CrewAI Agents Docker Image

• Motivation
• Prerequisites
• Installation
• Usage
• Tools
• Supported Versions
• Tips
• Example Projects
• Screen Captures
• Issues
• License

Motivation

This Docker image provides a convenient & secure way to create and run CrewAI agents without having to install any dependencies locally. I created the image to experiment with ollama code generation and execution.

Prerequisites

• Docker
• Basic knowledge of python
• Basic knowledge of crewAI
• Ollama installed locally or access to remote AI services like chatGPT
• Desire to learn and have fun

Usage

Starting the container

Run the following command after replacing container_name with the name of your container, project_name with the name of your project and tag with the tag of the image you want to use.

docker run -it --network host --name <container_name> -e P=<project_name> sageil/crewai:<tag> bash

Tip

if you leave out the P completely -e P=<project_name> from the command, a default crew will be created with the name default_crew.

Using locally installed Ollama

1. Changing the container local configuration.
   • Type v . to open neovim
   • Open Lazyvim Explorer using SPACE+e
   • Show hidden files using SHIFT+H
   • Change the model and the API_BASE to http://host.docker.internal:11434 and MODEL a model you have pulled on llama.
2. Running your crew
   • Open lazyvim terminal using CTRL+/
   • Run crewai run

Using remote services like chatGPT

1. Change your selected provider and model:
   • Type v . to open neovim
   • Open Lazyvim Explorer using SPACE+e
   • Show hidden files using SHIFT+H
   • Change the model and the OPENAI_API_KEY to your key and MODEL a model you wish to use.
2. Running your crew
   • Open lazyvim terminal using CTRL+/
   • Run crewai run

Changing model and provider using Code

1. Open your crew's crew.py
2. Add from crewai.llm import LLM to the imports
3. In your crew's CrewBase class create your LLM

 myllm = LLM (
        model='ollama/deepseek-r1:7b',
        base_url="http://localhost:11434",
        temperature=0.2)
    )

4. Assign the LLM to your agent by assigning it to the llm property

    @agent
    def researcher(self) -> Agent:
        return Agent(
            config=self.agents_config['researcher'],
            # tools=[MyCustomTool()], # Example of custom tool, loaded on the beginning of file
            verbose=True, # Print out all actions
            llm=self.myllm
        )

Tip

When working with remote services, you can also remove the --network host part of the command as its only required to allow the container access to the host's network.

Tools

• neoVim Latest stable version built from source neovim
• uv Dependency management tool for Python projects uv
• lazyVim A highly optimized Vim-like editor for Neovim lazyvim
• crewAI Platform for Multi AI Agents Systems official CrewAI documentation

Versions

Available Versions

• crewAI 0.108.0 crewai-tools 0.38.1
• crewAI 0.105.0 crewai-tools 0.37.0
• crewAI 0.102.0 crewai-tools 0.36.0
• crewAI 0.100.0 crewai-tools 0.33.0
• crewAI 0.98.0 crewai-tools 0.32.1
• crewAI 0.95.0 crewai-tools 0.25.8
• crewAI 0.85.0 crewai-tools 0.17.9
• crewAI 0.85.0 crewai-tools 0.17.0
• crewAI 0.83.0 crewai-tools 0.14.0
• crewAI 0.80.0 crewai-tools 0.14.0
• crewAI 0.79.4 crewai-tools 0.14.0
• crewAI 0.76.9 crewai-tools 0.13.4
• crewAI 0.76.2 crewai-tools 0.13.2
• crewAI 0.74.1 crewai-tools 0.13.2
• crewAI 0.70.1 crewai-tools 0.12.1
• crewAI 0.65.2 crewai-tools 0.12.1
• crewAI 0.64.0 crewai-tools 0.12.1
• crewAI 0.61.0 crewai-tools 0.12.1
• crewAI 0.55.2 crewai-tools 0.8.3
• crewAI 0.51.0 crewai-tools 0.8.3
• crewAI 0.41.1 crewai-tools 0.4.26
• crewAI 0.36.0 && crewai-tools 0.4.26

Tips

• v: alias v='nvim & alias vim='nvim'
• Running newcrew <project_name> will create a new crew project with the provided name, install dependencies and configure the project virtual environment.
• You can restart a container after stopping it by using docker container start -ai <container_name>

Example

Veterinary Assistant Crew

ScreenCaptures

Issues

Known issues:

1. Copying from nvim fails due to display driver
2. Icon fonts are not rendered correctly in the container's terminal? Watch. if the video peaked your interest in Wezterm, you can use my configuration from Wezterm configs

New Issues:

Please report other issues you encounter on the Issues including steps to reproduce them.

License

This project is licensed under the MIT License.