Corporate Knowledge Base

Project Description

This project is a Corporate Knowledge Base platform developed to centralize and facilitate internal knowledge sharing within a company. It evolves beyond a standard CMS by integrating a Private AI Assistant, robust permission systems, content moderation, version control, and real-time interactions.

Unlike cloud-based solutions, this platform runs 100% On-Premise, ensuring that sensitive corporate data (technical documents, internal blogs) never leaves the local server, making it fully compliant with strict data privacy regulations (GDPR/KVKK).

---

🚀 What's New in v3.0.0? (The AI Revolution)

v3.0.0 is the biggest milestone yet, transforming the application from a passive storage system into an active, intelligent assistant using Local LLMs and RAG Architecture.

🤖 1. Private AI Assistant (RAG Architecture)

• Local Intelligence: Powered by Microsoft Phi-3 (via Ollama) running locally. No API keys, no cloud costs, no data leaks.
• Hybrid Search Engine: Combines Vector Similarity (Nomic Embed Text) with Keyword Boosting. It understands both the meaning (Semantic) and specific terminology (Lexical) of your query.
• Multi-Source Indexing: The AI scans both Technical Documents and Blog Posts, finding the best answer from the entire corporate memory.
• Smart Context:
  • Auto-Embedding: Content is vectorized immediately upon creation or update.
  • Syntax Highlighting: Code blocks in AI responses are automatically formatted and colored (Prism.js).
  • Source Attribution: Every answer includes a direct link to the source document with a confidence score (e.g., "Source: EF Core Guide - 85% Match").

⚡ 2. Asynchronous Architecture (Hangfire)

• Fire-and-Forget Jobs: Heavy operations like PDF Generation no longer block the user interface. They are processed in the background, preventing browser timeouts.
• Dashboard: A dedicated monitoring panel (/hangfire) for tracking background tasks.

📡 3. Real-Time Collaboration (SignalR)

• Live Updates: Users receive instant notifications for completed reports, new comments, or AI analysis results without refreshing the page.

📜 4. Content Versioning (Git-Logic)

• Snapshot System: Automatic version history for every edit.
• Time Travel & Restore: Admins can view differences between versions and revert a document to any previous state with a single click.

---

📸 Screenshots

Smart AI Assistant	Rich Document Viewer
AI answering technical questions with code blocks	Markdown rendering, PDF export, and interactions
Version History	Hangfire Dashboard
Time-travel debugging for documents	Background job monitoring

---

🛠️ Architecture & Technologies

The system is built on a modular, scalable architecture designed for enterprise environments.

• Backend: ASP.NET Core 9.0 (MVC)
• Database: MS SQL Server (Entity Framework Core Code-First)
• AI & ML Stack (New):
  • Orchestrator: Microsoft.Extensions.AI
  • Inference Engine: Ollama (Local Server)
  • LLM Model: Microsoft Phi-3 Mini (3.8B)
  • Embedding Model: Nomic-Embed-Text-v1.5
  • Vector Operations: System.Numerics.Tensors
• Real-Time: SignalR (WebSockets)
• Background Jobs: Hangfire (SQL Storage)
• Frontend: Bootstrap 5, jQuery
• Key Libraries:
  • PuppeteerSharp: HTML-to-PDF conversion (Chromium headless).
  • Markdig: High-performance Markdown processor.
  • Prism.js: Advanced syntax highlighting for code blocks.
  • Tagify: Lightweight tagging interface.
  • EasyMDE: User-friendly Markdown editor.
  • DataTables.net: Advanced table controls for admin panels.
  • SweetAlert2: Professional, responsive popup replacements.

---

🔮 Roadmap: The "Universal Deployment" Vision

We are pausing feature development to focus on infrastructure and portability. The next major release aims to decouple the application from expensive enterprise hardware.

v4.0.0 - Universal Deployment (Raspberry Pi & Docker)

• 🐳 Containerization: Official Dockerfile and docker-compose support for "One-Click Deploy".
• 💾 Database Independence: Dropping the strict dependency on MS SQL Server. Adding support for PostgreSQL (Docker standard) and SQLite (Edge/IoT standard).
• 🍓 ARM64 Optimization: Optimizing the AI Inference stack to run efficiently on Raspberry Pi 5 (8GB) and other ARM-based edge devices.
• Goal: A fully functional, AI-powered corporate knowledge base that fits in your pocket and runs offline.

---

⚙️ Installation (v3.0 Local Setup)

1. Prerequisites:

   • .NET 9.0 SDK
   • SQL Server (LocalDB or Express)
   • Ollama installed.

2. Setup AI Models: The system is optimized for phi3 (performance) and nomic-embed-text (vectorization).

   ollama pull phi3
   ollama pull nomic-embed-text

   💡 Flexible Architecture: The system is model-agnostic. While Phi-3 is the default for efficiency, you can easily switch to larger models (e.g., Llama 3, Mistral, or Gemma) for enterprise-grade reasoning. Simply pull the desired model via Ollama and update the model name configuration in Program.cs.

3. Clone the Repository:

   git clone https://github.com/umitkrkmz/CorporateKnowledgeBase.git
   cd CorporateKnowledgeBase

4. Configuration (Crucial Step):

   • Open appsettings.json and update the ConnectionStrings section to match your local SQL Server instance.
   • (Optional) If using User Secrets for admin passwords, ensure you set them locally:

     dotnet user-secrets set "AdminSettings:DefaultPassword" "YourSecurePassword123!"

5. Run the Application: Apply database migrations and start the server:

   dotnet ef database update
   dotnet run

   • Default Admin User: admin@knowledgebase.com
   • Password: The one you set in Step 4 (or check DbInitializer.cs for defaults).

6. First Time Configuration (AI Indexing):

   • Log in as an Admin or Developer.
   • Navigate to the "Documents" page (/Document/Index).
   • Click the yellow "Re-Index AI" button (visible only to authorized roles) to generate vectors for existing content.

📄 License

Copyright (c) 2025 umitkrkmz Licensed under the MIT License.