CatEngine AI Hub — v2 Highlights

Multi-provider models: OpenAI (incl. GPT-5/4o), DeepSeek Chat, and Gemini 2.5 Flash with smart routing and vision-aware handling.
Live streaming replies: Tokens render in real time with a Stop button to cancel generation.
Richer inputs: Attach text/code files (auto-inlined in the prompt) and images (on vision models); previews, remove, and safe re-attach on the next turn.
Auto chat titles: First message generates a clean, short title; long titles are ellipsized in the sidebar; quick rename via ⋯ menu.
Token-safe conversations: Sliding context window trims oldest turns; GPT-5 uses max_completion_tokens and model-specific limits.
New sidebar UX: Desktop mini-rail + mobile off-canvas with scrim, scrollable chat list, and per-chat actions (rename / export / import / delete).
Polished theme: Unified gradient “glass” UI, consistent dark/light mode, themed scrollbars, and auto-contrasted logo.
Profile page: See account details, toggle dark mode, set default model, manage/show-hide API key, and clear local chats.
Accounts: Google Sign-In (popup) or Guest mode. Guest stores everything locally; Google syncs chats to Firebase and encrypts API keys client-side.
Help & Privacy: Rewritten, theme-aware page explaining storage, encryption, limits, and providers.
Mobile-first: Responsive login, sidebar, and chat; safe-area support; touch-friendly controls.
Stability & fallbacks: Better error messages, Gemini v1/v1beta fallbacks, and attachment flow fixes.
v2 focuses on real-time feel, safer long chats, cleaner UI, and simple account choices—use your own API keys, pick any model, and chat anywhere.