xml` | 80.9% | 169/209 |
|
||||
|
||||
##### grok-4-fast-non-reasoning
|
||||
|
||||
| Format | Accuracy | Correct/Total |
|
||||
| ------ | -------- | ------------- |
|
||||
| `toon` | 57.4% | 120/209 |
|
||||
| `json-pretty` | 55.5% | 116/209 |
|
||||
| `json-compact` | 54.5% | 114/209 |
|
||||
| `yaml` | 53.6% | 112/209 |
|
||||
| `xml` | 52.6% | 110/209 |
|
||||
| `csv` | 52.3% | 57/109 |
|
||||
|
||||
</details>
|
||||
|
||||
<details>
|
||||
<summary><strong>How the benchmark works</strong></summary>
|
||||
|
||||
#### What's Being Measured
|
||||
|
||||
This benchmark tests **LLM comprehension and data retrieval accuracy** across different input formats. Each LLM receives formatted data and must answer questions about it. This does **not** test the model's ability to generate TOON output – only to read and understand it.
|
||||
|
||||
#### Datasets Tested
|
||||
|
||||
Eleven datasets designed to test different structural patterns and validation capabilities:
|
||||
|
||||
**Primary datasets:**
|
||||
1. **Tabular** (100 employee records): Uniform objects with identical fields – optimal for TOON's tabular format.
|
||||
2. **Nested** (50 e-commerce orders): Complex structures with nested customer objects and item arrays.
|
||||
3. **Analytics** (60 days of metrics): Time-series data with dates and numeric values.
|
||||
4. **GitHub** (100 repositories): Real-world data from top GitHub repos by stars.
|
||||
5. **Event Logs** (75 logs): Semi-uniform data with ~50% flat logs and ~50% with nested error objects.
|
||||
6. **Nested Config** (1 configuration): Deeply nested configuration with minimal tabular eligibility.
|
||||
|
||||
**Structural validation datasets:**
|
||||
7. **Control**: Valid complete dataset (baseline for validation)
|
||||
8. **Truncated**: Array with 3 rows removed from end (tests [N] length detection)
|
||||
9. **Extra rows**: Array with 3 additional rows beyond declared length
|
||||
10. **Width mismatch**: Inconsistent field count (missing salary in row 10)
|
||||
11. **Missing fields**: Systematic field omissions (no email in multiple rows)
|
||||
|
||||
#### Question Types
|
||||
|
||||
209 questions are generated dynamically across five categories:
|
||||
|
||||
- **Field retrieval (33%)**: Direct value lookups or values that can be read straight off a record (including booleans and simple counts such as array lengths)
|
||||
- Example: "What is Alice's salary?" → `75000`
|
||||
- Example: "How many items are in order ORD-0042?" → `3`
|
||||
- Example: "What is the customer name for order ORD-0042?" → `John Doe`
|
||||
|
||||
- **Aggregation (30%)**: Dataset-level totals and averages plus single-condition filters (counts, sums, min/max comparisons)
|
||||
- Example: "How many employees work in Engineering?" → `17`
|
||||
- Example: "What is the total revenue across all orders?" → `45123.50`
|
||||
- Example: "How many employees have salary > 80000?" → `23`
|
||||
|
||||
- **Filtering (23%)**: Multi-condition queries requiring compound logic (AND constraints across fields)
|
||||
- Example: "How many employees in Sales have salary > 80000?" → `5`
|
||||
- Example: "How many active employees have more than 10 years of experience?" → `8`
|
||||
|
||||
- **Structure awareness (12%)**: Tests format-native structural affordances (TOON's [N] count and {fields}, CSV's header row)
|
||||
- Example: "How many employees are in the dataset?" → `100`
|
||||
- Example: "List the field names for employees" → `id, name, email, department, salary, yearsExperience, active`
|
||||
- Example: "What is the department of the last employee?" → `Sales`
|
||||
|
||||
- **Structural validation (2%)**: Tests ability to detect incomplete, truncated, or corrupted data using structural metadata
|
||||
- Example: "Is this data complete and valid?" → `YES` (control dataset) or `NO` (corrupted datasets)
|
||||
- Tests TOON's [N] length validation and {fields} consistency checking
|
||||
- Demonstrates CSV's lack of structural validation capabilities
|
||||
|
||||
#### Evaluation Process
|
||||
|
||||
1. **Format conversion**: Each dataset is converted to all 6 formats (TOON, JSON compact, JSON, CSV, YAML, XML).
|
||||
2. **Query LLM**: Each model receives formatted data + question in a prompt and extracts the answer.
|
||||
3. **Validate deterministically**: Answers are validated using type-aware comparison (e.g., `50000` = `$50,000`, `Engineering` = `engineering`, `2025-01-01` = `January 1, 2025`) without requiring an LLM judge.
|
||||
|
||||
#### Models & Configuration
|
||||
|
||||
- **Models tested**: `claude-haiku-4-5-20251001`, `gemini-2.5-flash`, `gpt-5-nano`, `grok-4-fast-non-reasoning`
|
||||
- **Token counting**: Using `gpt-tokenizer` with `o200k_base` encoding (GPT-5 tokenizer)
|
||||
- **Temperature**: Not set (models use their defaults)
|
||||
- **Total evaluations**: 209 questions × 6 formats × 4 models = 5,016 LLM calls
|
||||
|
||||
</details>
|
||||
|
||||
<!-- /automd -->
|
||||
|
||||
> [!NOTE]
|
||||
> **Key takeaway:** TOON achieves **73.9% accuracy** (vs JSON's 69.7%) while using **39.6% fewer tokens** on these datasets. The explicit structure (array lengths `[N]` and field lists `{fields}`) helps models track and validate data more reliably.
