RAG SaaS Backend

Production-ready .NET 8 RAG backend with:

• document upload (.pdf and .txt)
• chunking + embeddings generation
• Pinecone vector upsert/query
• question answering endpoint constrained to retrieved context
• unit tests and CI in GitHub Actions

Project Structure

• backend/RAGService/RAGService - API project
• backend/RAGService/RAGService.Tests - unit tests
• .github/workflows/ci.yml - CI pipeline (restore/build/test)

Prerequisites

• .NET SDK 8.x
• OpenAI API key
• Pinecone index and API key

Configuration

Set values in backend/RAGService/RAGService/appsettings.json or via environment variables:

• OpenAI__ApiKey
• OpenAI__BaseUrl (default: https://api.openai.com/v1/)
• OpenAI__EmbeddingModel (default: text-embedding-3-small)
• OpenAI__ChatModel (default: gpt-4o-mini)
• Pinecone__ApiKey
• Pinecone__BaseUrl
• Pinecone__DefaultTopK

Run Locally

cd backend/RAGService
dotnet restore
dotnet run --project RAGService/RAGService.csproj

Swagger UI is available in development mode.

Test Locally

cd backend/RAGService
dotnet test RAGService.sln

API Endpoints

• POST /api/documents/upload (multipart form-data, field file)
• POST /api/queries/ask (JSON body: { "question": "..." })
• GET /health