Mark VII

Enterprise-grade multi-provider AI chat platform with 45+ models, Firebase cloud configuration, and modern Material 3 design.

Features • Quick Start • Download • Documentation

---

Overview

Mark VII is a multi LLMs chat application for Android that provides unified access to 45+ state-of-the-art AI models from leading providers including Anthropic, OpenAI, Meta, Deepseek, Mistral, Google, and more through a single, elegant interface.

Key Capabilities

• 🤖 45+ AI Models - Gemini, GPT, Llama, Deepseek, Mistral, and more
• 🔥 Cloud-First Architecture - Firebase-powered configuration management and real-time sync
• ⚡ High Performance - <100ms startup, connection pooling, optimized streaming
• 🔗 Dual API Support - OpenRouter (40+ models) + Direct Gemini API
• 📱 Modern Material 3 UI - Dynamic theming, smooth animations, haptic feedback
• 🌐 Multilingual TTS - Automatic language detection for 15+ languages
• 🔒 Enterprise Security - Firebase Authentication, encrypted storage, HTTPS only
• 🐍 DevOps Tools - Python CLI for bulk model management via CSV

Visuals

Usage across models from OpenRouter

Why Mark VII?

Flexibility: Switch between providers and models instantly without code changes
Reliability: Automatic error recovery, exception handling, offline capability
Performance: 24x faster than v1.x of Mark-VII, real-time streaming, optimized rendering
Developer-Friendly: Complete Firebase integration, comprehensive documentation, open source

---

Features

🤖 AI Model Access

Dual API Architecture:

• OpenRouter Integration - Direct access to 100+ models with real-time catalog sync
• Gemini API Integration - Native Google Gemini support with vision capabilities
• Seamless Switching - Toggle between APIs mid-conversation

Supported Companies:

• Google - Gemini 2.0/2.5 Flash/Pro, Gemma 2/3
• Anthropic - Claude 3.5 Sonnet, Opus, Haiku
• OpenAI - GPT-4 Turbo, GPT-4o, GPT-3.5
• Meta - Llama 3.1/3.3/4 (8B to 405B parameters)
• Deepseek - Chat V3.1, R1, R1 Distill variants
• Mistral AI - Full lineup from Small to Large
• Qwen (Alibaba) - Qwen2.5, Qwen3 Coder
• xAI - Grok with vision support
• 30+ more - Cohere, AI21, Perplexity, and others

☁️ Firebase Cloud Integration

• Remote Model Management - Add/remove models without app updates
• Dynamic API Keys - Update credentials in real-time
• Instant Sync - Changes reflect on next app restart
• Secure Vault - API keys stored in Firebase, never in code
• Offline Support - Cached configuration for offline operation
• Exception Handling - Auto-detection and retry for :free suffix models
• Smart Recovery - Automatic 404 error handling with model correction

💬 Advanced Chat Features

• Real-Time Streaming - Server-sent events (SSE) for live responses
• Multi-Modal Support - Text and image understanding (vision models)
• Context Management - 6-message history for optimal performance
• Session Persistence - Cloud-synced chat history with Google Sign-In
• Brand Attribution - Clear provider identification (e.g., "Mark VII x Anthropic")
• Smart Retry - Re-run prompts with different models
• Stop Generation - Cancel responses instantly with red stop button
• Voice I/O - Speech recognition input + multilingual text-to-speech
• PDF Export - Professional formatting with syntax highlighting
• Copy & Share - Easy text extraction and sharing

🎨 User Experience

Material 3 Design:

• Dynamic theming (Light, Dark, System Default)
• iOS-style color palettes with smooth transitions
• Theme-aware status bar and navigation
• No white flash on startup

Performance Optimizations:

• <100ms startup time (24x faster than v1.x)
• Connection pooling for API requests
• Memoized rendering and lazy loading
• Optimized scroll performance

Interaction Design:

• Streaming cursor with haptic feedback
• 40dp touch targets for accessibility
• Smooth Lottie animations
• Auto-scroll to latest messages
• Syntax highlighting for code blocks
• Markdown support with inline formatting

Multilingual Support:

• MLKit language detection (15+ languages)
• Automatic TTS language switching
• Support for: Chinese, Japanese, Korean, Spanish, French, German, Italian, Portuguese, Russian, Arabic, Hindi, English

🛠️ Management Tools

Efficiently manage 45+ AI models using CSV-based workflows.

