lawremce-desigm/Dara-awarri

Dára Home is a multilingual, voice-first smart home assistant built for Nigeria. It uses N-ATLaS for Reasoning and intent understanding in English, Yoruba, Hausa, and Igbo, Whisper for speech-to-text, a Flutter app for interaction, and custom ESP32 firmware (C++) for real-world device control # Dára Home (Dara-awarri) 🇳🇬 **Dára Home** is a multilingual, voice-first smart home assistant designed specifically for the Nigerian context. It bridges the gap between modern smart home technology and local linguistic reality by supporting **English, Yoruba, Hausa, and Igbo**. ## 🏗️ Architecture Overview The system operates on a **Controller-Facilitator** model where the Mobile App acts as the central brain for orchestration, while the Backend handles intelligence and the ESP32 handles physical execution. ### 🔄 How It Works 1. **Voice Command**: The user speaks into the **Mobile App** (e.g., *"Tan ina yara mi"* - Turn on my room light). 2. **Processing**: The app sends the audio to the **Backend**. 3. **Intelligence**: * **STT**: Audio is transcribed to text (e.g., "Tan ina yara mi"). * **Reasoning**: The N-ATLaS model (finetuned LLM) analyzes the text, detects the language (`yo`), intent (`INSTRUCTION`), action (`TURN_ON`), and device (`LIGHT`). * **Response**: The system generates a natural text response in the same language and converts it to audio. 4. **Action**: The Backend returns the *Intent* and *Audio* to the App. 5. **Execution**: * The **App** plays the audio response to the user. * If an action is required, the **App** directly calls the **ESP32** over the local network (e.g., `POST http:// /light/turn/on`). --- ## 📂 Repository Structure This monorepo contains three distinct components: ### 1. `dara-mobile-app/` (Flutter) The user interface and central controller. * **Features/Roles**: Voice recording, audio playback, state management (`AppState`), and direct HTTP control of local ESP32 devices. * **Key Tech**: Flutter, Provider, HTTP, Audio Players. ### 2. `dara-backend/` (FastAPI / Python) The intelligence layer hosted in the cloud/locally. * **Features/Roles**: Speech-to-Text (Deepgram/Whisper), Intent Classification (N-ATLaS on Modal), Text-to-Speech (Spitch). * **Key Tech**: Python, FastAPI, Modal, Spitch. ### 3. ` …