LangChain Data Center Siting Agent with Helicone Monitoring
Build a data center site selection agent using LangChain and Helicone that evaluates grid capacity, water constraints, and local political sentiment. The system parses public zoning records to output a site feasibility score exceeding 85% accuracy.
What you are building
The core problem, expected build, and operating context for this challenge.
This challenge uses LangChain stateful agents combined with Helicone observability to evaluate site risk and zoning viability for energy-intensive data centers.
How work is evaluated
Evaluates the site risk score generation accuracy and Helicone telemetry integration.
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.
Schema Compliance Test
Ensures all required risk fields and trace IDs are present in output
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Risk Precision Absolute Error
Mean absolute error compared to expert gold standard risk evaluation • target: 5 • range: 0-15
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement a LangChain agent using stateful graph workflows for site selection analysis
Integrate Helicone to track agent latency, token usage, and output quality metrics
Parse unstructured zoning documents and municipal council sentiment signals
Calculate unified environmental and political feasibility risk metrics
Reference links and supporting material
Synthetic site selection dataset containing 50 municipal zoning records, power grid capacity metrics, and meeting transcripts.
Challenge setup
Evaluates the site risk score generation accuracy and Helicone telemetry integration.
Challenge input
JSON containing parcel metadata, grid capacity (MW), water availability, and meeting text
LangChain
Manages agent logic and graph state transitions.
Helicone
Tracks agent execution traces, latency, and LLM spending.
Evaluated output
JSON containing overall_risk_score (0-100), political_risk, utility_risk, and trace_id
- Ensures all required risk fields and trace IDs are present in output
- Mean absolute error compared to expert gold standard risk evaluation • target: 5 • range: 0-15
- Benchmark: BuiltEnvironmentFeasibility-v1
- Risk Precision Absolute Error target: 5
- 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
- Langchain · Optional
- Resemble AI · Optional
- Helicone · Required
- Langchain · Optional
Evaluation contract
- Schema Compliance Test · Weight 1
- Risk Precision Absolute Error · Weight 1
Metrics
- Risk Precision Absolute Error · Target 5
Benchmarks
- BuiltEnvironmentFeasibility-v1
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing parcel metadata, grid capacity (MW), water availability, and meeting text
Output: JSON containing overall_risk_score (0-100), political_risk, utility_risk, and trace_id
Execution harness: Python
Evaluation tools
- LangChain · Orchestration Framework
- Helicone · Observability System
Recipe freeze marker
No recipe freeze time is recorded.