AlcorAgent - AI-Powered Chat Platform

Intelligent chat agent powered by multiple RAG systems, MCP servers, and real-time knowledge base integration.

Tech Stack: Next.js , LangChain , Pinecone , MCP Protocol , Redis , Tailwind CSS , OpenAI , Confluence API , Notion API , TypeScript

Overview

AlcorAgent is a sophisticated chat application that combines multiple Retrieval-Augmented Generation (RAG) systems with Model Context Protocol (MCP) servers to create an intelligent agent capable of accessing diverse knowledge sources. Named after the binary star system Alcor-Mizar, it represents the interconnected nature of multiple data sources working in harmony.

The system leverages Next.js for the frontend, TypeScript for type safety, and integrates with various external services including Pinecone for vector storage, Confluence and Notion for internal knowledge bases, and multiple MCP servers for extended functionality.

Features

Multi-RAG Architecture

  • Document RAG: Processes PDFs, Word docs, and text files using Pinecone vector storage
  • Wiki RAG: Real-time integration with Confluence spaces and Notion databases
  • Smart Orchestration: Automatically routes queries to appropriate RAG systems
  • Extensible: Can be added any RAG into this easily

MCP Server Integration

  • Multiple Server Support: Connects to various MCP servers for extended capabilities
  • Tool Execution: Dynamic tool calling based on user queries and context
  • Official SDK: Built using @modelcontextprotocol/sdk for reliability
  • Type-Safe Communication: Full TypeScript support for all MCP interactions

Real-Time Synchronization

  • Confluence Webhooks: Instant updates when wiki pages are modified
  • Notion Polling: Regular synchronization with Notion databases and pages
  • Redis Caching: High-performance caching for frequently accessed content

Modern User Interface

  • Next.js Frontend: Server-side rendering with optimal performance
  • Tailwind CSS: Responsive, modern design system
  • Real-time Chat: Smooth conversational interface with typing indicators
  • Source Attribution: Direct links to original content sources
  • Context Preservation: Maintains conversation history and context

System Design

Architecture Design

Core Components

  1. Chat Agent Core: Central orchestrator that manages user interactions, maintains conversation context, and coordinates between different subsystems.

  2. RAG Orchestrator: Intelligent routing system that classifies queries and distributes them to appropriate RAG subsystems based on content type and context.

  3. MCP Client Layer: Handles communication with multiple Model Context Protocol servers, managing tool execution and external service integration.

  4. Vector Storage Layer: Pinecone-based vector database with multiple namespaces for different content types, enabling efficient semantic search across diverse data sources.

  5. Synchronization Engine: Manages real-time updates from external sources through webhooks and polling mechanisms, ensuring knowledge base freshness.

RAG Subsystems

  • Document RAG: Processes unstructured documents (PDFs, Word files, text documents) with intelligent chunking and metadata preservation
  • Communication RAG: Processes conversational data from various sources with context preservation and sentiment analysis

Technology Stack

  • Frontend: Next.js 15+ with App Router, TypeScript, Tailwind CSS
  • Backend: Node.js with TypeScript, Express.js middleware
  • Vector Database: Pinecone for high-performance similarity search
  • Caching: Redis for session management and query result caching
  • APIs: Confluence REST API, Notion API, OpenAI/Azure OpenAI
  • Protocols: Model Context Protocol (MCP) for extensible tool integration
  • Deployment: AWS Amplify
View on Github