[FEATURE]: AI-powered "Chat with your Quiz" — ask follow-up questions on quiz content

[hk2166]

Feature and its Use Cases

Problem

Currently, EduAid generates quizzes from content but provides no way
for users to interact with that content further. When a user gets a
question wrong or doesn't understand an answer, there is no way to
ask "why is this the correct answer?" or "explain this concept in
simpler terms". Users have to go back to the original source manually.

Proposed Solution

Add a conversational AI layer on top of the generated quiz that allows
users to:

• Ask follow-up questions like "explain why option B is correct"
• Ask "give me an example of this concept"
• Ask "summarize this topic in simple terms"
• Chat directly with an uploaded PDF — ask questions about its content
  beyond just the generated quiz

How It Works

Chat with Quiz

1. When a quiz is generated, the source content is stored as context
2. User can open a chat panel alongside the quiz
3. User asks a question — it gets matched against the source content
   using semantic search (RAG)
4. A language model generates a grounded answer from the source material

Chat with PDF

1. User uploads a PDF
2. PDF text is extracted and chunked
3. Chunks are embedded and stored in a FAISS vector index
4. User can ask any question about the PDF content
5. Relevant chunks are retrieved and passed to the LLM for a grounded answer

Tech Stack Suggestion

• Python, FAISS, Sentence Transformers (all-MiniLM-L6-v2)
• LLM: Google Gemini API or local open-source model
• Frontend: existing React/JS chat UI pattern

Why This Fits EduAid

EduAid's goal is to help students learn and retain information.
A chat interface transforms EduAid from a passive quiz tool into
an active learning companion — students don't just get tested,
they get to understand. This is a natural and powerful extension
of the existing quiz generation feature.

Additional Notes

I have prior experience building a RAG pipeline from scratch
(DocuMind — PDF Q&A using FAISS + Gemini) and would love to
implement this feature. Happy to discuss the approach before
starting.

Additional Context

No response

Code of Conduct

• I have joined the Discord server and will post updates there
• I have searched existing issues to avoid duplicates

[Tanushreesmallikarjuna]

Hi @hk2166 👋
This feature sounds really impactful for EduAid.
If you’re open to collaboration, I’d be happy to help with a subtask (frontend chat UI / FAISS integration / testing). Let me know!

[hk2166]

Hi @hk2166 👋 This feature sounds really impactful for EduAid. If you’re open to collaboration, I’d be happy to help with a subtask (frontend chat UI / FAISS integration / testing). Let me know!

Hi, I’m new to this organization. I wanted to ask about the contribution process — should I first get assigned to an issue before starting work, or can I go ahead and raise a PR directly for an open issue?

[Tanushreesmallikarjuna]

I'm new here too . The common practice is to comment first to show interest, then start working on a specific subtask.
Once you’re ready, you can open a PR for your work. Just try to keep your changes scoped and transparent so there’s no overlap with others

[mahek2016]

Hi 👋

This feature has strong potential to transform EduAid from a quiz generator into an interactive learning assistant.

I’m interested in contributing to this — particularly the RAG pipeline design (embedding + retrieval architecture).

Before starting implementation, I’d like to clarify a few architectural points to avoid overlap:

• Should embeddings be stored in-memory (FAISS) or persistently for session reuse?
• Are we leaning toward a local open-source LLM or an API-based model (e.g., Gemini)?
• Should the chat context be limited to quiz content only, or the entire uploaded document?

I can draft a structured implementation plan covering:

• PDF chunking strategy
• Embedding pipeline
• Vector index design
• Retrieval + grounding flow
• Backend API structure
• Frontend chat UI integration

Happy to collaborate and scope this properly.

[aryash45]

hi @hk2166 when you feel you're ready you can open a pr then ask the maintainers on discord to review it

[hk2166]

hi @hk2166 when you feel you're ready you can open a pr then ask the maintainers on discord to review it

Thanks for the clarification. I’ve started working on it and will be raising a PR shortly.

[hk2166]

@Zahnentferner @Aditya062003 @yatikakain Can you please review this issue and please let me know does it fit the project?