chatbot_setup.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# A Chatbot using GPT and a Database\n",
    "This allows multiple chatbot types (e.g. a health coach and a learning assistant) to be created. Multiple chatbot instances can be created per chatbot type (e.g. a health coach for user X and user Y, and a learning assistant for user P and user Q). Both, types and instances are stored with and referenced by an ID (e.g. a UUID) in the database.\n",
    "\n",
    "This notebook is a simplest possible tutorial guiding you to create one type with one instance. Have a look at the **chatbot_setup_advanced.jpynp** for more advanced features.\n",
    "\n",
    "### 1. Preparation\n",
    "\n",
    "##### 1.1. Set the OpenAI API Key and Model\n",
    "Rename the file **chatbot/openai_template.py** to **chatbot/openai.py** and set the following keys.\n",
    "- OPENAI_KEY = \"your OpenAI API key in quotes\"\n",
    "- OPENAI_MODEL = \"model name in quotes, e.g. gpt-3.5-turbo-16k\"\n",
    "\n",
    "##### 1.2. Install openai package\n",
    "<sup><sub>Note: If you are using GitHub Codespaces, the first time you execute a code cell, you might get asked questions from the development environment (IDE). You can say yes to recommended installations, and to installing/enabling reecommended extensions. When asked to **Select Another Kernel**, choose **Python Environments** and select the recommende environment, e.g., **Python 3.10.13**</sub></sup>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%pip install openai"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 1.3. Import Chatbot library"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from chatbot.chatbot import Chatbot"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Create a chatbot of type **health coach** for user **Daniel Müller**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 2.1. Prompt Engineering\n",
    "To create a chatbot, we need to provide prompts. There are three prompts that can be used to customise the behaviour of the chatbot: a ***role***, ***context***, and ***starter*** prompt.\n",
    "\n",
    "**Role Prompt (type_role)**\n",
    "\n",
    "This prompt will be put at the beginning of the utterances sent to openai for completion. If you create multiple chatbots of the same type, this prompt will be the same for each one of them.\n",
    "\n",
    "You can use this prompt to specify the general (user-independent) behaviour of the chatbot, for example specifying the kind of person to enact, the topical context it acts in, and the goal of conversations it conducts with its users."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "my_type_role = \"\"\"\n",
    "    As a digital therapy coach, check in daily with your patient to assess their well-being related to their chronic condition.\n",
    "    Use open-ended questions and empathetic dialogue to create a supportive environment.\n",
    "    Reflectively listen and encourage elaboration to assess the patient's detailed condition without directing the topic.\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Context Prompt (instance_context)**\n",
    "\n",
    "This prompt will be put right after the role prompt. If you create multiple chatbots of the same type, this prompt will be used only for one of them. <br />\n",
    "\n",
    "You can use this prompt to specify user-specific behaviour, for example information about the particular user that will use this chatbot instance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "my_instance_context = \"\"\"\n",
    "    Meet Daniel Müller, 52, who is tackling obesity with a therapy plan that includes morning-to-noon intermittent fasting, \n",
    "    thrice-weekly 30-minute swims, and a switch to whole grain bread.\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Starter Prompt (instance_starter)**\n",
    "\n",
    "This prompt will be appended after the role and context prompts and is meant to instruct GPT to create an initial message that opens the conversation with the user. <br />\n",
    "\n",
    "You can use this prompt to instruct GPT to welcome the user, ask initial questions, and how to proceed from there."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "my_instance_starter = \"\"\"\n",
    "    After reviewing the patient's profile, compose a single, very short message that the therapy coach would use to initiate today's check-in conversation with Daniel.\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 2.2. Create Chatbot\n",
    "The following code creates a new Chatbot where the chatbot type is identified by the type_id provided and the chatbot instance by the user_id. These two IDs will be used further below to construct the URL pointing to this chatbot once it is deployed.\n",
    "\n",
    "The type name is only used in the front-end and does not affect the conversational behaviour.\n",
    "\n",
    "Once this code is executed, the chatbot is stored in the database."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bot = Chatbot(\n",
    "    database_file=\"database/chatbot.db\", \n",
    "    type_id=\"c48a13d7\",\n",
    "    user_id=\"fe6d944d\",\n",
    "    type_name=\"Health Coach\",\n",
    "    type_role=my_type_role,\n",
    "    instance_context=my_instance_context,\n",
    "    instance_starter=my_instance_starter\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 2.3. Initiate Conversation Starter (Optional)\n",
    "\n",
    "If the chatbot should open the conversation with an initial message, the following code is executed. This code will execute a request to GPT, with the role, context and startet prompts specified above. The response from GPT (the initial message) will be stored in the database and therefore the user will see the opening message as soon as they access the chatbot."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(bot.start())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Deploy and Disseminate\n",
    "You can now deploy your chatbot. If you are using our guide for deploying to pythonanywhere.com, the URL to be handed out to the user is as follows.\n",
    "\n",
    "**Generic URL**\n",
    "\n",
    "https://[your pythonanywhere user name].pythonanywhere.com/[type id]/[user_id]/chat\n",
    "\n",
    "**For Example**\n",
    "\n",
    "https://monkey23.pythonanywhere.com/c48a13d7/fe6d944d/chat\n",
    "\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.4"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}