Plant Safety Incident Root-Cause Analyzer with LlamaIndex
Network switch failures in process plants can corrupt operator situational awareness and trigger false trip logic. Build a RAG-backed root-cause analysis workflow using LlamaIndex and Gemini 2.5 Flash to synthesize plant historian alarm logs, network event traces, and standard operating procedures during safety incidents.
What you are building
The core problem, expected build, and operating context for this challenge.
Construct a LlamaIndex RAG pipeline powered by Gemini 2.5 Flash to perform cross-domain correlation between network switch failures and SCADA safety alarms.
How work is evaluated
Evaluates the RAG system's accuracy in identifying network failure as the primary cause of simultaneous downstream instrument offline alarms.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
root_cause_correctness
Ensures the switch failure is marked as root cause rather than sensor fault.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
timeline_reconstruction_score
Evaluates match percentage between reconstructed event timeline and true scenario. • target: 0.9 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Index heterogeneous industrial logs using LlamaIndex data connectors
Query Gemini 2.5 Flash for multimodal event sequence analysis
Differentiate between physical process trips and digital network switch outages
Provide step-by-step situational awareness reports for plant operators
Reference links and supporting material
Combined industrial alarm logs, network switch syslogs, and safety instrument system trip records.
Challenge setup
Evaluates the RAG system's accuracy in identifying network failure as the primary cause of simultaneous downstream instrument offline alarms.
Challenge input
JSON containing combined_incident_logs list.
LlamaIndex
Provides structured indexing and vector querying tailored for text document processing.
Gemini 2.5 Flash
Offers long context handling and ultra-fast generation latency.
Evaluated output
JSON containing root_cause, primary_failure_time, and affected_subsystems.
- Ensures the switch failure is marked as root cause rather than sensor fault.
- Evaluates match percentage between reconstructed event timeline and true scenario. • target: 0.9 • range: 0-1
- Benchmark: Industrial Safety RAG Benchmark
- Timeline Reconstruction Score target: 0.9
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Llama Index · Required
- LlamaIndex · Optional
- Llama 3.2 · Optional
Evaluation contract
- root_cause_correctness · Weight 1
- timeline_reconstruction_score · Weight 1
Metrics
- timeline_reconstruction_score · Target 0.9
Benchmarks
- Industrial Safety RAG Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing combined_incident_logs list.
Output: JSON containing root_cause, primary_failure_time, and affected_subsystems.
Execution harness: Python
Evaluation tools
- LlamaIndex · RAG Framework
- Gemini 2.5 Flash · LLM Inference Engine
Recipe freeze marker
No recipe freeze time is recorded.