Forerunner® AI Live Chat
| 开发者 | designforonline |
|---|---|
| 更新时间 | 2026年7月11日 05:44 |
| 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:
Forerunner AI Live Chat provides a complete, modern chat solution for WordPress websites using n8n or another automation platform. Connect your visitors with AI-powered conversations through webhook integration while maintaining full control over your data and user experience.
Key Features
- AI-Powered Chat - Connect to any AI service via webhook
- Azure Functions Ready - Full support for Azure Functions with API Key authentication
- Custom API Headers - Add custom authentication headers (X-API-Key, Bearer tokens, etc.)
- Flexible Field Mapping - Configure request/response field names for any API format
- Custom Bot Icons - Choose from built-in icons or upload your own
- Mobile-Friendly Design - Responsive interface works on all devices
- Easy Setup - Simple configuration with no coding required
- Color Customization - Match your brand colors perfectly
- Webhook Integration - Connect to any external AI service (n8n, Azure, AWS, custom APIs)
- Configurable Timeout - Set custom timeout for slow AI agents (5-120 seconds)
- Voice Chat Ready - Compatible with voice chat extensions
- Install and activate the plugin
- Configure your webhook URL in the settings (from n8n or any other webhook-capable service)
- Customize the appearance, bot name, and messages to match your brand
- Visitors can start chatting with your AI assistant immediately
- Azure Functions - Full support with X-API-Key authentication
- n8n, Zapier, Make (Integromat), or Pipedream workflows
- OpenAI GPT models and Azure OpenAI Service
- AWS Lambda with API Gateway
- Google Cloud Functions
- Custom AI services and REST APIs
- Customer support systems
- CRM integrations
- Enter your Azure Function URL in the Webhook Endpoint field
- Set Authentication Type to "API Key Header"
- Set Header Name to "X-API-Key" (or your function's key name)
- Enter your Azure Function key as the Header Value
- Configure the Request/Response field names to match your function's format
安装:
- Upload the plugin files to
/wp-content/plugins/forerunner-ai-live-chat/ - Activate the plugin through the 'Plugins' screen in WordPress
- Navigate to 'Forerunner AI Live Chat' in your admin menu
- Configure your webhook URL (from n8n or another webhook-capable service) and customize settings
- Your chat widget will appear on your website frontend
- Webhook URL: Enter the endpoint where chat messages should be sent\ (e.g., your n8n production/test webhook URL, or another webhook-capable service such as Zapier, Make, or Pipedream)
- AI Name: Customize the name displayed for your AI assistant
- Messages: Set custom header, disclaimer, and greeting text
- Bot Icon: Choose from built-in icons or upload a custom image
- Colors: Customize colors to match your brand identity
- The Production URL (for live usage)
- The Test URL (for testing and development)
屏幕截图:
常见问题:
Do I need an AI service to use this plugin?
You need to provide a webhook URL that processes the chat messages. This could be an endpoint from n8n (recommended), Zapier, Make, Pipedream, a custom AI service, or any other system that can respond to HTTP requests.
Does this plugin store conversation data?
This plugin does not store conversation history by default. All messages are sent to your configured webhook URL. Data storage depends on your webhook implementation.
Can I customize the appearance?
Yes, you can customize colors, messages, bot icons, and the overall appearance through the admin settings. The plugin includes comprehensive color customization options.
Does it work on mobile devices?
Yes, the chat widget is fully responsive and optimized for mobile devices, tablets, and desktop computers.
Is voice chat supported?
Voice chat functionality can be added through our compatible extensions (such as the optional DFO Voice plugin). The base plugin focuses on text-based conversations.
What external services does this connect to?
The plugin only connects to the webhook URL you specify in the settings. It's designed to work with any AI service or webhook-capable platform. No third-party connections are made without your explicit configuration.
Can I test with a free workflow?
Yes. We provide a free starter workflow for n8n which you can import and adapt. See the "Example Workflows" section for the download link.
更新日志:
2.8.1
- SECURITY: Two defense-in-depth gaps found during a follow-up audit, both closed:
- The Active Visitors AJAX handler (
foreaili_active_visitors/foreaili_live_chat_visitors) verified a nonce but never independently checkedmanage_optionsinside the handler itself. Low practical risk (that nonce is only ever printed on an already admin-gated page), but a leaked nonce would have been enough on its own. Now checks capability explicitly. - The IP-debugging AJAX handler (
foreaili_debug_ip_detection) had no nonce check at all. Brought in line with its siblingdebug_chat_historyhandler, which already did this correctly.
