DIPESH AI — Mobile Voice Assistant starter A Hugging Face Spaces (Gradio) web app starter for Android/mobile browser. It uses a real hosted language model through the Hugging Face Inference API, hosted speech recognition, and text-to-speech. It is a starter implementation, not a complete Android system assistant. Deploy on Hugging Face Spaces Create a new Space at https://huggingface.co/new-space Choose Gradio as the SDK and a Python runtime. Upload app.py and requirements.txt (or upload the contents of this folder). In Space Settings → Secrets, add HF_TOKEN with a Hugging Face access token that has permission to use Inference Providers. Optional Space variables: HF_CHAT_MODEL: default Qwen/Qwen2.5-7B-Instruct HF_ASR_MODEL: default openai/whisper-large-v3 Wait for the Space to build, open its URL on Android Chrome, and use Add to Home screen if desired. Important limitations Hosted models/providers and availability can change. If the default chat or speech model is unavailable for your account, select an available model in Space Variables. Text-to-speech uses gTTS. It speaks aloud, but it does not guarantee a particular human female voice or a ChatGPT-identical voice. The app does not control all Android apps, run an always-on wake word, or continuously watch the screen. Android restricts background microphone use and cross-app control. Those features need a native Android app, foreground service, explicit permissions, and careful testing. Screenshot image analysis is not connected in this starter; a vision-capable model/API needs to be added for that feature. This starter does not fetch live market prices, place trades, or generate validated BUY/SELL signals. Do not treat outputs as financial advice. The chat history is session-only. It does not yet provide persistent memory, reminders, calendar integration, multi-AI routing, social auto-posting, or backtesting. Never put API tokens in source code or public files. Use Space Secrets.