Shiksha AI

Offline-first AI tutor app for Class 9 students, built with Expo React Native + a local-first AI/backend pipeline.

Repository Overview

• app/: Expo Router screens and navigation (index, setup-choice, model-manager, and tab routes)
• src/features/: feature modules (ai, chat, content, progress, user, auth, onboarding)
• src/core/: core services including sync and database initialization
• src/shared/: shared theme/design utilities
• backend/: Express API for tutor, OCR, translation, and dashboard stats
• android/, ios/: native mobile projects
• docs/: current developer documentation

Current Capabilities

• AI Tutor chat with Ollama-powered responses
• OCR pipeline for image-to-text extraction (English + Hindi trained data)
• Document post-processing (correct, summarize, qa, extract tasks)
• Translation support with NLLB microservice + cloud fallback
• Notes, flashcards, quiz, progress, profile/account screens
• Local-first persistence and sync hooks

Tech Stack

• Frontend: React Native 0.81 + Expo SDK 54 + Expo Router + TypeScript
• Storage: Expo SQLite + AsyncStorage
• Backend: Node.js + Express + Axios + Tesseract.js
• AI runtime: Ollama (default model: hf.co/Qwen/Qwen2.5-1.5B-Instruct-GGUF:Q4_K_M)

Quick Start

1. Install dependencies

npm install
cd backend && npm install
cd ..

2. Configure environment

Create .env in the project root:

EXPO_PUBLIC_API_URL=http://10.0.2.2:3000

Recommended backend .env (inside backend/):

PORT=3000
OLLAMA_HOST=http://localhost:11434
OLLAMA_MODEL=hf.co/Qwen/Qwen2.5-1.5B-Instruct-GGUF:Q4_K_M
TRANSLATOR_SERVICE_URL=http://localhost:3001

For physical devices, replace 10.0.2.2 with your machine LAN IP (example: http://192.168.1.25:3000).

3. Start Ollama

ollama pull hf.co/Qwen/Qwen2.5-1.5B-Instruct-GGUF:Q4_K_M
ollama serve

4. Start backend

npm run backend

5. Start app

npm start

Expo shortcuts in terminal: a (Android), i (iOS), w (web).

Scripts

Root:

• npm start: start Expo dev server
• npm run android: run Android build
• npm run ios: run iOS build
• npm run web: run web build
• npm run export-web: export web artifacts
• npm run backend: start backend (backend/server.js)

Backend (backend/package.json):

• npm start: start Express server
• npm run dev: start with nodemon

Backend API (Current)

• GET /: health check
• GET /dashboard/:userId: dashboard/progress stats
• POST /tutor: tutor responses from Ollama
• POST /process-document: OCR/document enhancement tasks
• POST /translate: translation endpoint
• POST /ocr: OCR extraction from base64 image

Project Structure

Shiksha-AI/
├── app/                  # Route screens (tabs + setup/model management)
├── src/
│   ├── features/         # Feature modules and domain logic
│   ├── core/             # Core infrastructure (db/sync)
│   └── shared/           # Theme + shared styles/utils
├── backend/              # Express backend + OCR/translation services
│   ├── db/               # Database services and stats
│   └── *.traineddata     # OCR language files
├── android/              # Android native project
├── ios/                  # iOS native project
├── assets/               # Static assets
├── scripts/              # Build/runtime helper scripts
└── docs/                 # Current + legacy docs

Documentation Map

Current docs:

• docs/GETTING_STARTED.md
• docs/PROJECT_STRUCTURE.md
• docs/ARCHITECTURE_STATUS.md
• docs/FEATURES.md

Legacy reference docs:

• docs/legacy/README.md
• docs/legacy/ARCHITECTURE.md
• docs/legacy/project_doc.md
• docs/legacy/TTS_IMPLEMENTATION.md

Security notes:

• SECURITY.md

Optional Translation Microservice

If you want to run the Python translator service locally:

pip install fastapi uvicorn transformers torch sentencepiece
cd backend
python translator_service.py

Troubleshooting

• Ollama unavailable: run ollama serve
• Missing model: run ollama pull hf.co/Qwen/Qwen2.5-1.5B-Instruct-GGUF:Q4_K_M
• Backend unreachable: ensure npm run backend is running on port 3000
• Android emulator networking: use http://10.0.2.2:3000
• Device networking: use LAN IP for EXPO_PUBLIC_API_URL (not localhost)

License

MIT