|
||||
|
||||
## Token Efficiency
|
||||
|
||||
Token counts are measured using the GPT-5 `o200k_base` tokenizer via [`gpt-tokenizer`](https://github.com/niieani/gpt-tokenizer). Savings are calculated against formatted JSON (2-space indentation) as the primary baseline, with additional comparisons to compact JSON (minified), YAML, and XML. Actual savings vary by model and tokenizer.
|
||||
|
||||
The benchmarks test datasets across different structural patterns (uniform, semi-uniform, nested, deeply nested) to show where TOON excels and where other formats may be better.
|
||||
|
||||
<!-- automd:file src="../../benchmarks/results/token-efficiency.md" -->
|
||||
|
||||
#### Mixed-Structure Track
|
||||
|
||||
Datasets with nested or semi-uniform structures. CSV excluded as it cannot properly represent these structures.
|
||||
|
||||
```
|
||||
🛒 E-commerce orders with nested structures ┊ Tabular: 33%
|
||||
│
|
||||
TOON █████████████░░░░░░░ 72,771 tokens
|
||||
├─ vs JSON (−33.1%) 108,806 tokens
|
||||
├─ vs JSON compact (+5.5%) 68,975 tokens
|
||||
├─ vs YAML (−14.2%) 84,780 tokens
|
||||
└─ vs XML (−40.5%) 122,406 tokens
|
||||
|
||||
🧾 Semi-uniform event logs ┊ Tabular: 50%
|
||||
│
|
||||
TOON █████████████████░░░ 153,211 tokens
|
||||
├─ vs JSON (−15.0%) 180,176 tokens
|
||||
├─ vs JSON compact (+19.9%) 127,731 tokens
|
||||
├─ vs YAML (−0.8%) 154,505 tokens
|
||||
└─ vs XML (−25.2%) 204,777 tokens
|
||||
|
||||
🧩 Deeply nested configuration ┊ Tabular: 0%
|
||||
│
|
||||
TOON ██████████████░░░░░░ 631 tokens
|
||||
├─ vs JSON (−31.3%) 919 tokens
|
||||
├─ vs JSON compact (+11.9%) 564 tokens
|
||||
├─ vs YAML (−6.2%) 673 tokens
|
||||
└─ vs XML (−37.4%) 1,008 tokens
|
||||
|
||||
──────────────────────────────────── Total ────────────────────────────────────
|
||||
TOON ████████████████░░░░ 226,613 tokens
|
||||
├─ vs JSON (−21.8%) 289,901 tokens
|
||||
├─ vs JSON compact (+14.9%) 197,270 tokens
|
||||
├─ vs YAML (−5.6%) 239,958 tokens
|
||||
└─ vs XML (−31.0%) 328,191 tokens
|
||||
```
|
||||
|
||||
#### Flat-Only Track
|
||||
|
||||
Datasets with flat tabular structures where CSV is applicable.
|
||||
|
||||
```
|
||||
👥 Uniform employee records ┊ Tabular: 100%
|
||||
│
|
||||
CSV ███████████████████░ 46,954 tokens
|
||||
TOON ████████████████████ 49,831 tokens (+6.1% vs CSV)
|
||||
├─ vs JSON (−60.7%) 126,860 tokens
|
||||
├─ vs JSON compact (−36.8%) 78,856 tokens
|
||||
├─ vs YAML (−50.0%) 99,706 tokens
|
||||
└─ vs XML (−66.0%) 146,444 tokens
|
||||
|
||||
📈 Time-series analytics data ┊ Tabular: 100%
|
||||
│
|
||||
CSV ██████████████████░░ 8,388 tokens
|
||||
TOON ████████████████████ 9,120 tokens (+8.7% vs CSV)
|
||||
├─ vs JSON (−59.0%) 22,250 tokens
|
||||
├─ vs JSON compact (−35.8%) 14,216 tokens
|
||||
├─ vs YAML (−48.9%) 17,863 tokens
|
||||
└─ vs XML (−65.7%) 26,621 tokens
|
||||
|
||||
⭐ Top 100 GitHub repositories ┊ Tabular: 100%
|
||||
│
|
||||
CSV ███████████████████░ 8,513 tokens
|
||||
TOON ████████████████████ 8,745 tokens (+2.7% vs CSV)
|
||||
├─ vs JSON (−42.3%) 15,145 tokens
|
||||
├─ vs JSON compact (−23.7%) 11,455 tokens
|
||||
├─ vs YAML (−33.4%) 13,129 tokens
|
||||
└─ vs XML (−48.8%) 17,095 tokens
|
||||
|
||||
──────────────────────────────────── Total ────────────────────────────────────
|
||||
CSV ███████████████████░ 63,855 tokens
|
||||
TOON ████████████████████ 67,696 tokens (+6.0% vs CSV)
|
||||
├─ vs JSON (−58.8%) 164,255 tokens
|
||||
├─ vs JSON compact (−35.2%) 104,527 tokens
|
||||
├─ vs YAML (−48.2%) 130,698 tokens
|
||||
└─ vs XML (−64.4%) 190,160 tokens
|
||||
```
|
||||
|
||||
<details>
|
||||
<summary><strong>Show detailed examples</strong></summary>
|
||||
|
||||
#### 📈 Time-series analytics data
|
||||
|
||||
**Savings:** 13,130 tokens (59.0% reduction vs JSON)
|
||||
|
||||
**JSON** (22,250 tokens):
|
||||
|
||||
```json
|
||||
{
|
||||
"metrics": [
|
||||
{
|
||||
"date": "2025-01-01",
|
||||
"views": 5715,
|
||||
"clicks": 211,
|
||||
"conversions": 28,
|
||||
"revenue": 7976.46,
|
||||
"bounceRate": 0.47
|
||||
},
|
||||
{
|
||||
"date": "2025-01-02",
|
||||
"views": 7103,
|
||||
"clicks": 393,
|
||||
"conversions": 28,
|
||||
"revenue": 8360.53,
|
||||
"bounceRate": 0.32
|
||||
},
|
||||
{
|
||||
"date": "2025-01-03",
|
||||
"views": 7248,
|
||||
"clicks": 378,
|
||||
"conversions": 24,
|
||||
"revenue": 3212.57,
|
