WP Manifestindependent plugin directory
manifest / seo / link-whisper-ai-bridge

Link Whisper AI Bridge

Routes Link Whisper Premium AI requests through an OpenAI-compatible external provider.

★ 0stars

Install

Download source zip No release zip yet

This is the repository archive. It installs, but the folder name will carry the branch suffix and updates will not flow.

Install straight from GitHub with WP-CLI:

wp plugin install https://github.com/inisoliman/link-whisper-ai-bridge/archive/refs/heads/main.zip
No built-in updater. New versions won't appear under Dashboard → Updates. Install new releases yourself, or let Git Updater track the repository.

Separate WordPress plugin for routing Link Whisper Premium AI calls to an OpenAI-compatible provider.

Install

Copy the link-whisper-ai-bridge folder into wp-content/plugins/ beside link-whisper-premium, then activate it from WordPress Plugins.

Configure

Open Settings > Link Whisper AI Bridge and set:

  • Enable bridge
  • Provider Base URL, with or without /v1
  • Provider API key
  • Chat model
  • Embedding model

The bridge stores the real provider key in its own encrypted option. It only writes a non-secret sentinel key to Link Whisper's OpenAI key option so Link Whisper keeps its AI features enabled.

The bridge settings page includes مساعد ضبط Link Whisper.

Use فحص إعدادات Link Whisper to generate an Arabic report of suggested Link Whisper settings. Each recommendation explains:

  • The Link Whisper option being changed
  • The current value
  • The recommended value
  • Why the change is useful
  • Whether the change is safe or should be reviewed

Only selected recommendations are applied. Before applying changes, the bridge stores a backup in lwai_bridge_last_lw_settings_backup, and استرجاع آخر تغييرات المساعد restores the most recent advisor-applied settings.

The advisor only updates an internal allowlist of Link Whisper options. It does not change WordPress core settings, Google Site Kit settings, API keys, or unrelated plugin options.

Embedding providers

For full semantic AI relation analysis, the provider needs a real OpenAI-compatible /v1/embeddings endpoint that returns a data array.

If your chat provider does not support embeddings, set Embedding source to Provider, then local fallback or Local lexical fallback only. The local fallback is not a true semantic embedding model, but it lets Link Whisper live embedding calls continue instead of failing with 404 Not Found. The bridge disables Link Whisper batch processing in fallback modes because local fallback cannot run through OpenAI Files/Batches.

If you have a second OpenAI-compatible provider just for embeddings, use Embedding Base URL and Embedding API key, then set Embedding source to Provider only.

Notes

If Link Whisper AI subscription mode is active, the bridge pauses and shows an admin notice. Disconnect Link Whisper AI inside Link Whisper before using the external provider.

The provider must support OpenAI-compatible chat completions, embeddings, and Files/Batches for full Link Whisper AI coverage.