feat: Security System for LLM

[thekishandev]

Issue(s)

Closes #53 - [Feat] : Security system for LLM

Acceptance Criteria fulfillment

• Implement security validation for AI requests with multiple security levels (STRICT, MODERATE, RELAXED)
• Add content filtering and pattern matching for inappropriate/harmful content
• Integrate security wrapper with existing AI providers (OpenAI, Gemini, Self-hosted)
• Add user preference configuration for security levels
• Implement response validation and filtering
• Add comprehensive logging for security events
• Add localized error messages for security violations

Proposed changes (including videos or screenshots)

Added

• New AISecurity class with three security levels
• Content validation patterns for illegal, inappropriate, and sensitive content
• Security level configuration in user preferences
• Response validation and filtering system
• Enhanced error handling and logging

Modified

• Updated AIHandler to use security wrapper
• Enhanced UserPreferenceModal with security level selection
• Updated translation files with security-related messages
• Improved API request formatting for all providers

Security Improvements

• Input validation before API requests
• Content filtering for harmful/inappropriate content
• Sensitive data detection (PII, credentials)
• Response validation and sanitization
• Security level-based restrictions
• Comprehensive security logging

Further comments

The security wrapper implementation provides three levels of security:

1. STRICT: Blocks most potentially harmful content and sensitive topics
2. MODERATE: Allows discussion of sensitive topics with appropriate context
3. RELAXED: Allows more open discussion while still blocking illegal content

The implementation includes:

• Pattern-based content validation
• Security level-specific restrictions
• Response filtering and sanitization
• Enhanced error handling with localized messages
• Comprehensive logging for security events

Testing can be performed using:

1. Configure security level using /quick config
2. Set AI Configuration and provider
3. Test with various inputs using /quick ai

Example test cases:

• Harmful content (should be blocked)
• Sensitive topics (handled based on security level)
• Normal queries (should pass through)

Screenshot

Demo Video

2025-04-21.00-50-40.mp4

[thekishandev]

@VipinDevelops Can please review the PR implementing #53 Security systems for LLM