agiresearch/AIOS

[Roadmap] AIOS Roadmap Q4 2024

evison opened this issue · 1 comments

Optimize the LLM kernel

  • Translate current implementation to more efficient implementation (more efficient and still cross-platform)
    • Multi-thread/Multi-process optimization (@om-raheja)
    • Memory optimization with low-level handling
    • Hierarchical storage

Support more agent frameworks in AIOS

  • ReAct
  • Reflection
  • AutoGen
  • Open Interpreter
  • MetaGPT
  • SwarmGPT

AIOS File System

AIOS Cache Mechanism

  • Prompt-Response Caching Mechanism (@samlin)

AIOS Semantic Scheduling

  • Efficiency-based Prompt-level Scheduling for AIOS
  • Priority-based Agent-level Scheduling for AIOS
  • Multi-core AIOS based on Multi-LLM Scheduling
  • AIOS Multi-task Collaborative Scheduling

LLM integrations

  • Groq
  • LiteLLm

Terminal/Shell UI

  • Ncurses terminal UI

Agent Hub

  • Agent packaging
  • Agent uploading and downloading
  • Agent distribution
  • One-click agent packing, uploading, distribution and deployment
  • Create a better categorization structure for agents

AIOS Personalization

  • Personalized memory mechanism for AIOS
  • Personalized memory retrieval
  • Personalized memory manager

AIOS Multimodality Support

  • Support Images/Videos on AIOS Chat
  • Support voice control on AIOS Chat

AIOS Hackathon

  • Standard interface for agents and LLM kernels
  • Memory optimization low level handling
  • https://www.openblas.net as cuda alternative

These would have to be completed before