EchoVerse - A Deeply Personal AI Companion

EchoVerse is a deeply personal AI companion app that transforms your existing chat histories into an interactive "Echo"—a digital twin of a person that communicates with their unique style, humor, and personality.

This is a private repository for the development of EchoVerse, a KMM application built with a clean, modular architecture designed for scalability and the integration of advanced AI agents.

---

Core Concept & How It Works

The primary goal of EchoVerse is to experience a new form of connection by interacting with an AI that genuinely feels and sounds like a specific person, based on their real conversations. This is achieved through a sophisticated, multi-stage AI workflow:

1. Ingestion & Pre-Analysis: A user exports a chat log (e.g., from WhatsApp) and shares it with EchoVerse. A lightweight Koog agent runs to extract the names of the participants.
2. Personality Synthesis: After the user selects a participant, a powerful Koog agent performs a deep analysis of their messages. It doesn't just summarize; it generates a structured Personality Profile, capturing everything from tone and emoji frequency to signature slang and the nature of their relationship with the user.
3. Dynamic Chat Interaction: When the user chats with an Echo, a third Koog agent comes online. For every single message, it performs a "Dynamic Context Funnel":
   • It uses the static Personality Profile as its core DNA.
   • It retrieves relevant long-term memories from a local database (RAG).
   • It considers the short-term history of the current conversation.
   • It uses this rich, multi-layered context to generate a reply that is nuanced, in-character, and feels alive.

Architectural Overview

This project is MultiPlatform built on a feature-based, multi-module clean architecture. This ensures high cohesion, low coupling, and makes the codebase easy to maintain and scale.

Tech Stack & Key Libraries

• Platform: Kotlin Multiplatform Mobile (KMM)
• UI: Compose Multiplatform (for both Android & iOS)
• AI Framework: JetBrains Koog
• LLM Provider: Google Gemini (easily swappable)
• Architecture: MVVM / MVI using a modular, clean architecture approach.
• Dependency Injection: Koin
• Navigation & Lifecycle: Decompose
• Database: SQLDelight
• Networking: Ktor

Getting Started

The project is fully configured and ready to build.

Prerequisites

• Android Studio (Jellyfish or newer recommended)
• Xcode and its Command Line Tools (for building the iOS target)
• A Google AI API Key for the Gemini model.

Installation & Setup

1. Clone the repository:

   git clone https://github.com/YazanAesmael/EchoVerse

2. Get your API Key:

   • Visit the Google AI for Developers page to generate your free API key.

3. Add your API Key:

4. Build and Run:

   • Open the project in Android Studio.
   • Wait for Gradle to sync all the modules and dependencies.
   • Select the androidApp configuration and run it on an emulator or a physical device.

✅ Project Status (Current)

The full MVP loop is complete and functional on Android.

• ✅ App successfully appears in the Android share sheet.
• ✅ Handles both .txt and .zip chat exports from WhatsApp.
• ✅ Phase 1 (Birth of the Echo): The KoogChatAnalyzer successfully generates a structured, multi-part PersonalityProfile.
• ✅ Phase 2 (Living Memory): The data layer is equipped to store and retrieve both persistent chat history and long-term "episodic" memories. The chat agent is using the RAG pattern to recall context.
• ✅ Full navigation between the Home, Create Echo, and Chat screens is implemented with Decompose.
• ✅ Chatting with an Echo is fully functional.
• ✅ Full DB Integration.