Cómo Crear Chatbots de IA Desde Cero (Tutorial Completo de Botpress para Principiantes 2024)

How to Create AI Chatbots Step by Step

Introduction to Chatbots

• The speaker, Nico, introduces the tutorial on creating AI chatbots that can save significant time in work tasks.

• Nico shares his experience helping businesses implement AI and promises viewers will learn to create personalized chatbots ready for business use.

Competitors in the Chatbot Space

• Two main competitors in chatbot creation are identified: Botpress and Voiceflow, likened to a rivalry similar to Barcelona vs. Madrid or Coca-Cola vs. Pepsi.

• Botpress is described as more technical, allowing code execution, while Voiceflow is simpler but lacks some advanced features.

Pricing Plans Overview

• The discussion shifts to pricing; Botpress offers a free plan and a paid team plan at $500 per month.

• The free plan operates on a "Pay As You Go" model, meaning users only pay for what they use rather than a fixed monthly fee.

Features of the Free Plan

• Users can create up to five chatbots with 2,000 messages/events monthly and three collaborators under the free plan.

• Additional features include 100 MB of vector database storage and data retention for 30 days.

Upgrading Features

• For production-ready bots, an "Always Live" feature is necessary so that chatbots remain active at all times.

• Users can add extra features like more messages/events or additional collaborators as needed; costs increase based on usage.

Integrations Available

Creating a Chatbot: Overview and Key Concepts

Initial Setup of the Chatbot

• The process begins with creating a chatbot named "chatbot udia," which sets up the working environment for development.

• Users can choose to start from scratch or use a template; in this case, the speaker opts to start from scratch.

Understanding Chatbots vs. AI Agents

• A distinction is made between chatbots and AI agents; chatbots guide users through predefined paths while utilizing AI at critical points.

• Chatbots are not fully autonomous; they rely on structured interactions rather than complete AI control, which can lead to errors in customer service scenarios.

• The risk of relying solely on AI is highlighted, as even a small error rate can result in significant customer dissatisfaction and lost revenue.

Features of the Chatbot Development Environment

• The chatbot's structure includes workflows for managing conversations, error handling, and timeouts when users are inactive.

• A knowledge base is essential for providing necessary information about the company to enable accurate responses from the chatbot.

Data Management within the Chatbot

• Tables function like integrated spreadsheets (Google Sheets/Excel), allowing storage of user data such as names and inquiries.

• Various types of agents exist within the platform, but focus will primarily be on the knowledge agent that interacts with documents relevant to business queries.

Interaction Design: Cards and User Engagement

• The interface allows for different types of cards (text, audio, image), with text cards being most commonly used for displaying information.

Understanding the Benefits of AI in Programming

The Role of Execute Code in WordPress vs. Voiceflow

• The speaker highlights the advantage of using "execute code" in WordPress, allowing users to integrate custom code without needing professional programming skills.

• Emphasizes that AI tools like ChatGPT can assist in generating this code, making it accessible for non-programmers.

Opportunities for Learning AI

• A new course on generative artificial intelligence is introduced, aimed at proactive individuals looking to master this technology without prior programming knowledge.

• Limited spots are available for the course, which focuses on accelerating careers or entrepreneurship within the AI field.

Types of Artificial Intelligence and Their Applications

• Discussion includes various types of AI technologies such as Antropic and OpenAI, highlighting their diverse applications.

• Introduces different logical cards used to identify user intentions and manage conversational flows effectively.

Capturing User Information

• The importance of capturing user data through various formats is discussed; examples include email addresses and phone numbers.

• Specific use cases for capturing unusual data points (like color or temperature) are mentioned but noted as less common.

Interaction Methods with Users

• Explains input methods such as single choice and multiple choice options for user interactions within a chatbot framework.

• Describes how these choices influence the flow of conversation based on user responses.

Building a Chatbot Flow

• Outlines the basic structure of a chatbot including start and end buttons, emphasizing flexibility in designing conversation flows.