||||
"bounceRate": 0.5
|
||||
},
|
||||
{
|
||||
"date": "2025-01-04",
|
||||
"views": 2927,
|
||||
"clicks": 77,
|
||||
"conversions": 11,
|
||||
"revenue": 1211.69,
|
||||
"bounceRate": 0.62
|
||||
},
|
||||
{
|
||||
"date": "2025-01-05",
|
||||
"views": 3530,
|
||||
"clicks": 82,
|
||||
"conversions": 8,
|
||||
"revenue": 462.77,
|
||||
"bounceRate": 0.56
|
||||
}
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
**TOON** (9,120 tokens):
|
||||
|
||||
```
|
||||
metrics[5]{date,views,clicks,conversions,revenue,bounceRate}:
|
||||
2025-01-01,5715,211,28,7976.46,0.47
|
||||
2025-01-02,7103,393,28,8360.53,0.32
|
||||
2025-01-03,7248,378,24,3212.57,0.5
|
||||
2025-01-04,2927,77,11,1211.69,0.62
|
||||
2025-01-05,3530,82,8,462.77,0.56
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
#### ⭐ Top 100 GitHub repositories
|
||||
|
||||
**Savings:** 6,400 tokens (42.3% reduction vs JSON)
|
||||
|
||||
**JSON** (15,145 tokens):
|
||||
|
||||
```json
|
||||
{
|
||||
"repositories": [
|
||||
{
|
||||
"id": 28457823,
|
||||
"name": "freeCodeCamp",
|
||||
"repo": "freeCodeCamp/freeCodeCamp",
|
||||
"description": "freeCodeCamp.org's open-source codebase and curriculum. Learn math, programming,…",
|
||||
"createdAt": "2014-12-24T17:49:19Z",
|
||||
"updatedAt": "2025-10-28T11:58:08Z",
|
||||
"pushedAt": "2025-10-28T10:17:16Z",
|
||||
"stars": 430886,
|
||||
"watchers": 8583,
|
||||
"forks": 42146,
|
||||
"defaultBranch": "main"
|
||||
},
|
||||
{
|
||||
"id": 132750724,
|
||||
"name": "build-your-own-x",
|
||||
"repo": "codecrafters-io/build-your-own-x",
|
||||
"description": "Master programming by recreating your favorite technologies from scratch.",
|
||||
"createdAt": "2018-05-09T12:03:18Z",
|
||||
"updatedAt": "2025-10-28T12:37:11Z",
|
||||
"pushedAt": "2025-10-10T18:45:01Z",
|
||||
"stars": 430877,
|
||||
"watchers": 6332,
|
||||
"forks": 40453,
|
||||
"defaultBranch": "master"
|
||||
},
|
||||
{
|
||||
"id": 21737465,
|
||||
"name": "awesome",
|
||||
"repo": "sindresorhus/awesome",
|
||||
"description": "😎 Awesome lists about all kinds of interesting topics",
|
||||
"createdAt": "2014-07-11T13:42:37Z",
|
||||
"updatedAt": "2025-10-28T12:40:21Z",
|
||||
"pushedAt": "2025-10-27T17:57:31Z",
|
||||
"stars": 410052,
|
||||
"watchers": 8017,
|
||||
"forks": 32029,
|
||||
"defaultBranch": "main"
|
||||
}
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
**TOON** (8,745 tokens):
|
||||
|
||||
```
|
||||
repositories[3]{id,name,repo,description,createdAt,updatedAt,pushedAt,stars,watchers,forks,defaultBranch}:
|
||||
28457823,freeCodeCamp,freeCodeCamp/freeCodeCamp,"freeCodeCamp.org's open-source codebase and curriculum. Learn math, programming,…","2014-12-24T17:49:19Z","2025-10-28T11:58:08Z","2025-10-28T10:17:16Z",430886,8583,42146,main
|
||||
132750724,build-your-own-x,codecrafters-io/build-your-own-x,Master programming by recreating your favorite technologies from scratch.,"2018-05-09T12:03:18Z","2025-10-28T12:37:11Z","2025-10-10T18:45:01Z",430877,6332,40453,master
|
||||
21737465,awesome,sindresorhus/awesome,😎 Awesome lists about all kinds of interesting topics,"2014-07-11T13:42:37Z","2025-10-28T12:40:21Z","2025-10-27T17:57:31Z",410052,8017,32029,main
|
||||
```
|
||||
|
||||
</details>
|
||||
|
||||
<!-- /automd -->
|
||||
299
docs/guide/format-overview.md
Normal file
299
docs/guide/format-overview.md
Normal file
@@ -0,0 +1,299 @@
|
||||
# Format Overview
|
||||
|
||||
TOON syntax reference with concrete examples. See [Getting Started](/guide/getting-started) for introduction.
|
||||
|
||||
## Data Model
|
||||
|
||||
TOON models data the same way as JSON:
|
||||
|
||||
- **Primitives**: strings, numbers, booleans, and `null`
|
||||
- **Objects**: mappings from string keys to values
|
||||
- **Arrays**: ordered sequences of values
|
||||
|
||||
### Root Forms
|
||||
|
||||
A TOON document can represent different root forms:
|
||||
|
||||
- **Root object** (most common): Fields appear at depth 0 with no parent key
|
||||
- **Root array**: Begins with `[N]:` or `[N]{fields}:` at depth 0
|
||||
- **Root primitive**: A single primitive value (string, number, boolean, or null)
|
||||
|
||||
Most examples in these docs use root objects, but the format supports all three forms equally ([spec §5](https://github.com/toon-format/spec/blob/main/SPEC.md#5-concrete-syntax-and-root-form)).
