Research
LATS-RCA: Multi-Agent LLM System for Microservice Root Cause Analysis
LATS-RCA replaces linear single-path diagnostic reasoning with a multi-agent tree search approach for root cause analysis in microservice-based systems. Prior LLM-based RCA tools follow a single diagnostic path and struggle with hallucinations that lead to irrelevant investigation steps. The multi-agent design lets specialized agents explore parallel hypotheses simultaneously, addressing the combinatorial complexity of debugging distributed systems. Directly useful for SRE teams considering LLM-assisted incident response.
Source
↳ Follow the thread