Chatbot with ChatGPT Wordpress

SmartSearchWP transforms your WordPress website into an intelligent AI-powered assistant. Instead of adding a basic chatbot widget, SmartSearchWP builds a structured semantic intelligence layer on top of your content — allowing visitors to ask questions and receive accurate, contextual answers based on your own data. Whether you run an eCommerce store, corporate site, blog, or documentation platform, SmartSearchWP improves engagement, reduces support workload, and enhances content discovery. Why SmartSearchWP is different Many AI chatbot plugins simply connect your site to OpenAI. SmartSearchWP is built differently.

• AI-powered semantic search engine
• Structured contextual content retrieval using embeddings
• Modular architecture with dedicated addons
• Advanced usage and access control
• Full prompt customization
• WooCommerce and PDF integration

It is not just an AI chatbot — it is a structured intelligence layer for WordPress. Built for Professionals SmartSearchWP is designed for developers, agencies, and serious WordPress site owners who need performance, control, and scalability — not just a basic AI widget. Core Features AI Chatbot Integration Connect your website to OpenAI's ChatGPT models to provide dynamic, contextual answers based on your own content. All available ChatGPT models are presented in the plugin settings depending on your API key and OpenAI account. Supported models include Chat models (gpt-4o, gpt-4o-mini, gpt-4.1, etc.), Reasoning models (gpt-5, o3, o4-mini, etc.), and legacy options — each with recommendations, estimated cost levels, and suitability guidance for chatbot use. Semantic Search Engine Improve content discovery with AI-driven semantic search instead of simple keyword matching. Contextual Content Indexing Index posts, pages, custom post types, and structured data to deliver accurate AI responses. Modular Architecture Extend functionality through dedicated addons without bloating your core installation. Advanced Control System Restrict AI usage by role, page, device, schedule, or quota to maintain full control over behavior and API costs. Prompt Customization Override and customize the system prompt to control tone, scope, and assistant behavior. WooCommerce Integration Turn your WooCommerce store into an intelligent AI-powered product assistant capable of answering customer questions contextually. PDF Integration Include PDF documents in your AI knowledge base to expand contextual understanding. Performance & Safety SmartSearchWP follows WordPress development standards and keeps a clean separation between core functionality and extensions.

• Lightweight architecture
• API-based processing via OpenAI
• No unnecessary front-end bloat
• Full administrative control
• Easy activation and deactivation

SmartSearchWP does not store unnecessary data and allows you to manage and limit API usage efficiently. Use Cases

• eCommerce stores providing instant product assistance
• Corporate websites reducing support requests
• Blogs with large content libraries improving discoverability
• Technical documentation sites offering intelligent Q&A
• Agencies deploying scalable AI solutions for clients

Addons (premium version [https://www.smartsearchwp.com/product/license-premium/]):

• WooCommerce integration
• Custom Post Type & ACF
• PDF integration
• Prompt customization The chatbot uses OpenAI as a third-party service to generate its answers. You can find more information about OpenAI here and consult their privacy policy.

2.8.4

• New: Email alerts on Error Logs — configure recipients and alert types (quota, invalid API key, rate limit, server errors) with 6-hour anti-spam throttle
• New: Test alert email and last-sent banner on the Logs tab
• New: Optional setting to hide the front chatbot when OpenAI reports insufficient quota / billing (default off; auto-restore after TTL or successful API / key validation)
• Developer: action wdgpt_error_logged after chatbot errors are persisted

2.8.3

• Fix: Contexts admin page no longer appears blank when OpenAI returns temporary API errors (5xx, rate limits, connectivity issues)
• Fix: OpenAI API validation distinguishes invalid keys from transient server errors instead of treating every non-200 response as an invalid key
• Fix: Settings page shows the correct warning during OpenAI outages; validation results are cached briefly to avoid repeated live checks on each admin load
• Developer: read-only filter wdgpt_rag_debug_retrieval_result after retrieval selection (including empty contexts); Chat Logs row link to RAG Debugger addon when active

2.8.2

• New read-only WDGPT_Logs_Repository for aggregated chat log metrics (conversations, messages, daily series, top pages, top first questions, no-answer patterns)
• Action wdgpt_conversation_logged fired after each persisted conversation turn (for Analytics addon cache invalidation)

2.8.1

• Front-end hooks for premium addons: filter wdgpt_chatbot_front_config (extends wdChatbotData) and action wdgpt_chatbot_ui_before_input with UI slot #wdgpt-addon-ui-slot for in-chat addon UI (e.g. Lead Capture)
• Plugins page: contextual license notice under active SmartSearchWP addon rows (links to the matching addon in Addons manager or License settings)