Features:

• Bulk Import/Export - Manage models in Excel/Google Sheets
• Validation - Automatic format and field checking
• Version Control - Track changes with Git
• Interactive CLI - Menu-driven operations
• Zero Downtime - Update models without app redeployment

Python CLI Usage:

cd update_models

# Import models from CSV (recommended)
python update_firebase_models.py --csv models.csv

# List current models in Firestore
python update_firebase_models.py --list

# Interactive mode
python update_firebase_models.py

# Export models to CSV
python update_firebase_models.py --export my_models.csv

CSV Format:

apiModel,displayName,isAvailable,order
google/gemini-2.0-flash-exp,Gemini 2.0 Flash,TRUE,1
deepseek/deepseek-chat-v3.1,Deepseek Chat V3.1,TRUE,2
anthropic/claude-3-5-sonnet-20241022,Claude 3.5 Sonnet,TRUE,3

Benefits:

• Edit 50+ models in Excel/Google Sheets
• Bulk enable/disable models
• Easy reordering with sort priority
• Version control friendly
• Automatic validation

---

Quick Start

Prerequisites

• Android Studio Ladybug or newer
• JDK 17+
• Firebase account (free tier sufficient)
• OpenRouter API key (free tier available at openrouter.ai/keys)
• Optional: Google Gemini API key for direct Gemini access

1. Clone Repository

git clone https://github.com/daemon-001/Mark-VII.git
cd Mark-VII

2. Firebase Setup (~5 minutes)

Create Firebase Project

1. Go to Firebase Console
2. Click "Add project" → Enter name: Mark-VII
3. Disable Google Analytics (optional) → "Create project"

Add Android App

1. Click Android icon in Project Overview
2. Enter package name: com.daemon.markvii
3. Download google-services.json → Place in Mark-VII/app/ directory

Configure Firestore Database

1. Navigate to Firestore Database → "Create database"
2. Select "Start in test mode" → Choose region → "Enable"

Add Configuration Data

Option A: Python Script (Recommended)

cd update_models
pip install firebase-admin

# Download service account key:
# Firebase Console → Project Settings → Service Accounts
# → Generate New Private Key → Save as mark-vii-firebase-service-account-key.json

# Import 45+ models from CSV
python update_firebase_models.py --csv models.csv

Option B: Manual Setup

1. In Firestore, create collection: app_config
2. Create document: models
3. Add field: list (type: array) with model objects:

{
  "id": "anthropic/claude-3.5-sonnet",
  "name": "Claude 3.5 Sonnet",
  "provider": "anthropic",
  "isEnabled": true,
  "sortOrder": 1,
  "apiType": "openrouter"
}

4. Create document: api_keys
5. Add fields:
   • openrouterApiKey (string) - Get from OpenRouter
   • geminiApiKey (string, optional) - Get from Google AI Studio

3. Build & Run

# In Android Studio:
# 1. File → Open → Select Mark-VII folder
# 2. File → Sync Project with Gradle Files
# 3. Build → Make Project (Ctrl+F9)
# 4. Run → Run 'app' (Shift+F10)

First Launch:

• Grant internet permissions
• Sign in with Google (creates cloud-synced chat sessions)
• Select a model and start chatting!

---

Installation

End Users

Android Smartphone:

1. Download APK from Releases
2. Enable "Install from Unknown Sources" in Settings → Security
3. Open APK file → Install → Open Mark VII

Android Emulator (PC):

1. Download APK from Releases
2. Drag APK into emulator window or use APK installer
3. Launch Mark VII from app drawer

Developers

See Quick Start above for source-based setup.

---

Troubleshooting

"Firebase not configured" in app

• Check Firestore structure: app_config/models document must have list field (not models)
• Verify app_config/api_keys document has openrouterApiKey field
• Ensure google-services.json is in app/ folder
• Check exp_models collection for exception models

HTTP 401 Error (Unauthorized)

• Invalid API key
• Get new key from OpenRouter
• Update in Firebase: app_config/api_keys/openrouterApiKey

HTTP 404 Error (Model Not Found)

• Model may require ":free" suffix
• App automatically adds failing models to exp_models collection
• Retry the request after error - it will use the correct format
• Check model exists on OpenRouter Models

HTTP 429 Error (Rate Limit)

• Too many requests
• Wait a few seconds and retry
• Consider upgrading OpenRouter plan

HTTP 502/503 Error (Server Issues)

• OpenRouter or model provider temporarily down
• Try a different model
• Wait and retry later

App Crashing

• Clear app data: Settings → Apps → Mark VII → Clear Data
• Reinstall the app
• Check for updates

Build Errors

# Clean and rebuild
./gradlew clean
./gradlew assembleDebug

---

Architecture

Overview

Mark VII uses a cloud-first architecture combining Firebase's real-time configuration management with dual AI API access (OpenRouter + Gemini), enabling zero-downtime model updates and enterprise-grade reliability.

