Summary

open-yachiyo provides an AI-native runtime and desktop shell optimized for production-like agent execution. It addresses the need for controllable, bounded ReAct loops, explicit tool-calling, and session-isolated desktop agents. The target audience is developers building complex client-heavy desktop applications where LLMs must safely drive local asynchronous tasks with auditable lifecycles and strict recovery boundaries. It offers a deterministic runtime core, distinct from multi-channel orchestration frameworks.

How It Works

The core approach centers on a bounded turn loop (ReAct: Reason -> Act -> Observe) with explicit tool-calling and memory operations. It employs session isolation and file-backed persistence, coupled with a desktop-first interaction model featuring Live2D avatars and streaming bubbles. This design prioritizes runtime controllability and auditable lifecycles over broad orchestration capabilities.

Quick Start & Requirements

Installation involves npm install and npm run dev. Configuration of LLM providers is managed via ~/yachiyo/config/providers.yaml, with detailed guides available for provider setup and voice cloning/TTS. A Windows installer (npm run desktop:dist:win) is provided for one-click deployment. Key resources include a demo video and documentation for architecture, testing, and usage cases.

Highlighted Details

• Controllable runtime loop with hard step boundaries and JSON-RPC/queue ingress.
• EventBus for decoupled tool dispatch and real-time streaming message pipeline.
• Real-time auto lipsync for streaming output, driving Live2D mouth parameters.
• Session persistence and long-term memory tools (memory_write, memory_search).
• Desktop rich rendering (Markdown, LaTeX, Mermaid) and multimodal image input.
• Full-chain debug lane via SSE for end-to-end request tracing.

Maintenance & Community

The project is maintained by sdyzjx and wkf16. Development status is tracked via GitHub issues and internal TODO lists, indicating active progress on core features and planned enhancements.

Licensing & Compatibility

The software license is not explicitly stated in the provided README. Compatibility notes for commercial use or closed-source linking are not detailed.

Limitations & Caveats

The project is in active development, with numerous features marked as TODO or in later development phases (e.g., Phase E stabilization). Key functionalities like macOS Accelerometer integration and advanced security/permission models are pending. The absence of a specified license may pose adoption risks.