A
Claude Code logoClaude CodePrivate bot, no API bills, data never leaves your box

Fully Local Private Telegram Bot With Ollama

setuproll@setuproll
86.0Overall score

A Telegram bot that talks to a Llama 3.x or DeepSeek model served locally by Ollama, so there are no per-token costs and no chat data sent to a cloud provider. Built for privacy-minded users and home-lab tinkerers who have a decent GPU or Apple Silicon Mac to spare.

86.0Score
2.0kVotes
5Components

Install this build

Export
terminal
ollama run llama3.3 && python bot.py

Components

Model

  • Llama 3.x 70B
  • DeepSeek V3
  • Qwen (local)

Stack

  • Ollama
  • aiogram
  • Python

Memory

  • In-process chat history
  • Optional SQLite persistence

How it works

  • Pull a model with Ollama and confirm it serves on localhost:11434
  • Point the bot at the Ollama chat endpoint instead of a cloud API
  • Forward Telegram messages to the local model and stream the reply
  • Keep everything offline, the token is the only external dependency

Deploy

  • Home server or Mac
  • Long polling
  • Tailscale for remote access

Summary

A Telegram bot that talks to a Llama 3.x or DeepSeek model served locally by Ollama, so there are no per-token costs and no chat data sent to a cloud provider. Built for privacy-minded users and home-lab tinkerers who have a decent GPU or Apple Silicon Mac to spare.

86.0 score 2.0k votes

0 Reviews

Your rating
Sign in to post

Loading discussion...

← All builds