┌─────────────────────────────────────────────────────────────┐
│                      Mark VII Android App                   │
├─────────────────────────────────────────────────────────────┤
│  UI Layer (Jetpack Compose + Material 3)                    │
│  ├─ MainActivity.kt         - Main UI orchestration         │
│  ├─ SettingsScreen.kt       - Theme & account management    │
│  └─ DrawerContent.kt        - Navigation & session list     │
├─────────────────────────────────────────────────────────────┤
│  ViewModel Layer (MVVM)                                     │
│  └─ ChatViewModel.kt        - State management + logic      │
├─────────────────────────────────────────────────────────────┤
│  Data Layer                                                 │
│  ├─ ChatData.kt             - OpenRouter API orchestration  │
│  ├─ GeminiClient.kt         - Direct Gemini API client      │
│  ├─ FirebaseConfigManager   - Remote model configuration    │
│  ├─ ChatHistoryManager      - Session persistence (Cloud)   │
│  ├─ AuthManager             - Google Sign-In + tokens       │
│  └─ ThemePreferences        - Local theme storage           │
├─────────────────────────────────────────────────────────────┤
│  Network Layer                                              │
│  ├─ Retrofit + OkHttp       - HTTP client with pooling      │
│  ├─ SSE EventSource         - Streaming response parser     │
│  └─ Connection Pool         - Persistent connections        │
└─────────────────────────────────────────────────────────────┘
              ↓                              ↓
    ┌───────────────────┐        ┌──────────────────────┐
    │ Firebase Services │        │   AI API Providers   │
    ├───────────────────┤        ├──────────────────────┤
    │ Firestore         │        │ OpenRouter API       │
    │ Authentication    │        │ (100+ models)        │
    │ Analytics         │        │ + Direct Gemini API  │
    └───────────────────┘        └──────────────────────┘

Tech Stack

Frontend:

• Kotlin with Jetpack Compose for declarative UI
• Material 3 Design System with dynamic theming
• Lottie for animations, Markdown with syntax highlighting
• StateFlow for reactive state management

Architecture Pattern:

• MVVM (Model-View-ViewModel) with repository pattern
• Coroutines + Flow for asynchronous operations
• Dependency injection via constructor parameters

Backend Integration:

• Firebase Firestore: Cloud configuration + chat history
• Firebase Authentication: Google Sign-In with OAuth
• Firebase Analytics: Usage tracking (optional)

Networking:

• Retrofit 2.11.0 with OkHttp 4.12.0
• Server-Sent Events (SSE) for streaming responses
• Connection pooling for <100ms startup time

AI Providers:

• OpenRouter API: 100+ models with unified interface
• Direct Gemini API: Native Google integration with vision
• MLKit Language ID: Automatic TTS language detection

Tools:

• Python 3.11+ with Firebase Admin SDK
• CSV-based model management workflow
• Git for version control

Project Structure

Mark-VII/
├── app/
│   ├── google-services.json              # Firebase config (download from console)
│   └── src/main/java/com/daemon/markvii/
│       ├── MainActivity.kt                # App entry + Firebase initialization
│       ├── ChatViewModel.kt               # MVVM state management
│       ├── data/
│       │   ├── Chat.kt                    # Data models (Message, ChatRequest, etc.)
│       │   ├── ChatData.kt                # OpenRouter API orchestration
│       │   ├── GeminiClient.kt            # Direct Gemini API client
│       │   ├── OpenRouterApi.kt           # Retrofit API interface
│       │   ├── FirebaseConfig.kt          # Model configuration models
│       │   ├── FirebaseConfigManager.kt   # Firestore config operations
│       │   ├── AuthManager.kt             # Google Sign-In + token management
│       │   ├── ChatHistoryManager.kt      # Cloud chat session storage
│       │   ├── ThemePreferences.kt        # Local theme persistence
│       │   └── Keys.kt                    # App metadata + build info
│       ├── ui/theme/
│       │   ├── Theme.kt                   # Material 3 theme + AppColors
│       │   └── Color.kt                   # Theme color definitions
│       ├── SettingsScreen.kt              # Theme selector + account UI
│       ├── DrawerContent.kt               # Navigation + session management
│       └── utils/
│           └── PdfGenerator.kt            # PDF export with syntax highlighting
├── update_models/
│   ├── update_firebase_models.py          # Model management CLI
│   ├── models.csv                         # Pre-configured 49 models
│   └── mark-vii-firebase-service-account-key.json  # Service account (download)
├── CHANGELOG.md                           # Git commit-based version history
├── FIREBASE_SETUP.md                      # Detailed Firebase setup guide
└── README.md                              # Project documentation

