Yi: The Opinionated Agent Adapter The "Bureaucracy" and "HR Department" of the 33GOD Ecosystem. Yi is the implementation framework that enforces 33GOD conventions on top of the Flume protocol. It serves as the "Adapter Layer" that wraps specific AI SDKs (Letta, Agno, Smolagents) into compliant Manager and Contributor nodes. 👔 The "Suit" for your Agents While Flume defines what a Manager is, Yi defines how a Manager behaves. Yi provides the machinery to: • Wrap Intelligence: Adapts raw LLM outputs from various providers into structured Flume WorkResults. • Enforce Memory: Implements the YiMemoryStrategy to synchronize context and "Memory Shards" across team members. • Manage Lifecycle: Connects to Agent Forge to handle "Recruitment" (creation) and "Onboarding" (context injection). ⚙️ Key Modules • YiManager: An abstract class that implements Flume's Manager interface, providing built-in logic for delegation strategies (e.g., Round Robin, LLM-Driven Selection). • OnboardingSpecialist: A service that injects TeamContext into raw agents before they are allowed to accept tasks. • Bloodbank Integration: Automatically emits standardized events (yi.agent.state.changed) to the system event bus. 🚀 Usage Yi is designed to be the bridge between "Brains" and "Jobs." import { YiManager, YiMemoryStrategy } from '@33god/yi'; import { LettaClient } from 'letta-sdk';

// Yi adapts the Letta Brain to the Flume 'Manager' Role class EngineeringManager extends YiManager { constructor(memory: YiMemoryStrategy) { super(memory); this.brain = new LettaClient(); } }