Chatbot Development for a Medical Clinic

Overview of Chatbot Functionality

• The chatbot operates by following a structured path from one node to another, allowing it to handle multiple tasks and inquiries indefinitely based on its design.

• An example is introduced where the chatbot is designed for a medical clinic, enabling clients to make reservations and ask questions.

• A real-time emulator allows developers to test the chatbot's responses without needing to deploy it fully, enhancing development efficiency.

Initial Interaction with Users

• When users initiate interaction by typing "Hello," the chatbot begins at the starting node and progresses through its workflow, displaying welcome messages.

• The importance of capturing user input is emphasized; AI transitions are utilized for gathering information effectively during conversations.

Information Gathering and Categorization

• Developers must set up mechanisms to capture user queries. This involves using variables to store user responses for later processing.

• Questions posed by the chatbot need categorization into predefined categories (e.g., appointment booking or general inquiries), which helps streamline responses.

Workflow Management

• Variables in chatbots can be categorized into three levels: workflow (specific conversation), user (across different conversations), and chatbot (general settings).

• The categorization process requires defining specific categories such as "Book Appointment" and "General Inquiry," which guide how the bot responds based on user input.

Advanced Features and User Experience

• The module includes advanced settings that allow developers to enforce category selection, ensuring users provide necessary information for accurate assistance.

• To enhance functionality, an autonomous node can be integrated within the chatbot framework. This acts like a mini-agent capable of providing concise answers based on a knowledge base.

Final Adjustments and Testing

Creating an Autonomous Agent for Knowledge Retrieval

Setting Up the Autonomous Node

• The speaker discusses the creation of an autonomous node that can perform actions based on given instructions, specifically to search for information in a knowledge base related to a clinic.

Establishing a Knowledge Base

• A new knowledge base named "clínica vuda" is created as there was no prior information uploaded about the business. The speaker emphasizes the need to populate this knowledge base with relevant data.

Uploading Information

• Various methods are available for uploading information, including URLs, PDF documents, and direct text input. The speaker opts to upload text directly from their own generated content about "clínica bía."

Testing Information Retrieval

• After setting up the knowledge base, a query is made to retrieve the clinic's address. The system begins searching for this information successfully.

Limitations of Autonomous Nodes

• While autonomous nodes provide flexibility in retrieving information, there are concerns regarding accuracy; even if they work correctly 99% of the time, errors in that 1% can be problematic.

Manual Query Input vs. Autonomous Search

Transitioning to Manual Queries

• Due to concerns over reliability with autonomous nodes, the speaker decides to manually input queries instead of relying solely on automated searches.

Structuring Queries Effectively

• A structured query format is introduced where specific prompts are used within brackets for clarity when using the knowledge agent.

Saving Responses

• Responses from queries must be saved appropriately; a variable named "respuesta cono" is created to store these responses effectively.

Challenges with Response Quality

Evaluating Response Accuracy

• There are instances where responses may not meet expectations or provide excessive detail. This highlights challenges in ensuring concise and relevant answers from automated systems.

Improving User Experience

Workflow for User Queries

Initial Setup and Response Creation

• The user is provided with information they requested, which is summarized into a variable named "respuesta." This variable serves as the foundation for further interactions.

• The system aims to print this summary on-screen. It attempts to refine the response through prompting, seeking a more concise output.

Enhancing User Interaction

• After resolving a user's query, the next step involves asking what else they would like to do. A decision node is introduced to facilitate this interaction.

• A single choice option is implemented, allowing users to indicate if they have further questions or wish to book an appointment.

Managing User Decisions

• The system prompts users about their needs: whether they want to reserve an appointment or still have questions. Unused variables are noted for clarity in workflow management.

• When users express ongoing doubts, the workflow branches accordingly. However, there are challenges in copying nodes within the interface.

Workflow Adjustments and Node Creation

• New nodes are created manually due to technical difficulties with copying existing ones. This includes adding text and AI responses.

• The process of connecting various components of the workflow is demonstrated through copy-pasting actions aimed at maintaining clarity in user queries.

Variable Management and Updates

• An error arises when attempting to connect variables without proper updates. It's crucial that user queries reflect current information accurately.