2.8.0

• RAG grounding: detect product identifiers (SKU, ref_xxx) and allow addons to force matching products into context before semantic ranking
• Prompt guardrails against inventing prices or statistics not present in the provided context
• Filters wdgpt_retrieval_product_identifiers, wdgpt_retrieval_force_contexts, wdgpt_answer_after_stream for WooCommerce and extensions
• Debug retrieval logging when SmartSearchWP debug mode is enabled

2.7.5

• Fix: Contexts (embeddings) admin tooltip no longer claims automatic generation on save — manual Generate Embeddings + Activate remains the default workflow
• Contexts page notice pointing to the premium RAG Auto Sync addon for automated indexing (with API cost awareness)
• New wdgpt_generate_embeddings_for_post() helper and wdgpt_before_generate_embeddings filter for addon integration

2.7.4

• Fix: proper handling of network/WAF errors on chatbot retrieve-prompt requests (no more raw Forbidden/error page displayed in the chat bubble)
• Automatic retry with backoff, admin-ajax.php fallback when /wp-json/ is blocked, localized user message, and console diagnostic logging
• FAQ note for hosts that must whitelist /wp-json/wdgpt/v1/retrieve-prompt

2.7.3

• Fix truncated or garbled assistant text in admin chat logs and scheduled PDF/email reports when using Chat Completions streaming (e.g. gpt-4.1-mini): buffer incomplete SSE lines server-side before JSON decode — cURL can split data: events mid-line, which previously dropped delta chunks while the live chat still looked correct
• Log storage: assistant messages use wp_kses_post() (safe HTML such as <br> and links); user messages use sanitize_textarea_field() for longer questions and UTF-8 text

2.7.2

• Fix chatbot response truncated or overwritten during streaming (only end of message displayed)
• More robust SSE parsing: use proper JSON.parse instead of fragile string extraction for stream chunks
• Simplified streaming display: single content span to avoid overwrites and duplication when links appear
• Safety net: on stream completion [DONE], re-render full answer to ensure correct display even if streaming glitched
• Cleanup of chunk queue and writing state on completion to prevent memory leaks

2.7.1

• Add "Download discussion" link in chat logs page to export a single conversation as .txt file (questions and answers)
• Fix multi-encoded HTML entities in streamed chatbot responses (&... on special chars, paths, code blocks)
• Use raw text buffer during streaming instead of re-processing rendered HTML through markdown (prevents cascading encoding)
• Fix gpt-5-nano stuck without response: use text.verbosity for OpenAI Responses API (parameter moved from root level in 2025)

2.7.0

• Added support for all available ChatGPT models (Chat, Reasoning, Frontier) depending on API key and OpenAI account
• New model selection with optgroups: "Recommended for chatbot" and "Other models"
• Model characteristics: recommendations, warnings, estimated cost level (Low/Medium/High)
• Icons and visual indicators for each model's suitability
• Support for reasoning models (gpt-5, o3, o4-mini, o3-mini, gpt-5.4, gpt-5-mini, gpt-5-nano) via Responses API

2.6.4

• Updated plugin description and readme content for WordPress.org
• Fix multi-encoded HTML entities in chatbot responses (paths, code, special chars displayed as &...)
• Add decodeHtmlEntities to restore readable display of over-encoded content in streamed responses and localStorage messages

2.6.3

• 100% compatibility with WordPress 6.9.1
• Fix undefined property and variable in answer generator (OpenAI client, error handling)
• Fix cosine similarity when embedding vectors have non-sequential keys (prevents PHP notices)

2.6.2

• Added export all conversations feature in chat logs page
• Fixed undefined array key error in logs table

2.6.1

• Fix welcome and greeting messages display issue
• Improve JavaScript handling for chat bubble text retrieval

2.6.0

• Add "Purge Transients" button in general settings to clear addons cache
• Improve error handling in API calls with try-catch blocks
• Enhanced stability when OpenAI API calls fail

2.5.6

• Corrected formatting of the URL for the 'get a free licence' button in the actions menu

2.5.5

• Official support for the Usage Control addon: restrict chatbot access by user role, page, post type, and device (mobile/desktop).
• Improved extensibility for future addons.
• Various fixes and optimisations.

2.5.4

• Support for the Prompt Override addon: ability to customise the prompt sent to the AI.
• Improved addon integration.
• Various fixes and optimisations.

2.5.3

• Modification of OpenAI API key verification.

