Pydantic AI Physical Process Anomaly Sentinel for Water Utilities
Cyber-physical attacks on water treatment facilities can alter valve states and dosage controls without generating network alerts. Build a type-safe anomaly detection agent using Pydantic AI to validate telemetry inputs against physics-based process bounds. The sentinel must detect physical parameter manipulation with zero false negatives while maintaining sub-second inference speed.
What you are building
The core problem, expected build, and operating context for this challenge.
Build a Pydantic AI agent with strict type validation to detect physical process parameter manipulation in water treatment networks.
How work is evaluated
Evaluates physical process attack detection accuracy and structured Pydantic model response validity.
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.
zero_false_negatives
Ensures no dangerous physical over-dosing events are missed
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
type_validation_rate
Percentage of agent outputs matching Pydantic response schema • target: 1 • 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
Implement a Pydantic AI agent with structured result types for physical safety validation
Define Pydantic models enforcing hydraulic and chemical mass balance constraints
Detect 'stealth' physical parameter tampering occurring below network alert thresholds
Provide deterministic diagnostic reasonings for utility operator intervention
Reference links and supporting material
Water utility telemetry dataset containing simulated physical attacks (e.g. valve override, chemical dosage spikes).
Challenge setup
Evaluates physical process attack detection accuracy and structured Pydantic model response validity.
Challenge input
JSON with flow_rate_gpm, chlorine_ppm, tank_level_ft, pump_status
Pydantic AI
Assigned framework for type-safe agent execution.
Evaluated output
Pydantic output model with physics_violated (bool), anomaly_type (string), threat_level (string)
- Ensures no dangerous physical over-dosing events are missed
- Percentage of agent outputs matching Pydantic response schema • target: 1 • range: 0-1
- Benchmark: Water Infrastructure Cyber-Physical Benchmark
- Type Validation Rate target: 100%
- 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
- Pydantic AI · Required
- Pydantic AI · Required
Evaluation contract
- zero_false_negatives · Weight 1
- type_validation_rate · Weight 1
Metrics
- type_validation_rate · Target 1
Benchmarks
- Water Infrastructure Cyber-Physical Benchmark
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with flow_rate_gpm, chlorine_ppm, tank_level_ft, pump_status
Output: Pydantic output model with physics_violated (bool), anomaly_type (string), threat_level (string)
Execution harness: Python
Evaluation tools
- Pydantic AI · Type-Safe Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.