json_repair

The json_repair library is designed to handle several types of malformed JSON input:

• Syntax Errors: Missing quotes, comma errors, unescaped characters, incomplete key-value pairs, and incorrect boolean/null values (e.g., fixing true/false/null).
• Broken Arrays and Objects: Completing unfinished arrays or objects by adding necessary elements (commas, brackets) or default values (null, "").
• Noise Removal: Cleaning strings that contain extra non-JSON characters like comments or miscellaneous text while preserving the structure.
• Auto-completion: Automatically providing reasonable default values (like empty strings or null) for missing fields to ensure the JSON remains parsable.

Install the json-repair package using pip to repair malformed JSON from LLMs, APIs, logs, and user input in Python. Requires Python 3.10 or higher.

pip install json-repair

To improve performance when you are 100% certain the input is invalid JSON, use the following flags:

• skip_json_loads=True: Skips the initial strict validation using the standard library, going directly to the repair parser.
• return_objects=True: Returns the parsed object directly, which is faster than returning a string.

Warning: Do not use skip_json_loads=True for valid JSON, as the repair parser might alter the structure or values.

from json_repair import repair_json

obj = repair_json(bad_json_string, return_objects=True, skip_json_loads=True)

Install the package using pip:

pip install json-repair

To use Schema-guided repair features, install the optional dependency:

pip install 'json-repair[schema]'

Use json_repair.loads() as a drop-in replacement for json.loads() to fix common JSON issues such as missing quotes, misplaced commas, unescaped characters, incomplete key-value pairs, and truncated values. It can also handle JSON containing comments or stray prose.

import json_repair

bad_json = '{"users":[{"name":"Ada","role":"admin",}],"ok":true'
decoded_object = json_repair.loads(bad_json)

# Result: {'users': [{'name': 'Ada', 'role': 'admin'}], 'ok': True}

You can use json_repair.loads() as a drop-in replacement for the standard library json.loads(). It automatically attempts to repair the JSON before parsing. This is the recommended way to handle potentially malformed JSON.

import json_repair

decoded_object = json_repair.loads(json_string)

By default, the library attempts to repair errors. If you want to enforce strict validation and raise a ValueError on structural issues (like duplicate keys, missing colons, or empty keys/values), use strict=True.

from json_repair import repair_json

repair_json(bad_json_string, strict=True)

json_repair.loads() is the recommended way to parse JSON. It first attempts a strict parse using the standard library json.loads(). If that fails, it automatically switches to the repair parser.

import json_repair

decoded_object = json_repair.loads(json_string)

You can guide repairs using a JSON Schema or a Pydantic v2 model. This allows the parser to fill missing values, coerce scalars, and drop disallowed properties.

Note: This requires the [schema] extra: pip install 'json-repair[schema]'. Schema guidance is mutually exclusive with strict=True.

Available schema_repair_mode values:

• standard (default): Basic schema guidance.
• salvage: Best-effort salvage of arrays/objects (e.g., dropping unrepairable array items or unwrapping single-item arrays).

from pydantic import BaseModel, Field
from json_repair import repair_json

class Payload(BaseModel):
    value: int
    tags: list[str] = Field(default_factory=list)

# Using a Pydantic model as a schema
repair_json(
    '{"value": "1", "tags": }',
    schema=Payload,
    skip_json_loads=True,
    return_objects=True,
)

To maintain stability when repairing JSON that is being streamed, pass stream_stable=True to repair_json() or loads().

stream_output = repair_json(stream_input, stream_stable=True)