Inflak organizes human and AI agent interactions with layered control
Inflak: A Unified Orchestration Architecture for Human-Agent Interaction
Human-Computer Interaction
Summary
Managing when and how AI agents talk to people can be complicated. The authors introduce Inflak, a system that breaks down this process into four clear steps to better handle different interaction needs. This approach helps make AI-human conversations more flexible, easier to update, and usable for many types of tasks. They tested Inflak in various situations and found it works well as a general solution for organizing these interactions.
What this means in practice
- •For conversational ai developers: Create flexible AI agents that handle complex user interactions across multiple tasks with modular plugins.
- •For customer support platform teams: Integrate AI agents that manage conversation flow and user interaction more dynamically for diverse support scenarios.
Authors
Lin Gao, Li Chen, Jianxun Lian, Ji Li, Haotian Li
Abstract
LLM-powered agents increasingly support open-ended activities, yet orchestrating when and how agents interact with humans remains challenging. Existing approaches often provide interaction mechanisms tied to particular tasks or execution patterns, lacking a unified architecture for diverse and evolving interaction needs. We introduce Inflak, a unified four-layer architecture for human-agent interaction orchestration that separates state reconstruction and routing, interaction and execution planning, interaction design, and interface rendering. Its layered design with progressive resolution enables general applicability, runtime flexibility, module composability, and orchestration maintainability. We implement Inflak as agent-native, skill-based plugins within a general-purpose agent framework and evaluate it through a gallery of various application scenarios and interaction paradigms, technical performance testing, and a comparative implementation case study. The evaluations show that Inflak can serve as a general and extensible foundation for orchestrating diverse human-agent interaction flows.