Recallly

AI-powered voice notes for field professionals
Record. Transcribe. Extract. Done.

---

About

Recallly is a native Android app built for Sales Reps, Field Engineers, and Insurance Adjusters who spend their day in the field and hate manual CRM data entry.

Tap the mic, talk naturally about your visit, and Recallly handles the rest — transcribing your voice note and extracting structured CRM fields using AI. Everything works offline-first with no cloud database dependency.

Key Features

Voice Recording & Dual-Mode Transcription

• Online mode — Android SpeechRecognizer with real-time partial results
• Offline mode — On-device whisper.cpp via JNI (no internet required)
• Automatic mode selection based on network connectivity
• Silence detection for hands-free auto-stop

AI Field Extraction

• Gemini 2.5 Flash extracts structured fields from transcripts (company name, contact info, next steps, deal value, etc.)
• Persona-specific prompts tailor extraction to your role
• Background extraction via WorkManager — close the app and results appear later
• Custom fields — define your own fields and they're included in extraction

Smart Onboarding

• Choose your persona (Sales Rep, Field Engineer, Insurance Adjuster)
• Toggle which CRM fields matter to you
• Set your work schedule and get reminder notifications
• Select your language (English, Arabic, Spanish, Turkish)

Export & Integrations

• PDF export with localized field names — share reports instantly
• Calendar integration for scheduling follow-ups from voice notes
• Full RTL support for Arabic

Offline-First Architecture

• Voice notes stored locally as JSON — no cloud sync required
• On-device Whisper model downloaded once (~142 MB), runs entirely offline
• DataStore preferences for all user settings
• Works without internet after initial setup

Tech Stack

Layer	Technology
UI	Jetpack Compose + Material 3
Architecture	Clean Architecture + MVVM + UDF
DI	Koin
Database	Room
Preferences	DataStore
Navigation	Navigation Compose (type-safe routes)
Auth	Firebase Auth + Credential Manager
AI	Google Generative AI (Gemini 2.5 Flash)
Speech	whisper.cpp (C/C++ via JNI) + Android SpeechRecognizer
Background	WorkManager
Billing	Google Play Billing
Language	Kotlin 2.3.10, 100% Kotlin

Building

Prerequisites

• Android Studio Ladybug or later
• JDK 11+
• NDK r26.3 (26.3.11579264) — install via SDK Manager
• CMake 3.22.1+

Setup

1. Clone the repository

   git clone https://github.com/7pak/Recallly.git
   cd Recallly
   git submodule update --init --recursive

2. Create local.properties in the project root with your keys:

   WEB_CLIENT_ID=your-firebase-oauth-client-id
   GEMINI_API_KEY=your-gemini-api-key

3. Add google-services.json from your Firebase project to app/.

4. Build

   ./gradlew assembleDebug

   On Windows, use gradlew.bat instead of ./gradlew.

Optional Build Config

These keys have working defaults (test IDs) but can be overridden in local.properties:

ADMOB_APP_ID=your-admob-app-id
ADMOB_REWARDED_PRE_RECORD_ID=your-ad-unit-id
ADMOB_REWARDED_POST_SAVE_ID=your-ad-unit-id
BILLING_SUBSCRIPTION_ID=your-subscription-product-id

Architecture

com.at.recallly/
├── core/           # DI, theme, utilities, Result<T>
├── data/           # Repositories, Room, DataStore, Whisper, Gemini, Billing, Ads
├── domain/         # Models, repository interfaces, use cases (pure Kotlin)
└── presentation/   # Compose screens, ViewModels, navigation

The app follows Clean Architecture with strict layer separation. The domain layer has zero Android dependencies. ViewModels expose a single StateFlow<UiState> and accept sealed UiEvent interfaces for unidirectional data flow.

See CLAUDE.md for detailed architectural documentation.

Supported Languages

Language	Code	Direction
English	en	LTR
Arabic	ar	RTL
Spanish	es	LTR
Turkish	tr	LTR

License

All rights reserved. This source code is provided for reference and educational purposes only. See the LICENSE file for details.

---

Built with Kotlin and Jetpack Compose