|
||||
|
||||
## Objects
|
||||
|
||||
### Simple Objects
|
||||
|
||||
Objects with primitive values use `key: value` syntax, with one field per line:
|
||||
|
||||
```yaml
|
||||
id: 123
|
||||
name: Ada
|
||||
active: true
|
||||
```
|
||||
|
||||
Indentation replaces braces. One space follows the colon.
|
||||
|
||||
### Nested Objects
|
||||
|
||||
Nested objects add one indentation level (default: 2 spaces):
|
||||
|
||||
```yaml
|
||||
user:
|
||||
id: 123
|
||||
name: Ada
|
||||
```
|
||||
|
||||
When a key ends with `:` and has no value on the same line, it opens a nested object. All lines at the next indentation level belong to that object.
|
||||
|
||||
### Empty Objects
|
||||
|
||||
An empty object at the root yields an empty document (no lines). A nested empty object is `key:` alone, with no children.
|
||||
|
||||
## Arrays
|
||||
|
||||
TOON detects array structure and chooses the most efficient representation. Arrays always declare their length in brackets: `[N]`.
|
||||
|
||||
### Primitive Arrays (Inline)
|
||||
|
||||
Arrays of primitives (strings, numbers, booleans, null) are rendered inline:
|
||||
|
||||
```yaml
|
||||
tags[3]: admin,ops,dev
|
||||
```
|
||||
|
||||
The delimiter (comma by default) separates values. Strings containing the active delimiter must be quoted.
|
||||
|
||||
### Arrays of Objects (Tabular)
|
||||
|
||||
When all objects in an array share the same set of primitive-valued keys, TOON uses tabular format:
|
||||
|
||||
::: code-group
|
||||
|
||||
```yaml [Basic Tabular]
|
||||
items[2]{sku,qty,price}:
|
||||
A1,2,9.99
|
||||
B2,1,14.5
|
||||
```
|
||||
|
||||
```yaml [With Spaces in Values]
|
||||
users[2]{id,name,role}:
|
||||
1,Alice Admin,admin
|
||||
2,"Bob Smith",user
|
||||
```
|
||||
|
||||
:::
|
||||
|
||||
The header `items[2]{sku,qty,price}:` declares:
|
||||
- **Array length**: `[2]` means 2 rows
|
||||
- **Field names**: `{sku,qty,price}` defines the columns
|
||||
- **Active delimiter**: comma (default)
|
||||
|
||||
Each row contains values in the same order as the field list. Values are encoded as primitives (strings, numbers, booleans, null) and separated by the delimiter.
|
||||
|
||||
> [!NOTE]
|
||||
> Tabular format requires identical field sets across all objects (same keys, order per object may vary) and primitive values only (no nested arrays/objects).
|
||||
|
||||
### Mixed and Non-Uniform Arrays
|
||||
|
||||
Arrays that don't meet the tabular requirements use list format with hyphen markers:
|
||||
|
||||
```yaml
|
||||
items[3]:
|
||||
- 1
|
||||
- a: 1
|
||||
- text
|
||||
```
|
||||
|
||||
Each element starts with `- ` at one indentation level deeper than the parent array header.
|
||||
|
||||
### Arrays of Arrays
|
||||
|
||||
When you have arrays containing primitive inner arrays:
|
||||
|
||||
```yaml
|
||||
pairs[2]:
|
||||
- [2]: 1,2
|
||||
- [2]: 3,4
|
||||
```
|
||||
|
||||
Each inner array gets its own header on the list-item line.
|
||||
|
||||
### Empty Arrays
|
||||
|
||||
Empty arrays have special representations:
|
||||
|
||||
```yaml
|
||||
items[0]:
|
||||
```
|
||||
|
||||
The header declares length zero, with no elements following.
|
||||
|
||||
## Array Headers
|
||||
|
||||
### Header Syntax
|
||||
|
||||
Array headers follow this pattern:
|
||||
|
||||
```
|
||||
key[N<delimiter?>]<{fields}>:
|
||||
```
|
||||
|
||||
Where:
|
||||
- **N** is the non-negative integer length
|
||||
- **delimiter** (optional) explicitly declares the active delimiter:
|
||||
- Absent → comma (`,`)
|
||||
- `\t` (tab character) → tab delimiter
|
||||
- `|` → pipe delimiter
|
||||
- **fields** (optional) for tabular arrays: `{field1,field2,field3}`
|
||||
|
||||
> [!TIP]
|
||||
> The array length `[N]` helps LLMs validate structure. If you ask a model to generate TOON output, explicit lengths let you detect truncation or malformed data.
|
||||
|
||||
### Delimiter Options
|
||||
|
||||
TOON supports three delimiters: comma (default), tab, and pipe. The delimiter is scoped to the array header that declares it.
|
||||
|
||||
::: code-group
|
||||
|
||||
```yaml [Comma (default)]
|
||||
items[2]{sku,name,qty,price}:
|
||||
A1,Widget,2,9.99
|
||||
B2,Gadget,1,14.5
|
||||
```
|
||||
|
||||
```yaml [Tab]
|
||||
items[2 ]{sku name qty price}:
|
||||
A1 Widget 2 9.99
|
||||
B2 Gadget 1 14.5
|
||||
```
|
||||
|
||||
```yaml [Pipe]
|
||||
items[2|]{sku|name|qty|price}:
|
||||
A1|Widget|2|9.99
|
||||
B2|Gadget|1|14.5
|
||||
```
|
||||
|
||||
:::
|
||||
|
||||
Tab and pipe delimiters are explicitly encoded in the header brackets and field braces. Commas don't require quoting when tab or pipe is active, and vice versa.
|
||||
|
||||
> [!TIP]
|
||||
> Tab delimiters often tokenize more efficiently than commas, especially for data with few quoted strings. Use `encode(data, { delimiter: '\t' })` for additional token savings.
