MultiAI ChatBot

Version 1.0.4 · WordPress plugin that adds an AI chat widget via the WordPress AI Client (Connectors), your own Google Gemini API key (Google IA), or Ollama, plus admin panel and usage telemetry.

Naming conventions (namespace)

The public widget uses the maicb-* class prefix and the #multch-plugin-root container with data-maicb-root. See docs/NAMING.md. Before publishing, run ./scripts/check-namespace.

Requirements

• WordPress 6.2+ (tested up to 7.0; 7.0+ recommended for cloud AI via Connectors)
• PHP 8.0+
• For WordPress AI: providers connected under Settings → Connectors
• For Google IA: a Google AI (Gemini) API key; WordPress 7.0+ recommended so the admin model list matches Settings → Connectors
• For Ollama: a server reachable from the WordPress host (e.g. http://127.0.0.1:11434)

Installation

WordPress ZIP (without .git)

WordPress does not allow uploading a ZIP that includes the .git folder. Generate the package from the repository:

./scripts/package-plugin

This creates multiai-chatbot.zip, ready for Plugins → Add New → Upload Plugin. The ZIP does not include scripts/ (development tools). For production, always use that ZIP or exclude scripts/ if you deploy via Git (e.g. WP Pusher). Translation compilation (./scripts/compile-languages) is for local development only.

Publishing and Plugin Check

Before submitting to WordPress.org, build and verify the production package:

./scripts/verify-plugin-package

This runs ./scripts/package-plugin, confirms the ZIP excludes scripts/, .github/, .git, and .env, and runs wp plugin check when WP-CLI is available.

Important: Run Plugin Check (or wp plugin check) on the unzipped contents of multiai-chatbot.zip, not on the full Git repository. Scanning the repo includes development-only files (scripts/, .github/) that are not shipped in the ZIP and will produce false positives.

1. Copy the multiai-chatbot folder to wp-content/plugins/ (or use the ZIP above).
2. Activate the plugin under Plugins.
3. Go to MultiAI ChatBot in the admin menu.
4. Under AI Model, choose WordPress AI, Google IA, or Ollama. For WordPress AI, connect providers under Settings → Connectors first. For Google IA, enter your Gemini API key and pick primary/fallback models from the Connectors catalog.
5. After activation, streaming rewrite rules are registered automatically. If the stream does not respond, visit Settings → Permalinks and save again.

Admin panel

Tab	Contents
General	Global widget, welcome message, system prompt, streaming, optional statistics/history
AI Model	Provider (WordPress AI, Google IA, or Ollama), models, API key (Google IA), Ollama URL
Security	Allowed origins, cache, telemetry, IP rate limits, and abuse suspension
Chat Style	CSS presets, custom colors, and widget position
Statistics	Totals, breakdown, and CSV export
History	Conversations in cards (ID CB-YYYY-MM-DD-HH-MM-SS), filters, and message detail

Choosing WordPress AI vs Google IA (Gemini)

You enable one provider at a time under MultiAI ChatBot → AI Model.

	WordPress AI	Google IA
Best for	Connectors already set up; OpenAI, Google Gemini, or Anthropic	Gemini only; key in this plugin or wp-config.php
API key	Settings → Connectors (not stored in this plugin)	AI Model tab or MULTCH_GEMINI_API_KEY
Request path	WordPress AI Client → Connectors	Direct HTTPS to generativelanguage.googleapis.com
WP version	7.0+ recommended	6.2+ (7.0+ recommended for model list from Connectors)
Same Gemini model IDs	Yes	Yes (catalog from Connectors when available)

Data sent to third parties: only when a visitor sends a chat message (message, recent context, system prompt). The API key is never sent to the browser. Local history and telemetry are optional (off by default). See docs/AI-PROVIDERS.md and readme.txt (== External services ==, == Privacy ==) for WordPress.org compliance detail.

Suggested site privacy policy text: Settings → Privacy (registered by the plugin).

AI providers

WordPress AI (Connectors)

• Provider ID: wordpress_ai
• Requires WordPress 7.0+ with the built-in AI Client
• Connect OpenAI, Google (Gemini), or Anthropic under Settings → Connectors (the connectors currently supported by WordPress 7.0+)
• Set a primary model and optional fallback model in MultiAI ChatBot → AI Model
• Optional Google automatic fallback when the configured models fail
• Optional wp-config.php overrides for model preference:

define( 'MULTCH_MODEL', 'gemini-2.5-flash' );
define( 'MULTCH_MODEL_CANDIDATES', 'gpt-4o-mini,claude-sonnet-4-6' );

Google IA (own Gemini API key)

• Provider ID: google_ia
• Uses the Google Generative Language API with an API key you supply (not WordPress Connectors credentials)
• Primary model and fallback model are chosen from the same Gemini model IDs listed by WordPress Connectors (Connectors is used as a catalog only; requests go to Google with your key)
• If the primary model fails (quota, rate limit, or unavailability), the plugin tries the fallback model
• Configure in admin under MultiAI ChatBot → AI Model, or in wp-config.php:

define( 'MULTCH_PROVIDER', 'google_ia' );
define( 'MULTCH_GEMINI_API_KEY', 'your-api-key' );
define( 'MULTCH_GEMINI_MODEL', 'gemini-2.5-flash' );
define( 'MULTCH_GEMINI_MODEL_CANDIDATES', 'gemini-2.5-flash-lite' );

Legacy names CHATBOT_PROVIDER, CHATBOT_GEMINI_API_KEY, CHATBOT_GEMINI_MODEL, and CHATBOT_GEMINI_MODEL_CANDIDATES are also supported.

