Feature

Multi-Agent RCA

When an anomaly is detected, a LangGraph-orchestrated pipeline of 4 specialised AI agents automatically investigates the fault, traces its root cause, and produces an actionable maintenance plan.

84.6%
RCA Success Rate
77s
Avg workflow time
0.862
Confidence score
4
Agents in pipeline
Start Free Trial →

The 4-agent pipeline

Agents execute in sequence. Each feeds its output into the next, building context progressively.

01

Diagnostic Agent

Symptom analysis

Receives the raw anomaly signal and ensemble score. Classifies the symptom cluster and queries the knowledge graph for matching failure patterns.

02

Reasoning Agent

Root cause inference

Uses Groq Llama 3.3 70B to reason over the symptom cluster, equipment history, and SWRL rules. Returns ranked probable root causes with confidence scores.

03

Planning Agent

Maintenance recommendations

Translates confirmed root causes into prioritised maintenance tasks — from immediate corrective actions to scheduled preventive measures.

04

Learning Agent

Continuous improvement

Records technician feedback and outcome data to refine future diagnoses. Accuracy improves with every confirmed or corrected RCA.

Technology

LangGraph

Directed agent orchestration with state management and conditional branching.

Groq Llama 3.3 70B

Ultra-fast inference for the reasoning agent — sub-second LLM responses even on complex fault trees.

Knowledge Graph

OWL ontology with 50+ entities provides the structured context agents reason over.