Attendant - AI Site Search & Content Finder
| 开发者 | sungraizfaryad |
|---|---|
| 更新时间 | 2026年6月19日 20:45 |
| PHP版本: | 8.0 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:
- Auto-discovers your site — scans post types, taxonomies, and custom fields (ACF, MetaBox, WooCommerce) automatically
- Function calling — uses OpenAI's tools API for guaranteed structured search parameters, not fragile text guessing
- Direct database access — queries WordPress directly (no HTTP crawling = faster, more accurate)
- Clickable source buttons — every answer links to the matching pages as buttons, so replies stay short and scannable
- Tap-to-answer chips — when the next step is a choice (refine the search, pick a time), the visitor taps an option instead of typing
- Guided help on no results — when nothing matches, the assistant offers real alternatives drawn from your own content and asks one question at a time to narrow things down
- Conversation history — the chat is remembered in the visitor's own browser, so it survives page changes and refreshes; a "New chat" button and a "Previous chats" list are built in
- Callback / lead capture (optional, off by default) — when the assistant cannot help, it can offer a callback, collect the email the visitor provides (and optionally a phone number and preferred time), and email the request to you
- RAG Q&A — embeds your content, searches by cosine similarity, answers with source links
- Custom Q&A pairs — admins can add specific answers that always take priority
- About-this-site context — a short description you write that helps the assistant understand what your site is for
- Manual or background indexing — index on demand with a live activity log, or let it run in the background; the widget stays hidden until the first index completes so visitors never see half-indexed answers
- Bring Your Own Key — works with your own OpenAI API key, no SaaS subscription required
- Privacy-first — conversation logging and lead capture are both OFF by default; IP addresses are never stored
- Real estate directories (find properties by price, bedrooms, location)
- Job boards (find jobs by type, salary, location, remote)
- WooCommerce stores (find products by price, category, attributes)
- Event sites (find events by date, location, category)
- Doctor/medical directories
- Any WordPress site with custom post types
- PHP 8.0 or higher
- WordPress 6.0 or higher
- OpenAI API key
安装:
- Upload the
attendantfolder to/wp-content/plugins/ - Activate the plugin in Plugins → Installed Plugins
- Go to Attendant → Settings
- Enter your OpenAI API key and click Save Settings
- Click Test Connection to confirm your key is working
- Go to Attendant → Content Indexing and click Start Indexing
- Add the chat widget with the shortcode
[attendant]or enable the floating widget (appears automatically in the footer on every page)
屏幕截图:
常见问题:
Do I need an OpenAI account?
Yes. You need an OpenAI API key from platform.openai.com. The plugin uses your key directly — we never see it.
How much will it cost to run?
With gpt-4o-mini (our recommended model), each conversation turn costs approximately $0.005 USD. 1,000 conversations per month ≈ $5–10. Content indexing for a 500-page site costs approximately $0.025 (one-time).
Is my data sent to OpenAI?
Only the user's chat message and relevant content excerpts (for Q&A mode) are sent to OpenAI. Your full database is never sent. API keys are encrypted before storage and never leave your server.
Will it slow down my site?
No. The chat widget JS is less than 10KB gzipped and loads asynchronously. Admin scripts only load on plugin admin pages. All AI processing happens via REST API calls triggered by the visitor — never on page load.
Does it work with ACF, MetaBox, WooCommerce?
Yes. The schema discovery engine automatically detects custom fields from ACF, MetaBox, and WooCommerce product attributes.
Is the chat history saved on my server?
No. The conversation is kept in the visitor's own browser (localStorage) so it survives page changes and refreshes. It is not stored on your server or in your database. The visitor can clear it with the "New chat" button at any time.
Can the assistant collect leads or callback requests?
Yes, but it is off by default. When you enable lead capture in Settings and a visitor agrees to be contacted, the assistant collects the email they provide (and optionally a phone number and preferred time) and emails the request to the address you configure. It is rate-limited to one request per conversation and a daily maximum, and nothing is stored in a database. See the Privacy section for details.
Can I keep a record of the conversations?
Yes, optionally. File logging is off by default. When enabled, each exchange is written to a dated file in a protected folder in your uploads directory, kept for 30 days, and downloadable only by an administrator. IP addresses are never stored and session identifiers are one-way hashed. See the Privacy section.
更新日志:
- Rebuilt as a structured-search-first site assistant.
- Natural-language queries are translated into a safe WP_Query over your post types, taxonomies, and custom fields.
- Embeddings / semantic Q&A is now optional and off by default.
- Short, scannable replies with clickable source buttons instead of links pasted into the text.
- Tap-to-answer quick-reply chips for refining a search, choosing a callback time, and other selections.
- Guided help when a search returns nothing: real alternatives from your own content, asked one question at a time.
- Conversation history kept in the visitor's browser (survives page changes and refreshes), with "New chat" and "Previous chats".
- Optional callback / lead capture: collects the email a visitor provides and emails the request to you, off by default, rate-limited.
- Optional file-based conversation logging in a protected uploads folder, admin-only download, 30-day rotation, off by default.
- "About this site" context setting to tune the assistant to your site.
- Manual and background content indexing with a live activity log; the chat widget stays hidden until the first index completes.
- Honest provider scope: OpenAI only in this release.
- Initial release.
- Settings page — encrypted API key storage, model selection, widget configuration, privacy/GDPR controls.
- OpenAI provider — chat completion (gpt-4o-mini / gpt-4o) and embeddings (text-embedding-3-small).
- Schema discovery — auto-detects post types, taxonomies, and custom fields (ACF, MetaBox, WooCommerce).
- Content indexing — chunker, embedder, background queue (WP-Cron), auto-sync on post save/delete.
- Chat engine — RAG retrieval (cosine similarity), OpenAI function-calling (search_posts), conversation handler with session history and token budgeting.
- Frontend chat widget — floating launcher, accessible dialog, brand colour + position overrides, [attendant] shortcode.
- Analytics page — monthly API cost history, index health stats, conversation stats (when logging enabled).
- Q&A Manager — admin-configured question/answer pairs matched semantically before RAG (threshold 0.92); REST CRUD API.