<<<<<<< HEAD
Let's be honest: Engineering syllabuses are huge, and old question papers (PYQs) are usually just messy PDFs or blurry photos. StudyBuddy AI is a smart tool I built to help students stop wasting time and start studying what actually matters.
It uses Gemini 2.5 Flash to "read" your documents and tell you exactly what to focus on.
Upload a photo of your syllabus, and the AI writes detailed notes for every unit. It even marks topics as [HIGH PRIORITY] so you know what's important.
Feed it your Previous Year Questions (PYQs). It finds patterns and predicts which 10-mark questions are most likely to come this year.
Automatically creates a 1-page revision guide for last-minute study.
Generates a 5-question MCQ quiz after every analysis to see if you actually understood the topic.
It can read text from clear photos, not just clean PDFs.
=======
The most advanced Multimodal AI Agent for engineering students. Most AI tools just "read" a file and forget it. StudyBuddy AI Pro is different. It builds a Persistent Knowledge Base from your PDFs and images using a Vector Database, allowing you to have a real-time, context-aware conversation with your entire syllabus.
Vectorized Knowledge Ingestion: Uses ChromaDB to embed and store document data as mathematical vectors for high-speed semantic retrieval.
Multimodal Intelligence: Powered by Gemini 2.5 Flash to analyze everything from clean PDFs to blurry handwritten exam papers.
Context-Aware Chatbot: An interactive tutor with Session Memory that remembers your previous doubts and provides "Genuine Help" based only on your uploaded materials.
Exam Pattern Analytics: Identifies high-frequency topics in Previous Year Questions (PYQs) and forecasts likely 10-mark questions using Chain-of-Thought reasoning.
Automated Academic Synthesis: Generates unit-wise notes, last-minute cheat sheets, and active-recall quizzes instantly.
I architected this using a modern AI Agent workflow:
f4da2b1 (chore: save local changes before rebase)
Language: Python 3.10+
<<<<<<< HEAD
- Python: The main language for all the logic.
- Google Gemini 2.5 Flash: The "Brain" that reads the images and understands the engineering concepts.
- LangChain: The bridge that connects the AI to our app's specific goals.
- RAG (In-Memory): A technique where the AI "retrieves" the specific text from your file to give accurate answers instead of just guessing.
- Streamlit: Used for the modern "Dark Mode" website interface.
- PyPDF2 & Pillow: To help the AI "see" and "read" PDFs and Images. ======= Orchestration: LangChain (LCEL) for building stable, modular retrieval pipelines.
Vector Database: ChromaDB for efficient document indexing and semantic search.
LLM & Embeddings: Google Gemini 2.5 Flash & text-embedding-004 for high-reasoning capabilities.
UI/UX: Streamlit with a custom Glassmorphism Dark Theme and responsive state management.
Data Processing: PyPDF2 for complex PDF parsing and Pillow for OCR-based image processing.
Unlike standard LLMs that "hallucinate," this project uses Retrieval-Augmented Generation.
Chunking: Documents are broken into 1,000-character segments with overlap.
Embedding: Text is converted into vectors using Google’s embedding models.
Retrieval: When a student asks a doubt, the system finds the exact 3 most relevant segments in the database to formulate an answer.
f4da2b1 (chore: save local changes before rebase)