Requirements

This plugin requires WordPress version 5.0 or higher and PHP 7.2 or higher.

Installation Steps

1. Download the MxChat plugin from the WordPress Plugin Directory or install it directly from your WordPress admin panel by searching for “MxChat” under Plugins > Add New.
2. Activate the plugin through the ‘Plugins’ menu in your WordPress dashboard.
3. Navigate to the ‘MxChat’ settings page to configure your OpenAI API key and other settings.

Shortcodes

MxChat provides shortcodes for easy placement of the chatbot on any page or post:

• [mxchat_chatbot floating="yes"]: Displays the chatbot as a floating widget on the page.
• [mxchat_chatbot floating="no"]: Embeds the chatbot directly within the page content.

These shortcodes allow you to choose between a floating or in-page chatbot, depending on your preference and page design.

Appending to Body

To make the chatbot accessible across your entire site, enable the “Append to body” option in the MxChat settings. This setting is located on the main MxChat plugin page, under the “Chatbot” tab. Simply scroll down until you find the toggle labeled “Append to body.” Enabling this option ensures that the chatbot appears on every page, providing site-wide accessibility without the need to add shortcodes manually to individual pages. This is the recommended setup for consistent chatbot presence across your website.

API Key

To configure the MxChat plugin, you will need API keys for the AI models you’d like to integrate. Here’s a breakdown of the API requirements:

• OpenAI API Key (Required): An OpenAI API key is necessary for core functionality, as OpenAI’s models are used for embeddings, intents, and content generation. This key is essential even if you plan to use other models. You can obtain an API key by signing up on the OpenAI platform. After signing up, you must add credits to your account—typically, $5 in credits is sufficient to get started.
• X.AI API Key (Optional, Pro Feature): For advanced conversational capabilities, you can also integrate models from X.AI. Obtain your X.AI API key to access models like `grok-beta` for human-like and context-aware responses. This is available in the Pro version.
• Claude API Key (Optional, Pro Feature): MxChat supports Anthropic’s Claude models for intelligent, nuanced interactions. To use Claude models, such as Claude 3.5 Sonnet, obtain your Claude API key. This feature is also part of the Pro version.

Once you have the necessary API keys, enter them in the MxChat settings page to start using the AI models of your choice.

AI Instructions

AI chatbots rely on instructions to ensure they provide accurate, contextually relevant responses. Clear instructions help the chatbot align with your goals and avoid generating incorrect or unrelated information.

Best Practices

• Avoid Hallucination: Add specific instructions such as:
  • “Do not assume information. Only use data directly from the knowledge base or provided URLs.”
  • “Do not make up links or information.”
• Use Domain-Specific Guidelines: If your chatbot is for a specific industry, instruct it to:
  • “Only respond to queries related to [your domain] or the provided knowledge base.”
• Encourage Transparency: For better trustworthiness, use instructions like:
  • “Always cite source URLs when responding with knowledge base information.”
• Clarify Unclear Queries: Prevent the chatbot from guessing:
  • “Ask the user a clarifying question if their query is unclear.”

By implementing these instructions, your chatbot will deliver more reliable, relevant, and helpful responses.

Customization

MxChat offers a wide range of customization options to ensure the chatbot matches your brand’s look and feel. The Pro version unlocks extensive customization, allowing you to adjust both the appearance and behavior of the chatbot through user-friendly settings within the WordPress panel.

These options enable you to craft a chatbot experience that is consistent with your brand while providing an engaging and personalized experience for users.

Rate Limits

The rate limit feature in MxChat helps you control how many messages users can send within a 24-hour period. This is particularly useful for managing costs and encouraging user engagement. You can configure separate rate limits for logged-in and not-logged-in users to suit your needs.

You’ve reached your daily message limit! Sign up for free to enjoy unlimited chats and personalized recommendations.

OpenAI Models

MxChat supports a range of OpenAI models that cater to different levels of conversational complexity. These models are ideal for both basic and advanced use cases, providing high-quality, intelligent responses.

• GPT-4o (Recommended): OpenAI’s flagship model for handling complex, multi-step tasks, suited for high-level AI interactions.
• GPT-4o-mini: A lightweight version of GPT-4o, balancing speed and intelligence, ideal for simpler conversations.
• GPT-4-turbo: An advanced model with high intelligence, designed for deep reasoning and complex queries.
• GPT-4: The previous generation model, known for its high performance in creative and complex tasks.
• GPT-3.5-turbo: A fast and cost-effective model suitable for straightforward tasks and quick responses.

