{
  "model": "gpt-4o",
  "version": 2024,
  "active": true,
  "temperature": 0.7,
  "system_prompt": null,
  "capabilities": [
    "text", "vision", "code"
  ],
  "limits": {
    "max_tokens": 128000,
    "rate_limit": 10000
  }
}

// Line 1 — one complete example
{"messages":[{"role":"system","content":"You are a helpful AI."},{"role":"user","content":"What is JSON?"},{"role":"assistant","content":"JSON is a lightweight data format."}]}

// Line 2 — next independent example
{"messages":[{"role":"user","content":"Explain JSONL"},{"role":"assistant","content":"JSONL is one JSON per line."}]}

// Line 3 — continues forever, one per line
{"messages":[{"role":"user","content":"What is Markdown?"},{"role":"assistant","content":"A plain text formatting language."}]}

import json

# ── Read JSONL ──────────────────
with open('data.jsonl', 'r') as f:
    for line in f:
        obj = json.loads(line.strip())
        print(obj)  # process each

# ── Write JSONL ──────────────────
records = [
    {"role": "user", "text": "Hello"},
    {"role": "assistant", "text": "Hi!"}
]
with open('out.jsonl', 'w') as f:
    for r in records:
        f.write(json.dumps(r) + '\n')

# Heading 1
## Heading 2
### Heading 3

Plain paragraph text here.

**bold text**  *italic text*
`inline code`

```python
# fenced code block
print("hello world")
```

- Unordered list item
- Another item
  - Nested item

1. Ordered list
2. Second item

[Link text](https://example.com)

> Blockquote text here

| Col 1 | Col 2 |  (GFM only)
|-------|-------|
| data  | data  |