sarish-json-extract

A lightweight utility to extract JSON from text — especially useful for parsing LLM responses that embed JSON inside prose or code fences.

Features

• Extract JSON from plain strings
• Extract JSON from ```json and ``` code fences
• Extract JSON embedded in prose (e.g. LLM responses like "Sure! Here's the data: {"name": "Sarish"} hope that helps")
• Handles nested objects and arrays
• Optional schema validation (type-check extracted fields)
• Zero dependencies — pure Python standard library
• Type hints with py.typed marker (PEP 561)

Installation

pip install sarish-json-extract

Quick Start

from sarish_json_extract import extract_json

# Plain JSON
result = extract_json('{"name": "Sarish", "age": 30}')
print(result)  # {'name': 'Sarish', 'age': 30}

# JSON in code fences
result = extract_json('''Here is the data:
```json
{"city": "Bangalore"}

''') print(result) # {'city': 'Bangalore'}

JSON embedded in LLM prose

result = extract_json('Sure! The answer is {"value": 42} as shown below.') print(result) # {'value': 42}

JSON arrays

result = extract_json('Items: [1, 2, 3] done.') print(result) # [1, 2, 3]

## Schema Validation

Pass an optional `schema` dict to validate that extracted JSON has the expected keys with the correct types:

```python
from sarish_json_extract import extract_json

text = 'The user data is: {"name": "RV", "age": 30, "active": true}'

result = extract_json(text, schema={"name": str, "age": int})
print(result)  # {'name': 'RV', 'age': 30, 'active': True}

# Returns None if schema doesn't match
result = extract_json('{"name": 123}', schema={"name": str})
print(result)  # None

API Reference

extract_json(text, schema=None)

Returns: dict | list | None — parsed JSON, or None if no valid JSON is found (or schema validation fails).

validate_schema(data, schema)

Returns: bool — True if all keys exist with correct types, False otherwise.

License

MIT