remember-all-built-in-ai

Use the Chrome built-in-ai to test your retention of information and to predict when to be tested again.

Comment

Inspiration

The primary inspiration for "Remember-All" comes from the universal challenge of retaining information over time, specifically the effective, data-driven principles of spaced repetition systems (SRS) like Anki and SuperMemo. Our goal was to create a modern, highly effective web application that makes spaced repetition accessible, runs entirely client-side for maximum privacy and portability, and leverages local AI for content generation. We built a tool that guides users to "remember it all" without relying on cloud services or external databases.

What it does

"Remember-All" is a powerful, client-side-only spaced repetition learning application designed to help users efficiently memorize and retain information over the long term, running entirely in the browser.

Key functionalities include:

  • Intelligent Spaced Repetition: It employs a robust algorithm that schedules reviews of learned material at optimally increasing intervals, based on the user's recall performance (Again, Hard, Good, Easy).
  • Offline LLM Integration: Uses the Chrome Built-in AI Prompt API to generate the "Answer" or "Long Description" for a flashcard based on a user-provided Hint/Content. This entire content creation process is offline-capable after the initial model download.
  • Vis.js Timeline Visualization: It uses the Vis.js library to create a highly effective, dynamic, and responsive visualization of the entire review schedule, allowing users to see their next review dates and memory curve.
  • Local Storage Persistence: The entire learning database (all flashcards, review history, and schedule data) is stored securely and privately in the user's browser Local Storage, ensuring 100% data autonomy and the ability to work offline.
  • Recall Comparison Tool: Features a simple comparison function to help the user self-score their written answer against the stored answer, contributing to the next spaced repetition interval calculation.

How we built it

"Remember-All" was developed as a powerful single-page, client-side application, using a highly efficient mix of simple code and specialized libraries:

  • HTML, Simple JavaScript, Inline CSS: These form the core structure. The code is structured to prioritize readability and transparency for teaching purposes.
  • Vis.js: The only major library used, it handles the dynamic creation and rendering of the sophisticated, interactive timeline visualization of the review schedule.
  • Chrome Built-in AI: Integrated using modern async/await JavaScript to handle the computationally intensive task of generating educational content (flashcard answers) on the user's local device.
  • Local Storage API: The native browser Local Storage API is used as the database for all flashcard data and user progress.
  • Coding Style Compliance: All core variables and functions are descriptively named with the my prefix and are in camelCase (e.g., mySaveToLocalStorage(), myGenerateAnswer()), adhering to the preferred simple coding style.

Challenges we ran into

  • Algorithm Accuracy & Client-Side Execution: Implementing a robust and highly accurate spaced repetition algorithm (like SM-2) entirely in client-side JavaScript, ensuring it calculates exponentially increasing intervals correctly across all browser sessions.
  • Content Generation Reliability: Crafting the prompt to the local LLM to consistently generate concise, accurate, and relevant flashcard answers from the hint text.
  • Data Serialization and Persistence: Reliably saving, loading, and parsing a potentially large amount of learning data (including the complex schedule history) using only the simple Local Storage API.
  • Integrating Vis.js: Properly integrating the data array with the Vis.js library to ensure the dynamic timeline accurately reflects the current state of the spaced repetition schedule.

Accomplishments that we're proud of

  • Fully Offline LLM-Powered Learning: Successfully created a sophisticated spaced repetition tool where both the data storage and the content creation (LLM generation) work entirely offline after the initial setup.
  • Visualizing the Learning Curve: The seamless integration of Vis.js to dynamically show the user's upcoming review schedule and the "exponentially increasing curve" of memory retention is a major achievement for user motivation and clarity.
  • Privacy and Portability: Delivered a powerful tool that offers complete data privacy (data never leaves the browser) and is highly portable, running on any modern Chrome browser.

What we learned

  • Client-Side Power: Reinforced the idea that highly complex, data-driven applications—including intelligent algorithms and LLM integration—can be built effectively using client-side web technologies and the Chrome Built-in AI.
  • Algorithm Tuning: Practical experience in implementing and fine-tuning the parameters of a spaced repetition algorithm for optimal user benefit.
  • Data I/O: The importance of robust serialization and deserialization routines when using Local Storage as the primary database for structured data.

What's next for Remember-All

  • Enhanced AI-Powered Testing: Exploring an AI feature that uses the local LLM to evaluate the user's typed answer for semantic accuracy, rather than just simple string comparison.
  • Advanced Content Generation Prompts: Refining the LLM prompt templates to allow users to generate different types of content (e.g., a multiple-choice question or a short summary).
  • Code Documentation: Enhancing the JavaScript with detailed comments, clearly explaining the logic of the spaced repetition algorithm and the LLM integration for teaching students.
  • Import/Export Refinement: Ensuring highly reliable export and import functionality for the data, to allow users to back up or share their learning decks.
Share this project:

Updates