What is ique.bot?
A task queue that sits behind a single Telegram bot and routes messages to different workers. You type a message, a router figures out which worker should handle it, and that worker fetches some data, thinks about it with an LLM, and sends you an answer.
That's it. No dashboard, no login, no app to install.
Why bother?
I kept building separate bots for separate things. One for SEC filings, one for grocery deals, one for meal planning. Each needed its own token, its own server process, its own deployment. And half the time I'd message the wrong one.
So I built a router. One bot, one chat thread. The router reads your message, figures out which skill you want, and hands it off. If you type /radar AAPL, the radar worker picks it up. If you type "what's for dinner," the meal planner gets it. You don't need to know or care how that works.
How a message moves through the system
User: "/radar AAPL"
│
▼
Router: slash command "/radar" → assign to "radar" boutique
│
▼
Queue: task sits in SQLite with status "queued"
│
▼
Radar worker: polls queue, sees the task
1. Fetch: GET SEC Form 4 filings for Apple from API
2. Reason: Send filings + system prompt to Haiku
3. Complete: Write the briefing back to the queue
│
▼
Adapter: polls for completed tasks, sends response to TelegramYou never see the queue or the router. You just get an answer in the same chat.
What's a "boutique"?
A boutique is a worker that does one thing. It has:
- A system prompt (
CLAUDE.md) that tells the LLM what this worker is and what rules to follow - A data source, like an API, a database, or just some local files
- A loop that polls the queue, does the work, and posts the result
I call them "boutiques" because they're small and opinionated. The opposite of a general-purpose agent that tries to do everything.
What this isn't
No conversation memory. No multi-turn dialogue. Each message is a standalone task that gets routed, processed, and answered. If you need a back-and-forth, this isn't the right tool.
No agents. Workers don't decide what to do next. They follow a fixed pipeline: fetch data, reason about it, respond. Boring on purpose.
No platform. It's a Node.js process and a SQLite file. You run it on whatever machine you want and shut it down when you're done.
Two ways to use it
Subscribe to someone else's bot. They run the server, you just chat. You send /radar AAPL to @ique_radar_bot and get a briefing. You didn't install anything.
Run your own. Clone the repo, get a bot token from @BotFather, add your API keys, run ./ique/start.sh. Now you have a private assistant with whatever skills you want. You can run the same radar worker that the public bot uses, plus personal stuff that touches your own data.
Both work at the same time. You can subscribe to someone's radar bot and also self-host your own bot with private boutiques. Different bots, same framework.
Next steps
- Use cases -- what people actually do with this
- Operator setup -- run your own instance
- Architecture -- how the pieces connect