v1.2.0 Latest
— 2026-06-11
Live model catalogues for every provider, voice dictation in the chat composer, and OpenAI "-pro" tier support.
Added
- Live model catalogues for every hosted provider (Anthropic, OpenAI, Gemini, DeepSeek, Grok, GitHub Models, and custom OpenAI-compatible gateways): the model picker now lists what your API key can actually use — queried from the vendor, cached for six hours, refreshed automatically when you change a key or endpoint. New models appear without a component update.
- Voice dictation: a microphone button next to Send turns speech into text right in the composer (browsers with the Web Speech API — Chrome, Edge, Safari). Nothing is sent to your AI provider until you review the text and press Send.
Fixed
- OpenAI "-pro" tier models (gpt-5.5-pro, o1-pro, o3-pro) now work: the o-series "-pro" models are routed to the Responses API instead of failing with "This is not a chat model", and Settings → Test respects their higher reasoning-effort floor.
Changed
- Model dropdowns sort flagship-family-first, newest-first, so the implicit "Default" model for a new conversation is the provider's current flagship.
- Cloud model lists are negatively cached on failure (ten minutes), so a revoked key or vendor outage never adds latency to every admin page render.
v1.1.0 — 2026-06-08
Ollama connector overhaul, AI-created articles now visible in the admin, fully responsive dashboard, and security hardening.
Added
- Ollama now lists the models actually installed on your server (queried live and cached) instead of a fixed list — the model picker and the Settings → Test button reflect what you have pulled.
- Settings → Test now uses the Endpoint URL you type, so you can test an as-yet-unsaved Ollama / OpenAI-compatible host. URLs are normalised (bare host, /v1, trailing slash, or a pasted /chat/completions URL all work).
- AI-created articles now record their Joomla workflow association, so they appear in Content → Articles immediately.
Fixed
- Ollama tool calling: tool-bearing turns now run as a non-streaming request (Ollama cannot stream structured tool calls), so tool-capable models such as qwen2.5 and llama3.1 actually invoke tools instead of describing them.
- Ollama reasoning models (e.g. deepseek-r1): the connection test now succeeds even when the model returns empty content, and the assistant degrades gracefully when a model does not support function calling.
- “Show Ollama in the chat picker” now saves from the Ollama tile (it was dropped by the per-tile save), so Ollama shows up in the conversation provider list.
- create_article: category is now optional and defaults to Uncategorised (or the first content category) instead of erroring; featured = front page is honoured.
- edit_article: backfills a missing workflow association (repairs earlier AI-made articles) and re-points the workflow stage when an article changes category.
- query_database: a LIMIT in your query no longer collides with the built-in row cap (previously a SQL syntax error).
- Safe install AND uninstall on a damaged admin menu: the component now repairs Joomla’s #__menu nested set (missing root, stale or overlapping rows) before adding its menu, and on uninstall removes only its own rows addressed by id, then rebuilds. This fixes the “Table::_getNode(1) failed / Error building Administrator Menus” install error and the missing Components → Seanos AI entry, and — more importantly — a case where uninstalling the component could delete OTHER components’ admin-menu items (Joomla core removes menu rows by tree position, which cascades on an already-corrupted menu).
- Conversations list: silenced a PHP 8.1+ deprecation notice (null passed to htmlspecialchars) that appeared when only a single page of conversations was present.
Changed
- The dashboard prompt library is fully responsive — tiles and category chips reflow instead of forcing a horizontal scrollbar.
- Security: model names sourced live from your server are escaped in the chat UI, and connector endpoint URLs are restricted to http(s) (SSRF hardening).
- Unreachable Ollama hosts are negatively cached so a slow / offline host no longer adds latency to every admin page.
- Verified on Joomla 6.1 with PHP 8.3 / 8.4.
