Quickstart

Requirements

Python 3.11+
Ollama installed
uv (recommended) or pip
A GPU with 16 GB+ VRAM (see Hardware)

Install

Clone and install

git clone https://github.com/Bumblebee-AGI/bumblebee.git
cd bumblebee
uv sync

Pull models

ollama pull gemma4:26b        # chat and reasoning
ollama pull nomic-embed-text   # memory similarity

Create your entity

Copy the example config:

cp configs/entities/canary.example.yaml configs/entities/canary.yaml

Or run the guided wizard:

bumblebee create

Talk

bumblebee talk canary --ollama

The --ollama flag starts the Ollama server if it isn’t running. Add --pull-models on a fresh machine to auto-download models.

Run as a daemon

For always-on mode with all configured platforms (Telegram, Discord, CLI):

bumblebee run canary --ollama

This starts the heartbeat daemon, soma body state, memory consolidation, proactive initiative, and every platform listed in your entity YAML.

Next steps

Entity configuration

Understand the full entity YAML structure.

Setup & onboarding

Guided bumblebee setup, hybrid stack, tunnel automation, and Railway.

Telegram setup

Connect your entity to a Telegram bot.

Local deployment

Tune Ollama, models, and local-only configuration.

Hybrid Railway

Home GPU + Railway worker + Postgres.

Architecture

How cognition, soma, identity, memory, and presence work together.

Getting started

Guides

Deployment

Agents

Build

Requirements

Install

Run as a daemon

Next steps

Entity configuration

Setup & onboarding

Telegram setup

Local deployment

Hybrid Railway

Architecture

​Requirements

​Install

​Run as a daemon

​Next steps

Entity configuration

Setup & onboarding

Telegram setup

Local deployment

Hybrid Railway

Architecture

Requirements

Install

Run as a daemon

Next steps