Chat with a personal AI assistant through Telegram that remembers context across sessions.
Draft and send email, with every message staged for your approval before it goes out.
Track contacts, deals, and tasks in a simple spreadsheet-based CRM, one folder per project.
Run the assistant fully on your own hardware using local Ollama models instead of paid cloud APIs.
| surfgolfski-dev/billy | 0xustaz/streamgate | a-bissell/unleash-lite | |
|---|---|---|---|
| Stars | 1 | 1 | 1 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | hard | hard |
| Complexity | 3/5 | 4/5 | 4/5 |
| Audience | developer | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires an Anthropic API key and Telegram bot token at minimum, cloud model usage costs pay-as-you-go money.
Billy is a personal AI assistant you chat with through Telegram instead of a separate app or website. It runs on your own computer, only responds to one Telegram account that you set up as its owner, and uses AI models, with Anthropic's Claude as the default brain, to actually complete tasks rather than just answer questions. Those tasks include reading and sending email through drafts you approve first, keeping a simple spreadsheet based CRM of contacts and projects, generating and editing documents, images, and video, doing web research with citations, and converting text to speech. Under the hood, Billy works as an orchestrator rather than one giant AI prompt trying to do everything at once. A cheaper model first decides what a given message actually needs, and only then does a more capable model step in to do the heavy work. Around eighty specialized tools stay dormant until a task calls for them, so most requests stay fast and inexpensive. Specific jobs, like writing, researching, or generating images, are handed off to their own smaller sub-agents that can run at the same time as each other. Over time, Billy can also write and reuse markdown based playbooks the project calls skills, letting it get better at repeated tasks. Two safety rules are enforced through automated tests: Billy cannot rewrite his own source code, and he cannot send an email unless a person actually asked him to in that conversation. It is built in Python 3.12 using a framework called LangGraph, and it can plug into different AI providers through a configuration file, including running entirely on your own hardware using a tool called Ollama instead of paying for cloud based models. Getting started means copying an example settings file, adding an Anthropic API key and a Telegram bot token, installing the Python dependencies, and starting the program. Every message sent to the cloud AI models costs a small amount of real money, controlled through each provider's own dashboard. Billy was built and tested primarily on Windows, though the core program is plain Python and should run wherever Python does. It is released under the MIT license.
Billy is a self-hosted personal AI assistant you talk to over Telegram, using Claude or local Ollama models to handle email, a lightweight CRM, research, documents, and skills it learns over time.
Mainly Python. The stack also includes Python, LangGraph, Telegram.
Permissive MIT license, free to use, modify, and distribute, including commercially.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.