LogAct: Enabling Agentic Reliability via Shared Logs

TL;DR

LogAct introduces a shared log abstraction for LLM-driven agents, enabling reliable, transparent, and recoverable agent actions with introspection and control capabilities.

Contribution

It proposes LogAct, a novel shared log framework that enhances agent reliability, transparency, and recoverability in complex environments.

Findings

Agents recover efficiently from failures.

Agents debug and optimize their performance.

Agents stop unwanted actions with minimal utility loss.

Abstract

Agents are LLM-driven components that can mutate environments in powerful, arbitrary ways. Extracting guarantees for the execution of agents in production environments can be challenging due to asynchrony and failures. In this paper, we propose a new abstraction called LogAct, where each agent is a deconstructed state machine playing a shared log. In LogAct, agentic actions are visible in the shared log before they are executed; can be stopped prior to execution by pluggable, decoupled voters; and recovered consistently in the case of agent or environment failure. LogAct enables agentic introspection, allowing the agent to analyze its own execution history using LLM inference, which in turn enables semantic variants of recovery, health check, and optimization. In our evaluation, LogAct agents recover efficiently and correctly from failures; debug their own performance; optimize token usage in swarms; and stop all unwanted actions for a target model on a representative benchmark with just a 3% drop in benign utility.