Description

App_Description.md
Wrist AI — A Flexible LLM Client for Pebble

Wrist AI lets you talk to any Large Language Model directly from your Pebble watch. Speak your question, get a response（Now with TTS） on your wrist. Your phone handles the API call over Bluetooth — no server, no cloud relay, no extra apps.

Works with OpenRouter (GPT, Claude, Gemini, Gemma, Llama, Qwen, and more) or your own custom API endpoint.

How to Use

Install the app and open its settings in the Pebble companion app
Paste your OpenRouter API key — or switch to Custom API mode and enter your own endpoint URL
Add models to the model list (e.g. google/gemma-3-27b-it, anthropic/claude-sonnet-4)
Launch the app on your watch and press SELECT to start talking
Button Controls Button | State | Action SELECT (short) | Idle | Start voice dictation SELECT (short) | Response | Return to main screen (ready for follow-up) SELECT (short) | TTS playing (Emery) | Stop reading aloud SELECT (long) | Any | Start a new conversation UP (short) | Response | Scroll up UP (short) | Other | Open chat list menu UP (long) | Response (Emery) | Open menu (Volume, Chat list, Web Search toggle, Model switch) UP (long) | Other | Open chat list menu DOWN (short) | Response | Scroll down DOWN (short) | Other | Open chat list menu DOWN (long) | Response (Emery) | Read reply aloud (Text-to-Speech) DOWN (long) | Idle | "Surprise Me" — random joke, fact, recipe, or question

Features Multi-Model Switching — Configure up to 5 models, switch between them directly from your watch Multi-Conversation — Up to 20 concurrent chats with LLM-generated titles, switchable from the watch menu with full context preserved Custom API Support — OpenRouter out of the box, or point it at your own backend / local models Conversation Export — Export any chat as structured JSON from the config page Text-to-Speech (Emery only) — Have AI replies read aloud through the built-in speaker via Google Cloud TTS. Automatic Chinese/English voice detection. Adjustable volume (10%–100%) and speaking rate. Long-form continuous playback with pre-encoding and Bluetooth flow control Health Context (Beta) — Optionally include your daily step count and active minutes in the AI's context Location Context (Beta) — Optionally include your approximate location (reverse-geocoded to city name) in the AI's context Web Search (Beta) — Let models search the web for up-to-date information (OpenRouter only). Toggle from the config page or the watch menu Font Size & Bold — Three text sizes (Normal / Large / Extra Large) plus bold toggle for readability "Surprise Me" Toggle — Long-press DOWN on idle for a random prompt, or disable it in settings Cross-Platform — Basalt, Chalk (fully optimized round layouts), Diorite, Emery Serverless & Private — Phone talks directly to the API. Conversations stored locally. Nothing leaves your device except the API call Dark-Themed Config Page — Modern settings interface for API keys, system prompts, model management, TTS key, and conversation history

Note Tested on a physical Pebble Time 2 (Emery) device. TTS voice playback is Emery-exclusive (the only Pebble with a speaker). If you run into any issues, please report them on GitHub.

Links Source: github.com/deusaw/Pebble-Wrist-AI Support: paypal.me/Asterwyn