---

Usage Examples

Basic Chat Workflow

1. Launch - Open Mark VII (cold start <100ms)
2. Sign In - Tap profile icon → "Sign in with Google" (optional, enables cloud sync)
3. Select Model - Tap dropdown → Choose model (e.g., "Claude 3.5 Sonnet")
4. Input - Type message or tap microphone icon for voice input
5. Send - Tap send button (▲) → Get real-time streaming response
6. Attribution - See response header "Mark VII x Anthropic"
7. Stop - Tap red stop button (■) to cancel streaming anytime

Switching Between API Providers

1. Open model dropdown
2. Select different provider mid-conversation to compare responses
3. Chat history preserves which API was used per message

Theme Customization

1. Open drawer (swipe right or tap menu) → Tap Settings icon (⚙️)
2. Under "Appearance", tap Theme selector
3. Choose theme:
   • System Default - Follows device settings (auto light/dark)
   • Light - iOS-inspired bright theme with gentle shadows
   • Dark - Eye-friendly with OLED-optimized blacks
4. Theme applies instantly without restart, status bar updates automatically

Managing Chat Sessions

1. Open navigation drawer (swipe right or hamburger menu)
2. View all sessions with preview of last message
3. Switch - Tap session to load conversation
4. Rename - Long-press → "Rename" → Enter new name
5. Delete - Long-press → "Delete" → Confirm
6. New Chat - Tap "+" button → Fresh session created
7. Cloud Sync - All sessions backup to Firebase when signed in

Image Understanding (Vision Models)

1. Tap plus icon (+) in input field
2. Select image from gallery or take photo
3. Type question about image (e.g., "What's in this image?")
4. Works with vision-capable models: Gemini 2.0 Flash, Claude 3.5, GPT-4o
5. Image preview shows pin indicator when attached
6. Send to get multi-modal analysis

Multilingual Text-to-Speech

1. Receive AI response in any language (Chinese, Spanish, Arabic, etc.)
2. Tap speaker icon on message
3. MLKit automatically detects language
4. TTS reads response in correct language/accent
5. Supports 15+ languages: Chinese (Mandarin), Japanese, Korean, Spanish, French, German, Italian, Portuguese, Russian, Arabic, Hindi, English, Dutch, Polish, Turkish

PDF Export

1. Complete conversation with AI model
2. Tap 3-dot menu → "Export to PDF"
3. PDF generates with:
   • Syntax-highlighted code blocks
   • Formatted Markdown rendering
   • Message timestamps
   • Model attribution (e.g., "Mark VII x Anthropic")
4. Share via any app or save to storage
5. Professional formatting for documentation/reports

Retry with Different Models

1. Send message to Model A (e.g., Claude 3.5 Sonnet)
2. Get response → Not satisfied with output
3. Tap model dropdown → Switch to Model B (e.g., GPT-4o)
4. Tap "Retry Last Prompt" button
5. Get alternative response from different AI
6. Compare responses side-by-side in chat history

Voice Input

1. Tap microphone icon in input field
2. Grant microphone permission (first time)
3. Speak your question naturally
4. Speech-to-text transcription appears automatically
5. Edit if needed, then send
6. Works in all languages supported by Android Speech Recognition

---

Performance

Current Metrics (v3.2.0)

Startup Performance:

• Cold Start: <100ms (24x faster than v1.x)
• Model Loading: <50ms (cached configuration)
• Theme Application: <10ms (instant visual feedback)
• Firebase Init: Asynchronous, non-blocking

Runtime Optimization:

• Streaming Latency: <500ms to first token
• Rendering: 60 FPS with memoized composables
• Memory: <50MB baseline, <150MB during streaming
• Network: Connection pooling reduces request overhead by 70%
• TTS Language Detection: <100ms with MLKit on-device processing

Note: Times vary based on network, server load, and prompt complexity.

Version Comparison

Metric	v1.x (Gemini)	v2.x (OpenRouter)	v3.0+ (Dual API)	Improvement
Startup Time	~2.65s	~110ms	<100ms	26.5x faster
Models Available	5-10	100+	100+ (dual APIs)	10-20x more
API Providers	1 (Google)	Multiple	OpenRouter + Gemini	Direct + unified
Configuration	Hardcoded	Cloud-based	Cloud + real-time	Instant updates
Streaming	No	Yes (SSE)	Yes (both APIs)	Real-time
Error Handling	Basic	Comprehensive	Auto-retry + 404 fix	Resilient
TTS Languages	1-2	1-2	15+ (MLKit)	7x more
Theme Support	Basic	Light/Dark	L/D/System + iOS-style	Polished
Offline Support	No	Limited	Cached config	Works offline
Stop Generation	No	Yes	Yes + haptics	User control

---

Contributing

We welcome contributions from the community! Whether you're fixing bugs, adding features, or improving documentation, your help makes Mark VII better.

