Install
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/blindseer316/wp-edit-with-ai/archive/refs/heads/main.zipNo built-in updater. New versions won't appear under Dashboard → Updates. Install new releases yourself, or let Git Updater track the repository.
Chat-driven content editing for non-technical WordPress clients. No MCP connector, no external app, no subscription required — just a chat box that makes the edit directly.
Description
Lets non-technical clients update their own WordPress content (schedules, headlines, copy, hours) by describing the change in plain language, instead of learning the block editor or a page builder.
Settings/dashboard: WP Admin > WP Edit With AI.
Content types supported, in build order:
- Plain HTML / Gutenberg blocks
- ACF fields and repeaters
- Elementor widgets (
_elementor_data)
Changelog
0.6.0
- Added conversation memory: the chat now sends prior turns with every message, so the AI has context instead of treating each message as an isolated request.
- Added update_post_title tool (previously only post content/body could be edited, not the title).
- Increased API timeouts (Gemini 30s -> 60s, kie.ai 45s -> 120s) and raised PHP's execution time limit on the chat endpoint, to accommodate slower reasoning-model responses and multi-turn tool loops.
0.5.1
- Added a "Thinking…" indicator in the chat while waiting on a reply.
- Assistant replies now render basic markdown (bold, italics, inline code, bullet/numbered lists, paragraphs) instead of a raw text blob.
0.5.0
- Added kie.ai (GPT-5.6 Luna) as a second AI provider, with a toggle in Settings to switch between Google Gemini and kie.ai.
- Added kie.ai API key + model fields to Settings, tested via the same "Test API Connection" button.
- Useful as a fallback when Gemini's free tier hits rate limits.
0.4.2
- Fix: functionCall parts from Gemini's current (thinking-enabled) models carry a thoughtSignature that must be echoed back unmodified on the next turn. The client now preserves the whole part instead of rebuilding it, fixing "Function call is missing a thought_signature" errors.
0.4.1
- Fix: tool-call results were sent back to Gemini with role "function", which the current API rejects ("Role 'function' is not supported"). Changed to role "user" per Gemini's current REST format, unblocking the find_pages/get_page_content/update_page_content loop.
0.4.0
- Removed the separate Settings submenu/page. Settings now render inline on the main WP Edit With AI page (visible only to users with manage_options), so there is only one registered menu item and one URL.
- Fix: on at least one live site, the Settings submenu link rendered as a broken/mismatched URL (likely a conflict with a menu-customization plugin) causing "not found". Collapsing to a single page removes the second menu registration entirely.
0.3.0
- Added the actual tool-calling loop: Gemini can now call find_pages, get_page_content, and update_page_content against the live database, looping until it gives a final reply.
- Chat UI now shows which tool calls ran and their result (success/error) for each message, so edits are visible/verifiable, not just the AI's text reply.
0.2.0
- Added Gemini API integration (generateContent, model configurable, defaults to gemini-flash-latest).
- Added Settings page (WP Admin > WP Edit With AI > Settings) for API keys and model.
- Added automatic key rotation (falls back to key 2 on auth/quota errors) for free-tier usage.
- Added "Test API Connection" check on the Settings page.
- Chat endpoint now sends messages to Gemini instead of returning a placeholder.
0.1.0
- Initial scaffold: admin chat UI, REST endpoint stub, tool-execution class stub.