Built as a browser-based academic assistant, Bard helps students ask better questions, bring in external context, generate quizzes, and turn task lists into practical daily study plans.
Open [+]Problem
Students often split revision across chat tools, search tabs, PDFs, task lists, and quiz apps, which makes focused study harder to sustain.
Approach
Bard organizes the workflow into QA, Web, Todo, and Quiz modes so learning, planning, source context, and revision stay connected.
Outcome
A lightweight, local-first full-stack study tool with modular vanilla JavaScript, serverless API routes, PWA caching, and practical export utilities.
System Notes
01 / Learning Modes
QA mode keeps persistent markdown-friendly conversations. Web mode searches Wikipedia, previews sources, and injects selected content into AI context. Todo mode turns timed tasks into an AI-prioritized day plan. Quiz mode generates MCQs from notes, PDFs, and OCR images for focused practice.
02 / Architecture
The vanilla JavaScript UI is split into feature modules for chat, web search, todos, quizzes, sidebar, dock, PDF handling, and shared interaction patterns.
03 / AI Layer
API routes handle AI prompting, planning, MCQ generation, and content retrieval through an OpenAI-compatible SDK configuration pointed at the DeepSeek endpoint.
04 / Persistence
Chat history, quizzes, preferences, and tasks are stored in browser storage for speed and continuity, with optional blob-backed persistence utilities. The service worker caches shell assets for repeat use while excluding live API responses to avoid stale AI output.
