johnnyq 80cdc4de41 AI stage 1: rebuild AI Providers and AI Models in the standard list pattern with ready-made provider kinds (OpenAI, Anthropic, Azure OpenAI, Gemini, Mistral, Groq, OpenRouter, Ollama, LM Studio, OpenAI-compatible), Test that lists a provider's models, Save anyway for a provider that doesn't answer, a first model offered when adding the first provider, the API key no longer shown, from-to audit lines, and delete confirms that say what goes and what stops; model window with suggested models, use case notes and Try it; new AI Usage page logging each request by feature, user, client, model, tokens, time and failure, with totals, views and filters, pruned with the audit log; passwords, keys, 2FA links and card numbers taken out before sending and put back in the answer, on by default per provider; new Use AI role permission; Reword shown only when set up and allowed, placed first in the toolbar, with its own instructions, undo in one step and a fixed endpoint on admin pages; ticket summary and template drafts purified; Settings overview warns while AI requests keep failing; database 3.3.8 adds provider kind and redact, ai_usage and module_ai

2026-10-20 17:45:00 -04:00

18 KiB

Raw Blame History

View Raw