PAIOS - Offline AI | Local AI analisi dell'app per 27 aprile

PAIOS - Personal AI Operating System A 100% offline, private, and powerful interface for AICore (and maybe something else!). Changelog You can find the full changelog here: https://github.com/Puzzaks/PAIOS/CHANGELOG.md Roadmap You can see what is planned for this project here: https://github.com/Puzzaks/PAIOS/ROADMAP.md Features • 100% Offline & Private • Full Model Control • Multiple chats! • Custom Instructions • Context-Aware • Total Transparency • A (Strong) Personality • 100% Offline & Private This app leverages local AI API. This means that your prompts and conversations should never leave your phone. The model this app uses runs entirely on-device using Google's AI Core. Your prompts are processed locally, and no data is ever sent to a server. What you ask the model stays between you and your phone. • Full Model Control You have full control over the model. Of course you cannot edit the System Prompt, but trust me, you don't want to play with it anyway. You can augment it by adding User Instructions that will be incorporated. You can also control the main parameters of the model: Temperature: Control the creativity vs. coherence of the AI (0.0 = robotic, 1.0 = chaotic). Max Response Tokens: Decide how much you want the AI to generate, from a short sentence to a longer paragraph. The AI usually generates less than 75% of the requested context size. This happens because not every token is used, but each one is counted. • Multiple chats! You now have ability to have multiple chats, each with own context. The global settings apply to each of them, but you can rename, pin, delete them at your will. The AI also can help you choose the name for the conversation! • Custom Instructions You can use [USER INSTRUCTIONS] to steer the AI to do what you want. You can create custom personalities, set specific rules ("always answer in pirate"), or tailor the AI's behavior for your specific tasks. • Context-Awareness Should you enable this function, the app will inject the current time/date and app language into the prompt, giving the model real-time awareness for more relevant answers. This is a proof of concept, if it works well there will be more integrations. • Total Transparency Unlike most AI apps, there are no hidden tricks. You can literally read the System Prompt used in the app from the settings. (https://github.com/Puzzaks/PAIOS/assets/system_prompt.txt) You see exactly what the model sees. • A (Strong) Personality Even though the model has it's own unchangeable prompt, a lot of time was spent on fixing model behavior. This does not work 100% times, but if you try to use the app without the prompt, you'll see that you are speaking with an older AI model. • Excercise caution, this model is in Alpha Just like the application, the model underneath is a developer preview of a new API for the model. The model requires Google's AI Core, which is currently only available on a very limited number of Android devices. If your device isn't supported, the app will still function, but the AI model would not be there for it. For this reason, the app will not allow the user to explore the app beyond the initiation screen if there is no support for the device. There are a handful of supported devices. This limitation is imposed by Google, not by me. You can learn more here: https://developers.google.com/ml-kit/genai#prompt-device • Disclaimer This is an independent, third-party app and is not officially affiliated with, endorsed by, or sponsored by Google. "Gemini" is a trademark of Google, LLC.