Repository files navigation

AIChatBot is a local and online AI-powered chatbot built using open-source language models. This project supports both locally hosted models (viaOllama) and cloud-based models viaOpenRouter. It demonstrates the integration of AI with a .NET 8 API and Angular 20 frontend.

Transform your chatbot experience withKnowledge-Based AI! Upload your documents and get contextually-aware responses based on your own content.

Key Highlights:

• 📄Multi-format Document Support: Upload.txt,.md, and.pdf files
• 🔍Intelligent Document Search: Advanced retrieval algorithms find relevant content
• 💬Context-Aware Responses: AI answers based on your uploaded documents
• 📊Source Attribution: See exactly which documents informed each response
• 🗂️Document Management: Easy upload, view, and delete capabilities
• 👤User-Specific Collections: Each user maintains their own document library

Experience AI that truly understands YOUR content!

Watch the AIChatBot in action on YouTube:
📺AIChatBot Demo

To run this project locally, ensure the following:

• Windows 10/11 with WSL support
• Installed Ubuntu 20.04.6 LTS (via Microsoft Store)
• .NET 8 SDK
• Node.js (v20+)
• Angular CLI (npm install -g @angular/cli)
• Ollama installed in Ubuntu for running local AI models
• Optional: Account onhttps://openrouter.ai

• Go to Microsoft Store → Search forUbuntu 20.04.6 LTS → Install.
• Open Ubuntu and create your UNIX user account.

curl -fsSL https://ollama.com/install.sh| sh

To pull and run the desired models:

# Pull modelsollama pull phi3:latestollama pull mistral:latestollama pull gemma:2bollama pull llama3:latest# Run modelsollama run phi3ollama run mistralollama run gemma:2bollama run llama3

# List all pulled modelsollama list# Stop a running modelollama stop phi3# View running modelsps aux| grep ollama

From Ubuntu terminal:

shutdown now

Or simply close the terminal window if you don’t need a full shutdown.

1. Go tohttps://openrouter.ai andsign up.

2. Navigate toAPI Keys in your profile andgenerate an API key.

3. Set this key as an environment variable in your API project:

   export OPENROUTER_API_KEY=your_key_here

4. Models used:

   • google/gemma-3-27b-it:free
   • deepseek/deepseek-chat-v3-0324:free

API requests are routed via OpenRouter using this key, supporting seamless AI chat.

1. Navigate toAIChatBot.API/
2. Run the following commands:

dotnet restoredotnet builddotnet run

• Ensureappsettings.json file includes:

ApiKey=YOUR_KEY_HERE

1. Navigate toAIChatBot.UI/
2. Run:

npm installng serve

• Access the chatbot UI athttp://localhost:4200/

The AIChatBot includes advanced RAG capabilities that allow AI models to answer questions based on your uploaded documents. This feature significantly enhances the AI's ability to provide contextually relevant and accurate responses.

The RAG system consists of several key components:

1. Document Processing Pipeline:

   • File upload handling for multiple formats
   • Text extraction from PDF, TXT, and MD files
   • Content chunking for efficient retrieval
   • In-memory indexing with similarity search

2. Retrieval System:

   • Semantic search across document chunks
   • Top-K retrieval (configurable, default: 3 chunks)
   • Relevance scoring and ranking
   • Source attribution and metadata tracking

3. Generation Enhancement:

   • Context-aware prompt construction
   • Integration with all supported AI models
   • Source citation in responses
   • Fallback to general knowledge when needed

The RAG system supports various AI models with different levels of effectiveness:

1. Upload Documents:

   # Supported formats- Plain text files (.txt)- Markdown files (.md) - PDF documents (.pdf)

2. Document Management:

   • View all uploaded documents
   • Delete individual documents
   • User-specific document collections
   • Automatic indexing upon upload

3. RAG-Enhanced Chat:

   • Select "Knowledge-Based (RAG)" mode
   • Ask questions about your uploaded content
   • Receive responses with source attribution
   • Contextual answers based on document content

Currently, the RAG system uses in-memory storage (InMemoryRagStore), which provides:

• Fast retrieval performance
• Simple deployment setup
• Automatic cleanup on application restart
• User-isolated document collections

Note: For production deployments, consider implementing persistent storage solutions.

