nl-rag-qdrant-legal

A sample application demonstrating Retrieval-Augmented Generation (RAG) for legal document Q&A using local vector search with QDrant, Ollama-hosted AI models, and Microsoft KernelMemory. Intended for legal, compliance, and technical professionals seeking to explore next-gen semantic search in the legal domain using C# and .NET 9.0.

Prerequisites

• .NET 9.0 SDK or later
• QDrant installed and running locally (via Docker or binary)
• Ollama installed and running locally
• Required AI models downloaded in Ollama:
  • nomic-embed-text:latest (for embeddings)
  • gemma3:1b (for legal Q&A/chat)

Installation

1. Clone this repository:

   git clone https://github.com/your-username/nl-rag-qdrant-legal.git
   cd nl-rag-qdrant-legal

2. Restore NuGet packages:

   dotnet restore

Setup

1. Start QDrant (Vector Database):

   • The easiest way is via Docker:

     docker run -p 6333:6333 -p 6334:6334 qdrant/qdrant

   • QDrant will run at http://localhost:6333

2. Install and Start Ollama:

   • Download Ollama from ollama.ai, install and launch it.
   • Download and prepare required models:

     ollama pull nomic-embed-text:latest
     ollama pull gemma3:1b

   • Ollama will run at http://localhost:11434

3. Verify Model & Service Configuration:

   • Ensure both services are running on their default endpoints:
     • QDrant: http://localhost:6333
     • Ollama: http://localhost:11434
   • Check the config in LegalDocConfig.cs for model versions and endpoints if you change ports/models.

Usage

Run the console application to start the legal RAG Q&A example:

cd RAGQdrantLegal
# Restore and build
dotnet restore
# (optional, if not yet run)
dotnet build
# Run
dotnet run

You will see interactive output and will be prompted to ask legal questions against the database. Example prompts:

• "What are the key confidentiality clauses in the NDA?"
• "How can I terminate the employment contract as per the document?"
• "What is the defined uptime in the SLA?"
• "Who is responsible for breach notification according to the DPA?"

Example Flow

1. Legal Document Import:
   • The app loads a set of sample legal documents, processes them into semantic embeddings using nomic-embed-text, and stores them in a QDrant vector collection.
2. Semantic Search & Q&A:
   • You enter your legal query; the app uses KernelMemory to search for the most relevant text chunks using vector similarity and the QDrant engine.
   • The app uses the gemma3:1b model (locally via Ollama) to generate a context-aware answer, optionally showing relevant sources from the vector DB.

Legal Document Types Included

• Non-Disclosure Agreement (NDA)
• Employment Contract
• Data Processing Addendum (DPA)
• Service Level Agreement (SLA)
• (Plus other sample agreements for demonstration - see LegalDocumentData.cs)

Key Technologies

• .NET 9.0 (C# Console App)
• QDrant (Vector similarity search)
• Ollama (Local AI model runtime)
• Microsoft KernelMemory (Microsoft.KernelMemory.Core)
• nomic-embed-text:latest (Nomic AI, for embeddings)
• gemma3:1b (Google, for chat/QA)

Dependencies

NuGet packages required (see .csproj):

<PackageReference Include="Microsoft.KernelMemory.Core" Version="0.98.250508.3" />
<PackageReference Include="Microsoft.KernelMemory.MemoryDb.Qdrant" Version="0.98.250508.3" />
<PackageReference Include="Microsoft.KernelMemory.AI.Ollama" Version="0.98.250508.3" />
<PackageReference Include="Microsoft.SemanticKernel" Version="1.61.0" />
<PackageReference Include="Qdrant.Client" Version="1.15.0" />
<PackageReference Include="Microsoft.Extensions.VectorData.Abstractions" Version="9.7.0" />
<PackageReference Include="Microsoft.Extensions.AI" Version="9.7.1" />

Project Structure

• LegalDocRagApp.cs: Orchestrates initialization, import, and chat
• Program.cs: Entry point
• LegalDocumentImporter.cs: Imports legal documents into the semantic memory
• LegalDocumentData.cs: Sample legal document database
• ChatService.cs: Interactive Q&A loop
• LegalDocConfig.cs: Contains configuration for endpoints and models

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contributing

Contributions are welcome! Please submit an issue or open a Pull Request if you find a bug or have an enhancement idea.