An advanced AI-powered system for analyzing Gujarat Technological University (GTU) Previous Year Questions (PYQs) to help students optimize their exam preparation strategy.

Live demo: https://gtu-pyqs-analyzer-cr7kag2ajmczjypucokxly.streamlit.app/

About this project

GTU PYQs Analyzer is a compact Streamlit app that processes Gujarat Technological University previous-year question PDFs, extracts and analyzes questions, and produces AI-powered study recommendations and a downloadable analysis PDF.

Technologies used

• Python 3.8+
• Streamlit (UI)
• PyMuPDF (fitz) for PDF text extraction
• NLTK for tokenization/stopwords
• scikit-learn for clustering and vectorization
• pandas, numpy for data handling
• reportlab for PDF generation
• plotly / matplotlib / seaborn for visualizations (Which is not on app, but in code)
• wordcloud, textstat, networkx for auxiliary analysis
• requests for API calls

Deployed website

https://gtu-pyqs-analyzer-cr7kag2ajmczjypucokxly.streamlit.app/

Code pipeline (high level)

1. User uploads one or more PDF files via the Streamlit UI (main.py).
2. src/pdf_processor.py extracts raw text and metadata from PDFs (PyMuPDF).
3. src/question_processor.py segments the raw text into individual questions and normalizes them.
4. src/nlp_analyzer.py runs NLP steps: keyword extraction, optional semantic clustering, frequency counts, and complexity/readability scoring.
5. src/llm_analyzer.py sends structured prompts + context to OpenRouter (LLM) and receives a comprehensive_analysis text and a personalized study_plan.
6. src/visualizer.py creates charts and visuals for the UI; src/utils.py contains helpers including create_pdf_from_text() which converts the analysis into a styled PDF.
7. Results are cached in AnalysisCache (session state) to speed repeated runs; users can download the analysis PDF from the results page.

How the website works (user & backend flow)

User flow:

• Open the deployed site and enter your OpenRouter API key (or set via Streamlit secrets).
• Upload PDF question papers for a single subject.
• Click "Start Advanced Analysis" to run the pipeline.
• View the analysis across section tabs (frequency, topics, question types, difficulty, recommendations).
• Download the full analysis as a PDF.

Backend flow:

• The app processes uploaded PDFs one-by-one, extracts text, and segments questions.
• Basic NLP (keywords, frequency, clustering) runs locally to produce summaries and context.
• A final comprehensive analysis and a personalized study plan are generated by an LLM via OpenRouter using the extracted questions and NLP context.
• The final text is parsed and displayed in tabbed sections; the PDF generator formats headings, lists, and inline bold/italic to preserve readability.

About OpenRouter

OpenRouter is a gateway/marketplace that provides access to LLM models (e.g., GPT-family and others) via a single API endpoint. This app uses OpenRouter as the LLM backend to generate the comprehensive_analysis and the study_plan.

Notes about using OpenRouter with this app:

• You must supply an OPENROUTER_API_KEY (stored in Streamlit secrets or as an environment variable).
• API usage may incur costs and rate limits; keep credentials private and monitor usage in your OpenRouter dashboard.
• The app sends only the minimal necessary context (extracted questions and NLP summaries) to the LLM to keep prompts focused and reduce token usage.

---

If you want this README expanded with screenshots or a short architecture diagram, tell me where to add them and I will update the file.

Key Features (as deployed)

• Upload multiple PDF question papers (same subject recommended)
• Automatic question extraction and basic preprocessing
• NLP analysis: key topics/keywords, optional semantic clustering, frequency and complexity summaries
• AI analysis via OpenRouter: comprehensive analysis text and a personalized study plan
• Download full analysis as a styled PDF (Markdown-like headings, bold/italic, bullets)
• Session caching to speed up repeated analysis runs

Quick Start (local)

1. Clone the repo

git clone https://github.com/yourusername/gtu-pyqs-analyzer.git
cd gtu-pyqs-analyzer

2. Install dependencies

pip install -r requirements.txt

3. (Optional) Add your OpenRouter API key locally for testing: Create .streamlit/secrets.toml (local only, do not commit):

OPENROUTER_API_KEY = "your_openrouter_api_key_here"

4. Run the app

streamlit run main.py

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

Deploy (Streamlit Community Cloud)

1. Push this repository to GitHub.
2. Go to https://streamlit.io/cloud and create a new app pointing to this repo and branch.
3. In the app settings, add OPENROUTER_API_KEY in Secrets.

The app reads the API key from st.secrets or the OPENROUTER_API_KEY environment variable.

Removed / Not included in the deployed app

• No public JSON/TXT download buttons are exposed on the site (PDF export only)
• No multi-LLM UI (app uses OpenRouter as the LLM back-end)
• Advanced deployment/CI examples were removed from this README — contact the maintainer if you need a Docker/GitHub Actions setup.

Project layout

• main.py — Streamlit application entry
• config/ — configuration constants
• src/ — core modules: PDF extraction, question processing, NLP, LLM integration, utilities
• assets/ — static assets (styles, logo)

Need a change?

If the live app is missing something you rely on, tell me which features or downloads you want visible and I will update the app and README to match.

---

• Caching: Store results for faster repeat analysis