• To address potential confusion regarding user intentions (e.g., wanting an appointment vs. having doubts), duplicating nodes helps clarify these paths.

Finalizing Workflow Logic

• By updating variables based on user input, the system ensures that it can effectively respond to repeated inquiries about services or appointments.

• The final structure allows seamless transitions between different types of user requests while ensuring all relevant information is captured correctly.

Testing and Iteration

Connecting Botpress to Make for Appointment Scheduling

Setting Up the Appointment Booking Process

• The process begins with defining a date and collecting user information such as name and email for appointment scheduling.

• A text input is created to ask users for their preferred date and time, which will be stored in a variable named "fecha."

• Additional inputs are requested from the user: their name (stored in "nombre") and email address (stored in "correo"), ensuring these variables maintain consistency across conversations.

Creating a Data Table for User Information

• A new table called "reservar" is established to store user details like name and email, facilitating organized data management.

• Columns are added to the table for both name and email, with adjustments made to ensure proper visibility of these fields.

Inserting Records into the Database

• An action card titled "Insert Records" is created, allowing the system to save user information into the previously defined table.

• The user's name and email are linked correctly within this action card, ensuring that data is accurately recorded.

Connecting with Make Automation

• To integrate with Make, an execution command is set up. Variables must be consistently named throughout this process.

• Key variables such as date, name, and email are prepared for sending via a webhook to Make's automation system.

Handling Webhook Responses

• A webhook will send collected data to Make; upon completion of tasks, it returns a response indicating success or failure.

Webhook and Google Calendar Integration

Setting Up the Webhook

• The speaker discusses changing the webhook, noting that each user will have a unique identifier (e.g., "ae9"). Users must copy and paste their specific webhook into the setup.

• A variable for date is created to adjust for WordPress's time zone discrepancies. If users do not face this issue, they can skip this step.

Creating Events in Google Calendar

• The process involves creating an event in Google Calendar with a start date and time, as well as an end time that adds one hour to the start time. No additional modifications are made at this stage.

• A response message is prepared to confirm session reservations, including sending a follow-up email with details about the reservation.

Email Notification Setup

• An email template is designed to thank users for their reservation and inform them of a reminder email sent prior to their consultation.

Chatbot Functionality Demonstration

Testing Chatbot Interaction

• The speaker demonstrates how users can interact with the chatbot by requesting an appointment, providing necessary details like name and email.

• Confirmation of successful session booking is shown through feedback from the chatbot, which also indicates that both calendar events and emails have been created.

Enhancing User Experience

• After booking confirmation, the chatbot prompts users if they have any further questions or wish to end the conversation. This ensures continuous engagement until all inquiries are addressed.

Finalizing Chatbot Design

Looping Conversation Structure

• The design includes a loop where users cannot exit without either making another appointment or confirming they have no further questions. This strategy aims to encourage bookings even from hesitant users.

Publishing on a Website

• Instructions are provided on how to publish the chatbot outside of WordPress Studio, allowing it to be integrated into external websites.

Styling Your Chatbot

Customizing Appearance

• The speaker emphasizes improving aesthetics using CSS stylesheets available online. This customization enhances user experience by making chatbots visually appealing.

Final Adjustments Before Launch

Chatbot Development and Integration

Overview of Chatbot Features

• The discussion begins with the introduction of a chatbot named "tu día," which can be customized to provide various information, such as health assistance.

• Users can upload images for the chatbot interface, but it is noted that images may stretch if not formatted correctly. A recommendation is made to avoid including contact information due to aesthetic concerns.

• The process includes setting up a state URL for the chatbot, which will change its appearance on the platform. This customization allows users to tailor their experience.

Technical Setup and Integration

• There is an emphasis on integrating the chatbot into a website, specifically mentioning WordPress and other HTML-based sites. The integration process involves using specific codes provided by the service.

• The speaker mentions using Hostinger for website management and describes how to copy integration code into their site settings. This step is crucial for enabling functionality.

Finalizing Chatbot Appearance