Skip to main content
Loading navigation
Back to challenges

BMS Telemetry Autonomous Maintenance Agent with LangChain & Gemini 2.5 Flash

Modern commercial smart buildings generate high-frequency Building Management System telemetry that humans struggle to monitor continuously. Build an autonomous operations assistant using LangChain and Gemini 2.5 Flash to transform BMS sensor telemetry into prioritized work orders. Maintain zero missed critical fault alerts across 24/7 telemetry streams.

Business OperationsHosted by Vera
Sign in free to start

What you are building

Construct a smart building predictive operations agent using LangChain graph workflows and Gemini 2.5 Flash model inference.

How work is evaluated

Evaluation

Evaluates BMS fault detection sensitivity and automated maintenance dispatch accuracy.

Datasets

Public datasets, plus the private datasets you have access to.

How submissions are scored

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

Critical Anomaly Detection

Must detect sudden temperature spike as fault

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

Fault Severity Precision

Precision score of severity classification • target: 0.98 • range: 0.85-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

  • Ingest BACnet/Modbus IoT time-series telemetry streams into LangChain processing nodes

  • Leverage Gemini 2.5 Flash low-latency inference to detect temperature and valve anomalies

  • Synthesize mechanical equipment diagnostics into structured facility work orders

  • Implement LangGraph stateful loops to track maintenance resolution states

Resources

Dataset notes

BACnet building management system telemetry logs covering HVAC, chiller plant, and VAV box performance.

Challenge setup

Evaluates BMS fault detection sensitivity and automated maintenance dispatch accuracy.

Challenge input

JSON with sensor_id, parameter, time_series_values, unit

LangChain

Provides stateful workflow graph capabilities via LangGraph

Gemini 2.5 Flash

Offers low-latency fast multimodal/structured reasoning

Evaluated output

JSON with fault_detected (boolean), severity, work_order_summary

Evaluation checks
  • Must detect sudden temperature spike as fault
  • Precision score of severity classification • target: 0.98 • range: 0.85-1
Configured evaluation evidence
  • Benchmark: BMSFaultBench
  • Fault Severity Precision target: 0.98
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • Google Gemini · Required
  • Langchain · Optional
Policy Serving
  • Gemini 2.5 Flash · Optional
Orchestration
  • Langchain · Optional

Evaluation contract

  • Critical Anomaly Detection · Weight 1
  • Fault Severity Precision · Weight 1

Metrics

  • Fault Severity Precision · Target 0.98

Benchmarks

  • BMSFaultBench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON with sensor_id, parameter, time_series_values, unit

    Output: JSON with fault_detected (boolean), severity, work_order_summary

    Execution harness: Python

Evaluation tools

  • LangChain · Agent Framework
  • Gemini 2.5 Flash · Inference Engine

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about BMS Telemetry Autonomous Maintenance Agent with LangChain & Gemini 2.5 Flash