2.5.2

• Debugging functionality

2.5.1

• Added the option of displaying the licence key in plain text in the back office
• Added new logs for licence management
• Better licence cache management

2.5.0

• New debugging functionality added to the plugin
• Better management of licence renewal dates

2.4.12

• Added identification of private pages in the list of pages and posts
• Improving contexts that are too long, mainly for pdf files
• Improved compatibility for version php8.1

2.4.11

• Update translations for all existing languages
• Improved licence key management for back office performance
• Add plugin version to general settings

2.4.10

• Update translations for all existing languages
• Add German (de_DE) translations
• Add Spanish (es_ES) translations
• Add bulk action to generate embeddings, activate or deactivate pages and posts
• Fix bug with 0 & 1 replacement in answers (Thanks to Robert Grygier)

2.4.9

• Update translations for all existing languages
• Add Italian (it_IT) translation
• Fix translation display and warnings in POT files
• Fixed default clause in switch statements

2.4.8

• Change deprecated function FILTER_SANITIZE to FILTER_SANITIZE_FULL_SPECIAL_CHARS
• Add norwegian langage (thanks to Robert Grygier)

2.4.7

• Fix embeddings generation if context is too big
• Add compatibility for generating context with pdf file (add-on with pdf compatibility will be required, soon available)

2.4.6

• Resolved an issue which might expose openAI key.

2.4.5

• Resolved an issue to SQL & XSS injection in the chatbot interface and user restriction to purge logs.

2.4.4

• Resolved an issue where the answer stream was truncate.

2.4.3

• Resolved an issue where the answer couldn't be written in the interface due to empty chunks.

2.4.2

• Resolved a problem where the Text-to-Speech function was incorrectly processing HTML and Markdown tags in the speech.
• Introduced notifications within the admin panel to alert users about pending updates for addons.
• Corrected a glitch causing the chat interface to scroll to the top upon submitting a prompt.
• Addressed an issue encountered when viewing the chatbot on a WooCommerce product page.

2.4.1

• Resolved a visual inconsistency issue with the Text-to-Speech (TTS) button on specific websites.
• Introduced a feature that allows users to halt the Text-to-Speech function by clicking the button a second time after it has been activated.
• Implemented a prevention mechanism to avoid the accumulation of Text-to-Speech tasks in the queue, enhancing the overall user experience.

2.4.0

• The chat now includes built-in Text-to-Speech and speech recognition features! A new button, located to the left of the input field, has been introduced to capture and transcribe user's spoken input. Additionally, each message can be audibly read out using the Text-to-Speech option.

2.3.2

• Resolved an issue where existing CSS rules on a WordPress site could potentially override the visual style of the chatbot plugin.

2.3.1

• The small notification bubble above the chat icon has been reinstated! It's now compatible with the customization addon and will display correctly based on the position set within the addon.
• We've eliminated the random number that used to appear on top of the chatbot icon, as the notification bubble now serves the same function.
• The greeting message is now customizable! If you don't alter them in the settings, it will default to the standard translations. This feature has also been implemented for the chatbot bubble.

2.3.0

• Cleaned up contexts' embeddings generation.
• Prevent the user from regenerating embeddings when there is no need to.

2.2.1

• Fixed a pagination issue inside the contexts' interface, and only show contexts based on the current language if the user is using WPML.

2.2.0

• Upgraded to gpt4-o from gpt-4 for faster answer generation and reduced token cost. This change is automatically applied within your plugin.
• With gpt4-o, the generated answers can be up to three times longer compared to gpt4 and gpt-3.5-turbo.
• Resolved an intermittent issue where the addons' catalog was not properly receiving data from the website.
• Fixed a problem with the premium license verification. It should now renew itself correctly. This issue did not affect your license but was not properly automated.
• Added a clickable permalink next to the context's title in the contexts' interface for easy page access.
• Integrated WPML support in the contexts' interface to identify the page language, useful when dealing with pages of the same name in different languages.
• Resolved an issue where Markdown was not always correctly applied when generating answers. This fix is also applied to the logs.
• Introduced a filter in the context's interface to display only posts or pages with the selected status, such as active or having embeddings.
• Fixed an issue where the search box was not functioning correctly under certain conditions.
• Corrected an issue where translations were not properly applied in the addons manager during premium license verification.

2.1.0

• Rectified a situation wherein the image and title of an addon within the catalog were mistakenly perceived as buttons, rather than directing to the appropriate addon.
• Resolved an issue occurring when the user's API key lacked credit, resulting in improper verification of results.
• Presenting a fresh look! The chat interface now boasts an enhanced default appearance compared to its predecessor.
• Addressed an issue where the disclaimer occasionally extended beyond the confines of the chat frame, subsequently augmenting its size.
• Implemented spacing between the input field and the send button for improved user experience.
• Enhanced aesthetics and space efficiency by substituting the chat's name with its icon when generating responses.
• Optimized the ellipsis used as a loading indicator for improved functionality.

