Nex Telegram Bot Deploy

What it does

• Ships base_bot.py with the NexBaseBot base class: token, bot_id, persona, welcome, command registration, default handlers (start, help, save, brain), and run(), so a new bot is config, not plumbing
• Includes assets/bot_template.py, a minimal bot built on the base: copy, set the persona and commands, done
• Adds an optional AI free-text handler (assets/smart_handler.py) that classifies a plain message into a task, follow-up, or question with one model call (Qwen or Claude), degrading to a plain reply on failure
• Deploys with scripts/setup.sh: dedicated non-root system user, venv, systemd unit (Restart=always, RestartSec=5), UFW, fail2ban (maxretry 3, bantime 3600), and SSH hardening (key-only, port 2222)
• Supports timezone-aware scheduled jobs via the JobQueue, like a daily morning briefing

Why use this

Most bot tutorials stop at polling on your laptop; this ships the way Nex AI actually runs bots in production on a Pi, as a hardened systemd service under a locked-down user, with every secret in env vars. The shared base means each next bot you build is a token, a persona, and a handful of handlers.

See it in action

Prompt:

I want a bot that takes my free-text notes, turns them into tasks with deadlines, and sends me a briefing every morning at 7:45.

Result:

bot = NexBaseBot(
    bot_id=ops,
    persona=Personal ops assistant. Direct, brief, Dutch or English.,
    commands={taken: list_tasks, klaar: complete_task},
)
bot.add_smart_handler(provider=qwen)   # free text -> task/followup/question
bot.add_daily_job(briefing, time=07:45, tz=Europe/Brussels)
bot.run()  # token from TELEGRAM_BOT_TOKEN, never hardcoded

[Service]
User=opsbot
Restart=always
RestartSec=5
EnvironmentFile=/opt/opsbot/.env

Deploy: SCP the folder, run setup.sh, set the token in .env.

Use cases

• A client-facing Q&A bot with a persona and custom commands
• A personal ops bot that classifies free-text notes into tasks and sends a morning briefing
• A reusable bot base an agency clones per client or use case
• Deploying any python-telegram-bot app as a hardened service on a Pi or VPS

Known limitations

• Telegram only; Discord and Slack use different SDKs and are out of scope.
• The full hardening stack targets a Raspberry Pi or Debian-like VPS with systemd; code-only delivery is supported for other hosts.
• The AI free-text handler needs a Qwen or Claude API key; without one it degrades to plain replies.

FAQ

What exactly is in the download?

SKILL.md, scripts/base_bot.py and setup.sh, assets/bot_template.py, smart_handler.py, bot.service, and .env.example, three reference docs (base-bot, deployment, ai-handler), manifest.json, README, changelog, and license.

Which agents does it work with?

Claude Code, Claude Cowork, Cursor, Codex CLI, and any agent that reads the SKILL.md format.

Do I need anything else?

A Telegram bot token from BotFather, and a Pi or VPS if you deploy. An AI API key only for the optional free-text handler.

Will the SSH hardening lock me out?

The setup moves SSH to port 2222 with key-only login, and the skill's delivery step explicitly warns you before it applies, so you reconnect on the new port with your key.