- SECURITY: The
/trackvisitor-tracking endpoint accepted requests from anywhere with no verification — a security review found it could be hit directly (e.g. with curl) without ever loading the site, and every outcome (disabled, bad params, bot, success) returned an identical200 OK, making it impossible to tell real tracking traffic from noise in server logs. The endpoint now requires a nonce generated when the page renders (so blind requests that never loaded the site are rejected) and returns a distinct HTTP status per outcome:403when tracking is disabled/bot-blocked/nonce-invalid,400for malformed requests,429when a single visitor is sending requests faster than the tracking interval allows, and200only for a genuinely accepted beat. - NEW: Added a per-visitor request throttle server-side (independent of the existing DB-write rate limit) so a stuck tab, a broken SPA-navigation watcher, or a replayed request can't repeatedly trigger full processing faster than the configured tracking interval.
- NEW: Tracking Interval setting (Settings → Performance) — admin-configurable heartbeat frequency, 20-300 seconds, default unchanged at 60s. Raise it on high-traffic sites to cut request volume further.
- FIXED: Visitor tracking could keep running after being switched off (or keep the old interval after an update) on any site using a page cache —
wp_localize_script()bakes settings straight into the cached page HTML, so a cached copy kept serving the old values until it happened to expire. Saving the Visitor Tracking toggle (and updating the plugin) now best-effort purges WP Rocket, WP Super Cache, W3 Total Cache, LiteSpeed Cache, SG Optimizer, WP Fastest Cache, and Cache Enabler so the change takes effect immediately. No effect on sites without a page cache. - IMPROVED: The tracking heartbeat now pauses while its browser tab is in the background (Page Visibility API) and resumes automatically when the tab becomes visible again — cuts request volume from visitors who leave a tab open without adding any delay for visitors actively on the site.
- IMPROVED: The xAI voice provider now applies a very short fade in/out to each streamed audio chunk before playback, removing occasional clicking/popping at the boundary between chunks (a natural side effect of scheduling many small separate audio buffers back-to-back over a raw WebSocket, which WebRTC otherwise handles for OpenAI automatically).
- IMPROVED: Both voice providers now show a clear "Voice chat disconnected unexpectedly" message if the connection drops on its own (network issue, timeout, etc.), instead of silently returning to text chat with no explanation. Intentionally stopping the voice chat yourself is unaffected and shows no message.
- Minor cleanup of the xAI audio-decoding code (no functional change).
- REMOVED: The custom client-side "noise gating" audio processing (a compressor/filter/volume-gate chain we ran on the microphone before sending it to either voice provider) has been removed entirely. It was a blunt, one-size-fits-nobody amplitude threshold duplicating what the browser's own microphone processing and each AI provider's own server-side voice detection already do properly, and it could simultaneously gate out quiet speech while still letting louder background noise through. The "Background Noise Sensitivity" setting is removed from Voice Settings as it no longer has any effect.
- FIXED: The xAI voice provider could send an unnecessary cancel request when background noise falsely triggered speech detection, causing a visible "Cancellation failed: no active response found" error and occasionally ending the call. xAI's own documentation states interruption is already automatic in this mode; we now only stop local audio playback immediately and no longer send our own cancel request.
- CHANGED: The xAI voice provider no longer sends any interruption-sensitivity value of its own — it fully defers to whatever is configured on the saved agent at console.x.ai. The "AI Interruption Sensitivity" setting on the Voice Settings page now only applies to the OpenAI provider, which has no equivalent per-agent setting of its own.
- Removed two unused, unreferenced duplicate admin page files left over from an earlier reorganisation (assets/admin/voice-settings.php, assets/admin/colors.php) — the real pages have always lived under admin/.
- FIXED: Voice chat could stop responding after the first exchange when using the xAI provider. The previous release removed our own turn-detection override entirely so the saved agent's console settings would apply, but this meant conversation broke completely if the agent's default turn-detection mode wasn't automatic. Turn detection is now always explicitly enabled (required for the conversation to continue turn after turn) while still not overriding any other agent-side tuning.
- FIXED: The "AI Interruption Sensitivity" setting on the Voice Settings page was saved but never actually used anywhere. It's now wired into both the OpenAI and xAI voice providers as the actual interruption/VAD sensitivity, so raising it (e.g. in a noisy room) genuinely makes the assistant harder to interrupt with background noise, and lowering it makes it easier to interrupt.
