LibreChatLab - Personal AI Chat Configuration

A curated collection of custom configurations, security enhancements, and documentation for running a personal LibreChat instance.

🎯 Purpose

This repository contains only the custom modifications for a personal LibreChat deployment, without the full LibreChat codebase. Use these files to enhance your own LibreChat installation.

📁 Contents

Core Configuration

• librechat.yaml - Optimized multimodal configuration for RTX 4080 16GB VRAM
• rag.yml - Vector database and RAG services setup
• docker-compose.yml - Custom Docker configuration

Security Enhancements

• api/middleware/rateLimiter.js - Comprehensive rate limiting
• api/utils/pathValidator.js - Path traversal protection
• Security patches for CodeQL vulnerabilities

Documentation & Guides

• CLAUDE.md - AI assistant guidance for the codebase
• VECTOR_SETUP.md - Complete vector database setup guide
• SECURITY_FIXES.md - Security vulnerability analysis and fixes
• mock_ocr.py - Mock OCR service for testing

🚀 Quick Start

1. Install LibreChat following the official documentation

2. Apply these configurations:

   # Copy configuration files
   cp librechat.yaml /path/to/your/librechat/
   cp rag.yml /path/to/your/librechat/
   
   # Apply security middleware
   cp -r api/middleware/ /path/to/your/librechat/api/
   cp -r api/utils/ /path/to/your/librechat/api/

3. Configure your environment:

   • Update .env with your API keys
   • Modify librechat.yaml with your specific model endpoints
   • Set up vector database using VECTOR_SETUP.md

🔧 Key Features

AI Model Configuration

• Multiple Ollama endpoints optimized for different use cases
• Multimodal support with various parameter sizes
• Memory-optimized for RTX 4080 16GB VRAM

Security Hardening

• Rate limiting on all API endpoints
• Path traversal protection for file operations
• ReDoS vulnerability fixes in regex patterns
• Input validation and sanitization

Vector Database & RAG

• PostgreSQL + pgvector for semantic search
• File processing with multiple format support
• Memory management for conversation context

📊 Model Configurations

Multimodal Models (Vision + Text)

• qwen2.5vl:72b - Flagship multimodal (like GPT-4V)
• qwen2.5vl:7b - Balanced multimodal (like Claude 3)
• qwen2.5vl:3b - Fast multimodal (like Gemini Flash)

Specialized Models

• qwen2.5-coder:32b - Advanced coding (like GitHub Copilot)
• gemma3:27b - Google's flagship model
• qwen2.5:3b - Ultra-fast text processing

🔒 Security Features

• ✅ Rate limiting (auth: 5/15min, API: 100/15min, files: 50/15min)
• ✅ Path validation prevents directory traversal attacks
• ✅ ReDoS protection with regex length limits
• ✅ Input sanitization on all endpoints
• ✅ No sensitive data exposed in repository

🛠️ Installation Notes

Prerequisites

• LibreChat base installation
• Docker and Docker Compose
• GPU with sufficient VRAM for your chosen models

Environment Variables

Set these in your .env file:

RAG_PORT=8000
RAG_API_URL=http://localhost:8000
POSTGRES_DB=mydatabase
POSTGRES_USER=myuser
POSTGRES_PASSWORD=mypassword

📚 Documentation

• CLAUDE.md - Complete project guidance for AI assistants
• VECTOR_SETUP.md - Step-by-step RAG setup instructions
• SECURITY_FIXES.md - Detailed security analysis and fixes

🤝 Contributing

This is a personal configuration repository. Feel free to:

• Fork for your own modifications
• Submit issues for configuration questions
• Share improvements via pull requests

📄 License

These configurations are provided as-is for educational and personal use. The underlying LibreChat project has its own license terms.

🔗 Related Projects

• LibreChat: github.com/danny-avila/LibreChat
• Official Docs: docs.librechat.ai

---

Note: This repository contains only configuration files and enhancements. You need a separate LibreChat installation to use these configurations.