StudyRAG

Inspiration

StudyRAG was born from the need to streamline chaotic study workflows. Students juggle scattered notes and documents, losing time searching for insights. We envisioned a smart, ML-driven app to unify knowledge, boost productivity, and make studying intuitive.

What it does

StudyRAG enhances an existing study app with a RAG system. Users upload docs, PDFs, or text notes to a SQLite database. Every 30 seconds, a cronjob extracts and embeds content using Gemini API, storing it in a vector database. Users can search across their knowledge base with NLP or generate concise summary for quick insights, increasing productivity by 10x.

How we built it

• Database: SQLite for storing user-uploaded docs/notes.
• Cronjob: Scheduled every 30 seconds to extract content.
• Embedding: Gemini API for text embedding and chat feature.
• Storage: Chroma Vector database for efficient retrieval.
• API Server: Built with FastAPI to handle interaction with VectorDB.
• ML Pipeline: Integrated multiple ML models for embedding and summarization.

📂 SQLite -> ⏰ Cronjob -> 🧠 Gemini API -> 📈 VectorDB -> 🔍 API

Challenges we ran into

• Syncing cronjobs with real-time uploads.
• Optimizing Gemini API for large-scale embedding.
• Balancing vector database performance with cost.
• Ensuring NLP search accuracy across diverse formats.

Accomplishments that we're proud of

• Working RAG for medium-scale notes.
• Chat with your documents!

What we learned

• NLP Techniques is crucial for efficient RAG workflow.
• Vector databases demand careful control.

What's next for StudyRAG

• Add real-time embedding for instant updates.
• Add Graph feature to makes your docs look lively.

Built With

• fastapi
• gemini
• langchain
• python
• rag
• sqlite
• vectordb