Reporting Issues:

1. Open an issue
2. Include: Android version, app version, steps to reproduce
3. Add screenshots if helpful

Feature Requests:

1. Open a Feature Request with tag enhancement
2. Describe the feature and its use case
3. Provide mockups/examples if relevant
4. Discuss with maintainers before implementing

Code Contributions:

1. Fork the repository
2. Create feature branch: git checkout -b feature/your-feature-name
3. Make changes following Kotlin code style guidelines
4. Test thoroughly on multiple devices/Android versions
5. Commit with clear messages: git commit -m "Add: Feature description"
6. Push to your fork: git push origin feature/your-feature-name
7. Open Pull Request with detailed description

---

Dependencies

Core Libraries

// Firebase (BOM manages versions)
firebase-bom:33.7.0
firebase-firestore-ktx          // Cloud configuration + chat history
firebase-analytics-ktx          // Usage analytics
firebase-auth-ktx               // Google Sign-In authentication

// Google Services
play-services-auth:21.3.0       // Google Sign-In UI
mlkit-language-id:17.0.6        // Multilingual TTS detection

// Networking
retrofit:2.11.0                 // Type-safe HTTP client
okhttp:4.12.0                   // Connection pooling + SSE
gson:2.10.1                     // JSON serialization

// PDF Generation
itext7-core:7.2.5               // PDF document creation
html2pdf:4.0.5                  // HTML to PDF conversion

// Markdown & Syntax Highlighting
compose-markdown:0.5.4          // Rich text rendering
code-highlight:2.0.0            // Syntax highlighting for code blocks

// UI & Animation
androidx.compose.bom:2024.12.01 // Jetpack Compose
androidx.material3:*            // Material 3 components
lottie-compose:6.0.0            // Lottie animations
coil-compose:2.4.0              // Image loading

Minimum Requirements

• Android SDK: 24 (Android 7.0 Nougat) or higher
• Target SDK: 35 (Android 15)
• Kotlin: 2.1.0
• Gradle: 8.7
• JDK: 17+

---

Security & Privacy

Data Protection

• Encrypted Storage: API keys secured in Firebase, never in source code or APK
• HTTPS Only: All API communication uses TLS 1.3 encryption
• OAuth 2.0: Google Sign-In with secure token management
• Local Storage: Chat history cached locally, synced to Firebase when signed in
• No Tracking: Zero third-party analytics or ad networks
• Firestore Rules: Read/write access restricted to authenticated users

Configuration Security

• google-services.json excluded from Git via .gitignore
• mark-vii-firebase-service-account-key.json excluded (Python tool only)
• Firebase project rules restrict access to authenticated users
• API keys rotatable without app updates (cloud configuration)

User Control

• Sign Out Anytime - Settings → Sign Out → Clears local cache
• Delete Chat Sessions - Long-press session → Delete (removes from cloud)
• Theme Preferences - Stored locally, never synced
• Optional Cloud Sync - Chat history syncs only when signed in with Google
• Offline Mode - Works with cached configuration when no internet

Best Practices

• Never commit google-services.json or service account keys to public repos
• Rotate Firebase API keys periodically in Firebase Console
• Use Firebase Authentication rules to restrict Firestore access
• Enable Firebase App Check for production builds to prevent API abuse

---

Support & Contact

Get Help

Bug Reports:

• GitHub Issues
• Include: Android version, app version, steps to reproduce

Feature Requests:

• Feature Request Form
• Describe use case and expected behavior

Documentation:

• CHANGELOG.md - Version history with git timestamps
• FIREBASE_SETUP.md - Detailed Firebase configuration
• README.md - Comprehensive project documentation

Email Support:

• Developer: nitesh.kumar4work@gmail.com
• Response time: 1-12 hours

Connect

Developer:

• Name: Nitesh Kumar
• GitHub: @daemon-001
• LinkedIn: @daemon001

Project:

• Repository: Mark-VII
• Stars: Give us a ⭐ if you find this useful!
• Forks: Welcome - see Contributing section

Every contribution, no matter how small, helps make Mark VII better for everyone!

---

Built with ❤️ by Nitesh

⬆ Back to Top ⬆

---

Home • Download • Report Bug • Request Feature

Enjoy your advanced multi-LLMs AI chatbot experience!