Description
Vexalyn AI Summary adds a sidebar panel to the Gutenberg editor that lets you generate a summary of your article with a single click using the LLM provider of your choice: Ollama (self-hosted or Ollama Cloud), Claude (Anthropic) or ChatGPT (OpenAI).
Once generated, the summary is instantly inserted at the top of your article as native Gutenberg blocks (group, heading, paragraphs, signature). You can validate, cancel, or regenerate before publishing.
Features
- Choose your LLM provider: Ollama (Cloud or self-hosted), Claude (Anthropic) or ChatGPT (OpenAI)
- WordPress 7.0+ Connectors API support: if an Anthropic or OpenAI key is configured in Settings > Connectors, it is used automatically — no need to enter it twice
- Choose the model from a live list fetched from the provider API
- Fully customizable prompt with built-in examples
- Summary inserted as native blocks: group or details > heading/summary > content > signature
- Choice between Group block and Details/Summary block for the summary wrapper
- Generate a dedicated excerpt with configurable word count, review/edit/validate flow
- Visual highlight of inserted blocks while awaiting validation
- Test your connection and your prompt directly from the settings page
- CSS class on the group block is configurable for easy theme styling
- Developer filter to customize the thinking messages displayed during generation
Requirements
- An Ollama instance (local or remote), an Ollama Cloud account, an Anthropic API key or an OpenAI API key
No build step required. No Composer, no npm.
External services
This plugin connects to a third-party AI API to generate summaries using language models. Only the provider selected in the settings is contacted. For every provider, the following data is sent:
- Article content entered by the user for summary or excerpt generation
- Selected model name
- Configuration parameters (prompt)
When: only when the user manually clicks the “Generate summary” / “Generate excerpt” button, or starts a batch excerpt generation from the settings page — no automatic or background processing occurs otherwise.
No tracking of any kind is performed by this plugin. API keys are stored on your site and are never sent to the browser.
Ollama (https://ollama.com)
- Cloud: your requests are sent to Ollama’s hosted infrastructure at https://api.ollama.com — an API key is needed.
- Self-hosted: Ollama runs entirely on your own server or machine; no data ever leaves your infrastructure.
Privacy policy: https://ollama.com/privacy
Terms of service: https://ollama.com/terms
Anthropic (Claude) — requests are sent to https://api.anthropic.com
Privacy policy: https://www.anthropic.com/privacy
Terms of service: https://www.anthropic.com/legal/commercial-terms
OpenAI (ChatGPT) — requests are sent to https://api.openai.com
Privacy policy: https://openai.com/policies/privacy-policy
Terms of service: https://openai.com/policies/terms-of-use
Installation
- Upload the
vexalyn-ai-summaryfolder to/wp-content/plugins/. - Activate the plugin in Plugins > Installed Plugins.
- Go to Vexalyn > AI Summary to choose your provider and configure the connection.
Screenshots
Changelog
1.0.0
- Initial release
1.1.0
- Add page hub in the Vexalyn menu
- Add notice when settings are saved
- All text in English
1.2.0
- Add excerpt generation: dedicated AI call, configurable word count, review/edit/validate flow
- Add choice between Group block and Details/Summary block as summary wrapper
- Details block opens by default when inserted
1.3.0
- Add batch excerpt generation: select post types, generate missing excerpts in background via WP Cron
- Background processing persists across page reloads — leave and come back to see progress
- Per-post-type progress bar with processed/total count
- Show/hide list of items modified by batch (links to edit screen, capped at 200)
- Rollback: reset excerpts and remove marker for all batch-modified items of a post type
- Validate: keep excerpts, remove rollback marker when satisfied
- Stop button to interrupt a running batch
- Anti-concurrent lock per post type (HTTP 409 if batch already running)
- New option
vexalyn_resume_post_types: select which public post types support batch - New post meta
_vexalyn_excerpt_batch_modifiedto track batch-generated excerpts - Refactor excerpt prompt into shared private method (reused by single and batch generation)
1.4.0
- Plugin renamed “Vexalyn AI Summary” (slug unchanged)
- New LLM provider choice: Ollama (Cloud or self-hosted), Claude (Anthropic) or ChatGPT (OpenAI)
- WordPress 7.0+ Connectors API support: Anthropic and OpenAI keys configured in Settings > Connectors (database, PHP constant or environment variable) are used automatically and take precedence over the plugin fields
- Model selection stored per provider: switching provider keeps the model chosen for each one
- API keys are no longer printed in the settings page HTML: masked placeholder, keep-if-empty behaviour and a “Remove the saved key” checkbox
- Test the connection with unsaved values before saving
- Gutenberg sidebar now uses
wp.editor.PluginSidebar(WordPress 6.6+) with a fallback for older versions - New filter
vexalyn_resume_max_tokensto cap the generated output (default 4096) - Security hardening: LLM output filtered server-side (wp_kses_post, block delimiters stripped) and restricted to a safe block list in the editor; HTML-to-text conversion via an inert DOMParser document
/generateand/generate-excerptnow require apost_idthe user can edit, cap content size and rate-limit generations per user (filtersvexalyn_resume_generate_capability,vexalyn_resume_rate_limit)- Connection test never sends the saved Ollama key to a URL different from the saved one
- Batch excerpts skip password-protected posts and no longer re-filter post content through kses
- New
uninstall.phpremoves options (including API keys), batch states and post metas on deletion - Tested up to WordPress 7.1



