AI RAG .NET + Python Monorepo

AI Q&A system: Python RAG (FastAPI + OpenAI embeddings) + .NET 10 Clean Architecture API + SQL Server logging + xUnit/pytest + Docker Compose + GitHub Actions CI.

Portfolio project showing how to integrate AI capabilities into a production-style .NET backend.

Highlights

• Python Retrieval-Augmented Generation with OpenAI embeddings (keyword fallback).
• .NET 10 Web API structured with Clean Architecture.
• API key authentication, ProblemDetails, rate limiting, and health checks.
• HTTP orchestration with Polly-based resilience on the RAG client.
• SQL Server persistence for consultation audit logs.
• Unit/integration tests and CI coverage artifacts.
• One-command local stack via Docker Compose.

Built by

Mario Ramb - Systems Engineer with 18+ years in enterprise .NET development. This project demonstrates AI integration on top of a production-grade .NET backend.

LinkedIn · GitHub

Repository Structure

dotnet-rag-api/
|-- rag-service/        # Python FastAPI RAG service
|-- DotnetRagApi/       # .NET 10 API + Clean Architecture layers + tests
|-- docker/             # SQL init scripts for Compose
|-- docs/               # Architecture diagram
|-- docker-compose.yml
`-- LICENSE

Tech Stack

• Python, FastAPI, OpenAI SDK (embeddings + chat), python-dotenv, pytest
• .NET 10, ASP.NET Core Web API, Dapper
• SQL Server
• Docker Compose
• xUnit + coverlet, GitHub Actions

Architecture Overview

1. Client sends a question to the .NET endpoint (POST /api/ask) with X-Api-Key.
2. Application use case invokes the Python RAG service (POST /rag/ask).
3. Python retrieves relevant context with embeddings (or keyword fallback) and calls an LLM.
4. Response returns to .NET and the use case persists an audit row in SQL Server.
5. Logs can be queried from GET /api/logs (authenticated).

Architecture

Quick Start with Docker Compose

Prerequisites: Docker Desktop, an OpenAI API key.

copy .env.example .env
# edit .env and set OPENAI_API_KEY

docker compose up --build

Then open:

• Scalar: http://localhost:5000/scalar/v1
• Health: http://localhost:5000/health
• Readiness: http://localhost:5000/health/ready

Default API key (override with API_KEY in .env):

change-me-demo-api-key

Example request:

curl -X POST http://localhost:5000/api/ask \
  -H "Content-Type: application/json" \
  -H "X-Api-Key: change-me-demo-api-key" \
  -d "{\"pregunta\":\"What is retrieval augmented generation?\",\"usuario\":\"demo\",\"rol\":\"user\",\"topN\":2}"

Manual Local Setup

Prerequisites

• .NET SDK 10+
• Python 3.10+
• SQL Server LocalDB / SQL Server / Docker SQL
• OpenAI API key

1) Run the Python RAG Service

From rag-service:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
copy .env.example .env

Start:

uvicorn app:app --host 127.0.0.1 --port 8000 --reload

2) Create the SQL database

Run:

• DotnetRagApi/database/create-database.sql

3) Run the .NET API

Copy DotnetRagApi/DotnetRagApi.Api/appsettings.example.json to appsettings.json and set:

• ConnectionStrings:DefaultConnection
• Security:ApiKey
• RagApi:BaseUrl

cd DotnetRagApi
dotnet run --project .\DotnetRagApi.Api\DotnetRagApi.Api.csproj

API Endpoints

Method	Route	Auth	Description
POST	/api/ask	X-Api-Key	Ask a question through RAG
GET	/api/logs?top=50	X-Api-Key	Latest audit logs (top clamped 1..100)
GET	/health	No	Liveness
GET	/health/ready	No	SQL + Python readiness

Tests

.NET:

cd DotnetRagApi
dotnet test .\DotnetRagApi.slnx

Python:

cd rag-service
pytest -q

Why It Works As A Portfolio Project

• Demonstrates practical AI integration instead of isolated toy logic.
• Applies Clean Architecture so business orchestration stays testable.
• Shows cross-language integration (Python + .NET) with resiliency and health checks.
• Includes authentication, Docker packaging, automated tests, and CI.

Notes

• Domain/DTO fields intentionally use Spanish naming (pregunta, respuesta) to match the business vocabulary.
• Retrieval defaults to OpenAI embeddings (RAG_RETRIEVAL_MODE=embeddings) with keyword fallback.
• For cloud hosting, the Compose services map cleanly to Azure Container Apps / App Service + Azure SQL.

License

MIT - see LICENSE.