v2.0.1

Important

Android APK notice: The APK is a Termux bootstrap + WebView shell, not a native Android server build. It opens an already-running local Marinara server, and on first launch it can download Termux from F-Droid, hand it to Android's installer, and start Marinara through Termux after Android permission prompts. Follow the Android wrapper guide if Android blocks the bootstrap handoff.

Fixed

• Fixed the Android APK first-launch bootstrap so the app checks the local Marinara server before showing the WebView, keeps the Install / Start screen visible while Termux or Android permission prompts are active, and no longer strands users on a raw 127.0.0.1:7860 connection error page.
• Fixed Conversation Mode Active Context previews changing lorebook entries while idle by making preview-only probability and weighted group selection deterministic for unchanged chat state.
• Fixed Conversation Mode input lag and slow DM switching in large chats by reducing per-keystroke work, throttling draft sync, and limiting rendered transcript work to the visible window.
• Fixed Conversation setup flows where connection or semantic-search selectors could fail to persist selected connections, leave the picker at None, or keep the setup wizard/sidebar layered over the wrong newly created chat on mobile.
• Fixed Conversation presence/status wording so away messages use the character's actual name, and improved multilingual character-name matching for avatars, lookup, search, and command matching.
• Fixed Professor Mari chat behavior after v2.0.0: home-page sessions now survive refresh until manually reset, mobile layout starts below the top bar, the mobile CTA says "Ask Professor Mari", and tool/db command instructions are less likely to surface as plain text for local models.
• Fixed Roleplay and generation parameter handling so chats using connection custom defaults respect reasoning/output settings, custom provider parameters continue to be sent, and stored provider reasoning remains routed correctly.
• Fixed built-in local model generation so chat/preset Advanced Parameters control max output tokens instead of being capped by the local runtime fallback value.
• Fixed suppressed/unknown-model parameter handling so max output tokens are still sent while sampler-specific parameters remain gated.
• Fixed OpenRouter service tier handling so Flex/Priority and custom service_tier values still reach OpenRouter when unknown-model parameter suppression is active.
• Fixed the post-release issue sweep for chat metadata cache corruption, mobile Characters panel scroll restoration, mobile notification bubbles, Bubble-style multi-speaker messages, Professor Mari mobile restart access, memory-recall embedder retries, YouTube player default visibility, display-size overflow, mobile panel layering, local textual tool calls, new/delete chat failure handling, and visible extension/Mari workspace import errors.
• Fixed Roleplay agent toggles so enabled agents stay enabled after switching to persona, lorebook, or other editor screens instead of being overwritten by stale chat metadata.
• Fixed Roleplay chat-settings presets so metadata-only actions like Advanced Parameters, Translation, Lorebooks, Memory Recall, Tool Use, tracker actions, and agent toggles no longer reset the preset selector back to custom settings.
• Fixed mobile tab/library drag-and-drop ergonomics by requiring the explicit drag handle for touch dragging, restoring normal scrolling elsewhere, improving touch auto-scroll, and preserving folder tap open/close behavior.
• Fixed browser/source dropdown layering, chat window layering, Chat Settings/Gallery mutual closing, and mobile topbar/sidebar stacking so popovers and side panels no longer hide underneath or cover the wrong UI region.
• Fixed display-size scaling regressions where large/huge text caused topbar icons, settings buttons, regex rows, preset rows, and other tab controls to overlap or escape their containers.
• Fixed notification/toast behavior so stacked notifications fade/dismiss consistently and special Professor Mari toast variants use the unified toast styling.
• Fixed Game and Roleplay UI edge cases around widgets, branches, author-note style popovers, gallery image opening, pinned-image depth, YouTube player coloring, and chat toolbar buttons.
• Fixed Game Illustrator wiring so Gallery → Illustrate, scene illustrations, NPC portraits, and background generation use the game chat's selected image connection and scene image instructions consistently.
• Fixed Game Gallery → Illustrate so manual illustrations use the Game Illustrator image connection and asset pipeline directly, preventing false "No connection configured" errors from the retry-agent path.
• Fixed NovelAI reference-image generation so uploaded/data-URL references are normalized to the base64 payload NovelAI expects before being sent.
• Fixed ComfyUI reference-image avatar generation regressions, bot-browser import/delete flows, SillyTavern bulk import mappings, tracker field-lock serialization, and lorebook/import/export edge cases found during the post-2.0.0 stabilization pass.
• Fixed Docker Compose onboarding documentation so the root docker-compose.yml location is linked clearly.
• Fixed Marinara's Universal Preset v12 so the bundled language choice defaults to English instead of Polish.
• Fixed legacy persona Extended Descriptions migration so old persona description blocks become persona-linked lorebook entries just like character Extended Descriptions.
• Fixed version metadata for the v2.0.1 hotfix release across packages, the homepage-visible app version, Windows installer sources, PWA manifest, and Android APK metadata.

Platform Notes

• Android versionName is 2.0.1 with versionCode 20; users need a rebuilt v2.0.1 APK for the bootstrap/WebView fix.
• Windows, macOS/Linux, Termux, Docker, and PWA users can update through their usual v2 updater paths.