|
||||
|
||||
## Key Folding (Optional)
|
||||
|
||||
Key folding is an optional encoder feature (since spec v1.5) that collapses chains of single-key objects into dotted paths, reducing tokens for deeply nested data.
|
||||
|
||||
### Basic Folding
|
||||
|
||||
Standard nesting:
|
||||
|
||||
```yaml
|
||||
data:
|
||||
metadata:
|
||||
items[2]: a,b
|
||||
```
|
||||
|
||||
With key folding (`keyFolding: 'safe'`):
|
||||
|
||||
```yaml
|
||||
data.metadata.items[2]: a,b
|
||||
```
|
||||
|
||||
The three nested objects collapse into a single dotted key `data.metadata.items`.
|
||||
|
||||
### When Folding Applies
|
||||
|
||||
A chain of objects is foldable when:
|
||||
- Each object in the chain has exactly one key (leading to the next object or a leaf value)
|
||||
- The leaf value is a primitive, array, or empty object
|
||||
- All segments are valid identifier segments (letters, digits, underscores only; no dots within segments)
|
||||
- The resulting folded key doesn't collide with existing keys
|
||||
|
||||
::: details Advanced Folding Rules
|
||||
**Segment Requirements (safe mode):**
|
||||
- All folded segments must match `^[A-Za-z_][A-Za-z0-9_]*$` (no dots, hyphens, or other special characters)
|
||||
- No segment may require quoting per §7.3 of the spec
|
||||
- The resulting folded key must not equal any existing sibling literal key at the same depth (collision avoidance)
|
||||
|
||||
**Depth Limit:**
|
||||
- The `flattenDepth` option (default: `Infinity`) controls how many segments to fold
|
||||
- `flattenDepth: 2` folds only two-segment chains: `{a: {b: val}}` → `a.b: val`
|
||||
- Values less than 2 have no practical effect
|
||||
|
||||
**Round-Trip with Path Expansion:**
|
||||
To reconstruct the original structure when decoding, use `expandPaths: 'safe'`. This splits dotted keys back into nested objects using the same safety rules ([spec §13.4](https://github.com/toon-format/spec/blob/main/SPEC.md#134-key-folding-and-path-expansion)).
|
||||
:::
|
||||
|
||||
### Round-Trip with Path Expansion
|
||||
|
||||
When decoding TOON that used key folding, enable path expansion to restore the nested structure:
|
||||
|
||||
```ts
|
||||
import { decode, encode } from '@toon-format/toon'
|
||||
|
||||
const original = { data: { metadata: { items: ['a', 'b'] } } }
|
||||
|
||||
// Encode with folding
|
||||
const toon = encode(original, { keyFolding: 'safe' })
|
||||
// → "data.metadata.items[2]: a,b"
|
||||
|
||||
// Decode with expansion
|
||||
const restored = decode(toon, { expandPaths: 'safe' })
|
||||
// → { data: { metadata: { items: ['a', 'b'] } } }
|
||||
```
|
||||
|
||||
Path expansion is off by default, so dotted keys are treated as literal keys unless explicitly enabled.
|
||||
|
||||
## Quoting and Types
|
||||
|
||||
### When Strings Need Quotes
|
||||
|
||||
TOON quotes strings **only when necessary** to maximize token efficiency. A string must be quoted if:
|
||||
|
||||
- It's empty (`""`)
|
||||
- It has leading or trailing whitespace
|
||||
- It equals `true`, `false`, or `null` (case-sensitive)
|
||||
- It looks like a number (e.g., `"42"`, `"-3.14"`, `"1e-6"`, or `"05"` with leading zeros)
|
||||
- It contains special characters: colon (`:`), quote (`"`), backslash (`\`), brackets, braces, or control characters (newline, tab, carriage return)
|
||||
- It contains the relevant delimiter (the active delimiter inside an array scope, or the document delimiter elsewhere)
|
||||
- It equals `"-"` or starts with `"-"` followed by any character
|
||||
|
||||
Otherwise, strings can be unquoted. Unicode, emoji, and strings with internal (non-leading/trailing) spaces are safe unquoted:
|
||||
|
||||
```yaml
|
||||
message: Hello 世界 👋
|
||||
note: This has inner spaces
|
||||
```
|
||||
|
||||
### Escape Sequences
|
||||
|
||||
In quoted strings and keys, only five escape sequences are valid:
|
||||
|
||||
| Character | Escape |
|
||||
|-----------|--------|
|
||||
| Backslash (`\`) | `\\` |
|
||||
| Double quote (`"`) | `\"` |
|
||||
| Newline (U+000A) | `\n` |
|
||||
| Carriage return (U+000D) | `\r` |
|
||||
| Tab (U+0009) | `\t` |
|
||||
|
||||
All other escape sequences (e.g., `\x`, `\u`) are invalid and will cause an error in strict mode.
|
||||
|
||||
### Type Conversions
|
||||
|
||||
Numbers are emitted in canonical decimal form (no exponent notation, no trailing zeros). Non-JSON types are normalized before encoding:
|
||||
|
||||
| Input | Output |
|
||||
|-------|--------|
|
||||
| Finite number | Canonical decimal (e.g., `1e6` → `1000000`, `1.5000` → `1.5`, `-0` → `0`) |
|
||||
| `NaN`, `Infinity`, `-Infinity` | `null` |
|
||||
| `BigInt` (within safe range) | Number |
|
||||
| `BigInt` (out of range) | Quoted decimal string (e.g., `"9007199254740993"`) |
|
||||
| `Date` | ISO string in quotes (e.g., `"2025-01-01T00:00:00.000Z"`) |
|
||||
| `undefined`, `function`, `symbol` | `null` |
|
||||
|
||||
Decoders accept both decimal and exponent forms on input (e.g., `42`, `-3.14`, `1e-6`), and treat tokens with forbidden leading zeros (e.g., `"05"`) as strings, not numbers.
