Problem
A wellness/coherence platform needs public content, secure role-based member areas and AI assistance without sending sensitive user data to third-party cloud LLMs.
PrototypeSchool portfolio project, full-stack implementation, Clean Architecture and local AI integration
TheWayOfCoherence
A UCL Datamatiker school portfolio project: a Blazor/.NET wellness platform with SQL, role-based auth and local Ollama AI assistants.
Project snapshot
Status
Prototype
Role
School portfolio project, full-stack implementation, Clean Architecture and local AI integration
Main focus
The project combines a Blazor Server public site, ASP.NET Core Identity roles, SQL Server data, admin content management and three role-specific AI assistants powered by a local Ollama model through a Python FastAPI microservice.
Stack
Blazor ServerC# / .NET 9ASP.NET Core IdentitySQL ServerEntity Framework CoreMediatR
Solution
The project combines a Blazor Server public site, ASP.NET Core Identity roles, SQL Server data, admin content management and three role-specific AI assistants powered by a local Ollama model through a Python FastAPI microservice.
Technology
Blazor ServerC# / .NET 9ASP.NET Core IdentitySQL ServerEntity Framework CoreMediatRPython FastAPIOllamaClean ArchitectureAzure
Product walkthrough
Click the dots or arrows to move through the main GoTorz flows.
01 / 04
Public landing page
The Danish public landing page presents the platform message, hero section and first entry point for visitors. It shows the public face of the school project before login.
Architecture
- Built with Clean Architecture split into Domain, Application, Infrastructure and Web layers.
- Uses Blazor Server, C#/.NET 9, ASP.NET Core Identity, MediatR and Entity Framework Core.
- SQL Server stores users, roles, content and application data.
- A separate Python FastAPI microservice connects the .NET app to a local Ollama model.
- The local dotnet server was run during development; the local port is currently stopped/free again.
AI and roles
- Admin AI supports platform/admin tasks, content and member-related guidance.
- Owner/member and patient/guest areas can receive role-specific assistant behavior.
- Chat requests include the user's role so the Python service can apply the correct system prompt.
- The design keeps sensitive user data in the .NET/SQL layer while the LLM runs locally instead of in a third-party cloud.
- Prompts are designed for concise, GDPR-aware responses that match the user's responsibility level.