home Codeware
← Back to Portofolios

AI Knowledge App

Intelligent Knowledge Management with Autonomous Agents AI Alpha

🤖 Project Overview

An AI-powered knowledge management system that leverages autonomous agents and LLM integration to automatically collect, enrich, and maintain a growing knowledge base. This alpha-stage project demonstrates innovative approaches to knowledge management through vector embeddings, MCP integration, and AI-driven dossier management.

  • Autonomous Agents - .NET workers for automatic knowledge collection and updates
  • Lean Storage Model - Metadata + embeddings instead of full document copies
  • Multi-Source Integration - RSS feeds, email labels, GitHub issues, and more
  • Vector Embeddings - AI-powered semantic search and relevance matching
  • MCP Integration - Model Context Protocol for Claude Desktop integration
  • RAG Capability - Retrieval-Augmented Generation for enhanced AI responses
  • Pluggable Storage - Support for JSON, SQL, and cloud storage backends
  • Topic-Based Organization - Alpha focused on pension compliance topics

🎯 Key Features

  • 🤖
    Autonomous Agents - .NET workers that automatically update and enrich knowledge base on schedule
  • 📡
    Multi-Source Integration - Collect knowledge from RSS feeds, email, GitHub, and custom sources
  • 🔍
    Vector Search - Semantic search using embeddings for intelligent knowledge discovery
  • 🧠
    AI Enrichment - Automatic tagging, categorization, and knowledge extraction using LLMs
  • 🔌
    MCP Integration - Model Context Protocol support for Claude Desktop and other LLM tools
  • 📚
    RAG Capability - Retrieval-Augmented Generation for enhanced AI-powered responses
  • 💾
    Lean Storage - Efficient storage with metadata, embeddings, and snippets instead of full copies
  • 📋
    Dossier Management - Organize related knowledge items into coherent dossiers with AI assistance

🏗️ Architecture & Technology

Technology Stack

.NET Ecosystem

.NET Workers C# Async/Await Hosted Services

AI & LLM

Claude Integration Embeddings MCP Server RAG Architecture

Data & Storage

Vector Database Metadata Store Pluggable Storage Efficient Indexing

Agents & Workflows

Watch Agent Merge Agent Research Loop Scheduled Updates

Core Components

  • RSS Worker - Automated RSS feed parsing and content collection
  • Email Integration - Process labeled emails for knowledge extraction
  • GitHub Source - Collect issues and discussions as knowledge items
  • AI Enrichment Pipeline - Use Claude for categorization, tagging, and extraction
  • Vector Embedding Engine - Generate and store semantic embeddings
  • Metadata Store - Efficient metadata management for all knowledge items
  • MCP Server - Model Context Protocol implementation for LLM integration
  • RAG API - Retrieval-Augmented Generation endpoints for queries
  • Dossier Manager - Organize related knowledge items into coherent groupings

📊 Lean Storage Model

Storage Philosophy

Instead of storing full articles and documents, the system stores only essential information:

  • Source URL - Link to original content for reference
  • Metadata - Author, publish date, tags, topics, and categories
  • Snippet - First 200-300 characters for preview
  • Embeddings - Vector representation for semantic search
  • Topics Weights - AI-generated topic relevance scores

Benefits

  • Minimal storage costs compared to full-text storage
  • Reduced bandwidth for knowledge updates and syncing
  • Faster semantic search using vector operations
  • Easy referencing to original sources
  • Scalable to thousands of knowledge items

🤖 Autonomous Agents

Agent Types

  • Watch Agent - Monitors sources for new content and updates
  • Merge Agent - Combines related knowledge items into coherent dossiers
  • Research Loop Agent - Deep research on topics with Claude assistance
  • Update Agent - Refreshes existing knowledge items with new information

Autonomous Workflows

  • Scheduled Collection - RSS feeds and sources polled every 4-8 hours
  • AI Enrichment - Automatic categorization and tagging using Claude
  • Deduplication - Intelligent detection of duplicate content
  • Knowledge Merging - Combining related items into topical dossiers
  • Continuous Refinement - Updates existing knowledge with new findings

🔌 Claude Desktop Integration

The application implements Model Context Protocol (MCP) to integrate seamlessly with Claude Desktop:

MCP Capabilities

  • searchKnowledge - Query knowledge base with semantic search
  • getTopic - Retrieve complete topic dossiers and related items
  • appendToDossier - Add new findings to existing dossiers
  • listTopics - Browse available topics and categories

Research Workflow

  • User queries knowledge base through Claude Desktop
  • MCP retrieves relevant knowledge items using vector search
  • Claude analyzes and synthesizes findings
  • Results are rich context for Claude's responses
  • Claude can append new findings back to dossiers

📋 Alpha Scope: Pension Topic

Focus Area

The initial alpha implementation focuses on pension compliance and regulations covering Belgium, Netherlands, and EU jurisdictions. This domain was chosen to fully test end-to-end functionality before expanding to other topics.

Knowledge Sources

  • Pension regulatory updates and announcements
  • Belgium and Netherlands pension law blogs
  • EU directive publications and guidance
  • Tax authority announcements
  • Pension fund guidance documents

🎓 What This Project Demonstrates

AI Knowledge App showcases expertise in:

  • ✅ Autonomous Agents - .NET-based worker agents with independent intelligence
  • ✅ AI Integration - Claude API, embeddings, and semantic search
  • ✅ MCP Protocol - Model Context Protocol implementation for LLM integration
  • ✅ Vector Databases - Semantic search and similarity operations
  • ✅ RAG Architecture - Retrieval-Augmented Generation for enhanced responses
  • ✅ Lean Design - Efficient storage and minimal overhead
  • ✅ Multi-Source Integration - RSS, email, GitHub, and custom sources
  • ✅ Knowledge Management - Organization and curation of information at scale

📋 Project Status

AI Knowledge App is an alpha-stage project demonstrating innovative approaches to AI-driven knowledge management. The architecture is proven and extensible, with a clear roadmap for expansion to additional domains.

  • ✅ Core architecture designed and documented
  • ✅ Agent framework implemented
  • ✅ Lean storage model proven
  • ✅ MCP integration operational
  • 🔄 Alpha domain (pension) in active development
  • 📋 Ready for expansion to additional knowledge domains