Skip to content

Latest commit

 

History

History
52 lines (48 loc) · 2.61 KB

File metadata and controls

52 lines (48 loc) · 2.61 KB

Project Roadmap


Phase 1 AI Core - Basic

  • Release 1.0.0
    • Text-only interface for the AI Core API (Gemini Nano / Flan T-5) implementation, Play Store and source release
  • UI and UX 1.0.1
    • Multi-chat application with finalized MD3 style and state control for model download
  • Basic features 1.1.0
    • Multichat, chat pinning, basic per-chat settings
  • Hotfixes and monitoring 1.1.1, 1.1.2
    • Optional analytics and crash reporting, logging, issue fixes in UI and backend
  • Additional features
    • Per-chat personas, temperature and token limit settings, etc.
  • Generation extension
    • Workaround for 25s response generation time that ends answers prematurely
  • Master Prompt fixes
    • Finalizing the master prompt so it delivers more consistency
  • Monetization
    • Monetize the project to allocate more time to the project
  • Finalization
    • Polishing and fixing issues, supporting users and optimisations

Phase 2 AI Core - Improved

  • Agentism
    • Add agentic and RAG capabilities to the model with tool use loops, tool usage protocol
  • Context expansion
    • Find a way to allow model to use more data simultaneously, without overflowing the context window
  • Multimodality
    • Add image recognition that is supported by Gemini Nano
  • Ephemeral Context
    • Implementing a "Short-Term Memory" buffer for heavy data (files/logs) that doesn't pollute the long-term chat history.

Phase 3 Model-agnosticizm

  • Custom models
    • Allow users to set up and run custom models (like Gemma 2B) besides Gemini Nano for redundancy
  • Connectivity
    • Create a companion app that allows bigger models to run in other environment to connect the PAIOS to better hardware (prolly via Ollama)
    • This has to be made in a way that allows even users who are not tech-savvy to just install the model and run it
  • Failover
    • Auto-switch to local model if the remote connection drops.

Phase 4 PAIOS

  • File System Access
    • Read access to files on the device, in the network or online for more capabilities
  • Scripting
    • Allow the AI to use scripts to interact with other devices and use better tools like smart home integration
  • Extenstions "market"
    • Create a way for users to create extensions and publish them for other users to use
  • Voice chat
    • Real-time, low-latency voice interaction (running locally). Possibly, ability to join chats on social media apps or act on behalf of user
  • And much, much more