- FIXED: Ending a voice chat could throw an uncaught error ("Cannot close a closed AudioContext") if the close button's cleanup and the normal voice-chat teardown both tried to close the same audio context. Now guarded so this can't happen.
- FIXED: The xAI voice provider no longer sends its own turn-detection settings on connect, which could silently override interruption-related settings already configured on the saved agent (e.g. "Caller can interrupt" in the xAI console). It also now explicitly cancels an in-progress response as a safety net when the caller starts talking, in addition to stopping local audio playback immediately.
- NEW: The voice chat widget now supports xAI's Grok Voice Agent as an alternative to OpenAI Realtime (Pro feature, configured in the companion voice plugin). Since xAI has no public browser WebRTC endpoint, this connects the browser directly to xAI over a secure WebSocket using a short-lived token, with its own audio streaming and playback path, while OpenAI continues to use its existing WebRTC connection unchanged. Which provider is used is chosen in settings; existing OpenAI voice setups are completely unaffected.
- NEW: Added a forerunner_ai_visibility_request_headers filter so add-ons can attach visitor identity headers to the Chat Visibility check. The Pro add-on uses this to forward the same WordPress User Context (and SSO) headers it already sends with the chat webhook, so your endpoint can decide show/hide per user type. No effect unless the visibility check is configured.
- IMPROVED: The optional Chat Visibility check now runs through a secure server-side relay. The endpoint URL and an optional auth header (e.g. API key) are stored on your server and never exposed to visitors, and the check is same-origin so there are no cross-origin (CORS) issues. Add a configurable auth header name and value under Settings. Behaviour is unchanged when the check is left disabled.
- NEW: Optional Chat Visibility check. Set an endpoint URL under Settings and the widget will only display when that service returns {"showChatbot": true}. Leave it empty (the default) and the chat always shows, so existing sites are unaffected. If the service is unreachable, times out, or returns anything else, the chat stays hidden.
- COMPATIBILITY: Confirmed compatible with WordPress 7.0. Updated Tested up to header.
- CLEANUP: Remove debug console.log statements added during OpenAI GA API migration.
- FIXED: Voice waveform bars: right half of bars were frozen because they mapped to 12–24 kHz (inaudible). Now all bars map to the voice frequency range (0–8 kHz) with a symmetric left-right waveform.
- FIXED: JavaScript TDZ error (Cannot access variable before initialization) in minified build caused by single-letter variable name collision.
- FIXED: Voice chat broken after OpenAI retired the Realtime Beta API. Updated WebRTC endpoint from deprecated
/v1/realtime?protocol=webrtcto the GA endpoint/v1/realtime/calls.
- FIXED: Active visitors flickering in and out of the Pro dashboard — idle heartbeats (same-page pings) were being dropped by
forerunner_ai_persist_visitor_trackingbecause$data_changedonly flipped on page navigation (a 2.4.5 perf optimisation).last_activityupdated in memory but never persisted, so visitors vanished after ~90 s and reappeared on their next navigation. Heartbeats now always refresh the transient cache. - CHANGED: DB write rate-limit tightened from 30 s to 15 s. The expensive serialised-options write is still rate-limited (preserving the 2.4.5 "no SETEX on every heartbeat" intent for the DB path), but the in-memory transient is now refreshed per heartbeat so the Pro Active Visitors view never sees stale data.
- NEW: Persistent chat transcripts — messages now write to a dedicated
wp_foreaili_messagesdatabase table instead of one-hour transients, so conversation history survives object-cache evictions and server restarts. - NEW: Read-only REST endpoints (
GET /wp-json/forerunner-ai/v1/conversationsand/conversations/{visitor_id}) feed the Pro inbox without any admin-ajax round-trips. Both requiremanage_options. - NEW: WordPress Heartbeat integration — when a Pro admin tab is open, new messages are delivered via the native 15 s Heartbeat tick instead of a bespoke polling loop.
- NEW: 90-day retention — a daily WP-Cron job (
forerunner_ai_messages_purge) prunes old rows automatically so the table never grows unbounded. - Upgrade: First load after upgrade runs
dbDeltato create the new table; existing in-flight transient chats continue to work and get added to the table from the next message onwards.
