Lira

Lira is a sleek Flutter mobile app (iOS/Android) that acts as your always-on voice buddy—like ChatGPT’s voice mode but cozier.
It uses a cloned "grandma" voice to provide empathetic advice on daily life, emotional check-ins, quick planning (e.g., "Remind me about that meeting in Amharic?"), or just venting sessions.

• Hands-free, real-time chat: Speak naturally (even with Ethiopian accents), it listens live, thinks via AI, and responds in a warm, storytelling tone.
• Privacy-first (mostly on-device), with optional integration with Neuroviate vibes for multicultural empathy.
• Monetization can be added later via premium voices or third-party integrations.
• Target audience: Busy individuals craving low-key wisdom, starting in Ethiopia/global diaspora.

---

🔥 Tech Stack & Free LLM Options

Frontend

Backend

AI & LLMs (Free)

Speech Processing (Free)

---

📸 Screenshots

---

🏗️ UI Components

1. Home / Dashboard Screen (lib/screens/home_screen.dart)

• User greeting with profile picture
• "Good Morning" prompt
• Main "Talk to AI assistant" card with Start Talking button
• Voice and Image feature cards
• Topics section with pill-shaped buttons
• Information cards (Blood pressure, Sleep)
• Bottom navigation bar with AI sparkle button

2. Voice Analysis Screen (lib/screens/voice_analysis_screen.dart)

• "Listening..." indicator
• Animated 3D orb visualizer with gradient colors
• Live transcript display
• Bottom control bar with timer, microphone button, and cancel button

3. Smart Chat Screen (lib/screens/smart_chat_screen.dart)

• Chat interface with AI and user message bubbles
• Sparkle icons for AI messages
• Audio message bubbles with waveform visualization
• Text input field with mic and add buttons
• Pre-populated sample conversation

4. Shared Components

• Gradient Background (lib/utils/gradient_background.dart) — Purple/pink gradient
• Status Bar (lib/widgets/status_bar.dart) — Time, signal, WiFi, battery
• Orb Visualizer (lib/widgets/orb_visualizer.dart) — Animated 3D sphere with swirling patterns

---

🎨 Design Features

• Purple/pink gradient backgrounds matching app visuals
• Rounded corners on all UI elements
• Modern, clean aesthetic
• Smooth animations on the orb visualizer
• Consistent color scheme using #9B7EDE purple

---

🧠 Lira MVP Workflow

flowchart TD
    A[User speaks into Flutter app] --> B[Flutter captures audio]
    B --> C[Speech-to-Text (Whisper / Vosk / Coqui STT)]
    C --> D[Text sent to Python FastAPI backend]
    D --> E[Backend queries Free LLM (Mistral / LLaMA / OpenRouter)]
    E --> F[AI generates agentic response (grandma voice style)]
    F --> G[Text returned to Flutter app]
    G --> H[Text-to-Speech (Coqui TTS / flutter_tts)]
    H --> I[Flutter plays AI voice response]
    I --> A[User continues conversation]

Loading

Workflow explanation:

1. User speaks → Flutter captures audio
2. Audio → text via STT
3. Python backend receives text → queries free LLM
4. LLM generates empathetic, agentic response
5. Text-to-speech converts AI text → voice
6. Flutter plays voice back to user
7. Conversation continues naturally

---

🛠️ Free Backend Setup (Python + LLM)

• Python with FastAPI for REST API endpoints
• Free LLM options: OpenRouter, HuggingFace Inference (Mistral, LLaMA, Grok, Qwen)
• Speech-to-Text: Whisper (local) or Vosk
• Text-to-Speech: Coqui TTS or flutter_tts
• Conversation memory: store last 3–5 messages in RAM (privacy-first)

Fully free, no subscription required, and privacy-friendly MVP

---

📂 Project Structure

Lira/
│── lib/
│   ├── screens/
│   │   ├── home_screen.dart
│   │   ├── voice_analysis_screen.dart
│   │   └── smart_chat_screen.dart
│   ├── widgets/
│   │   ├── status_bar.dart
│   │   └── orb_visualizer.dart
│   └── utils/
│       └── gradient_background.dart
│── assets/
│── backend/
│   ├── app/
│   │   ├── main.py (FastAPI factory)
│   │   ├── routers/ (chat, stt, tts routes)
│   │   ├── schemas.py (Pydantic models)
│   │   └── services/ (LLM provider abstractions)
│   └── requirements.txt
│── README.md

---

▶️ Getting Started

1. Clone the repository

git clone https://github.com/your-username/lira.git
cd lira

2. Install Flutter dependencies

flutter pub get

3. Run the app

flutter run

4. Backend setup

cd backend
pip install -r requirements.txt
cp .env.example .env
# Edit .env and add your LLM API key
uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

5. Configure environment

Edit backend/.env:

LLM_API_BASE_URL=https://openrouter.ai/api/v1
LLM_API_KEY=sk-...
LLM_MODEL=mistralai/mistral-7b-instruct

6. Configure Flutter backend URL

Edit lib/config/api_config.dart and set your backend URL:

• Local: http://localhost:8000
• Android Emulator: http://10.0.2.2:8000
• Physical device: http://YOUR_COMPUTER_IP:8000

📖 For detailed setup instructions, see SETUP.md

---

📌 Roadmap for MVP → Full App

• Multi-language support (Amharic, English)
• Premium voices & AI personality options
• Push notifications & reminders
• Advanced conversation memory & reasoning
• Integrations with Neuroviate for multicultural empathy
• Polished UI animations and orb visualizer

---

🗣️ Voice + Agentic Integration Checklist

• Audio capture (Flutter): use record or flutter_sound to stream PCM via WebSocket to the /stt endpoint. Buffer 1–2s chunks for responsiveness.
• Speech-to-Text (Python): replace the stub with Whisper (faster-whisper) or Vosk. Emit partial transcripts to the client so voice_analysis_screen.dart can display live text.
• Conversation hand-off: send the latest transcript plus last 5–10 turns to /chat. The backend keeps persona prompts + temperature settings server-side.
• LLM provider config: switch models using LLM_MODEL env var without touching Flutter. Supports OpenRouter, HuggingFace or local inference once you point LLM_BASE_URL accordingly.
• Text-to-Speech: call /tts with the assistant reply. Implement Coqui TTS (offline) or gTTS for a quick cloud option; Flutter plays via just_audio.
• Memory + tools: use conversation payload to pass lightweight memory now; later extend backend to persist slots and emit tool_calls for reminders, journaling, etc.

---

🤝 Contributing

Pull requests welcome! Please open an issue for major changes.

---

📄 License

MIT License