Auto-GPT-DockerSetup

Welcome to the Auto-GPT-DockerSetup repository! This project aims to provide an easy-to-use starting point for users who want to run Auto-GPT using Docker. This setup separates runtime configuration from the actual Auto-GPT repository by providing a Docker Compose file tailored to running an instance from a Docker image. The official Docker Compose file in the Auto-GPT source includes building the image, but with it always building, not mounting, no solution for plugin-requirements its not very pleasant to maintain.

Note: This template is designed to make it easier for users to get started with Auto-GPT and Docker. It is assumed that users have some experience with Docker and an understanding of what they're doing.

Features

• An example docker-compose.yml file for running Auto-GPT in a Docker container.
• Mounts multiple directories into the container for ease of use. For more serious setups, users should modify the Dockerfile to copy directories instead of mounting them.
• Includes a personas folder with an example YAML file. Users should replace this with their own persona YAML files.
• Provides a clear structure to start from, making it easy for users to adapt the template to their specific needs.
• Uses embedded Weaviate for long-term memory storage and retrieval.
• The Dockerfile extends a specific tag, allowing for easy plugin integration without having to modify the Auto-GPT requirements file or Dockerfile.

Getting Started, Step by Step:

1. clone this repository (or download as zip)

2. remove the .git folder (ideally git init . for your own)

3. copy your .env file from auto-gpt, if you don't have one yet, use their template

4. while you can setup a different memory backend, the easiest setup with long term memory is using embedded weaviate, to do so, you can configure it in your .env file as follows:

   MEMORY_BACKEND=weaviate
   WEAVIATE_HOST="127.0.0.1"
   WEAVIATE_PORT=8080
   WEAVIATE_PROTOCOL="http"
   USE_WEAVIATE_EMBEDDED=True
   WEAVIATE_EMBEDDED_PATH="/app/aviate"

5. if you want to run any plugins

   • copy their zip files into the plugins folder
   • add their requirements to requirements.txt
   • don't forget to add them to the allowed list in your .env

6. Now we get to the docker part:

   • you will have to build a base image from Auto-GPT's source first, named autogpt-upstream, you can do that by cloning their repo and running docker build -t autogpt-upstream . in the root of the repo
   • the Dockerfile in this repo extends from autogpt-upstream and uses the name my-own-autogpt, so the next step is to build it: docker build -t my-own-autogpt .

7. let the bot battle begin: docker-compose run --rm auto-gpt -C personas/Entrepeneur-GPT.yml

   • you can add --debug if you want to see more output
   • you can add --continuous (if you dare) to launch in continuous mode

Help!

• If you encounter a bug, please open an issue
• If you don't know what the commands described above do, don't just run them, read the documentation first
• If you get stuck, it doesn't work, but you know about things like Docker, docker-compose, git.. feel free to reach out, for example in the related thread on Auto-GPT's discord.
• If you don't know about those things, please refer to their documentation first. (yes I know I wrote it twice)

Shameless plug

• The first ever Auto-GPT instance to write mastodon toots is AgentOneAI, you should follow it at https://mstdn.social/@AgentOne

Contributors

• Peter "ppetermann" Petermann (https://devedge.eu)