Refinery Repair Schedule Diagnostician with Pydantic AI and Fireworks AI
Restoring Kuwait refinery operations after unexpected disruption requires rapid equipment repair scheduling. Build a Python agent using Pydantic AI and Fireworks AI high-speed inference to process refinery inspection logs, validate Pydantic output schemas, and sequence critical repair tasks under tight safety constraints.
What you are building
The core problem, expected build, and operating context for this challenge.
Leverage Pydantic AI type-safe agent validation backed by Fireworks AI fast inference to sequence refinery emergency repair operations.
How work is evaluated
Evaluates Pydantic schema validation accuracy and task sequence safety order for refinery repair operations.
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.
Safety Dependency Test
Ensures dependent repair tasks strictly follow prerequisite tasks.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Schema Conformance Rate
Percentage of generated outputs strictly passing Pydantic validation. • 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 Pydantic AI type-safe agent dependency injection and model binding
Utilize Fireworks AI fast inference runtime for low-latency emergency decision pipelines
Structure maintenance log parsing into validated Pydantic data schemas
Order refinery repair sequences according to safety dependencies and operational priority
Reference links and supporting material
Simulated emergency repair logs and safety dependency graphs for Kuwait's Al-Zour and Mina Al-Ahmadi refineries.
Challenge setup
Evaluates Pydantic schema validation accuracy and task sequence safety order for refinery repair operations.
Challenge input
JSON object with damaged_units list, safety_dependencies, and available_crews count.
Pydantic AI
Type-safe Python agent output validation
Fireworks AI
Ultra low-latency LLM execution
Evaluated output
JSON object with repair_sequence list and estimated_hours_total.
- Ensures dependent repair tasks strictly follow prerequisite tasks.
- Percentage of generated outputs strictly passing Pydantic validation. • target: 1 • range: 0-1
- Benchmark: GCC Refinery Maintenance Benchmark
- Schema Conformance 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
- Sacred · Optional
- Fireworks AI · Optional
- Fireworks AI · Optional
- Pydantic AI · Required
Evaluation contract
- Safety Dependency Test · Weight 1
- Schema Conformance Rate · Weight 1
Metrics
- Schema Conformance Rate · Target 1
Benchmarks
- GCC Refinery Maintenance Benchmark
- energy
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON object with damaged_units list, safety_dependencies, and available_crews count.
Output: JSON object with repair_sequence list and estimated_hours_total.
Execution harness: Python
Evaluation tools
- Pydantic AI · Agent Framework
- Fireworks AI · Inference Engine
Recipe freeze marker
No recipe freeze time is recorded.