|
||||
|
||||
For complete rules on quoting, escaping, type conversions, and strict-mode decoding, see [spec §2–4 (data model), §7 (strings and keys), and §14 (strict mode)](https://github.com/toon-format/spec/blob/main/SPEC.md).
|
||||
239
docs/guide/getting-started.md
Normal file
239
docs/guide/getting-started.md
Normal file
@@ -0,0 +1,239 @@
|
||||
# Getting Started
|
||||
|
||||
## What is TOON?
|
||||
|
||||
**Token-Oriented Object Notation** is a compact, human-readable encoding of the JSON data model for LLM prompts. It provides a lossless serialization of the same objects, arrays, and primitives as JSON, but in a syntax that minimizes tokens and makes structure easy for models to follow.
|
||||
|
||||
TOON combines YAML's indentation-based structure for nested objects with a CSV-style tabular layout for uniform arrays. TOON's sweet spot is uniform arrays of objects (multiple fields per row, same structure across items), achieving CSV-like compactness while adding explicit structure that helps LLMs parse and validate data reliably.
|
||||
|
||||
Think of it as a translation layer: use JSON programmatically, and encode it as TOON for LLM input.
|
||||
|
||||
### Why TOON?
|
||||
|
||||
Standard JSON is verbose and token-expensive. For uniform arrays of objects, JSON repeats every field name for every record:
|
||||
|
||||
```json
|
||||
{
|
||||
"users": [
|
||||
{ "id": 1, "name": "Alice", "role": "admin" },
|
||||
{ "id": 2, "name": "Bob", "role": "user" }
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
YAML already reduces some redundancy with indentation instead of braces:
|
||||
|
||||
```yaml
|
||||
users:
|
||||
- id: 1
|
||||
name: Alice
|
||||
role: admin
|
||||
- id: 2
|
||||
name: Bob
|
||||
role: user
|
||||
```
|
||||
|
||||
TOON goes further by declaring fields once and streaming data as rows:
|
||||
|
||||
```yaml
|
||||
users[2]{id,name,role}:
|
||||
1,Alice,admin
|
||||
2,Bob,user
|
||||
```
|
||||
|
||||
The `[2]` declares the array length, enabling LLMs to answer dataset size questions and detect truncation. The `{id,name,role}` declares the field names. Each row is then a compact, comma-separated list of values. This is the core pattern: declare structure once, stream data compactly. The format approaches CSV's efficiency while adding explicit structure.
|
||||
|
||||
For a more realistic example, here's how TOON handles a dataset with both nested objects and tabular arrays:
|
||||
|
||||
::: code-group
|
||||
|
||||
```json [JSON (235 tokens)]
|
||||
{
|
||||
"context": {
|
||||
"task": "Our favorite hikes together",
|
||||
"location": "Boulder",
|
||||
"season": "spring_2025"
|
||||
},
|
||||
"friends": ["ana", "luis", "sam"],
|
||||
"hikes": [
|
||||
{
|
||||
"id": 1,
|
||||
"name": "Blue Lake Trail",
|
||||
"distanceKm": 7.5,
|
||||
"elevationGain": 320,
|
||||
"companion": "ana",
|
||||
"wasSunny": true
|
||||
},
|
||||
{
|
||||
"id": 2,
|
||||
"name": "Ridge Overlook",
|
||||
"distanceKm": 9.2,
|
||||
"elevationGain": 540,
|
||||
"companion": "luis",
|
||||
"wasSunny": false
|
||||
},
|
||||
{
|
||||
"id": 3,
|
||||
"name": "Wildflower Loop",
|
||||
"distanceKm": 5.1,
|
||||
"elevationGain": 180,
|
||||
"companion": "sam",
|
||||
"wasSunny": true
|
||||
}
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
```yaml [TOON (106 tokens)]
|
||||
context:
|
||||
task: Our favorite hikes together
|
||||
location: Boulder
|
||||
season: spring_2025
|
||||
friends[3]: ana,luis,sam
|
||||
hikes[3]{id,name,distanceKm,elevationGain,companion,wasSunny}:
|
||||
1,Blue Lake Trail,7.5,320,ana,true
|
||||
2,Ridge Overlook,9.2,540,luis,false
|
||||
3,Wildflower Loop,5.1,180,sam,true
|
||||
```
|
||||
|
||||
:::
|
||||
|
||||
Notice how TOON combines YAML's indentation for the `context` object with inline format for the primitive `friends` array and tabular format for the structured `hikes` array. Each format is chosen automatically based on the data structure.
|
||||
|
||||
### Design Goals
|
||||
|
||||
TOON is optimized for specific use cases. It aims to:
|
||||
|
||||
- Make uniform arrays of objects as compact as possible by declaring structure once and streaming data.
|
||||
- Stay fully lossless and deterministic – round-trips preserve all data and structure.
|
||||
- Keep parsing simple and robust for both LLMs and humans through explicit structure markers.
|
||||
- Provide validation guardrails (array lengths, field counts) that help detect truncation and malformed output.
|
||||
|
||||
## When to Use TOON
|
||||
|
||||
TOON excels with uniform arrays of objects – data with the same structure across items. For LLM prompts, the format produces deterministic, minimally quoted text with built-in validation. Explicit array lengths (`[N]`) and field headers (`{fields}`) help detect truncation and malformed data, while the tabular structure declares fields once rather than repeating them in every row.
|
||||
|
||||
::: tip Production Ready
|
||||
TOON is production-ready and actively maintained, with implementations in TypeScript, Python, Go, Rust, .NET, and more. The format is stable, but also an idea in progress. Nothing's set in stone – help shape where it goes by contributing to the [specification](https://github.com/toon-format/spec) or sharing feedback.
