Description

• Uses strict=False in all json.loads() calls that parse LLM responses in mem0/memory/main.py.
• This allows control characters (newlines, tabs, etc.) inside JSON string values, which LLMs occasionally produce.
• Applied to both sync (Memory) and async (AsyncMemory) code paths, covering new_memories_with_actions and new_retrieved_facts parsing.

Fixes #2194

Type of change

• Bug fix (non-breaking change which fixes an issue)

Testing

Verified the fix handles the exact scenario from the issue:

import json

# Reproduces issue #2194 - literal newline in JSON string value
bad_json = '{"memory": [{"id": "0", "text": "app\n                         store", "event": "NONE"}]}'

# Before fix: raises JSONDecodeError: Invalid control character
json.loads(bad_json)  # FAILS

# After fix: parses successfully
json.loads(bad_json, strict=False)  # SUCCESS: {'memory': [{'id': '0', 'text': 'app\n                         store', 'event': 'NONE'}]}

• strict=False only relaxes control character handling — all other JSON validation remains unchanged.
• No behavior change for valid JSON responses (which is the common case).
• Fallback extract_json() paths also updated to ensure consistent handling.

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• My changes generate no new warnings
• New and existing unit tests pass locally with my changes

Maintainer Checklist

• closes When special characters such as line breaks exist in "new_memories_with_actions", json conversion fails #2194
• Made sure Checks passed