---
title: CelachoBot
emoji: 🤖
colorFrom: gray
colorTo: blue
sdk: docker
app_port: 7860
pinned: false
---

# CelachoBot — Hugging Face Space + llama.cpp

Minimal Telegram bot using a local GGUF model via `llama.cpp`.

## Secrets

Set these in **Space Settings → Variables and secrets**:

- `TELEGRAM_BOT_TOKEN` = your BotFather token
- `MODEL_REPO` = GGUF repo id, e.g. `bartowski/Qwen3.5-0.8B-GGUF`
- `MODEL_FILE` = GGUF filename, e.g. `Qwen3.5-0.8B-Q4_K_M.gguf`
- `BOT_SYSTEM` = optional system prompt
- `PUBLIC_URL` = optional Space URL, e.g. `https://YOUR-SPACE.hf.space`

## One-time webhook setup

If `PUBLIC_URL` is set, the app tries to set the webhook on boot.

Manual fallback:

`https://api.telegram.org/bot<TELEGRAM_BOT_TOKEN>/setWebhook?url=https://YOUR-SPACE.hf.space/webhook`

## Notes

- No history. Each Telegram message is answered independently.
- No database. No persistence.
- If the Space sleeps, the first webhook after idle may be slow.