RAG Demo with Local Embeddings and Qdrant

This project demonstrates how to build a simple Retrieval-Augmented Generation (RAG) system using:

• Local Text Embeddings via SmartComponents.LocalEmbeddings
• Vector Storage & Search via Qdrant
• Ollama to serve a local language model (e.g., mistral:7b)

💡 Why this project?

The biggest challenge in RAG systems is text embedding — converting meaningful text into high-dimensional vectors for retrieval. While many solutions rely on cloud APIs or heavy frameworks like ONNX, this project demonstrates how to:

• Use local embeddings with minimal setup
• Achieve fast and accurate semantic search
• Keep everything offline and privacy-preserving

⚙️ Technologies Used

Component	Tool/Library
Embeddings	SmartComponents.LocalEmbeddings
Vector DB	Qdrant (running locally)
LLM	Ollama (mistral:7b)
API Layer	ASP.NET Core Web API

🧠 How It Works

[User Question]
     ↓
Local Embedder (Microsoft's smartcomponents)
     ↓
Qdrant Search (Vector DB)
     ↓
Top Document Retrieved
     ↓
Mistral via Ollama (LLM)
     ↓
Final Answer

Setup

1. Clone the repo.
2. Install Qdrant locally or run via Docker.
3. Run Ollama with:

   ollama run mistral

4. Launch the ASP.NET API project.

License

MIT License.