2.0.0

• Premium addons are now available! You can access them via the "Addons" submenu within the admin page. There's also a free license available for any potential free addons.
• We've resolved an issue where the chatbot container would occasionally hide behind other elements. The container will now consistently maintain its priority.
• Fixed an issue that could sometimes prevent the user from deactivating the plugin.
• Added a verification for the required 'allow_url_fopen' server setting.

1.9.0

• Added a new disclaimer below the chat interface to inform users that artificial intelligence can occasionally make mistakes. Users are advised to verify critical information.
• Improved answer generation for a more seamless reading experience for users.
• Fixed an issue that prevented links from being clickable while the chatbot was generating a response.
• Resolved an issue where certain segments of words could be omitted during answer generation, leading to potential problems in the output.
• Addressed an issue where certain characters were displayed as their HTML codes instead of their readable forms (e.g., € was displayed as €).
• Corrected a problem where the general settings sliders did not update their values correctly upon page reload.
• Enhanced visual appeal! The answers now support markdown and parse them for users to enjoy a better experience, such as bold or italics. This is also in effect inside the chat logs and the reporting.

1.1.8.4

• Fixed an issue that could occur if the posts had a special encoding.

1.1.8.3

• Fixed an issue with the counting of tokens when using GPT-4.

1.1.8.2

• Added a verification for users that had updated to 1.1.8 before.

1.1.8.1

• Fixed an issue that prevented some users from correctly updating the database.

1.1.8

• Added a new validation inside the admin page to verify that the API key is valid.
• Added a new validation inside the admin page to verify if your current API key is able to use gpt-4.
• Added a reporting that the current API key has insufficient credits to be used.
• Added a new validation inside the admin page to let the user know if the OpenAI API is having issues.
• Added a toggle to enable/disable the chatbot on the front-end.
• Fixed an issue with translations not being loaded correctly in the administration page.
• Added a new color inside the administration interface for contexts that have been activated, but their embeddings are not "up to date" with the content of the page.
• Added a legend on top of the contexts table to explain the different colors.
• Added a new chat bubble next to the interface to better show the chatbot to the users. This bubble can be hidden directly, or 5 seconds after the page has been loaded.
• Added an option to hide the chat bubble completely.
• Added a new feature to receive, by mails, a PDF that will contain the conversations made by users in the last day or week. This report will run through a cron every day.

1.1.7.1

• Fixed an issue with error reporting.

1.1.7

• Fixed an issue with error reporting on some PHP versions.
• Fixed an issue with the answer generation with bad formatting of posts.

1.1.6

• Fixed an issue with bigger contexts using the wrong model.

1.1.5

• Added missing vendor files.

1.1.4

• Added a new feature to have the chatbot answer "word by word" instead of waiting for the full answer to be generated.
• The new feature will also make the chatbot answer faster, especially on long answers.
• The chatbot will no longer automatically scroll to the bottom when a new answer is generated. This allows the user to start reading the answer while it is being generated.
• Further fixed the clickable links in the chatbot interface to prevent them from not being interpreted correctly.
• Fixed an issue with the chatbot logs delete function on some PHP versions.
• Fixed an issue with the chatbot interface in some resolutions.
• Fixed an issue when holding a conversation with the chatbot for a long time.
• Fixed an issue with the chatbot interface when the user is typing a message while the chatbot is generating an answer.
• Fixed an issue when trying to find the most relevant page to answer a question on certain posts.
• Fixed an issue when using the previous messages in the conversation when trying to generate an answer.

1.1.3

• Fixed an issue with previous versions of PHP.
• Updated minimum PHP version to 7.4.

1.1.2

• Fix an issue with some model versions for retrieving the chatbot's answer.

1.1.1

• Fix issue with some PHP versions.

1.1.0

• Answer time should now be faster.
• Optimized the chatbot interface for better performance.
• Improved the answer generation algorithm for better quality, especially with longer answers.
• Added a new "reset conversation" button to the chatbot interface.
• Introduced an "Expand/Compress" button to adjust the width of the chatbot interface.
• Added colour to better differentiate activated and deactivated pages in the summaries page.
• Fixed display issues with certain links and line breaks in the chatbot interface.
• Fixed display issues with certain links and line breaks inside the chatbot logs admin page.

1.0.3

• Fix issue on the chatbot interface on certain sites.

1.0.2

• Add link to the SmartSearchWP website in the description of the plugin.

1.0.1

• Add chatgpt Model selection in settings (chatgtp3.5-turbo or chatgpt-4)
• Some CSS fixe

1.0.0

• Initial release.