|
||||
:::
|
||||
|
||||
## When Not to Use TOON
|
||||
|
||||
TOON is not always the best choice. Consider alternatives when:
|
||||
|
||||
- **Deeply nested or non-uniform structures** (tabular eligibility ≈ 0%): JSON-compact often uses fewer tokens. Example: complex configuration objects with many nested levels.
|
||||
- **Semi-uniform arrays** (~40–60% tabular eligibility): Token savings diminish. Prefer JSON if your pipelines already rely on it.
|
||||
- **Pure tabular data**: CSV is smaller than TOON for flat tables. TOON adds minimal overhead (~5-10%) to provide structure (array length declarations, field headers, delimiter scoping) that improves LLM reliability.
|
||||
- **Latency-critical applications**: Benchmark on your exact setup. Some deployments (especially local/quantized models) may process compact JSON faster despite TOON's lower token count.
|
||||
|
||||
> [!NOTE]
|
||||
> For data-driven comparisons across different structures, see [benchmarks](/guide/benchmarks). When optimizing for latency, measure TTFT, tokens/sec, and total time for both TOON and JSON-compact and use whichever performs better in your specific environment.
|
||||
|
||||
## Installation
|
||||
|
||||
### TypeScript Library
|
||||
|
||||
Install the library via your preferred package manager:
|
||||
|
||||
::: code-group
|
||||
|
||||
```bash [npm]
|
||||
npm install @toon-format/toon
|
||||
```
|
||||
|
||||
```bash [pnpm]
|
||||
pnpm add @toon-format/toon
|
||||
```
|
||||
|
||||
```bash [yarn]
|
||||
yarn add @toon-format/toon
|
||||
```
|
||||
|
||||
:::
|
||||
|
||||
### CLI
|
||||
|
||||
The CLI can be used without installation via `npx`, or installed globally:
|
||||
|
||||
::: code-group
|
||||
|
||||
```bash [npx (no install)]
|
||||
npx @toon-format/cli input.json -o output.toon
|
||||
```
|
||||
|
||||
```bash [npm]
|
||||
npm install -g @toon-format/cli
|
||||
```
|
||||
|
||||
```bash [pnpm]
|
||||
pnpm add -g @toon-format/cli
|
||||
```
|
||||
|
||||
```bash [yarn]
|
||||
yarn global add @toon-format/cli
|
||||
```
|
||||
|
||||
:::
|
||||
|
||||
For full CLI documentation, see the [CLI reference](/cli/).
|
||||
|
||||
## Your First Example
|
||||
|
||||
The examples below use the TypeScript library for demonstration, but the same operations work in any language with a TOON implementation.
|
||||
|
||||
Let's encode a simple dataset with the TypeScript library:
|
||||
|
||||
```ts
|
||||
import { encode } from '@toon-format/toon'
|
||||
|
||||
const data = {
|
||||
users: [
|
||||
{ id: 1, name: 'Alice', role: 'admin' },
|
||||
{ id: 2, name: 'Bob', role: 'user' }
|
||||
]
|
||||
}
|
||||
|
||||
console.log(encode(data))
|
||||
```
|
||||
|
||||
**Output:**
|
||||
|
||||
```yaml
|
||||
users[2]{id,name,role}:
|
||||
1,Alice,admin
|
||||
2,Bob,user
|
||||
```
|
||||
|
||||
### Decoding Back to JSON
|
||||
|
||||
Decoding is just as simple:
|
||||
|
||||
```ts
|
||||
import { decode } from '@toon-format/toon'
|
||||
|
||||
const toon = `
|
||||
users[2]{id,name,role}:
|
||||
1,Alice,admin
|
||||
2,Bob,user
|
||||
`
|
||||
|
||||
const data = decode(toon)
|
||||
console.log(JSON.stringify(data, null, 2))
|
||||
```
|
||||
|
||||
**Output:**
|
||||
|
||||
```json
|
||||
{
|
||||
"users": [
|
||||
{ "id": 1, "name": "Alice", "role": "admin" },
|
||||
{ "id": 2, "name": "Bob", "role": "user" }
|
||||
]
|
||||
}
|
||||
```
|
||||
|
||||
Round-tripping is lossless: `decode(encode(x))` always equals `x` (after normalization of non-JSON types like `Date`, `NaN`, etc.).
|
||||
|
||||
## Where to Go Next
|
||||
|
||||
Now that you've seen your first TOON document, read the [Format Overview](/guide/format-overview) for complete syntax details (objects, arrays, quoting rules, key folding), then explore [Using TOON with LLMs](/guide/llm-prompts) to see how to use it effectively in prompts. For implementation details, check the [API reference](/reference/api) (TypeScript) or the [specification](/reference/spec) (language-agnostic normative rules).
|
||||
139
docs/guide/llm-prompts.md
Normal file
139
docs/guide/llm-prompts.md
Normal file
@@ -0,0 +1,139 @@
|
||||
# Using TOON with LLMs
|
||||
|
||||
TOON is designed for passing structured data to Large Language Models with reduced token costs and improved reliability. This guide shows how to use TOON effectively in prompts, both for input (sending data to models) and output (getting models to generate TOON).
|
||||
|
||||
This guide is about the TOON format itself. Code examples use the TypeScript library for demonstration, but the same patterns and techniques apply regardless of which programming language you're using.
|
||||
|
||||
## Why TOON for LLMs
|
||||
|
||||
LLM tokens cost money, and JSON is verbose – repeating every field name for every record in an array. TOON minimizes tokens especially for uniform arrays by declaring fields once and streaming data as rows, typically saving 30-60% compared to formatted JSON.
|
||||
|
||||
TOON adds structure guardrails: explicit `[N]` lengths and `{fields}` headers make it easier for models to track rows and for you to validate output. Strict mode helps detect truncation and malformed TOON when decoding model responses.
