ai-engine
Last commit date
app
4 days ago
archives
6 months ago
classes
4 days ago
common
4 weeks ago
constants
4 days ago
images
1 month ago
labs
4 days ago
languages
4 months ago
themes
4 days ago
vendor
11 months ago
CHANGELOG-PROPOSAL.txt
4 days ago
ai-engine.php
4 days ago
readme.txt
4 days ago
uninstall.php
10 months ago
CHANGELOG-PROPOSAL.txt
24 lines
| 1 | Proposal for the 3.6.3 entry, merging the six lines already drafted in readme.txt with the rest |
| 2 | of the release. Nothing here has been written into readme.txt. |
| 3 | |
| 4 | IMPORTANT: the drafted line "Workspace moved to the free core" is no longer accurate. The |
| 5 | Workspace module toggle was gated behind Pro in commit 8cf28a27, so that line is replaced below. |
| 6 | |
| 7 | = 3.6.3 (2026/07/27) = |
| 8 | * Add: Workspace (Pro): a full-screen chat surface in wp-admin, with a model picker across all your environments, conversation history, folders, a prompt library, per-user themes, and edit-and-branch on any message. |
| 9 | * Add: The Workspace can generate and edit images, search the web with each provider's own search tool (OpenAI, Anthropic, Google, OpenRouter), and work on the site itself through the MCP tools, with an approval dialog before any site-changing tool runs. |
| 10 | * Add: QR-code pairing for the Workspace mobile companion app, with a revocable Application Password. |
| 11 | * Add: Internal (WordPress DB) vector environment that stores embeddings locally, now the default for new installs. |
| 12 | * Add: Discussions now show the tool calls and their parameters alongside the embeddings usage, and can be searched server-side by content and title. |
| 13 | * Add: Claude Opus 5. |
| 14 | * Fix: Public chat and form endpoints no longer trust client-supplied models, environments, instructions or messages, and internal errors are hidden from visitors; the stateful-response restore is now owner-scoped, so another user's chat ID can no longer resume their conversation. |
| 15 | * Fix: Content-aware chatbots no longer inject a post the visitor cannot read, so a client-supplied context ID can no longer leak draft or private content. |
| 16 | * Fix: Chatbot error and notice messages are sanitized before being rendered as HTML, and configuration errors are shown to editors only instead of being printed publicly. |
| 17 | * Fix: Function calling is more reliable: Claude no longer re-calls tools with the previous message's parameters, multi-round calls on stateless providers keep their feedback blocks instead of clearing them each round, and the Gemini 2.x models work again. |
| 18 | * Fix: Google model names and capability detection, so image, thinking and web-search models are recognised and preview variants are no longer ambiguous. |
| 19 | * Fix: OVH replies arriving empty, caused by an oversized max_tokens and a hijacked catalog fetch. |
| 20 | * Fix: Embeddings re-sync destroying a vector when the new content comes back empty, and guest file listing crashing with a 500. |
| 21 | * Fix: The usage-limit lock never firing on streaming chatbots, and the Discussions table spinning forever when opened in a background tab. |
| 22 | * Update: MCP server hardening: a failing tool now returns a readable error instead of killing the server for the client, and the tools list is capped at 500 results. |
| 23 | * Update: Settings Reset now asks you to type the site host and says plainly that it deletes the API keys, environments, chatbots and themes; more native browser dialogs were replaced with NekoModal across Discussions, Query Logs, Knowledge, Templates and Limits. |
| 24 |