Lanlokun/mkunga

Mkunga (Swahili for "midwife") is an AI-powered call center providing real-time antenatal care guidance for pregnant women. This repository offers quickstart examples using the Gemini Multimodal Live API integrated with MedGemma as a medical knowledge backend. # Mkunga: Swahili Antenatal Care AI Call Center Mkunga (Swahili for "midwife") is an AI-powered call center providing real-time antenatal care guidance for pregnant women. This repository offers quickstart examples using the **Gemini Multimodal Live API** integrated with **MedGemma** as a medical knowledge backend. It enables a voice-to-voice service delivering empathetic pregnancy advice in **Swahili and English**. > **Note:** This is an experimental API. Gemini Live and MedGemma integration is custom, and the implementation uses API version `v1beta1` for VertexAI and `v1alpha` for Gemini Developer API. Documentation from Google is still evolving. ## Features - **Voice-to-voice chat**: Speak directly to Mkunga and receive audio responses. - **Medical expertise**: MedGemma handles all antenatal and pregnancy-related queries. - **Prompt engineering**: MedGemma behaves as an empathetic Swahili antenatal care assistant. - **Language support**: Responses in **Swahili** or English (via TTS). - **Clarification loop**: Detects unclear or incomplete queries and prompts for more details. - **[MedGemma] labeling**: Responses from MedGemma are clearly indicated. - **Customizable session flow**: Includes introduction, query handling, and new-call simulation. ## Examples ### 1. Voice-to-Voice Chat (`audio-to-audio-liveapi.py`) - Speak directly into your microphone - Receive audio responses through speakers - MedGemma handles all pregnancy-related queries - Press `ctrl+c` to quit --- ## Requirements ``` google-genai pyaudio ``` Python 3.11+ is required for the voice-to-voice example, as it uses `asyncio.TaskGroup` for concurrent task management. ## Setup 1. Clone the repository: ```bash git clone github.com cd mkunga ``` 2. Install dependencies: ```bash pip install -r requirements.txt ``` 3. Choose your authentication method: ### Option A: Google AI Studio API Key - Visit aistudio.google.com to get your API key - Set the environment var …