Doctors-Office-A2A

# Doctor's Office Agent to Agents Welcome to **Doctor Office A2A**. I created this project to demonstrate a distributed multi-agent system where each agent runs on its own device, but they all communicate via Agent-to-Agent (A2A) protocols. ## The Agents ### 1. Doctor Office Agent (`doctor_office/`) - **Model:** `functiongemma:270m` (via Ollama) - **Hardware:** Extremely lightweight. I actually ran this on my old phone via Termux! It's perfect for an older clinic PC. - **Role:** Registers patients and serves a local dashboard. Keeps everything private and local. ### 2. MedGemma Agent (`medgemma/`) - **Model:** `medgemma-4b` (via Ollama) - **Hardware:** Resource intensive. Needs a PC with >4GB VRAM GPU, an M-series Mac, or a cloud instance like Google Colab. - **Role:** Provides preliminary medical guidance based on symptoms. - **Note:** If your hardware can't handle it, you can skip running this agent. Just update the WhatsApp agent's instructions to remove the consultation step. ### 3. WhatsApp Orchestrator Agent (`whatsapp/`) - **Model:** `gemini-3-flash-preview` (via Google AI Studio) - **Platform:** Served via Agno's AgentOS WhatsApp interface. - **Role:** The user-facing interface. It chats with the user and coordinates the other two agents. - **Deployment:** Can be deployed on a server, cloud function, or locally. ## How to Run You can run these agents on different devices (e.g., Doctor Office on a clinic PC, MedGemma on Colab, WhatsApp Agent on a server). **Crucial Step:** If you're splitting them up, you **must** update the base addresses in `whatsapp/agent.py` to point to the actual IP addresses of the other agents. ### Running Locally (The Easy Way) To save time, you can just create one virtual environment for the whole project instead of three separate ones. 1. **Setup the Environment** ```bash # Create one venv in the root python -m venv .venv . .venv/Scripts/activate # Windows (PowerShell) # source .venv/bin/activate # Mac/Linux # Install dependencies for everyone pip install -r medgemma/requirements.txt pip install -r doctor_office/requirements.txt pip install -r whatsapp/requirements.txt ``` 2. **Run the Agents** (Open 3 separate terminals, activate the venv in each, and go!) **Terminal 1: MedGemma** (Port 7778) ```bash cd medgemma python medgemma.py ``` **Terminal 2: Doctor Office** (Port 7779) ```bash cd doctor_office python doctor_office_agent.py ``` **Terminal 3: WhatsApp Orchestrator** (Port 7777) ```bash cd whatsapp $env:GOOGLE_API_KEY="<your_api_key>" python agent.py ```