X.AI Models (Pro Feature)

For advanced conversational needs, MxChat includes models from X.AI, available in the Pro version. X.AI models are designed for human-like interactions and excel in delivering context-aware, natural conversations.

• grok-beta (Early Beta): X.AI’s latest model in early beta, offering advanced language comprehension for highly personalized and context-driven responses.

Claude (Anthropic) Models (Pro Feature)

Anthropic’s Claude models, available in the Pro version, are optimized for high-intelligence, nuanced interactions. They are ideal for users seeking responsive and detailed conversational capabilities.

• Claude 3.5 Sonnet (Most Intelligent): Designed for complex conversations, capable of handling in-depth queries with a high degree of understanding.
• Claude 3 Opus (Highly Complex Tasks): Ideal for intricate tasks that require detailed responses and advanced reasoning.
• Claude 3 Sonnet (Balanced): A versatile model that combines speed and intelligence, fitting most use cases effectively.
• Claude 3 Haiku (Fastest): Optimized for fast responses with reliable accuracy, perfect for quick interactions.

Overview of Intents

In MxChat, intents define specific actions or responses that the chatbot can execute based on user input. Intents allow the chatbot to perform tasks such as capturing emails, answering product inquiries, and assisting with checkout. Each intent is designed to meet specific user needs and enhance the chatbot’s interactivity and functionality. Some intents are available to all users, while others are exclusive to Pro users, offering advanced eCommerce and engagement features.

To determine the user’s intent, MxChat uses vector embeddings, a technique that helps the chatbot understand the meaning behind user messages by comparing sentence similarities. This allows MxChat to recognize and respond to a wide variety of phrases for each intent, even if the phrasing differs. For example, the chatbot can understand that phrases like “I’m ready to checkout” and “go to checkout” mean the same thing and trigger the checkout intent accordingly. This capability enables MxChat to deliver accurate and relevant responses, making interactions feel natural and intuitive for users.

Setting Up the User-to-Agent Slack App

Follow these steps to create and configure the User-to-Agent Slack app:

1. Download and Edit the Manifest

• Download the manifest_website_to_user file: Download User-to-Agent Manifest.
• Open the file in a text editor (e.g., Notepad).
• Replace example.com with your website’s domain (e.g., yourdomain.com).
• Save the changes.

2. Import the Manifest into Slack

• Visit api.slack.com/apps.
• Click “Create New App.”
• Select “From an app manifest.”
• Choose your workspace and paste the contents of the edited manifest file.
• Click “Next” and then “Create.”

3. Gather Required Credentials

After creating the app, collect the following details:

• Signing Secret: Found under “Basic Information” in the app settings. Paste it into the “Slack Secret Key” field in the MxChat settings.
• Bot User OAuth Token: Found under “OAuth & Permissions” after installing the app. Paste it into the “Slack Bot OAuth Token” field in the MxChat settings.
• Webhook URL: Found under “Incoming Webhooks” after activating webhooks. Paste it into the “Slack Webhook URL” field in the MxChat settings.

4. Configure Live Agent and Switch to Chatbot Intents

To enable seamless switching between live agents and the chatbot, configure the following intents in your MxChat settings:

• Live Agent Intent:
  • Go to the “Intents” section of your MxChat settings.
  • Click “Add Intent” and give it a label of your choice, such as Live Agent.
  • Enter key phrases that should trigger this intent (e.g., talk to a person, need help from an agent, live support), separated by commas.
  • In the dropdown menu labeled Action, select Live Agent.
  • Save the intent.
• Switch to Chatbot Intent:
  • Go to the “Intents” section of your MxChat settings.
  • Click “Add Intent” and give it a label of your choice, such as Switch to Chatbot.
  • Enter key phrases that should trigger this intent (e.g., back to bot, no longer need agent, use chatbot), separated by commas.
  • In the dropdown menu labeled Action, select Switch to Chatbot.
  • Save the intent.

These intents allow users to easily transition between live agents and the chatbot during a conversation, ensuring a seamless experience.

