Kodee-demo: Hostinger's AI-powered chat assistant

Welcome to the Kodee-demo repository! This open-source project provides an example of how you can implement your own LLM-based AI chat assistant like Kodee from Hostinger. Below, you will find detailed instructions on how to set up, run, and interact with it.

Table of contents

1. Introduction
2. Features
3. Installation
4. Setup
5. API endpoints

Introduction

Kodee is an AI assistant developed by Hostinger, a web hosting company, to help users manage their websites and hosting services effortlessly. It provides instant support, answers, and guidance across various Hostinger products and platforms, including hPanel, Hostinger Website Builder, WordPress, VPS hosting, and more.

Kodee enhances user experience by addressing a wide range of queries, performing tasks, and even solving issues automatically. Whether you're a beginner or an experienced webmaster, Kodee is your go-to companion for building and maintaining a website.

You can read the whole story of how we built Kodee and find Kodee demo set up instructions in our blog post: How we built one of the most advanced LLM-based chat assistants: Lessons learned and tips

Installation steps

1. Clone the repository: Start by pulling the Kodee-demo repository to your local machine using Git.

2. Set up environment variables:

   1. Copy the .env.example file to a new file named .env in the same directory. This file contains template environment variables needed for the project.
   2. Obtain the necessary .env values. These may include API keys, database credentials, and other sensitive configurations.
   3. In this example project, DB_HOST is set to 'db' and REDIS_HOST set to 'redis' as they are the names of the Docker containers. You can change these values if needed.

3. Run the setup command: With Docker running, execute the following command in the terminal within the project directory. This command builds and starts the necessary Docker containers for Redis, Postgres, and the application itself. It also runs automatic migrations to set up your Postgres database.

   make setup
   

4. For subsequent runs: Once the initial setup is complete, starting the project is even simpler. Just use the following command:

   make up
   

This command starts all the necessary services. Your Kodee-demo environment should now be up and running, ready for use.

API endpoints

Initialize chat session

Endpoint: POST /api/chat/initialization

Example request:

{
  "user_id": "test",
  "metadata": {
    "domain_name": "hostinger.com"
  }
}

Example response:

{
  "conversation_id": "7ef3715a-2f47-4a09-b108-78dd6a31ea17",
  "history": []
}

Respond to chat

Endpoint: POST /api/chat/respond

Example request:

{
  "user_id": "test",
  "role": "user",
  "content": "Hello",
  "chatbot_label": "chatbot"
}

Example response:

{
  "conversation_id": "76f9c06b-b6ce-498e-8eb4-050c1eeaa358",
  "message": {
    "role": "assistant",
    "content": "Hello! How can I assist you today?"
  },
  "handoff": {
    "should_handoff": false
  }
}

Restart chat

Endpoint: POST /api/chat/restart

Example request:

{
  "user_id": "test"
}

Example response:

{
  "conversation_id": "7263698e-9ff5-4bb2-b8cf-b7fa7aedbb6d",
  "history": []
}

Query events

Endpoint: GET /api/history/events?conversation_id=<conversation_id>

Example response:

{
  "status": "success",
  "data": [
    {
      "id": 22,
      "conversation_id": "7263698e-9ff5-4bb2-b8cf-b7fa7aedbb6d",
      "event_type": "user",
      "payload": {
        "content": {
          "role": "user",
          "content": "Hello"
        }
      },
      "message_part_id": "test-a8f5995c-f244-4f71-9f53-77287f463d14",
      "created_at": "2024-09-15T15:01:13.748932"
    },
    {
      "id": 23,
      "conversation_id": "7263698e-9ff5-4bb2-b8cf-b7fa7aedbb6d",
      "event_type": "cs_chatbot_label",
      "payload": {
        "content": {
          "message": "Setting active chatbot label for conversation",
          "label": "out_of_scope"
        }
      },
      "message_part_id": "test-5f864306-65f3-4fd7-890d-f8ed59d84b72",
      "created_at": "2024-09-15T15:01:15.279001"
    },
    {
      "id": 24,
      "conversation_id": "7263698e-9ff5-4bb2-b8cf-b7fa7aedbb6d",
      "event_type": "assistant",
      "payload": {
        "content": {
          "status": "success",
          "message": "Hello! How can I assist you today?",
          "chatbot_label": "out_of_scope_bot"
        }
      },
      "message_part_id": "test-5f864306-65f3-4fd7-890d-f8ed59d84b72",
      "created_at": "2024-09-15T15:01:15.908485"
    }
  ]
}

Query messages

Endpoint: GET /api/history/messages?conversation_id=<conversation_id>

Example response:

{
  "status": "success",
  "data": [
    {
      "id": 14,
      "conversation_id": "7263698e-9ff5-4bb2-b8cf-b7fa7aedbb6d",
      "author_type": "user",
      "message": "Hello",
      "chatbot_label": "chatbot",
      "message_part_id": "test-a8f5995c-f244-4f71-9f53-77287f463d14",
      "created_at": "2024-09-15T15:01:13.752007"
    },
    {
      "id": 15,
      "conversation_id": "7263698e-9ff5-4bb2-b8cf-b7fa7aedbb6d",
      "author_type": "assistant",
      "message": "Hello! How can I assist you today?",
      "chatbot_label": "out_of_scope_bot",
      "message_part_id": "test-5f864306-65f3-4fd7-890d-f8ed59d84b72",
      "created_at": "2024-09-15T15:01:15.910829"
    }
  ]
}