SaM (Study agentic Model) transforms standard passive learning into a proactive agentic pipeline. Drop your syllabus or textbook PDF once—SaM orchestrates study schedules, parallel flashcards, e-book summaries, and dynamic mock tests in the cloud background.
SaM replaces manual conversation prompts with autonomous background workflows designed for deep focus.
Multi-modal content ingestion supporting raw text snippets and complex PDF documents. Items are structured locally and packed intelligently within strict token boundaries.
Powered by Google Genkit, this agent parses syllabus materials in parallel. With a single trigger, it compiles Flashcards, TTS Chapters, and Mock Exams concurrently.
Reel-style vertical swipe navigation for timed or untimed mock tests. Built with active anti-cheat overlays, precision timers, and instant score evaluation telemetry.
How SaM processes high-complexity multimodal data while maintaining sub-second client responsiveness.
React Native client bundles PDF & text documents inside Shelf.
Google GenAI countTokens API evaluates token payload prior to model dispatch.
Gemini 3.5 Flash-Lite orchestrates cards, notes, and exams via schema enforcement.
Payload hydrates local UID sandbox; triggers native system push notification.
Engineered on Google enterprise infrastructure for high resilience, security, and low latency.
Powers autonomous structured flows with rigid Zod schema validation, ensuring zero hallucinated JSON structures across study resources.
Utilizes Gemini 3.5 Flash-Lite on global multi-region load balancers, completely eliminating localized 404/429 endpoint locks.
Serverless compute locked to asia-south1 (Mumbai) to minimize data latency to the database layer.
Real-time state synchronization, remote God File configurations, and transactional audit logging with atomic increment guards.
Provides zero-leak enterprise security by delivering API and cryptographic keys in-memory on demand without APK hardcoding.
Smooth cross-platform client with instant-hydration navigation, Zustand state management, and offline-first storage.
SaM enforces strict multi-tenant isolation. Your study notes, generated cards, and personal materials are cached locally using sandboxed AsyncStorage keys (@sam_local_*_${uid}). Non-critical logs, pomodoro session histories, and to-do lists never leave your device without explicit consent.
The SaM app is currently available to hackathon judges only for evaluation. The public release is not yet available.