|
||||
|
||||
## Sending TOON as Input
|
||||
|
||||
TOON works best when you show the format instead of describing it. The structure is self-documenting – models parse it naturally once they see the pattern.
|
||||
|
||||
Wrap your encoded data in a fenced code block (label it ` ```toon` for clarity):
|
||||
|
||||
````md
|
||||
Data is in TOON format (2-space indent, arrays show length and fields).
|
||||
|
||||
```toon
|
||||
users[3]{id,name,role,lastLogin}:
|
||||
1,Alice,admin,2025-01-15T10:30:00Z
|
||||
2,Bob,user,2025-01-14T15:22:00Z
|
||||
3,Charlie,user,2025-01-13T09:45:00Z
|
||||
```
|
||||
|
||||
Task: Summarize the user roles and their last activity.
|
||||
````
|
||||
|
||||
The indentation and headers are usually enough – models treat TOON like familiar YAML or CSV. The explicit array lengths (`[N]`) and field headers (`{fields}`) help the model track structure, especially for large tables.
|
||||
|
||||
> [!NOTE]
|
||||
> Most models don't have built-in TOON syntax highlighting, so ` ```toon` or ` ```yaml` both work fine. The structure is what matters.
|
||||
|
||||
## Generating TOON from LLMs
|
||||
|
||||
For output, be more explicit. When you want the model to **generate** TOON:
|
||||
|
||||
- **Show the expected header** (e.g., `users[N]{id,name,role}:`). The model fills rows instead of repeating keys, reducing generation errors.
|
||||
- **State the rules**: 2-space indent, no trailing spaces, `[N]` matches row count.
|
||||
|
||||
Here's a prompt that works for both reading and generating:
|
||||
|
||||
````md
|
||||
Data is in TOON format (2-space indent, arrays show length and fields).
|
||||
|
||||
```toon
|
||||
users[3]{id,name,role,lastLogin}:
|
||||
1,Alice,admin,2025-01-15T10:30:00Z
|
||||
2,Bob,user,2025-01-14T15:22:00Z
|
||||
3,Charlie,user,2025-01-13T09:45:00Z
|
||||
```
|
||||
|
||||
Task: Return only users with role "user" as TOON. Use the same header format. Set [N] to match the row count. Output only the code block.
|
||||
````
|
||||
|
||||
**Expected output:**
|
||||
|
||||
```toon
|
||||
users[2]{id,name,role,lastLogin}:
|
||||
2,Bob,user,2025-01-14T15:22:00Z
|
||||
3,Charlie,user,2025-01-13T09:45:00Z
|
||||
```
|
||||
|
||||
The model adjusts `[N]` to `2` and generates two rows.
|
||||
|
||||
### Validation with Strict Mode
|
||||
|
||||
When decoding model-generated TOON, use strict mode (default) to catch errors:
|
||||
|
||||
```ts
|
||||
import { decode } from '@toon-format/toon'
|
||||
|
||||
try {
|
||||
const data = decode(modelOutput, { strict: true })
|
||||
// Success – data is valid
|
||||
}
|
||||
catch (error) {
|
||||
// Model output was malformed (count mismatch, invalid escapes, etc.)
|
||||
console.error('Validation failed:', error.message)
|
||||
}
|
||||
```
|
||||
|
||||
Strict mode checks counts, indentation, and escaping so you can detect truncation or malformed TOON. For complete details, see the [API reference](/reference/api#decode).
|
||||
|
||||
## Delimiter Choices for Token Efficiency
|
||||
|
||||
Use `delimiter: '\t'` for tab-separated tables if you want even fewer tokens. Tabs are single characters, often tokenize more efficiently than commas, and rarely appear in natural text (reducing quote-escaping).
|
||||
|
||||
```ts
|
||||
const toon = encode(data, { delimiter: '\t' })
|
||||
```
|
||||
|
||||
Tell the model "fields are tab-separated" when using tabs. For more on delimiters, see the [Format Overview](/guide/format-overview#delimiter-options).
|
||||
|
||||
## Tips and Pitfalls
|
||||
|
||||
**Show, don't describe.** Don't explain TOON syntax in detail – just show an example. Models learn the pattern from context. A simple code block with 2-5 rows is more effective than paragraphs of explanation.
|
||||
|
||||
**Keep examples small.** Use 2-5 rows in your examples, not hundreds. The model generalizes from the pattern. Large examples waste tokens without improving accuracy.
|
||||
|
||||
**Always validate output.** Decode generated TOON with `strict: true` (default) to catch errors early. Don't assume model output is valid TOON without checking.
|
||||
|
||||
## Real-World Example
|
||||
|
||||
Here's a complete workflow: send data to a model and validate its TOON response.
|
||||
|
||||
**Prompt with TOON input:**
|
||||
|
||||
````md
|
||||
System logs in TOON format (tab-separated):
|
||||
|
||||
```toon
|
||||
events[4 ]{id level message timestamp}:
|
||||
1 error Connection timeout 2025-01-15T10:00:00Z
|
||||
2 warn Slow query 2025-01-15T10:05:00Z
|
||||
3 info User login 2025-01-15T10:10:00Z
|
||||
4 error Database error 2025-01-15T10:15:00Z
|
||||
```
|
||||
|
||||
Task: Return only error-level events as TOON. Use the same format.
|
||||
````
|
||||
|
||||
**Validate the response:**
|
||||
|
||||
```ts
|
||||
import { decode } from '@toon-format/toon'
|
||||
|
||||
const modelResponse = `
|
||||
events[2 ]{id level message timestamp}:
|
||||
1 error Connection timeout 2025-01-15T10:00:00Z
|
||||
4 error Database error 2025-01-15T10:15:00Z
|
||||
`
|
||||
|
||||
const filtered = decode(modelResponse, { strict: true })
|
||||
// ✓ Validated – model correctly filtered and adjusted [N] to 2
|
||||
```
|
||||
Reference in New Issue
Block a user