Optional: ModSecurity Exceptions

If your hosting provider uses ModSecurity, you may need to add exceptions to your .htaccess file to ensure proper functionality. Add the following:

# BEGIN ModSecurity Exceptions
<IfModule mod_security.c>
    <If "%{REQUEST_URI} =~ m#^/wp-json/mxchat/v1/stream#">
        SecRuleEngine Off
    </If>
    <If "%{REQUEST_URI} =~ m#^/wp-json/mxchat/v1/agent-response#">
        SecRuleEngine Off
    </If>
    <If "%{REQUEST_URI} =~ m#^/wp-json/mxchat/v1/slack-interaction#">
        SecRuleEngine Off
    </If>
</IfModule>
# END ModSecurity Exceptions

Save your changes and verify that Slack communication is working correctly.

Troubleshooting

• Ensure the correct manifest file is used for each app.
• Double-check that all keys (Signing Secret, Bot User OAuth Token, and Webhook URL) are correctly copied into the MxChat settings.
• If issues persist, ensure the .htaccess configuration is applied if using ModSecurity.
• Reach out to support if the issue persists.

Overview

Important Note: MxChat Agents are currently optimized for use with the floating widget implementation only. Support for embedded chatbots will be added in future updates.

The agents will automatically scrape your website’s main landing page to gather context for conversations. Future updates will include more customizable content sources and knowledge base integration.

Agent Groups

General Users

• Detail-Oriented User: Focuses on specific details and asks precise, thorough questions.
• Tech-Savvy User: Tests technical understanding and complex scenarios.
• Casual Browser: Simulates casual website exploration with general inquiries.
• Overly Friendly User: Tests chatbot’s handling of informal interactions.

Customer Types

• Price-Sensitive Buyer: Focuses on pricing, deals, and value propositions.
• Skeptical Buyer: Challenges claims and seeks verification.
• Eco-Conscious Consumer: Inquires about sustainability and ethical practices.
• Custom Order Requester: Tests handling of complex, customized requests.

Specialized Personalities

• Feature Hunter: Explores detailed product specifications and features.
• Researcher: Conducts thorough information gathering.
• Parent/Caregiver: Focuses on safety and family-oriented concerns.
• Policy Enforcer: Tests knowledge of terms, policies, and compliance.

Running Tests

1. Navigate to the MxChat Agent Deployment Center.
2. Enter the URL where your chatbot is implemented.
3. Select the number of messages per agent (4, 6, or 8).
4. Deploy individual agents by clicking their respective “Deploy Agent” buttons.

Coming Soon

• Customizable knowledge base integration
• Multiple page scanning for context
• Embedded chatbot support
• Custom agent creation
• Batch testing capabilities

Submission Methods

There are several ways to add content to the Knowledge Database:

• Manual Submission: Add specific content directly through the MxChat admin dashboard.
• URL Submission: Submit individual URLs, and MxChat will automatically pull relevant content from each URL to include in the knowledge database.
• Sitemap Submission: Add your site’s sitemap, and MxChat will extract content from each listed URL, keeping your database updated with your latest site content.
• WooCommerce Integration: If using WooCommerce, products can be automatically added, updated, or removed in the Knowledge Database, ensuring accurate and up-to-date product information.
• PDF Upload via URL: Submit URLs linking directly to PDF files, and MxChat will extract relevant content to add to the Knowledge Database. (Support for Word documents coming soon!)

With these flexible submission methods, MxChat ensures your most important content is always available and up-to-date in the Knowledge Database.

Query Matching Process

MxChat uses vector embeddings to process user queries and identify the most relevant content in the knowledge database. Here’s how the process works:

• Vector Embedding for User Query: When a user submits a query, MxChat uses vector embeddings to convert it into a numerical format, capturing the essential meaning and context of the input.
• Comparison with Database Entries: The chatbot then compares this query vector with vectors generated for all entries in the knowledge database. By measuring similarity, MxChat determines which content is most relevant to the user’s question.
• Relevant Response Generation: MxChat retrieves the best-matching content from the database and uses it to generate an accurate and context-aware response for the user.

This process enables MxChat to recognize and respond appropriately to various phrasing styles and queries, ensuring a seamless experience for users.

Managing Content

You can easily search, edit, and delete content within the Knowledge Database directly through the MxChat dashboard. This feature allows you to refine, update, and manage the content that the chatbot references, helping keep information accurate and relevant.