mindroot 5.6.0

An agent framework using LLMs

Table of Contents

• Gallery
  • Admin Interface
  • Plugin Management
  • 3D Graph Visualization
  • Chain Rule Explanation
  • Character Generation
  • Fantasy Character Creation
  • Morgan's Method
  • Heygen Integration Demo
• Installation
• Configuration
• Starting the Server
• Installing Plugins
• Developers: Creating Plugins and Tool Commands etc.
• Adding Commands to Agents
• API Documentation
• Python SDK
• Key Features

Mindroot

Gallery

Admin Interface

Plugin Management

Computer Use

3D Graph Visualization

Chain Rule Explanation

Character Generation

Fantasy Character Creation

Morgan's Method

Heygen Integration Demos

---

Mindroot is a powerful plugin-based Python framework for creating, deploying, and sharing AI agents and related models. It offers a flexible architecture with indices and a public registry (coming soon) for easily sharing and finding plugins, agents, personas, services, knowledgebases, and apps.

Installation

You can install Mindroot using pip:

(You probably want to create a virtual environment first: python -m venv venv and source venv/bin/activate)

pip install mindroot

For development, you can install the package in editable mode:

git clone https://github.com/mindroot/mindroot.git
cd mindroot
pip install -e .

Configuration

You will need to set environment variables like JWT_SECRET_KEY (anything you want) and LLM provider keys etc.

To enable email verification: env. variable REQUIRE_EMAIL_VERIFY=true

See SMTP core plugin readme

Starting the Server

Go to the directory with the virtual environment and run:

mindroot

or if you want to start the server on a different port:

mindroot -p 8001

Note that you will always need to start from that working directory if you want your settings to be preserved.

Installing Plugins

Most of the commands or services are in plugins that you have to install. Go to the /admin page and find the Plugins section Click on the 'default' Index It should show a list of Plugins I suggest installing all of them.

The first would be either Anthropic (recommended) or OpenAI

Important: You will need either ANTHROPIC_API_KEY or OPENAI_API_KEY set in your environment variables.

To install a plugin just click the Install button and wait 10-20 seconds and you will see a confirmation pop up.

Important: You will need to restart the server to see the changes.

For now I recommend not using the Server control section on the Admin page but rather just hit Control-C in the terminal and then run mindroot again.

Knowledgebase

Go to /admin | Plugins | Install from Github | runvnc/mr_kb

Walkthrough: Custom Agent Using a Knowledgebase

Walkthrough

Developers: Creating Plugins and Tool Commands etc.

See the Plugin Documentation

Adding Commands to Agents

Go to Admin -> Agents Select an agent from the drop down, such as Assistant

There are toggle switches for all of the commands. I recommend avoiding the 'replace_' commands in files for now because they are not reliable.

API Documentation

MindRoot provides a REST API for programmatic interaction with agents. This allows you to integrate AI capabilities into your applications without using the web interface.

For detailed information on available endpoints, authentication, and usage examples, see the API Documentation.

Python SDK

There is also an SDK for Python that interacts with the API. You can install it with pip:

pip install mrsdk

Usage example:

from mrsdk import MindRootClient

client = MindRootClient(api_key="your_api_key_here",
                        base_url="http://localhost:8010")

result = client.run_task(
    agent_name="Assistant",
    instructions="What is the square root of 256? Show your work."
)

print(result["results"])

Further details including how to get a full trace of the agent activity, see the repo here: mrsdk

Key Features:

• Public registry for sharing and finding plugins, agents, personas, models, and knowledgebases
  • https://registry.agenthost.org (work in progress)
  • Customizable and swappable for user-specific registries
• Extensible plugin architecture for adding services, commands, and building arbitrary web apps
• Customizable AI agents with persona definitions
• Intelligent service management based on agent requirements
• Flexible service providers for various AI capabilities
• Plugins can add/use hooks and pipelines such as for modifying prompts, running startup commands, or anything you want
• Easily customizable UI built on Jinja2 and Lit Web Components
• Support for both local and remote AI services
• RAG: easily share, find and use pre-generated embeddings and documents for topic knowledgebases