Features

Telesphorus leads ⚕ A⚕clepiu⚕ into the City

Carl Jung leads the Emperor into the City – by Dall.E3

Cacbot

Cacbot is a WordPress plugin that allows interaction with artificially intelligent agents directly on a WordPress site using the native comments system to record chats. It is written entirely in PHP and works on any site that has WordPress. Cacbots can transform and generate text, create images, execute functions, and perform actions. Since the conversations are captured in the WordPress comment system, you can use the responses any way you can think of! You can provide system instructions and keep them hidden from the user. You can create multi-step reasoning and conversations of any complexity, and keep the data entirely within your WordPress system.

It is intended for developers and admins who want to build WordPress applications that use AI. Try it for free, the Pro version is only $20 a month.

Getting Started

  1. Download the ZIP file.
  2. Upload Cacbot.zip to your WordPress site and activate the plugin.
  3. In the WordPress admin backend, click the Cacbot Conversations tab and select “Add New.”
  4. Give a name to the conversation (e.g., “test”).
  5. Set the conversation visibility to Private and click Publish.
  6. View the conversation on the frontend.
  7. Type something into the comment section. When you publish a comment, the Cacbot Assistant will respond as if spoken to, and the response will be saved as a comment in the database.

Cacbot User Role

Users on WordPress websites have roles and capabilities, and Cacbots are no different. The plugin introduces a user role called Cacbot. In WordPress, users can have multiple roles.

By default, Cacbots have both Cacbot and editor roles. Upon installation, a Cacbot can perform any actions that a WordPress editor can perform. You can adjust this manually, or a Cacbot can modify its own capabilities with an Administrator’s permission.

Cacbot Conversation Custom Post Type

The cacbot-conversation custom post type (CPT) is the basic data structure for communicating with Large Language Models (LLMs) using Cacbot. Conversations (aka “chats”) are mapped into the default WordPress comment system. Since Cacbot Conversations are regular WordPress CPTs, you can use any technique to manipulate them, such as PHP.

Cacbots will intelligently respond and act based on comments published on a cacbot-conversation.

Conversation Structure

A cacbot-conversation is a conversation between two users on a WordPress site. The first user to speak is the Interlocutor. The second speaker is the Respondent. You can set the allowed users by setting _cacbot_interlocutor_user_id and _cacbot_respondent_user_id. One, or both, speakers must be a Cacbot with the user role Cacbot. The other user can be any allowed to comment on the post. If a human user initiates a comment on a cacbot-conversation, the Cacbot Assistant will automatically be assigned as the Respondent.

_cacbot_action_enabled_archive Post Meta

accepts: bool
default: false

A trash-can icon appears if there is at least one comment. Pressing the button archives the Conversations.

_cacbot_action_enabled_fetch_image Post Meta

accepts: bool
default: false

An image icon appears if there are no comments yet. Pressing the button submits the comment field to the Dalle-e-3 model.

Your Cacbot can generate an image and save it directly to the Media Library on your WordPress site. To use this feature:

  1. Install the plugin LH Add Media from URL from the WordPress.org repository.
  2. Open a new Cacbot Conversation.
  3. Create a custom field with the name _cacbot_action_enabled_fetch_image and set it to true.
  4. Describe the image you want.
  5. Click the “Fetch Image” icon. The generated image will be saved to your WordPress media library.

_cacbot_max_replies Post Meta

The maximum number of comments allowed before closing comments.

_cacbot_max_archives Post Meta

The total number of times the user can click the “archive” button.

_cacbot_system_instructions Post Meta

Typical API calls to Large Language Models [LLMs] often have two general parts: the system instruction and the context window. Both of these together constitute the prompt that is passed to the LLM model. In Cacbot, the post’s comments are passed as the context, and the post meta data with the key _cacbot_system_instructions is given to the LLM as the system instruction. If you do not set the Cacbot system instructions, “You are a helpful assistant.” is passed.

You can pass very complex instructions to the LLM. Set the post custom field _cacbot_system_instructions to whatever you want in plain English. If you enter only an integer, then the content of the post with that ID will be used. This includes expanded shortcodes in the post.

You can create system instructions that call WordPress shortcodes. Since shortcodes are functions, you can pass arbitrary information into the LLM this way.

Tip: Use a plugin like Insert Pages to put the content of one post into another post, and then use that as the system instructions. This allows you to create very complex and dynamic instructions. i.e., include all the instructions with a particular tag, or a particular category or user.

_cacbot_interlocutor_user_id Post Meta

The first speaker in a Cacbot Conversation.

_cacbot_respondent_user_id Post Meta

The second speaker in a Cacbot Conversation. If a non-Cacbot user speaks first on a Cacbot Conversation, the Cacbot Assistant will automatically be assigned as the Respondent.

_cacbot_anchor_post Post Meta

When a user makes a comment on an anchor post, the system will create a new Cacbot conversation in the background, record the conversation on that new CPT, and display the comment thread on the anchor post in the frontend. This allows many users to chat with a particular Cacbot on a particular URL.

You can designate any post type as an anchor post. This allows you to enable a Cacbot on any WordPress post or page.

To enable an anchor post, set the custom field post meta _cacbot_anchor_post to 1.

_cacbot_comment_form_placeholder_text Post Meta

The placeholder text in the comment form.

Talk to the Cacbot.

Please log in to speak to the Assistant.