Build a Gas Field Expansion Evaluator with AutoGen Multi-Agent Systems
Evaluating complex field expansion bids for Saudi Arabia's Jafurah gas project requires coordinated specialized expertise. Build a multi-agent conversational system using Microsoft AutoGen to simulate joint evaluations between drilling, environmental, and financial analyst agents.
What you are building
Construct an AutoGen multi-agent debate and consensus pipeline to assess bidder capability for the Jafurah gas expansion project.
How work is evaluated
Evaluates multi-agent consensus formation and correct ranking of gas field expansion contractors.
Datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
consensus_check
Verify agents reached explicit consensus
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
ranking_accuracy
Accuracy of frontrunner identification • target: 0.95 • range: 0.85-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
Learning goals
Configure AutoGen UserProxyAgent and AssistantAgent instances with domain prompts
Implement group chat orchestrators to mediate consensus between conflicting bidder ratings
Evaluate unconventional gas drilling capabilities and environmental footprint claims
Output consolidated bidding rank matrices with transparent reasoning traces
Resources
Contractor technical proposals, fracking rig capabilities, and cost models for Saudi Jafurah field development.
Challenge setup
Evaluates multi-agent consensus formation and correct ranking of gas field expansion contractors.
Challenge input
JSON containing contractor profiles
AutoGen
Microsoft framework built specifically for multi-agent conversational patterns
Evaluated output
JSON consensus ranking
- Verify agents reached explicit consensus
- Accuracy of frontrunner identification • target: 0.95 • range: 0.85-1
- Benchmark: MultiAgentConsensusBench
- Ranking Accuracy target: 0.95
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Shows the first task, up to three tools, and up to three evidence items.
View technical recipe
Configured tools
- AutoGen · Required
- Microsoft AutoGen · Optional
- Zed · Optional
Evaluation contract
- consensus_check · Weight 1
- ranking_accuracy · Weight 1
Metrics
- ranking_accuracy · Target 0.95
Benchmarks
- MultiAgentConsensusBench
- general
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON containing contractor profiles
Output: JSON consensus ranking
Execution harness: Python
Evaluation tools
- AutoGen · Agent Framework
Recipe freeze marker
No recipe freeze time is recorded.