Local Power & Water Consumption Predictor for Data Centers with OpenAI Agents SDK & LocalAI
Data center permitting faces fierce political headwinds due to regional water usage and electrical load surges. Build a local resource impact prediction system using OpenAI Agents SDK backed by an offline LocalAI inference runtime to compute projected daily water consumption and peak grid demand. Ensure model execution runs fully local with less than 5% prediction error.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an offline data center resource consumption modeling tool using OpenAI Agents SDK configured with a LocalAI inference runtime.
How work is evaluated
Evaluates local resource consumption predictions against physical thermodynamics ground truth.
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.
power_exact
Verifies peak power calculation exactness.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
water_mae
Mean Absolute Error in predicted daily water consumption. • target: 5000 • range: 0-100000
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Configure OpenAI Agents SDK to route requests through LocalAI OpenAI-compatible endpoints
Build resource demand modeling tool functions for IT megawatt capacity
Compute cooling system water evaporative loss based on local dry-bulb weather inputs
Ensure zero external network calls during offline policy evaluation
Reference links and supporting material
Resource consumption dataset containing weather conditions, cooling system specifications, and measured water/power consumption.
Challenge setup
Evaluates local resource consumption predictions against physical thermodynamics ground truth.
Challenge input
JSON object with datacenter_it_mw, cooling_type, ambient_temp_f, pui_factor
OpenAI Agents SDK
Mandatory assigned agent SDK framework.
LocalAI
Mandatory offline local inference runtime integration.
Evaluated output
JSON object with daily_water_gallons, peak_power_mw
- Verifies peak power calculation exactness.
- Mean Absolute Error in predicted daily water consumption. • target: 5000 • range: 0-100000
- Benchmark: DataCenterEnergyBench
- Water Mae target: 5000
- 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
- LocalAI · Required
- OpenAI · Optional
- aiXplain · Optional
Evaluation contract
- power_exact · Weight 1
- water_mae · Weight 1
Metrics
- water_mae · Target 5000
Benchmarks
- DataCenterEnergyBench
- energy
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON object with datacenter_it_mw, cooling_type, ambient_temp_f, pui_factor
Output: JSON object with daily_water_gallons, peak_power_mw
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Framework
- LocalAI · Local Inference Server
Recipe freeze marker
No recipe freeze time is recorded.