Academic Oracle is a learning-focused AI platform designed to maximize understanding, not passive consumption.
Instead of immediately giving answers, Academic Oracle follows a scientifically grounded flow:
Ask → Think → Hint → Attempt → Feedback → Pattern → Insight → Mastery
The goal is not memorization — it’s deep, durable learning.
Most AI tools optimize for speed.
Academic Oracle optimizes for retention, intuition, and reasoning.
- Active recall before answers
- Progressive hinting instead of instant solutions
- Error-correction loops
- Pattern discovery over rote explanation
- Minimal UI disruption to maintain cognitive flow
You don’t just learn faster — you learn properly.
- Hint-based reasoning flow (Ask first, reveal progressively)
- Structured thinking prompts
- Pattern extraction instead of answer dumping
- Follow-up suggestion system (NEW v2.3.0)
- Context-aware follow-up buttons appear on text selection
- Enables deeper exploration without breaking learning flow
- Reduces friction between curiosity → action
- Auto-generated concept-specific quizzes
- Multi-question adaptive testing
- Mastery popups & performance feedback
- Reinforcement-based correction
- Mid-session language switching
- Unified Chat + Quiz UI system
- Multi-mode execution pipeline:
- Standard
- Fast
- Balanced
- Agentic
- Web Search
- Real-time system state visibility via Loading Status Text Bar
- Displays current processing stage
- Improves transparency of AI behavior
- Dynamic routing based on:
- Query complexity
- Latency conditions
- System load
Academic Oracle doesn’t just respond — it decides how to think first.
- JigsawStack-powered search pipeline
- Designed for:
- Real-time knowledge retrieval
- SPA / dynamic site parsing
- Activated only when needed (cost-efficient routing)
- Hybrid reasoning:
- AI + live data synthesis
- All AI API calls moved to Supabase backend
- No direct client exposure of sensitive keys
- Production-grade architecture
- Secure Edge Function orchestration
- AES-GCM-256 encryption for sensitive data
- Supabase-backed session continuity
- Jailbreak detection & filtering system
- Prompt sanitation before model execution
- Controlled response shaping to prevent misuse
Academic Oracle is no longer just a frontend AI tool —
it is a secured distributed AI system.
- Robust Markdown rendering
- Math (KaTeX)
- Tables
- Code blocks
- Dark / Light mode
- Responsive design (desktop & mobile) <<<<<<< HEAD
- Fail-in-console architecture (UI never crashes)
- AES-GCM-256 encryption for sensitive keys
- Supabase-backed session continuity
- Arcade-style interactive onboarding demo
-
Dynamic model routing based on real-time traffic conditions <<<<<<< HEAD
-
Promise-race orchestration during high-load periods
-
Automatic fallback to most efficient single-model pipeline under normal conditions
-
Intelligent compatibility matching per user request type
-
Stepfun-3.5 integration as high-load inference offloader
=======
- Promise-race orchestration during high-load periods
- Automatic fallback to most efficient single-model pipeline under normal conditions
- Intelligent compatibility matching per user request type
- Stepfun-3.5 integration as high-load inference offloader
refs/rewritten/merge-resolve-README-conflict
- Automatic cancellation of non-winning model responses to preserve cost efficiency
Academic Oracle adapts not just to learners - but to system conditions.
- Frontend: React 19 + TypeScript
- Backend (AI): Google GenAI models (Gemini-3, Gemini-2.5) and StepFun-3.5
- Backend (Auth): Supabase & Google OAuth
- Build Tool: Vite 6
- Styling: Tailwind CSS
- Math Rendering: KaTeX
- State & UX: Custom lightweight logic (no heavy frameworks)
- Security: AES-GCM-256 encryption for sensitive keys
- AI Provider: Gemini API (user-supplied key) =======
- Non-blocking UI architecture
- Failures never crash the interface
- Graceful degradation on errors
4a83297 (change: README)
- Node.js (v18+ recommended)
- Install dependencies:
npm install- Setup your Supabase project
- Create database
- Configure auth
- Deploy Edge Functions
- Configure environment variables:
VITE_SUPABASE_URL=YOUR_SUPABASE_URL
VITE_SUPABASE_ANON_KEY=YOUR_SUPABASE_ANON_KEY
# Only public key needed (web search)
VITE_JIGSAWSTACK_KEY=YOUR_JIGSAWSTACK_API_KEY❌ Removed:
- VITE_GEMINI_KEYS
- VITE_STEPFUN_KEY
✅ All AI provider keys are now handled securely in Supabase backend
- Start development server:
npm run dev- Frontend: React 19 + TypeScript
- Backend (AI Orchestration): Supabase Edge Functions
- Models:
- Google GenAI (Gemini-3, Gemini-2.5)
- Stepfun-3.5 (fallback / high-load routing)
- Web Search: JigsawStack API
- Auth & Database: Supabase (Postgres + OAuth)
- Build Tool: Vite 6
- Styling: Tailwind CSS
- Math Rendering: KaTeX
v2.3.0 marks a major shift:
From: Client-heavy AI calls
→ To: Backend-controlled AI orchestration
This enables:
- Real production deployment
- Key security
- Scalable request routing
- Advanced filtering & control layers
Academic Oracle aims to redefine how AI integrates into education:
- Not as a solver.
- Not as a shortcut.
But as a structured reasoning partner.
The long-term goal is to build a universal academic cognition system that scales from secondary education to research-level inquiry.
Academic Oracle was designed and built by Vo Tan Binh.
This project represents original work in:
- Learning-science–driven AI interaction design
- Progressive reasoning and hint-based pedagogy
- Closed-feedback AI tutoring systems
- Secure, minimal, and distraction-free educational UX
If you build upon this work, attribution is appreciated.
If Academic Oracle helps your learning:
-
⭐ Star the repository
-
☕ Support via Buy Me a Coffee
-
🧠 Use it, break it, and learn from it
Recognition matters. Impact matters more.