Study Canvas AI

Transform raw PDFs and handwritten notes into structured study guides, interactive quizzes, and generates custom visual aids instantly.

Comment

Inspiration

Students are constantly overwhelmed by dense textbooks, messy lecture slides, and disorganized hand-drawn notes. While traditional AI tools can summarize text, they often miss the bigger picture—failing to map out how complex ideas visually connect or testing whether a student actually understands the material.

We wanted to create an all-in-one AI revision companion that bridges the gap between raw, unstructured study material and active visual learning. StudyCanvas AI was born to turn cluttered screenshots, PDFs, and lecture diagrams into structured revision packs in just seconds.

What it does

StudyCanvas AI transforms raw educational material into interactive, high-yield study assets calibrated to your selected difficulty level (School, College, or University):

  • Smart OCR Text Extraction: Accepts PDFs, scanned documents, and images (PNG, JPG, JFIF). It extracts text natively or leverages multi-modal vision AI for image-based materials.
  • Visual Concept Flowcharts: Automatically converts complex topics into clean, readable logical diagrams rendered via Kroki (Mermaid.js).
  • High-Yield Summaries & Misconception Tracking: Distills lengthy documents into key takeaway points while identifying and correcting common mistakes.
  • Interactive Self-Assessment Quizzes: Generates multiple-choice quizzes tailored to the student's academic level with real-time feedback.

How we built it

We engineered StudyCanvas AI using a full-stack, pure-Python approach built on Reflex:

  • Frontend & Backend UI: Developed entirely in Python using Reflex for real-time, responsive UI updates and state management.
  • OCR & Document Processing: Powered by PyPDF2 for fast native extraction, backed by Google Gemini Vision API for processing scanned documents and screenshots.
  • Resilient Multi-Model LLM Engine: Built a fallback pipeline routing primary generation through Gemini (gemini-3.8-flash with cascading fallback to gemini-2.5-flash) and Groq (llama-3.1-8b-instant) to eliminate downtime and rate-limit bottlenecks.
  • Diagram Rendering: Generates raw Mermaid.js diagram syntax via LLMs and renders crisp, clear PNG flowcharts through the open-source Kroki API.
  • Parallel Execution: Leveraged Python's ThreadPoolExecutor to handle concurrent multi-file processing seamlessly.

Challenges we ran into

  • Handling Mixed File Inputs: Distinguishing between highlightable text PDFs and flat image-based PDFs required setting up dynamic extraction fallbacks that seamlessly switch between local parsing and vision models.
  • API Rate Limits & Server Demand: High server demand (503 errors) during testing pushed us to architect a robust multi-key, multi-provider API cascading engine with automatic model degradation handling.
  • Cloud Deployment & Secret Management: Managing environment variable injection across cloud containers on Reflex Cloud required careful orchestration of cloud build hooks and state re-initialization.

Accomplishments that we're proud of

  • Zero-Downtime Reliability: Designed an API engine that gracefully recovers from individual model failures or rate limits without crashing the user session.
  • Pure Python Stack: Built a seamless, highly reactive web application with interactive UI components purely in Python without writing custom JavaScript.
  • Accurate Diagramming: Replaced unreliable AI image diffusion with deterministic Mermaid.js code generation rendered cleanly via Kroki.

What we learned

  • Designing Resilient AI Applications: Building production-grade AI tools requires designing defensive backend pipelines with multi-key pools, automatic model fallbacks, and strict concurrency locks.
  • Reflex Architecture: Mastered managing state synchronization, async event handlers, and web-socket locks across modern full-stack Python architectures.

What's next for Study Canvas AI

  • Audio & Video Summarization: Expanding input options to accept recorded lectures or YouTube links.
  • Exportable Flashcard Decks: Adding one-click exports to Anki and Quizlet formats.
  • Collaborative Study Rooms: Enabling real-time multi-user canvas sessions for group study and revision.

Built With

Share this project:

Updates

Submission history