China Clean Energy CO2 Trajectory Monitor with OpenAI Agents SDK and Meta Llama
In a historic shift, rapid expansion of solar, wind, and nuclear generation in China has driven carbon emissions into structural reverse for the first time. Verifying these emissions reductions requires cross-referencing satellite remote sensing signals of power plant plumes, thermal coal burn indexes, and state grid telemetry. In this challenge, you will implement an autonomous energy tracking agent using the OpenAI Agents SDK fine-tuned with Meta Llama open-weights models. Your agent will continuously ingest regional thermal anomaly datasets, coal consumption rates, and solar generation telemetry to quantify CO2 trajectory shifts. Your solution will classify whether thermal power generation declines are structural or temporary weather anomalies, providing real-time carbon intelligence for climate risk analysts and carbon offset markets.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an emissions trajectory analysis agent using OpenAI Agents SDK and Meta Llama to verify clean energy replacement of thermal generation.
How work is evaluated
Evaluates the system's ability to process grid capacity feeds and accurately determine monthly CO2 delta.
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.
math_integrity_check
Emissions reduction formula correctly derived from generation mix
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
co2_estimation_error
Absolute percentage error on calculated CO2 delta • target: 0.02 • 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
Build an autonomous tracking agent with OpenAI Agents SDK.
Integrate open-weights Meta Llama models for specialized domain classification.
Analyze multi-source energy grid and satellite carbon plume data.
Calculate real-time CO2 avoidance metrics relative to historical baselines.
Reference links and supporting material
Synthetic time-series dataset of China's regional generation output by fuel source (coal, gas, hydro, solar, wind) based on Carbon Brief data.
Challenge setup
Evaluates the system's ability to process grid capacity feeds and accurately determine monthly CO2 delta.
Challenge input
JSON containing monthly megawatt-hours by generation source
OpenAI Agents SDK
Mandatory framework for agent tools and workflow creation.
Meta Llama
Open-weight foundational model for energy domain classification.
Evaluated output
JSON with co2_change_mt, clean_energy_share_pct, structural_reversal (bool)
- Emissions reduction formula correctly derived from generation mix
- Absolute percentage error on calculated CO2 delta • target: 0.02 • range: 0-1
- Benchmark: EnergyTransitionBench
- Co2 Estimation Error target: 0.02
- 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
- OpenAI · Required
- Meta · Optional
- RunPod · Optional
Evaluation contract
- math_integrity_check · Weight 1
- co2_estimation_error · Weight 1
Metrics
- co2_estimation_error · Target 0.02
Benchmarks
- EnergyTransitionBench
- energy
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing monthly megawatt-hours by generation source
Output: JSON with co2_change_mt, clean_energy_share_pct, structural_reversal (bool)
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Framework
- Meta Llama · Language Model
Recipe freeze marker
No recipe freeze time is recorded.