- CHANGED: Removed all human-agent takeover functionality — the plugin is now AI-only, matching how most sites actually use it.
- Performance: Dropped the 500 ms agent-message polling loop and the 5-second visitor-status polling loop from the chat widget — visitor browsers now make ~7 fewer requests/minute per open chat.
- Performance: Removed 15 legacy AJAX action handlers and ~600 lines of server-side agent-routing code — fewer admin-ajax hits, smaller plugin footprint.
- Performance: chat-widget.min.js reduced from ~43 KiB to ~40 KiB; source from ~101 KiB to ~93 KiB.
- Upgrade: One-shot cleanup runs on first load after upgrade — purges the legacy
forerunner_ai_live_chat_agentsoption, drops pending chat-request transients, and strips stalechatStatus/agentId/agentNamefields from visitor records. - Note: Conversation/transcript storage and the visitor-tracking dashboard are unchanged — Pro monitoring continues to work.
- Performance: Visitor tracking REST endpoint now returns instantly — all Redis/DB writes deferred to shutdown hook
- Performance: IP geolocation results cached per-IP (1 hour) — eliminates blocking HTTP call on repeat visits
- Performance: Stale visitors (>5 min inactive) pruned on every write to keep serialised payload small
- Performance: Heartbeat-only pings (same page, same visitor) skip the write entirely
- Performance: Geolocation HTTP timeout reduced from 5 s to 2 s; failed lookups cached for 5 min
- Fix: Legacy AJAX tracking handler replaced with no-op stub — prevents duplicate DB writes from cached JS
- Performance: Frontend assets (JS and CSS) now ship minified for production — reduces chat-widget.js from ~101 KiB to ~43 KiB (57% smaller)
- Performance: Added chat-widget.min.css — reduces CSS payload by ~21%
- Improved: Uses SCRIPT_DEBUG constant — developers get unminified files for debugging, production sites get optimised builds
- Improved: Better Core Web Vitals scores (LCP, TBT) for sites using PageSpeed Insights
- Fixed: Added REST API authentication bypass for Forerunner endpoints - ensures proxy works even when third-party security plugins or server configurations restrict REST API access for non-logged-in users
- Fixed: Resolves "You are not currently logged in" error in environments with SSO, Basic Auth, or REST API restrictions
- Fixed: Removed X-WP-Nonce from proxy REST API requests - resolves "You are not currently logged in" error for SSO and non-WordPress-authenticated users
- Fixed: Proxy endpoints now work correctly for all visitors regardless of WordPress login status
- Improved: Visitor tracking moved to lightweight REST API endpoint (replaces admin-ajax.php)
- Improved: Visitor tracking interval increased to 60 seconds (reduced server load)
- Improved: Bot filtering added - crawlers and automated agents skip DB work entirely
- Improved: Transient-based visitor reads (benefits from Redis/Memcached if installed)
- Improved: DB writes rate-limited to once per 30s to reduce wp_options contention
- New: Admin toggle to enable/disable visitor tracking on the main settings page
- Fixed: Reverted credential changes from 2.3.8/2.3.9 that caused connection issues with direct webhook mode
- Fixed: Resolved CORS issue when using direct webhook mode (non-proxy) introduced in 2.3.8
- Improved: Added credentials support to API requests for SSO compatibility
- FIXED: New chat button now properly generates a fresh conversation ID
- Improved: AI text streaming now preserves paragraph spacing during animation for better readability
- Improved: Reduced line-height from 1.5 to 1.2 for more compact, readable text
- NEW: Font Weight setting - Control text thickness (Light 300 to Bold 700)
- FIXED: ChatKit duplicate message issue when using reset chat button
- FIXED: Widget sizing step values now support 425px and 675px defaults
- Improved: Removed widget border for cleaner appearance
- Improved: Disclaimer tertiary text spacing adjustment
- NEW: Chat persistence - Conversations now persist across page navigation for seamless user experience
- NEW: Reset chat button in header with rotate icon for quick conversation restart
- FIXED: Markdown elements (lists, tables, headings) now respect custom font size settings
- FIXED: Widget max-width now respects viewport to prevent overflow on mobile devices
- Improved: Chat widget automatically restores open state when navigating between pages
- NEW: Full markdown support - tables, headers, bullet/numbered lists, code blocks, inline code
- NEW: Markdown link syntax support text alongside raw URL detection
- NEW: Pro plugin maximize/expand chat button support
- Improved: Code blocks rendered with syntax-friendly dark theme
- Improved: Tables render with horizontal scroll for large datasets
- FIXED: Visitor tracking now works with page caching (nonce removed for tracking endpoint)
- Improved: Visitor tracking approach now consistent with industry standard (Tidio, Intercom, GA)
