Bhagavad Gita – The Sacred Verses

Screenshots

Role Developer

Platform Android · iOS (TestFlight)

Built with Flutter, Dart

Status Live on Play Store · iOS TestFlight

The Problem

Many Bhagavad Gita apps are cluttered with ads, complicated navigation, and poor readability. Users seeking spiritual guidance deserve a distraction-free experience that respects the sacred nature of the text while being accessible on modern devices.

The Approach

Engineered an offline-first mobile app using Flutter with a clean-architecture layered design. All 700 verses ship embedded in a pre-seeded SQLite database — the entire reading experience works without a single network call. FTS5 indexing enables full-text search across all verses and translations in under 50ms. The latest update brings a refined UI, AI-powered conversations with Lord Krishna, ambience mode for a spiritual atmosphere, brief explainers for complex verses, expanded language support, private reflections, and richer chapter summaries.

System Architecture

Offline-first architecture with three distinct layers. The presentation layer handles UI state via Flutter's provider pattern, the domain layer encapsulates all business logic (verse lookups, bookmarking, search indexing), and the data layer manages local persistence and optional API sync.

Presentation

Flutter UI + Provider

Refined UI, ambience mode, multi-language renderer, AI chat interface

↕

Domain

Business Logic

Verse service, search indexer, bookmark manager, reflection manager, AI conversation engine

↕

Data

SQLite + Shared Prefs

700 verses, multi-language translations, bookmarks, reflections, preferences

External

Auth + AI Services

Authentication, Talk with Krishna AI, crash reporting via Supabase

Data Design

All 700 verses ship embedded in the APK via a pre-seeded SQLite database — zero network dependency for core content. Translations are stored as separate columns per language rather than normalized rows, trading storage for read speed since users read far more than they write.

verses

id INTEGER PK
chapter INTEGER
verse_num INTEGER
sanskrit TEXT
transliteration TEXT
lang_en TEXT
lang_hi TEXT
lang_gu TEXT

bookmarks

id INTEGER PK
verse_id FK → verses
created_at TIMESTAMP
note TEXT NULL

reflections

id INTEGER PK
verse_id FK → verses
content TEXT
created_at TIMESTAMP

preferences

language STRING
font_size FLOAT
theme STRING
reminder_time TIME NULL
last_read INTEGER

Technical Decisions & Tradeoffs

Flutter vs Native

Chose: Flutter (single codebase)

Shipping to Android first but architected for iOS expansion. Flutter's widget system enabled custom typography controls for Sanskrit rendering without platform-specific code.

Alternative: Kotlin-only — better Play Store integration, but doubles effort for iOS later.

Embedded DB vs API-first

Chose: Pre-seeded SQLite, API optional

Spiritual reading apps must work offline — temples, retreats, and rural areas often lack connectivity. Embedding all verses in the APK adds ~2 MB but removes the network dependency entirely.

Alternative: Remote API — smaller APK, but fragile UX when offline.

Denormalized Translations

Chose: One row per verse, columns per language

Read-heavy workload (users read verses thousands of times, never write). Avoids JOINs on every verse load. Five language columns add negligible disk overhead for 700 rows.

Alternative: Normalized language table — cleaner schema, but slower reads for the primary use case.

Local vs Push Notifications

Chose: Local scheduled notifications

Daily verse reminders don't need a server. Local scheduling via flutter_local_notifications avoids running a push infrastructure for a simple cron-like feature.

Alternative: FCM push — server-controlled, but overkill for user-set daily reminders.

Performance & Optimization

Designed for low-end Android devices. Lazy-loads chapter content, caches rendered widgets, and defers audio loading until playback is requested. Search uses a pre-built FTS5 index so full-text queries across all 700 verses complete in under 50ms.

<50ms

Full-Text Search

FTS5 index across all verses and translations

~80 MB

Total App Size

Embedded DB + assets + podcasts, optimized with tree shaking

0

Network Calls for Core

100% offline reading, audio streams on demand

What It Does

Full 700 Verses

Talk with Krishna (AI)

Ambience Mode

Multi-Language

Authentication

Brief Explainers

Chapter Summaries

Private Reflections

Dharma Scenarios

Sacred Art

Natural Search

Refined UI

Podcasts

Daily Reflection Sheet

What I Learned

Building an offline-first system with AI features forced me to think differently about data architecture — every byte matters when it ships in the APK, and every AI interaction needs to feel contextual, not generic. The decision to denormalize translations into columns instead of rows cut verse-load time significantly for the 99% read-heavy use case. Adding Talk with Krishna taught me how to design AI conversations that feel meaningful within a spiritual context. This project taught me that the best engineering decisions come from deeply understanding how users actually interact with the product, not from following schema textbooks.

What's Next

iOS Release

Available via TestFlight

The iOS version is now available for testing through Apple TestFlight — download it and experience the app before the full App Store release.

Adaptive Layouts

Tablets & larger screens

Expanding the reading experience to tablets and foldables with adaptive layouts — multi-pane views for chapter navigation alongside verse reading, optimized for the screen real estate that larger devices offer.

Now Available

Live on Google Play Store · iOS TestFlight

View on Play Store Download on TestFlight