Case study · 2026
Codivus
An offline AI programming-education platform combining local language models with retrieval-augmented learning workflows.
- Discipline
- AI education · Offline-first software
- Technology
- React · FastAPI · Ollama · RAG · Local LLMs
CODIVUS
Programming help, running locally.
Overview
Codivus brings together React, FastAPI, Ollama, and retrieval-augmented generation to make programming support available without depending on a cloud-hosted model.
System map
How the pieces connect
The challenge
What made the problem worth solving
Programming education tools often assume reliable connectivity and treat a model answer as the entire learning experience. Codivus explores how local AI can support learning without making cloud access a prerequisite.
The approach
Turning the problem into an engineering plan
- 01
Used Ollama to run language models locally instead of routing every interaction through a hosted service.
- 02
Connected a React learning interface to a FastAPI backend for a clear separation between product and inference layers.
- 03
Added retrieval-augmented generation so assistance could be grounded in relevant learning material.
Result
Codivus extends my local-first AI work into programming education, combining accessibility, practical deployment, and structured learning in one system.