User Manual

The full user manual is hosted in the project README on GitHub, covering navigation, material generation, session management, and classroom controls.

Deployment Manual

Full setup instructions are available in the Getting Started section of the project README on GitHub. A summary of prerequisites and quick-start steps is included below.

Prerequisites

GDPR and Data Privacy

Manuscripta is designed with privacy by design as a foundational principle. The system operates entirely over a local area network with no cloud connectivity, meaning no student data is ever transmitted to external servers or third parties.

Key GDPR Principles Addressed

1. Lawfulness, Fairness and Transparency

The system does not collect personally identifiable student data. Students are identified only by anonymised device IDs that cannot be traced back to an individual. Teachers are informed through the application interface that all data remains on the local network. This approach was confirmed as a requirement by project partners including the National Autistic Society and is consistent with UK GDPR Article 5.

2. Purpose Limitation

Data processed by Manuscripta is used solely for the purpose of delivering educational materials and monitoring classroom engagement during a live session. No data is repurposed, shared or used for any secondary purpose including model training or analytics.

3. Data Minimisation

The system collects only what is necessary for classroom operation:

Data Type	Purpose	Identifiable?
Anonymised device ID	Identifies which tablet is which	No
Device status (on task, needs help, disconnected)	Real-time classroom monitoring	No
Student responses to quizzes and polls	Teacher marking and feedback	No
AI-generated material content	Educational delivery	No

No names, photographs, ages or any personally identifiable information are stored at any point.

4. Storage Limitation

Session data is held in memory during the lesson and cleared when the session ends. Material content saved to the teacher's library is stored in a local SQLite database on the teacher's laptop only. No data is retained on student devices beyond the active session.

5. Integrity and Confidentiality

All data is processed on the teacher's laptop within the school's local network. No internet connection is required at runtime. AI inference runs locally via Ollama using the Qwen3 and IBM Granite models. No student responses are sent to external AI providers.

6. Accountability

The teacher retains full control over all data at all times. No student response reaches a student as feedback until the teacher has reviewed and approved it. This oversight mechanism ensures the teacher remains the data controller throughout the session.

Privacy Safeguards

Data Handling

• No Cloud Processing: Manuscripta operates entirely within the school's local Wi-Fi network. No data leaves the school premises at any point during normal operation.
• Local Storage: All persistent data including materials and session records is stored in a SQLite database on the teacher's Windows laptop. No server-side storage or third-party database is used.
• AI and Privacy: The on-device AI models process teacher-provided content only. Student responses are never passed to the AI without teacher initiation and review.

User Rights

Teachers can delete any material or session data at any time through the application interface. No student login is required and no student account is created, meaning there is no personal data subject to access or deletion requests under UK GDPR.

Legal Contact

For data protection or legal enquiries related to this project:

• University Supervisor: Professor Dean Mohamedally (d.mohamedally@ucl.ac.uk)
• Module: COMP0016 Systems Engineering, UCL Computer Science
• Project Period: October 2025 to March 2026

This privacy statement has been prepared by the student project team to the best of our knowledge. We are not legal professionals. Schools deploying Manuscripta should seek independent legal advice to ensure compliance with their own data protection obligations.

Source Code License

Manuscripta is released under the MIT License. The full license text is in the Licence section of the project README on GitHub.

Note that the Windows application's material editor modal includes various icons from Google's official icon set "Material Symbols". Use of these icons is licensed under the Apache License, Version 2.0.

Monthly Videos

December Update

In December, we made significant progress connecting the front-end prototype to the .NET back-end, introducing persistent data storage and a rich text material editor that supports LaTeX equations and PDFs. We successfully deployed the interactive student interface to an e-ink tablet for quizzes and viewing materials. Our next priorities include detailing LLM integration, designing the Retrieval-Augmented Generation (RAG) pipeline, and evaluating compatible on-device AI models.

January Update

In January, we introduced our new rich text editor to the teacher portal, which supports formatting, tables, KaTeX equations, and embedded PDFs. On the student side, we deployed our interactive UI onto an e-ink tablet, allowing students to view tasks and answer questions in a distraction-free environment. Moving forward, we are focusing on LLM integration for AI generation and marking, designing our RAG pipeline, and benchmarking on-device AI models on Intel and Qualcomm hardware.

February Update

In February, we focused on website design, PDF rendering, and hardware integration. We updated the styling of the Manuscripta website to match the design of our Windows application to maintain visual consistency. We improved core material distribution by implementing native PDF rendering, allowing teachers to export any lesson material as a PDF document for traditional classroom use. Finally, we successfully integrated the platform with reMarkable tablets, enabling the deployment of questions and embedded PDFs directly to paired reMarkable devices.

Project Presentation

In our final project presentation, we showcase Manuscripta's complete classroom orchestration platform, bridging the gap between distraction-free e-ink tablets and real-time teacher control over a local network. The walkthrough highlights our four core requirements in action: generating lesson materials and marking schemes using on-device LLMs (Qwen and IBM Granite), distributing content to student devices via our .NET teacher dashboard, monitoring live progress, and returning immediate or manual feedback. We also detail our clean architecture and extended support for reMarkable and Kindle tablets.

External Resources