AIChatBot/│├── AIChatBot.API/           # .NET 8 API for chatbot│   ├── Controllers/         # API endpoints│   │   └── DocumentsController.cs    # RAG document upload/management│   ├── Services/           # Business logic services│   │   ├── RagChatService.cs         # RAG-enabled chat functionality│   │   ├── InMemoryRagStore.cs       # Document indexing and search│   │   ├── AgentService.cs           # AI tool integration│   │   └── ChatService.cs            # Standard chat functionality│   ├── Interfaces/         # Service contracts│   │   └── IRagStore.cs              # RAG storage interface│   └── Migrations/         # Database schema updates for RAG├── AIChatBot.UI/           # Angular 20 UI frontend│   └── src/app/components/│       ├── document-upload/          # RAG document upload component│       ├── chat/                     # Main chat interface│       └── model-selector/           # AI model and mode selection└── README.md               # Project documentation

The AIChatBot supports three advanced operation modes beyond simple chat:

In this mode, the AI can recognize specific tasks in user prompts and use internal tools (functions) to perform actions. Integrated tools include:

These functions are executed server-side in.NET, with input parsed from natural language prompts.

NEW FEATURE: The RAG (Retrieval-Augmented Generation) mode enables AI to answer questions based on your uploaded documents. This powerful feature allows you to:

• Upload Documents: Support for.txt,.md, and.pdf files
• Intelligent Retrieval: Automatically finds relevant content from your documents
• Source Attribution: AI responses include references to source documents
• Document Management: View, organize, and delete uploaded documents
• User-Specific Storage: Each user has their own document collection

• Text Files:.txt - Plain text documents
• Markdown Files:.md - Formatted markdown documents
• PDF Documents:.pdf - Portable document format

1. Upload Documents: Use the drag-and-drop interface or browse to upload documents
2. Select RAG Mode: Choose "Knowledge-Based (RAG)" from the chat mode dropdown
3. Ask Questions: Query your documents using natural language
4. Get Contextual Answers: Receive AI responses enriched with your document content

User: "What are the key findings in the latest market report?"AI: Based on the provided context from "market_analysis_2025.pdf", the key findings include:1. Consumer spending increased by 15% in Q42. Digital transformation investments rose by 23%3. Supply chain disruptions decreased significantly*Sources: 1 document(s) referenced*

The AI agent is capable of:

• Understanding high-level tasks
• Selecting and invoking appropriate tools
• Providing intelligent responses based on the outcome

This is powered by anAgentService that works with bothlocal LLMs (via Ollama) andcloud models (via OpenRouter) to determine the right function to execute and handle the response.

You can toggle between AI modes via the UI:

• Chat-Only Mode
• AI + Tools Mode
• Knowledge-Based (RAG) Mode
• Agent Mode (multi-step planning, coming soon)

1. Choose your preferred model type (local or online)
2. Start the backend using.NET 8
3. Start the frontend using Angular CLI
4. Access AIChatBot athttp://localhost:4200/

5. Upload Documents:

   • Click the "Documents for RAG" section in the sidebar
   • Drag & drop or browse to upload.txt,.md, or.pdf files
   • Wait for successful upload confirmation

6. Enable Knowledge-Based Mode:

   • Select "Knowledge-Based (RAG)" from the chat mode dropdown
   • Choose an AI model with good RAG support (⭐⭐⭐⭐ or ⭐⭐⭐⭐⭐)

7. Start Asking Questions:

   Examples:"Summarize the key points from my uploaded documents""What does the report say about market trends?""Find information about [specific topic] in my files"

8. Review Source Attribution:

   • AI responses will include "Sources: X document(s) referenced"
   • Responses are enriched with content from your uploaded documents

Pull requests and suggestions are welcome! Feel free to fork the repo and enhance it.

This project is open-source and available under theMIT License.