ClimaSafe · Telegram bot

Heat and cold risk,
right in your chat.

The ClimaSafe Telegram bot answers with your personal risk level, manages your profile and routines, and sends daily alerts before dangerous days arrive. Responses are drafted by a local LLM running on your machine.

MCP server github.com/ANFAIA/ClimaSafe ← back to project
01
create

Create the bot with @BotFather

→ Open @BotFather in Telegram and send /newbot.
→ Choose a name (what users see) and a username ending in bot.
→ Copy the API token that BotFather returns. It looks like 123456:ABC-DEF....
02
token

Where the token goes

The token goes in a .env file at the root of the ClimaSafe repository: 
TELEGRAM_BOT_TOKEN=123456:ABC-DEF...  # never commit this
Start with .env.example as a reference: copy it to .env and fill in your token. The token is a secret: it never goes in the code, never in the repository, never in a commit.
03
run

Run it locally

The bot is started from the repository root with the Makefile targets: make spacebot (first-time setup) and make spacebot-start (run the bot). The daily alert and prediction pipeline runs on a schedule.
El LLM del bot es local (Ollama). El bot redacta sus respuestas con un LLM local — Ollama sirviendo qwen3:climasafe, un modelo fine-tuneado — no con una API en la nube. Para que el bot funcione, ejecuta el repositorio mientras Ollama sirve el modelo. Lo demás — el formulario, la predicción de riesgo y las alertas diarias — es determinista y funciona sin el LLM.
Source and setup details: github.com/ANFAIA/ClimaSafe.