An automated, AI-powered tool that extracts student marks from unstructured report files (like PDFs and images) to generate analytical charts and presentation-ready PowerPoint decks.
Extracting marks and performance data manually from unstructured student reports is time-consuming and prone to human error. Report2Plot solves this by providing a fully automated workflow. It uses Google Gemini and Docling for intelligent data extraction and Model Context Protocol (MCP) servers to manage modular processing tasks like charting and rendering.
- Automated Data Extraction: Convert unstructured PDFs/images to structured data using Docling and Google Gemini Vision.
- Intelligent Caching (
Local_Mem): Bypasses redundant API calls for previously processed reports, drastically reducing API costs and latency. - Automated Visualization: Dynamically generates statistical charts using Matplotlib.
- Presentation Ready: Automatically builds downloadable PowerPoint (
.pptx) decks summarizing student performance. - Web UI: A simple, browser-based user interface for uploading files and tracking the processing pipeline.
- History Tracking: Keeps a log of past assignments and generated reports.
- Modular Architecture: Built on the Model Context Protocol (MCP) using
fastmcp.
Report2Plot/
├── agents/ # Core Python logic (orchestrator, docling converter, local memory manager)
├── mcp_servers/ # MCP servers (file_watcher, vision_extractor, plot_renderer)
├── UI/ # Static HTML/JS pages for the web interface
├── Input/ # Source directory for uploaded files (PDFs, images)
├── Output/ # Destination for generated charts and .pptx files
├── Local_Mem/ # Local file system cache and history storage
├── my-agent/ # Scaffolding for generated ADK agent projects
├── scan-agent/ # Scaffolding for scanning-specific ADK agent
├── tests/ # High-level workflow tests
├── run.bat # Windows batch script to launch the application
└── factory_reset.py # Utility to clear Input, Output, and Local_Mem directories
- Upload: User places report files in the
Input/folder or uploads them via the Web UI. - Document Conversion:
docling_converter.pyconverts source PDFs into structured Markdown. - AI Extraction:
mcp_servers/vision_extractor.pyuses Gemini to intelligently extract student marks and details into structured JSON. - Caching: Extracted data is cached in
Local_Mem/to prevent redundant API usage on future runs. - Rendering:
mcp_servers/plot_renderer.pytransforms the data into Matplotlib charts and a PowerPoint presentation. - Delivery: The final output is saved to the
Output/directory and presented in the UI.
- Python 3.10+ (Recommended)
- Windows OS (Due to the
run.batlauncher, though Python scripts are cross-platform) - A Google Gemini API Key (Make sure to configure it in your
.envfile)
- Clone the repository (if you haven't already).
- Configure your environment: Ensure you have an active
.envfile containing your Gemini API key (e.g.,GEMINI_API_KEY=your_key_here). - Launch the application:
Simply double-click or run
run.batfrom your terminal:This script will automatically verify Python, install missing dependencies fromrun.bat
requirements.txt, start the local orchestrator server, and open the Web UI in your default browser.
- Factory Reset: If you need to wipe all data (clear
Input/,Output/, andLocal_Mem/), you can run the factory reset script:python factory_reset.py
- Languages: Python, HTML, JavaScript
- AI & ML: Google Gemini,
google-genai - Architecture:
fastmcp(Model Context Protocol), built-in Pythonhttp.server - Document Processing:
docling,pymupdf,pillow - Data Visualization:
matplotlib - Reporting:
python-pptx
- Cloud Database: Migrate from local JSON file caching to a cloud database like Firebase or PostgreSQL.
- Broader Exports: Add support for Excel exports and analytical PDF reports.
- Advanced Analytics: Class-wide trend dashboards and predictive analytics.