- Note: Tracking data (page URL, user agent) is non-sensitive and doesn't require authentication
- FIXED: Webhook calls now work correctly for simple setups (n8n, Zapier, etc.) without authentication
- FIXED: Proxy mode only activates when API Key or Query Parameter authentication is configured
- FIXED: Direct browser-to-webhook calls restored for unauthenticated webhooks (faster, simpler)
- Improved: Better error handling and response formatting in proxy mode
- SECURITY: All webhook requests now route through secure server-side proxy
- SECURITY: API keys and authentication tokens are no longer exposed in frontend JavaScript
- NEW: Pro integration hooks - Full support for Pro version advanced webhook features
- Added: Server-side proxy endpoint for secure webhook communication
- Added: Rate limiting integration with Pro rate limiting feature
- Added: Retry logic support with configurable retries and exponential backoff
- Added: Enhanced request body mapping with page URL, conversation ID, and message history
- Improved: JavaScript helper functions for better modularity and Pro compatibility
- Improved: Internationalization - more strings now translatable
- NEW: Webhook Authentication - Add API Key headers or query parameters for secure API connections
- NEW: Azure Functions support - Full compatibility with Azure Functions using X-API-Key authentication
- NEW: Request Field Mapping - Configure which JSON field name to send messages in (chatinput, message, input, query, prompt, text, content)
- NEW: Response Field Mapping - Configure which JSON field to read responses from (auto-detect, reply, output, message, response, text, answer, content, result)
- NEW: Configurable Timeout - Set custom request timeout (5-120 seconds) for slow AI agents or orchestration services
- Improved: Better error messages for timeout and connection failures
- Improved: More robust request handling with abort controllers
- Enhanced: Admin settings UI with collapsible authentication sections
- Fixed visitor tracking compatibility with Redis/Memcached object cache
- Visitor data now stored directly in database to ensure persistence
- Improved reliability for real-time visitor monitoring
- Enhanced performance for high-traffic sites with object caching
- Added extensibility hook for Pro plugin integration
- Enhanced conversations page with action hook for additional features
- Improved architecture for free/pro plugin compatibility
- Made webhook endpoint field optional - can now be left empty to hide chat widget
- Chat widget automatically hidden when webhook field is empty
- Plugin remains active for future configuration
- Improved field description with usage instructions
- Improved webhook compatibility with multiple AI service providers
- Added support for various response field formats (reply, output, message, response)
- Enhanced integration with n8n, Make, Zapier, and custom webhook implementations
- Backward compatible with existing webhook configurations
- Fixed JavaScript null reference error in chat widget
- Added defensive null checks for DOM elements to prevent crashes
- Improved stability when elements are not yet loaded or unavailable
- Enhanced compatibility with theme and plugin conflicts
- Enhanced color customization system with proper admin controls
- Improved chat icon shimmer effect and hover states
- Fixed border and background color consistency
- Better CSS override system for theme compatibility
- Performance optimizations and code cleanup
- Restored original chat icon shimmer effect
- Fixed hover state behavior and visual consistency
- Improved CSS layering and z-index management
- Enhanced color application system
- Major version update with comprehensive improvements
- Enhanced shimmer effects and visual polish
- Improved color customization interface
- Better mobile responsiveness
- Code optimization and security enhancements
- Enhanced chat messages area with improved background effects
- Added backdrop blur for modern appearance
- Updated container styling for better visual hierarchy
- Improved transparency effects throughout interface
- WordPress.org directory submission version
- Complete color customization system
- Added header gradient color controls
- Comprehensive CSS override system
- Removed debug output for production
- Enhanced security with proper nonce verification
- Full WordPress coding standards compliance
- WordPress.org compliant version
- Removed external dependencies for privacy compliance
- Uses WordPress default fonts for better compatibility
- Clean, unrestricted core features
- Improved admin interface design
- Added custom bot icon upload functionality
- Introduced quick action buttons
- Enhanced color customization options
- Improved admin interface usability
- Better error handling and validation
- Initial public release
- Core chat functionality implementation
- Webhook integration system
- Basic customization options