Future and Roadmap

Pheron Agent (Pheron Agent) is constantly evolving to push the boundaries of native, on-device AI orchestration. Here is the upcoming feature and performance optimization roadmap.

Upcoming Milestones

1. Vision Optimization (v11.0)

  • Compact VLMs: Support for next-generation on-device vision models designed for 16 GB unified memory systems without disabling the semantic_vision layer.
  • Metal-Accelerated Frame Processing: Direct pipeline streaming from macOS ScreenCaptureKit into MLX GPU tensors for real-time visual monologue loops.

2. Multi-Agent Swarms

  • XPC Actor Swarms: Spawning separate Distributed Actor instances in isolated processes to collaborate on parallel tasks (e.g. one agent coding while another runs test suites).
  • Federated On-Device Memory: Encrypted local sync of SkillVault procedurals across multiple owned Apple devices.

3. Energy and Power Profiling

  • Ultra-Low Battery Modes: Adaptive inference budgets that adjust token generation speeds and context lengths dynamically when the device is running on battery power.
  • NE-Only Classification: Transitioning all routing decisions exclusively to the Apple Neural Engine to keep the high-performance GPU cores completely asleep during background idle loops.

4. Deep App Integrations

  • Native Slack and Xcode Hooks: Direct APIs for popular development and communication tools without relying on browser scraping or system automation falls.
  • Enhanced Siri Shortcuts Bridge: Automated compilation of typed Swift extensions from Siri Shortcuts payloads.