WinaltChat for WooCommerce
| 开发者 |
winalttechnologies
fahadali158 |
|---|---|
| 更新时间 | 2026年6月2日 04:02 |
| PHP版本: | 7.4 及以上 |
| WordPress版本: | 7.0 |
| 版权: | GPLv2 or later |
| 版权网址: | 版权信息 |
详情介绍:
WinaltChat adds a beautiful, fast AI-powered chat widget to your WooCommerce store. The intelligent assistant helps customers browse your catalogue, answers common questions about shipping and returns, and — when a product is out of stock — can suggest relevant Amazon affiliate alternatives to keep the sale within reach.
Key Features
- Dual AI Modes — Use your own API key from Groq — LLaMA (Free Tier), ChatGPT (OpenAI), Claude (Anthropic), or Gemini (Google) for full conversational AI. No API key? The built-in Basic AI mode works entirely on your server with zero external calls.
- Smart Product Search — Semantic search across your WooCommerce catalogue with intent detection, so shoppers find what they mean, not just what they type.
- Conversation Context — The assistant remembers context within a session (stored in the visitor's browser) so follow-up questions ("which one comes in blue?") work naturally.
- Amazon Affiliate Fallback — When a product is unavailable, optionally surface an Amazon affiliate link so you still earn a commission.
- Fully Brandable — Customise colours, fonts, icons, welcome message, and header text to match your store.
- Mobile-Optimised — Designed for a smooth experience on all screen sizes.
- WordPress 6.0 or later
- WooCommerce 7.0 or later (must be installed and active)
- PHP 7.4 or later
- An API key from Groq, OpenAI, Anthropic, or Google (optional — Basic AI mode requires no key)
- Support Forum: wordpress.org/support/plugin/winaltchat-for-woocommerce
- Email: support@aisalesmanchatpro.com
- GitHub Issues: github.com/fahadali158/winaltchat/issues
安装:
- Upload the
winaltchatfolder to/wp-content/plugins/, or install via the WordPress Plugins screen directly. - Activate the plugin through the Plugins screen in WordPress.
- Make sure WooCommerce is installed and active — this plugin requires it.
- Navigate to WinaltChat in the WordPress admin sidebar.
- On the Settings tab, choose your preferred AI provider and enter your API key (or leave blank to use Basic AI mode).
- Optionally enter your Amazon Associate tag if you want the affiliate fallback feature.
- Customise the chat appearance on the Branding tab.
- The chat widget will now appear on the front end of your store for all visitors.
- Groq — LLaMA (free tier, recommended for getting started): console.groq.com
- OpenAI: platform.openai.com
- Claude (Anthropic): console.anthropic.com
- Gemini (Google): aistudio.google.com
常见问题:
Do I need an AI API key?
No. The plugin ships with a Basic AI mode that runs entirely on your server using WooCommerce product data and a customisable knowledge base. You only need an API key if you want the full conversational AI experience.
Which AI provider do you recommend?
Groq — LLaMA is recommended for most stores. It is free to use, extremely fast, and supports the Llama 3 family of models. ChatGPT (OpenAI) gives the most natural conversations but has usage costs.
Does the plugin store customer messages?
Chat messages are stored only in the visitor's own browser using localStorage, tied to an anonymous session ID that is generated client-side. No conversation data is written to the server database. No cookies are set by this plugin. Data is cleared when the visitor ends the chat or clears their browser storage.
Will the plugin slow down my site?
No. All JavaScript is loaded in the footer, stylesheets are versioned for browser caching, and the chat widget is not rendered until a visitor interacts with the chat icon.
How do I remove all plugin data on uninstall?
Deleting the plugin via the WordPress Plugins screen will automatically remove all WinaltChat settings and options from your database.
Does it work with my theme?
Yes. The chat widget is injected into the page footer and uses its own scoped CSS so it does not conflict with theme styles.
更新日志:
1.1.01
- Fix: Provider dropdown now correctly locks and Remove Key button stays visible after page refresh — fixed hasSavedKey detection in change handler to read data-has-saved-key attribute, and changed .remove() to .hide() so the Remove button is never permanently deleted from the DOM.
