Agent Chat - Home
Enterprise LibreChat Replacement with Multi-Model Routing and LLM Platform Integration
Overview
Agent Chat is a production-grade enterprise chat platform that provides a LibreChat-compatible interface with deep LLM Platform ecosystem integration. It combines multi-model AI routing, vector search, agent orchestration, real-time collaboration, and comprehensive observability into a unified chat experience.
Core Function: Provides an enterprise-grade conversational AI interface with intelligent model routing, semantic search, continuous learning, and seamless integration with agent workflows. Replaces LibreChat with enhanced capabilities for production deployments.
Quick Start
# Install
npm install @bluefly/agent-chat
# Start development server
npm run dev
# Access services
# Chat UI: http://localhost:3080
# GraphQL: http://localhost:3080/graphql
# Health: http://localhost:3080/health
# MCP Server: stdio (for Claude Desktop)
Key Features
💬 Enterprise Chat Interface
- LibreChat-compatible API for seamless migration
- Modern React UI with dark mode and responsive design
- Real-time streaming responses via WebSocket
- Multi-turn conversations with context management
- Conversation history and advanced search
🤖 Multi-Model Routing
- Intelligent routing via LLM Gateway integration
- Support for Claude, GPT-4, Mistral, Gemini, Ollama
- Automatic model fallback and cost optimization
- Custom model preferences per user/organization
- A/B testing across models
🧠 Agent OS Integration
- 5-layer enterprise memory system (Redis, PostgreSQL, Qdrant)
- VORTEX v3 token optimization (30-50% reduction)
- Continuous learning from user feedback
- Pattern recognition and improvement generation
- Agent deployment and lifecycle management
🔍 Vector Search & RAG
- Semantic search across conversation history
- Retrieval Augmented Generation (RAG)
- Document ingestion and indexing via Qdrant
- Similarity-based recommendations
- Knowledge graph integration
👥 User & Team Management
- Role-based access control (RBAC)
- Team and organization support
- Usage tracking and quotas
- SSO integration (Drupal, LDAP)
- Multi-tenant architecture
🔌 MCP Integration
- MCP server for Claude Desktop integration
- Tool execution within chat sessions
- Context sharing across tools
- Custom tool registration
- Bidirectional communication
🛡️ OSSA 1.0 Compliance
- Standardized agent protocol implementation
- Agent capability discovery
- Health monitoring and metrics
- Distributed tracing support
📊 Observability
- Phoenix Arize AI-specific tracing
- Prometheus metrics export
- Grafana dashboards
- Structured logging with Winston
- Real-time performance monitoring
Wiki Navigation
Core Documentation
- Architecture - System design, WebSocket layer, state management
- API Reference - REST and WebSocket endpoints
- Integration Guide - Embedding chat in applications
- Development Guide - Local setup and contribution
Advanced Topics
- Agent OS Features - Enterprise memory and continuous learning
- LibreChat Migration - Compatibility and migration guide
- Vector Search - Semantic search and RAG implementation
- Multi-Model Routing - Intelligent model selection strategies
- MCP Protocol - Claude Desktop integration
Operations
- Deployment - Kubernetes, Docker, production setup
- Monitoring - Phoenix, Prometheus, Grafana configuration
- Security - Authentication, authorization, compliance
- Troubleshooting - Common issues and solutions
Current Focus (Active Features)
Enterprise Memory System
- 5-Layer Architecture: Conversation (Redis), User (PostgreSQL), Knowledge (Qdrant), Learning (PostgreSQL), Performance (Prometheus)
- Token Optimization: VORTEX v3 reduces token usage by 30-50%
- Continuous Learning: Automatic improvement from user feedback and corrections
- Priority: Production-ready
ROCKETSHIP Features (13 capabilities)
- Autonomous Agent Swarms
- Self-Healing System
- Federated Learning Coordinator
- Agent Marketplace
- Autonomous Research Agent
- AI Security Threat Detector
- Predictive Scaling
- Cross-Modal Understanding
- Autonomous Code Generation
- Edge Computing Coordinator
- Reinforcement Learning Orchestrator
- Quantum-Inspired Optimizer
- Collaborative Agent Network
LibreChat Enhanced Integration
- Kagent: Knowledge graph-based agent interactions
- OssA: OSSA 1.0 compliance layer
- Agent Studio: Multi-platform agent development (Electron, macOS, iOS, web)
- Voice Assistant: Echo Voice integration for voice-controlled chat
Performance Metrics
- Response Time: <200ms p50, <1s p99 (excluding LLM time)
- Throughput: 5,000+ concurrent WebSocket connections
- Availability: 99.9% uptime with multi-instance deployment
- Token Efficiency: 30-50% reduction via VORTEX v3
- Memory Retrieval: <50ms for context assembly
- Learning Cycle: <2s for pattern recognition and improvement
Technology Stack
- Runtime: Node.js 20+, TypeScript 5.0+
- Web Framework: Express 4.18+
- Real-time: Socket.IO (WebSocket)
- GraphQL: Apollo Server with subscriptions
- Databases:
- PostgreSQL (Prisma) - persistent storage
- MongoDB - conversation archive
- Redis - session cache and pub/sub
- Qdrant - vector embeddings
- Authentication: JWT, bcrypt, Drupal SSO
- Observability: Phoenix Arize, Prometheus, Winston
- MCP: Model Context Protocol SDK 1.0+
- UI: React 18+, modern CSS
Related Projects
- @bluefly/agent-router - Multi-provider LLM routing
- @bluefly/agent-mesh - Agent coordination layer
- @bluefly/agent-brain - Knowledge graph and reasoning
- @bluefly/agent-protocol - OSSA protocol implementation
- @bluefly/agent-tracer - Phoenix observability
- @bluefly/agent-buildkit - CLI and orchestration
Quick Links
- Repository: https://gitlab.bluefly.io/llm/common_npm/agent-chat
- Issues: https://gitlab.bluefly.io/llm/common_npm/agent-chat/-/issues
- CI/CD: https://gitlab.bluefly.io/llm/common_npm/agent-chat/-/pipelines
- Package Registry: https://gitlab.bluefly.io/llm/common_npm/agent-chat/-/packages
- OpenAPI Specs: Technical Guide Registry
Use Cases
Customer Support
- Deploy intelligent support agents with knowledge base integration
- Track customer satisfaction and continuously improve responses
- Escalate complex issues to human agents
- Multi-language support with model routing
Internal Knowledge Management
- Semantic search across company documentation
- RAG-powered answers from internal knowledge base
- Team collaboration with shared chat sessions
- Integration with existing enterprise systems
Development Workflows
- Code review assistance via Claude Desktop MCP
- BuildKit integration for CI/CD workflows
- Agent swarm coordination for complex tasks
- Real-time collaboration on technical challenges
Research & Analysis
- Autonomous research agents with web search
- Multi-model comparison and analysis
- Data extraction and summarization
- Collaborative investigation workflows
Support
- Issues: https://gitlab.bluefly.io/llm/common_npm/agent-chat/-/issues
- Documentation: This wiki
- Team: LLM Platform Team llm-platform@bluefly.io
Last Updated: 2025-11-02 Maintainer: LLM Platform Team License: GPL-2.0-or-later