GitHub

BeeBot is your personal worker bee, an Autonomous AI Assistant designed to perform a wide range of practical tasks
autonomously.
Status
Development of BeeBot is currently on hold. I've decided that LLMs as they are now (late 2023) aren't up to the task of generalized autonomous AI. I will revive the project if either:
- LLMs get significantly better at structured thinking, reliable outcomes, and obeying instructions
- I can develop or fine tune a custom model which is trained specifically for Autonomous AI
- I figure out a particular subset of tasks that BeeBot is acceptably good at that I can focus on. (Hint: It's not coding)
Check back here, hopefully this will get re-started.
Features
- Tool selection via AutoPack and the ability to acquire more tools during task execution
- Built-in persistence
- REST API conforming to the e2b standard.
- A websocket server to publish all events that occur within BeeBot
- Swappable filesystem emulation so that files can be stored in-memory, on-disk, or in a database
- A Web UI for managing your tasks (coming very soon)
- Dynamic manipulation of history during task execution
- Built-in caching with Helicone if enabled.
Installation
To get started with BeeBot, you can clone the repo to your local machine and install its dependencies using poetry.
These instructions may vary depending on your local development environment.
Windows is officially unsupported but it may work. PRs are welcome for Windows compatibility but will not be a primary
focus.
Persistence
Persistence is _required_. While SQLite is officially supported and is used in tests, it is highly recommended that
you use Postgres via docker, simply by executing docker compose up -d.
Running
CLI
To use the CLI run:
API (via e2b)
To start the server run:
If you're doing development on BeeBot itself, you may want to use this command:
and then you can call the API using the following commands:
To **create a task** run:
You will get a response like this:
Then to **execute one step of the task** copy the task_id you got from the previous request and run:
Websocket Connection
_Note: Notifications are currently undergoing a rework and may not work at the moment_
To receive a stream of changes to all the data models in BeeBot, you can subscribe to the websocket connection at
the /notifications endpoint with the same host/port as the web api, e.g. ws://localhost:8000/notifications. Use your
favorite websocket testing tool to try it out. (I like Insomnia)
Web Interface
We are working on a web interface using Node.js (Remix)
Philosophy
BeeBot's development process is guided by a specific philosophy, emphasizing key principles that shape its development
and future direction.
Priorities
The development of BeeBot is driven by the following priorities, always in this order:
1. Functionality: BeeBot aims to achieve a high success rate for tasks within its range of _expected_ capabilities.
2. Flexibility: BeeBot strives to be adaptable to a wide range of tasks, expanding that range over time.