Build Your Own Open-Source Dots with Pi and Telegram
- Source
- youtube.com
- Author
- Hugging Face
- Date
Context
In a Hugging Face tutorial, the presenter shows how to build an open-source version of what they describe as OpenAI's dots, always-on agents with specialized jobs. Each agent is just a directory on an always-on machine, such as a rented server or home desktop. The directory holds an instructions file and installed skills, and runs under Pi, a coding-agent harness. One agent does research; another reads Gmail and a calendar through the Google Workspace .
To reach the agents away from the terminal, a background gateway the presenter wrote listens to Telegram, accepting messages only from the owner's user ID. It builds an from the source, the agent name and the Pi session, stores it in a local SQLite database, and routes each message to the matching session in that directory, so that agent's skills and apply. Sending /new starts a fresh session. The design is inspired by the message gateways in Hermes and OpenClaw, and other chat platforms could be added.
- Repository instructions — Versioned guidance that tells agents how a repository is structured and verified.
- CLI tool — A command-line capability an agent can discover and compose through shell commands.
- Identifier — A value that names exactly one record within a stated system, such as S-003 in the committee's table.
- Context and context window — Context is the information available for a generation; the context window is the capacity available to hold it.
Checking sign-in…
Loading comments…






