QPredict

Turn past papers into smarter preparation

Comment

QPredict

Turn past papers into smarter preparation.

QPredict is an AI-powered Streamlit app that analyzes previous-year examination question papers. Upload PDFs, and QPredict identifies recurring topics, historical trends, and study priorities — all grounded in evidence from your own papers.

Live demo: https://qpredict-snke7a4smrvtccsawgji8m.streamlit.app

Features

Page What it does
📤 Upload Upload exam PDFs, extract questions automatically
📚 My Papers View and delete uploaded papers
📊 Dashboard Priority-ranked topic clusters with year heatmap and trend charts
🔍 Deep Analysis TF-IDF + semantic similarity analysis per subject
📖 Syllabus Map questions to syllabus units
📅 Study Plan Auto-generated day-by-day study schedule
❓ Quiz Practice MCQ + short-answer quiz from your papers
🤖 Study Assistant RAG-powered chat — ask questions about your uploaded papers

How it works

  1. Upload previous-year PDFs — supports KTU and most Indian university formats
  2. Questions are extracted automatically using pattern matching (Part A / Part B / Module sections)
  3. Embeddings are computed using sentence-transformers (falls back to TF-IDF if unavailable)
  4. Clustering groups semantically similar questions into topic clusters
  5. Priority scoring ranks topics by frequency, year coverage, recency, and consistency
  6. Dashboard displays ranked topic cards with charts, year timelines, and score breakdowns

Setup (local)

1. Clone the repo

git clone https://github.com/Athmana/QPredict.git
cd QPredict

2. Create a virtual environment

python -m venv venv
# Windows
venv\Scripts\activate
# macOS/Linux
source venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Run

streamlit run app.py

Open http://localhost:8501 in your browser.

Project structure

QPredict/
├── app.py                        Home page + app entry point
├── requirements.txt
│
├── pages/
│   ├── 1_📤_Upload.py
│   ├── 2_📚_My_Papers.py
│   ├── 3_📊_Dashboard.py
│   ├── 4_🔍_Deep_Analysis.py
│   ├── 5_📖_Syllabus.py
│   ├── 6_📅_Study_Plan.py
│   ├── 7_❓_Quiz.py
│   └── 8_🤖_Study_Assistant.py
│
├── src/
│   ├── pdf_parser.py             PDF text extraction (PyMuPDF)
│   ├── question_extractor.py     KTU + general question parsing
│   ├── text_cleaner.py           NLP normalization
│   ├── embeddings.py             Sentence Transformer / TF-IDF embeddings
│   ├── similarity.py             Cosine similarity analysis
│   ├── clustering.py             Agglomerative + DBSCAN clustering
│   ├── scoring.py                Historical Priority Score (4 components)
│   ├── trend_analyzer.py         Year-over-year trend classification
│   ├── ui_helpers.py             Reusable Streamlit components + Plotly charts
│   ├── syllabus_mapper.py        Syllabus unit mapping
│   ├── study_planner.py          Study plan generation
│   ├── quiz_generator.py         MCQ + short-answer quiz generation
│   ├── chunker.py                Text chunking for RAG
│   ├── rag_retriever.py          FAISS / NumPy vector index
│   └── rag_assistant.py          RAG-powered LLM assistant
│
├── database/
│   └── database.py               SQLite schema and query functions
│
├── data/                         Runtime only — not committed
│   ├── uploads/                  Uploaded PDFs
│   ├── processed/                Extracted text files
│   └── qpredict.db               SQLite database (auto-created)
│
└── tests/                        Unit tests

Optional: AI features (API key required)

The Quiz and Study Assistant pages can use an LLM to generate better questions and synthesized answers:

Enter your API key in the sidebar on those pages. Without a key, both pages still work in offline mode using your uploaded papers.

Tech stack

Layer Technology
UI Streamlit
PDF parsing PyMuPDF (fitz)
Embeddings sentence-transformers / scikit-learn TF-IDF
Vector search FAISS / NumPy fallback
Clustering scikit-learn AgglomerativeClustering
Charts Plotly
Database SQLite (built-in, no setup needed)
LLM Groq / OpenAI (optional)

Disclaimer

QPredict''s Historical Priority Score is based on patterns in your uploaded papers only. It is not a prediction or guarantee of future exam content. All topics in your syllabus remain important.

Built With

Share this project:

Updates