This project is built on a single core principle: user-controlled, privacy-first architecture. All user data is stored exclusively in local IndexedDB within the browser—no login, no cloud sync, no tracking, and no server-held data. The only data to ever leave the device is via explicit user-triggered LLM queries or manual export.
Core Problem
Lack of User Control and Data Privacy
Modern AI tools often require cloud sync and server-side data storage, creating privacy risks and vendor lock-in for users' personal or sensitive information.
Fragmented, Single-Purpose Tools
Users often need to switch between multiple specialized AI applications (e.g., a document analyzer, a writing coach, a research tool), leading to a disjointed workflow.
Poor Data Portability
Migrating data, chat history, and context between devices or services is often impossible, forcing users to start from scratch or remain locked into a single ecosystem.
Strategic Objectives
- Data is stored locally in IndexedDB by default; document backup/export options and device-loss implications.
- Eliminate the need for user accounts, logins, or any form of cloud synchronization.
- Ensure the only data that leaves the device is through explicit user actions like LLM queries or manual data export.
Key Initiatives
- Features six distinct coaching modes (Finance, Health, etc.) with dynamic questionnaires.
- Generates tailored documents and funnels user-provided context directly into a domain-scoped chat agent.
- Demonstrates how to build a versatile, context-aware AI assistant on the front end.
- Allows selection from over 12 analytic modules (e.g., SWOT, Market Analysis) to generate structured reports.
- Uses a licensed, privacy-compliant API for real-time web search with strict source prioritization and citation.
- Exports fully formatted documents with all source links included for traceability.
- Users play against a competent AI that provides real-time analysis and coaching from its perspective after every move.
- Integrates a contextual chat for move rationale, rules clarification, and post-game analysis.
- Uses a resilient fallback model system (Gemini/GPT-4o) to ensure consistent availability.
- Enables secure, no-server upload and processing of .zip files directly into the browser's IndexedDB.
- Automatically summarizes, keywords, and maps each document to a persistent, locally stored chat history.
- Strictly scopes the LLM to the extracted document content to ensure verifiable, grounded answers.
- Implements hands-free, voice-to-voice interaction with silent-detection across all modules.
- Functions as a fully installable Progressive Web App (PWA) for a native-like experience on desktop and mobile.
- Ensures the entire application, including all its complex features, works seamlessly offline.
Executive Summary
AIDemo.tech is a personal project built in Google Firebase Studio that serves as a multi-modal AI toolkit designed with a strict privacy-first architecture. It offers a suite of tools—including multi-mode AI coaching, a business report generator, a chess coach, and a secure document manager—all of which run entirely in the browser using IndexedDB for storage. This ensures no personal data is ever stored on a server, giving the user complete control. The platform demonstrates how to leverage modern AI (Gemini and OpenAI) and web technologies (PWAs, voice-to-voice) to build powerful, user-centric applications without compromising on privacy or data portability.
- ➤Privacy-First by Design: All data lives in the browser's IndexedDB, ensuring zero server-side data storage.
- ➤User-Controlled Portability: A global export/import feature allows users to back up and move their entire profile as a single file.
- ➤Frictionless UX: Features responsive, installable PWA behavior with integrated voice-to-voice interfaces for hands-free operation.