MediPal

AI-powered medical communication relay that converts urgent health needs into speakable audio for non-native speakers abroad.

Built on nanobot with OpenRouter for LLM and ElevenLabs for TTS.

How it works

You message MediPal (via Telegram) with a medical need — pharmacy visit, ER situation, allergy warning, dietary restriction. MediPal produces a short, clear phrase in the local language and generates audio you can play directly to staff.

Setup

1. Install uv:

brew install uv

2. Configure environment:

cp .env.example .env
# Fill in OPENROUTER_API_KEY, TELEGRAM_BOT_TOKEN, ELEVENLABS_API_KEY

3. Install dependencies:

uv sync

4. Run:

uv run --env-file .env nanobot gateway

Deploy on Railway

1. Connect this repo to a Railway project.
2. Set the three environment variables from .env.example in the Railway dashboard.
3. Deploy — Railway auto-detects the Dockerfile.

Project layout

├── config.json              # Nanobot runtime config
├── pyproject.toml           # Python dependencies (uv)
├── Dockerfile               # Railway-compatible container
├── .env.example             # Required environment variables
└── workspace/
    ├── AGENTS.md            # Behavior rules and output format
    ├── IDENTITY.md          # Who MediPal is
    ├── TOOLS.md             # TTS tool documentation
    ├── USER.md              # User profile template
    └── tools/
        └── tts.py           # ElevenLabs text-to-speech tool

Environment variables

Variable	Required	Description
OPENROUTER_API_KEY	Yes	OpenRouter API key for LLM
OPENROUTER_BASE_URL	No	Defaults to https://openrouter.ai/api/v1
TELEGRAM_BOT_TOKEN	Yes	Telegram bot token from @BotFather
ELEVENLABS_API_KEY	Yes	ElevenLabs API key for TTS