CloudMentor is a Retrieval-Augmented Generation (RAG) revision chatbot designed to help prepare for the AWS Certified Developer – Associate exam. It simulates an exam-style experience using MCQ-based questions generated from handwritten notes, making it a personalised and effective revision tool.
The knowledge base is built from handwritten study notes located in the notes directory. These notes are chunked, embedded using the nomic-embed-text model via Ollama, and stored in a local ChromaDB instance for fast semantic retrieval.
-
After cloning the project, you'll need to create a
.envfile in thebackendfolder. Within this file, create two variables: theOPENAI_API_KEYwhich you'll need to configure via OpenAI, andNUM_QUESTIONS, which determines the number of questions in the exam, 65 being the standard. -
Next, make sure you have Ollama installed, and the embedding model mentioned above pulled locally.
-
Next, create a virtual environment for the backend dependencies and install the dependencies from the
requirements.txtfile. From within thebackendfolder, runpython main.pyto start the backend API. Finally travel to thefrontendfolder, runnpm installto install all the required frontend dependencies andnpm run devto run the frontend. Go tohttp://localhost:5173/to see the project dashboard. -
The Dashboard shows a set of statistics about past exam attempts, including a visual chart of your score by exam domain. The Exam page will load questions sequentially, this encourages you to read the question carefully while generating the questions one at a time. Questions are generally 1-answer only, altough some 2- or even 3- answer questions might be occasionally generated. Questions can additionally be marked for review. Once the exam is submitted, a review will be available to let the user see where they went wrong and how to improve.
- If I plan to undertake more certification exams, I will incorporate the notes made while preparing within CloudMentor. Generalising this project to support materials for mutliple MCQ-based certifications should be a straight-forward task as long as the quality of the knowledge-base for the chatbot is good enough.