- Fix: API key UI now correctly shows Remove Key button and locks the provider dropdown after page refresh — root cause was PHP not setting data-original-value on the input, so JS could not detect the saved key state after reload.
- Fix: Added data-has-saved-key attribute to the key input in PHP — JS initialization now reads this directly instead of relying on trigger('change') which could lose the disabled state.
- Fix: Provider dropdown now correctly locks (disabled) on page load when a key is saved.
- Fix: Remove Key and Show Key buttons now correctly appear on page load when a key is configured — JS now reads PHP-rendered data-has-key attribute instead of inferring state from masked input value.
- Fix: Quick API Setup Guide now hidden on page load when any provider has a saved key.
- Fix: Remove Key and Show Key buttons now correctly appear when an API key is saved — added explicit page-load initialisation that reads the provider from the selected option element directly, fixing a disabled-select edge case where trigger('change') could leave the row hidden or buttons in the wrong state.
- Fix: "Error saving API key" message after successful save eliminated — removed dead call to winalt_fetch_provider_models() that was deleted in the previous release but not cleaned up from the save handler, causing a PHP fatal error.
- Fix: Removed dynamic model selector from admin UI — replaced with fixed smart defaults per provider (Groq: llama-3.3-70b-versatile, OpenAI: gpt-4o, Claude: claude-haiku-4-5, Gemini: gemini-2.0-flash).
- Fix: Removed winalt_fetch_models AJAX handler and all provider model fetch functions — no longer needed.
- Fix: Updated Gemini test and chat models from deprecated gemini-1.5-flash/pro to gemini-2.0-flash.
- Fix: Gemini chat model now reads from saved option, falling back to gemini-2.0-flash.
- Fix: Added support contact section to plugin description.
- Fix: Gemini API key validation now accepts Google AI Studio keys (AQ.Ab prefix) in addition to legacy AIza keys.
- Fix: Provider names updated in readme for clarity (Groq — LLaMA, ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google)).
- Fix: Corrected Tested up to version to 6.7.
- Fix: Cleaned up changelog entry HTML.
- Fix: Contributor username updated.
- Fix: Removed blank gaps between product cards by hiding extra line-break elements injected between card HTML blocks.
- Fix: "Thanks" detection fixed — thx/ty aliases added so acknowledgement fires correctly.
- Fix: Blank gaps between product cards collapsed in PHP and system prompt updated.
- Fix: Variant follow-up after discount response now finds the last single product, not a bulk listing.
- Fix: Markdown headers in AI/KB responses converted to bold text for correct chat display.
- Optimization: Default Claude model changed to claude-haiku-4-5 (10x cheaper). Catalog browse routes to Basic AI. Prompt size reduced ~25%. Estimated 94% cost reduction.
- Fix: Claude models auto-migrated to 4.5 family on first admin page load.
- Feature: Live Model Selection dropdown in admin fetches available models from the provider API with 24-hour cache and manual refresh button.
- Feature: Models auto-fetched on API key save; manual model ID entry option added.
- Fix: Basic AI fallback no longer exposes internal mode details to users.
- Fix: Product searches during fallback return clean unavailable-product responses.
- Feature: Real AI falls back to Basic AI on provider failures (rate limits, timeouts, errors).
- Fix: Failures logged server-side only; never exposed in chat responses.
- Fix: AI responses always use WooCommerce store currency symbol.
- Fix: Duplicate product name bullet and wrong About Us page routing both corrected.
- Fix: Negative sentiment detection expanded for standalone "expensive" and Urdu/English patterns.
- Fix: Cheapest product and price filter routing expanded with more natural phrasings.
- Fix: Standalone budget number replies now correctly trigger the price filter.
- Fix: Price hallucination guard strips false AI price claims; price filter includes result type.
- Fix: Purchase intent, variant query, and location/delivery intent patterns all expanded.
- Fix: Duplicate product name bullets stripped from product card responses.