ganeshnikhil/J.A.R.V.I.S.2.0 ? reverse-engineered prompt

Build me a voice based Jarvis style assistant app that runs from a simple web interface and can understand spoken commands, reply out loud, and handle everyday tasks like weather, news, emails, basic data analysis from a CSV, and chatting with a more personal tone.

It should also be able to look at images from an upload, a URL, or the camera, then do basic detection, object detection, segmentation, and resizing. I want it to use local smaller models when possible, with Gemini as a fallback for smarter tool use, and it should include a memory and knowledge lookup system so it can answer from saved markdown notes without making things up.

Please add Android phone automation through ADB so it can make calls, open apps, and do simple device actions. Keep it easy to set up, with a single start screen, clear status, and support for both text and voice. If you need current docs for any APIs or models, look them up online.