v1.3.1

Bug fixes

• Fixed: OpenAI-compatible local servers (LM Studio, LocalAI, text-generation-webui, etc.) were incorrectly treated as cloud backends, causing "API key not configured" errors
• Fixed: AI buttons blinking as unconfigured when only an OpenAI-compatible local server was set up
• Fixed: Default model picker in Local settings now shows both Ollama and OpenAI-compatible models
• Fixed: Cloud fallback cleanup when disabling the setting
• Fixed: Local OpenAI-compatible models no longer appear in the Cloud model picker

v1.3

What's new in v1.3

Home screen widgets

• New individual widgets for Voice, Chat, and Execute (1x1 size)
• Voice and Execute widgets have a settings button to configure model and voice mode directly from your home screen
• Combined widget also updated with settings buttons

Update notifications

• Check for updates button on the main settings screen
• Auto-check for updates toggle in Settings > About (checks GitHub daily)
• Notification permission is requested when enabling auto-check

Tips carousel

• 15 tips on the settings screen with inline button previews
• Random starting tip each time you open settings
• Covers all AI features, toolbar customization, widgets, and more

v1.2.1 - Polish & bugfixes

What's New

UX feedback

• AI Assist and AI Tone buttons pulse when no provider is configured
• "Type something first" hint when tapping AI buttons without text
• Test Connection button in AI Settings for all providers

Light mode fixes

• AI preview text, tone chips, scrollbar, and clipboard label now properly visible
• Close button moved to top-right of AI preview panel

Reliability

• All in-keyboard feedback messages now show inside the keyboard itself, so they remain visible even when system toasts are disabled

---

Hotfix (re-uploaded same version, redownload the APK)

• Onboarding wizard now uses dark styling regardless of system theme (fixes poor visibility in light mode)
• Custom OpenRouter model IDs: paste any slug in AI Settings > Cloud > Advanced (one per line or comma-separated) and it shows up in the model picker alongside the built-ins

v1.2.0

What's new

• AI Tone Chips: Rewrite your text as Formal, Casual, Friendly, Shorter, Longer, fix Grammar, translate to Dutch or English. Tap the new AI Tone button on your toolbar. Fully customizable in Settings > AI > General > Advanced > Tone Chips.
• AI Preview Panel: See AI results before they replace your text. Apply, retry, copy, or dismiss with one tap. Includes slide-up animation, typewriter effect, and shimmer loading indicator.
• Inline Commands: Type shortcuts like //formal, //grammar, //summarize, //shorten, or //reply at the end of your text, then tap the AI button. Manage aliases in Settings > AI > General > Advanced > Prompt Aliases.
• Command Chaining: Chain multiple commands in one go, e.g. //formal //english first makes your text formal, then translates it.
• Clipboard Context: Copy a message you received, type your reply, then tap AI Tone. The AI understands the context and adjusts your reply accordingly.
• What's New dialog: See what changed after each update, with video demos.

v1.1.1

What's new

• Fix OpenRouter models: 3 of 4 free models had outdated IDs and returned "model not found". Updated to current free tier models (Gemma 4 26B, Gemma 3 27B, Llama 3.3 70B, Qwen3 Coder).

v1.1.0

What's new

• Settings reorganization: All AI settings tabs (General, Local, Cloud) now split into Essentials and Advanced sections. Advanced is collapsed by default.
• Clearer labels: Renamed settings for clarity (Base Model Instruction, Inline command model, Model sources, Default voice model)
• Shorter helper texts: All descriptions are now concise and scannable
• Cloud tab: Groq and Gemini promoted to Essentials, other providers under Advanced
• Local tab: Ollama URL, connection test, and model picker in Essentials; Model Wizard, other servers, ONNX under Advanced

v1.0.3

What's new

• Quick Start now includes a simple Groq API key step to enable AI features right away
• Streamlined Advanced setup by removing the personalization step (available in Settings)

v1.0.2

What's new

• Redesigned onboarding wizard with Quick Start and Advanced Setup paths
• Quick Start: get started instantly, no API key needed
• Advanced: choose between Cloud (Groq/Gemini) or Local (Ollama) with guided setup
• Live AI demo during setup: see AI rewrite your text before you start
• Have-fun demo video on completion screen
• Ollama setup: improved labels, alternate connection for Tailscale/LAN

v1.0.1

Bug fixes

• Fix model picker saving wrong model: The slot, instruction, clipboard, and voice dialogs could save the wrong model when the background model refresh reordered the list. Selection is now tracked by model identifier instead of list index.
• Fix custom prompt not persisting: Custom prompts entered for shortcut keys 1-4 could be silently overwritten or lost due to the same index drift issue.
• Fix instruction field lock state: The instruction field now correctly locks/unlocks when switching between custom and regular models after a background refresh.

Landing page

• Updated hero headline for clarity.

Deskdrop 1.0

Initial release of Deskdrop — an Android keyboard with built-in AI.

Highlights:

• Connect to Ollama, LM Studio, or any OpenAI-compatible server
• Cloud providers: Gemini, Groq, OpenRouter, Anthropic, OpenAI
• 17 built-in AI tools (calendar, navigation, calls, SMS, contacts, and more)
• MCP support for external tool servers
• Voice transcription via self-hosted Whisper
• Multi-turn conversation interface
• 4 configurable AI shortcut slots
• Home screen widget

Requires Android 9.0+