Skip to content

ClassMate.md

Latest commit

 

History

History
77 lines (48 loc) · 3.1 KB

ClassMate.md

File metadata and controls

77 lines (48 loc) · 3.1 KB
Raw

ClassMate: Ask your course repo

Inspiration

Students learn faster when answers are close to the source: the slides, labs, and notes their lecturer already wrote. I wanted a simple way for a class to ask questions in natural language and get answers grounded in the course repo—not random web results.

Reality

Course material lives in a GitHub repository (Markdown, PDFs, notebooks). Students ask on Teams or by email. Replies repeat or arrive late. There’s no single place to search exactly what the course says.

Problem

  • Students can’t quickly find the right answer in a technical question.
  • TAs and lecturers answer the same questions many times.
  • Generic chatbots hallucinate or quote sources that are not part of the course.

Solution

Build a small app that reads material from a GitHub course repository, indexes the content, and answers questions using an LLM that is restricted to the repo.

Students ask; the app retrieves the most relevant passages and returns a short answer with citations to file + section.

school-of-rock

🎬 From School of Rock, starring Jack Black.

Key behaviors

  • Pull and parse files from a GitHub repo (Markdown, PDFs, TXT, notebooks).
  • Chunk and index content as a 7–10 sentence “knowledge blocks”.
  • Retrieve top matches per question and cite the exact files/sections.
  • If an answer is not inside the repo; suggest where to look instead.
  • Optional “Instructions” repo: house rules, grading policy, lab etiquette used in answers.
  • Simple admin refresh to re-index after material updates.

Example (better — auth key generation)

Student asks:

“How do I set up Lab 2 on Windows with Python 3.11?”

Answer (grounded):

  • Install Python 3.11 from the link in labs/lab-02/setup.md.
  • Run python -m venv .venv and activate it (Windows steps shown).
  • Install requirements using pip install -r requirements.txt.
  • Launch the tests with pytest -k lab02.

References: [1] labs/lab-02/setup.md — Windows steps, [2] labs/lab-02/requirements.txt

Student asks:

“When is the deadline for assignment 1?”

Answer (grounded):

  • Assignment 1 (30%) deadline is on 12/12/2025. Please double check the moodle submission space.

References: [1] moodle.bbk.ac.uk/acc/assigment-1

Minimal tech plan / stack

  • Data: GitHub API (repo content), simple cron/trigger for refresh.
  • Storage: Lightweight vector index (e.g., FAISS) + small metadata store.
  • Model: LLM for responses with retrieval (embeddings + top-k context).
  • Backend: Python service exposing /ask and /refresh.
  • Frontend: Streamlit app with Q&A box, citations, and a “Sources” view.
  • Safeguards: “Answer only from provided context”; show sources; log unanswered topics.

Goal

Ship a minimal MVP that:

  1. Connects to a GitHub course repo and builds a searchable index.
  2. Answers student questions using only that material, with clear citations.
  3. Gives staff a one-click refresh after pushing new content.
  4. Reduces repeat questions and keeps learning aligned with the course.