# Telegram Bot Split for Local vs Hosted Models

Use this when the user wants a separate Telegram bot for the local model instead of switching models inside one bot.

## Why this is the cleanest setup

- One bot/profile pair per model avoids confusion.
- The hosted bot stays untouched.
- The local model can be exposed independently.
- `/model` remains useful inside each profile, but it is no longer asked to solve cross-profile switching.

## Operational shape

- Hosted bot:
  - Hermes profile: default / hosted ChatGPT-backed profile
  - Telegram gateway service: default profile service

- Local bot:
  - Hermes profile: `local-ollama`
  - Telegram gateway service: `hermes-gateway-local-ollama.service`

## Steps

1. Create or reuse a Telegram bot via BotFather.
2. Add the bot token to the target profile's `.env` as `TELEGRAM_BOT_TOKEN`.
3. Set `TELEGRAM_ALLOWED_USERS` for the intended user list.
4. Install the gateway service for that Hermes profile.
5. Start the gateway and verify it is running.
6. Keep the default profile/gateway separate from the local one.

## Verification commands

```bash
hermes profile show local-ollama
hermes gateway list
hermes --profile local-ollama gateway status
```

## Caveats

- Never assume `/model` can cross from one profile to another; it only switches within the active profile/session.
- If a Telegram bot token was pasted in chat, treat it as exposed and rotate it after confirmation.
- Use separate tokens and services when the user wants a hard split between local and hosted behavior.