Links
Comment on page
🧪

Advanced Mode (Experimental)

Chat with a containerized application with custom tools

Overview

Advanced mode allows you to define a containerized application with your own set of tools. You can prepare your use cases on that containerized environment and allow your team members to chat with it from within Kubiya.
Each containerized application is defined as a custom profile, utilizing stateless container technology and AI agents to execute a wide array of tasks. This document provides a technical overview of the architecture and functionalities of Kubiya’s custom copilots.

Built in tools

Managed containers (available in the free trial and sandbox) are equipped with the following tools:
  • Terraform CLI
  • AWS CLI
  • GH (GitHub CLI)
  • Bash
  • Python (3.11)

Architecture

Custom Copilots are built on top of stateless container technology. These containers serve as isolated execution environments where a set of tools and applications can be deployed.
Custom profiles can be managed using a source control for easy maintenance of the core instructions for the AI on the specified profile environment
Example profile with Terraform enabled - creating an sqs queue by reading thourgh the needed variables

Key Components:

  1. 1.
    Stateless Containers: The containers are ephemeral, initializing with a predefined configuration and terminating after the execution is complete. This ensures a consistent environment for every execution cycle.
  2. 2.
    Can run on your own environment using a local runner (Kubernetes)
  3. 3.
    AI Agent: An AI-powered agent resides within the container, acting as the executor of tasks. It interprets natural language commands and interacts with the tools and applications within the container to perform the specified actions.

Functionalities

  1. 1.
    Tool Integration: Containers can be provisioned with a wide variety of tools according to the use case requirements, such as DevOps tools like Terraform or cloud interfaces like AWS CLI.
  2. 2.
    Natural Language Interface: Users can interact with the copilots using natural language commands, making the system accessible to users with varying levels of technical expertise.
  3. 3.
    Ability to plan and execute code
  4. 4.
    Easy definition of use cases using simple yaml specification

Deployment Options

  1. 1.
    On-Premises: For users requiring bespoke configurations and control over the execution environment, custom copilots can be deployed on private infrastructure.
  2. 2.
    SaaS Platform: For ease of use and maintenance, Kubiya offers a SaaS platform where custom copilots with pre-configured tools are available. This option provides streamlined access with some limitations in tool customization.

Usage

  1. 1.
    Command Execution: Users issue commands in natural language, which the AI Agent interprets and executes within the container environment after starting a Slack thread
  2. 2.
    Session Lifecycle: Each session initializes a stateless container with the necessary tools, executes the commands, and then safely terminates the container after a limited amount of time. default is 90 minutes

Limitations

  1. 1.
    Core Tool Customization: While on-premises deployment allows full customization, the SaaS version limits the ability to customize core tools within the container.
​