Ollama

• Provider ID: ollama
• No API key in this plugin
• Default base URL: http://127.0.0.1:11434
• Model: name of the model installed in Ollama (e.g. llama3)

Site usage

Global widget

Enable Show site-wide on the General tab. The widget loads on wp_footer.

Shortcode

[multch_widget]
[multch_widget mode="inline"]

• floating (default): floating button + panel
• inline: panel embedded in the page

Styles

Presets available on the Chat Style tab (visual selector with preview):

ID	Name
default	Sapphire
dark-glass	Midnight
obsidian	Obsidian
minimal	Monochrome
ocean	Aqua
sunset	Ember
forest	Emerald
lavender	Amethyst
plum	Plum

Positions: bottom-right, center-right, bottom-left, center-left, bottom-center.

Optional overrides: primary color, accent, background, text, radius, panel max width and height, font, z-index, animations, and automatic theme via prefers-color-scheme.

Per-page style via shortcode:

[multch_widget preset="ocean" position="bottom-left"]
[multch_widget mode="inline" primary="#059669"]

Export/import theme JSON from the admin (Chat Style tab).

Translations (i18n)

• Source language: English in PHP/JS code (__(), esc_html_e()).
• Spanish: languages/multiai-chatbot-es_ES.po and languages/multiai-chatbot-es_CO.po.
• After editing .po files, compile .mo: ./scripts/compile-languages (or php scripts/compile-languages.php).

REST API

Endpoint	Method	Description
/wp-json/multch/v1/chat	POST	JSON response { answer, meta }
/multch/v1/chat/stream	POST	Simulated streaming (text/plain)

Required headers:

• X-WP-Nonce: REST nonce (wp_rest)
• X-Chat-Session-Id: anonymous session identifier (optional)

Example body:

{
  "message": "Hello",
  "history": [
    { "role": "user", "content": "..." },
    { "role": "assistant", "content": "..." }
  ],
  "currentPath": "/",
  "currentUrl": "https://example.com/"
}

The API key is never exposed to the frontend.

Plugin structure

multiai-chatbot.php
includes/
  class-plugin.php
  admin-settings.php
  api-handler.php
  rest-api.php
  telemetry.php
  enqueue.php
  providers/
    class-provider-wordpress-ai.php
    class-provider-google-ia.php
    class-provider-ollama.php
assets/
  css/
    admin.css
    chatbot.css
  js/
    chatbot.js
uninstall.php

Conversation history

When Store statistics and history is enabled under General (disabled by default), each user/assistant exchange is stored in {prefix}multch_conversations and {prefix}multch_messages.

• Public ID: CB-2026-05-29-14-35-42 (date and time in the site timezone)
• Internal ID: auto-increment number for administration
• Grouped by visitor session (30 minutes of inactivity starts a new conversation)
• The frontend sends conversationId in the body to continue the same thread

Telemetry

When statistics and history are enabled, each chat request logs an event in the {prefix}multch_events table:

• Provider, model, status, latency, error code
• Session hash (no plain IP address)

CSV export from the Statistics tab. On plugin uninstall, the table and options are removed.

Security

• Do not commit API keys to the repository.
• For Google IA, prefer MULTCH_GEMINI_API_KEY in wp-config.php in production instead of storing the key only in the database.
• WordPress AI keys stay in Settings → Connectors; they are never sent to the browser.
• IP rate limiting uses WordPress transients.
• Rotate keys if they were accidentally exposed.

See docs/env.example and docs/AI-PROVIDERS.md for configuration and provider choice.

WordPress.org submission

Before uploading to the plugin directory:

1. Run ./scripts/verify-plugin-package and Plugin Check on the unzipped ZIP, not the full Git repo.
2. Ensure readme.txt documents all optional external services (Connectors: OpenAI, Anthropic, Google Gemini; plus Google IA and Ollama), when data is sent, and privacy links — see == External services == and == Privacy ==.
3. Confirm Stable tag and Version in multiai-chatbot.php match the release you submit.
4. Set Tested up to in readme.txt to the latest stable WordPress major version (currently 7.0). WordPress.org rejects values below the current minimum (for example, 6.8 when 6.9+ is required).

Author

J. Santiago Ravelo Velasco

• GitHub: github.com/JunniorRavelo/multiai-chatbot
• GitHub Sponsors: github.com/sponsors/JunniorRavelo
• LinkedIn: linkedin.com/in/jsravelo

License

This project is distributed under the GNU General Public License v2.0 or later (GPL-2.0-or-later), compatible with the WordPress.org plugin directory requirements.