by-hypothesis-debate

What problem does it solve?

Adversarial multi-agent strategy selection for protein/antibody design campaigns. Spawns competing hypothesis agents (conservative / aggressive / diverse), then ranks proposals with a reflection agent before committing GPU compute. Use when starting a novel-target campaign, when research findings are contradictory, or when multiple design modalities are viable.

Core Features & Use Cases

• Spawns Conservative, Aggressive, and Diverse agents in parallel to explore different design spaces.
• Uses a Reflection agent to rank proposals against a fixed rubric and resolve ties.
• Outputs an updated campaign_config.yaml and a decision_summary.md for governance and handoff.
• Validates all proposals against a strict schema and logs audit trails for compliance and troubleshooting.

Quick Start

Invoke the orchestrator after by-research results are ready to spawn three hypothesis agents, run the debate, and write the winning campaign plan to campaign_config.yaml.