ArchonConvert Operator (generators -> critic -> fuser)

[yash94404]

• Created new ArchonConvert operator that runs a convert prompt through a layer of multiple generators, which have all their outputs critiqued in the critic layer, and fused together into an optimal response in the fuser layer
• Implemented ArchonConvert class, new prompt strategies for Archon operations, new prompts for Archon
• Integrated into optimizer with ArchonConvertRule, allowing the optimizer to use any subset of models for the generator layer (can modify LLM architecture if needed)

Next Steps:

• Modifying LLM Architecture to support multiple critics/fusers, adding verifier/ranker?
• Correspondingly to above, improving optimizer using DAG approach talked about with Siva
• Any other ideas!