Gas Expansion Procurement Evaluator with OpenAI Agents SDK and Windsurf
L&T's Jafurah fourth expansion award requires automated multi-vendor EPC contract verification. Build an OpenAI Agents SDK tool pipeline inside Windsurf IDE workflows to extract contract clauses, cross-reference Aramco procurement specifications, and flag budget deviation risks.
What you are building
The core problem, expected build, and operating context for this challenge.
Develop an agentic compliance parser using OpenAI Agents SDK within the Windsurf AI IDE environment to evaluate massive EPC expansion bids.
How work is evaluated
Evaluates contract extraction precision and compliance flagging accuracy for EPC project bids.
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.
Contract Value Extraction Test
Checks if extracted USD amount matches gold ground truth.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Contract Parsing Precision
Ratio of correctly identified contract parameters. • 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
Build OpenAI Agents SDK function-calling pipelines for contract verification
Leverage Windsurf AI IDE workflows for rapid agent prompt and code refactoring
Extract key financial and schedule deliverables from Jafurah phase four contract text
Evaluate EPC bid compliance against Saudi Aramco standards
Reference links and supporting material
Synthetic EPC bidding documents and scope statements for the Jafurah gas expansion phase four awards.
Challenge setup
Evaluates contract extraction precision and compliance flagging accuracy for EPC project bids.
Challenge input
JSON object with contract raw text and target compliance rule set.
OpenAI Agents SDK
Function calling and dynamic tool selection
Windsurf
AI-assisted code generation and agent workflow debugging
Evaluated output
JSON object with extracted total_value_usd, lead_contractor, and compliance_status.
- Checks if extracted USD amount matches gold ground truth.
- Ratio of correctly identified contract parameters. • target: 1 • range: 0-1
- Benchmark: GCC EPC Contract Parsing Benchmark
- Contract Parsing Precision target: 1
- 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
- Windsurf · Required
- OpenAI · Optional
- Cartesia · Optional
- Windsurf · Required
Evaluation contract
- Contract Value Extraction Test · Weight 1
- Contract Parsing Precision · Weight 1
Metrics
- Contract Parsing Precision · Target 1
Benchmarks
- GCC EPC Contract Parsing Benchmark
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON object with contract raw text and target compliance rule set.
Output: JSON object with extracted total_value_usd, lead_contractor, and compliance_status.
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Framework
- Windsurf · AI Engineering Tooling
Recipe freeze marker
No recipe freeze time is recorded.