iteration-manager

Manage CFN Loop iterations by calling prepareIteration to compute the next iteration and capture feedback, then use wakeAgents to generate wake signals that trigger agents for the subsequent iteration. This orchestrates container-based execution, collects confidence scores, and applies gate checks across dockerized workflows.

Each iteration performs task analysis, stores context in Redis, spawns agents, monitors loop execution, applies gate checks, collects consensus scores, and executes GOAP-driven decision flows with configurable thresholds. The iteration manager coordinates these steps and prepares feedback for the next cycle.

Yes, iteration-manager applies to both Loop 2 validators and Loop 3 implementers. It manages execution progression, collects confidence scores and consensus from both validator and implementer agents, and triggers Product Owner decisions based on configurable iteration limits and thresholds.

Call prepareIteration with the current iteration number and feedback data as input. The function computes the nextIteration value and structures feedback to pass to wakeAgents, which then generates signals that notify agents to resume with updated context.

Iteration-manager stores agent context in Redis during loop execution. This includes task analysis results, agent state, confidence scores, and consensus data needed for gate checks and GOAP-driven decision flows across the iteration cycle.

Use configurable thresholds when you need fine-grained control over consensus requirements or confidence score gates specific to your workflow. Use default iteration limits when standard progression is sufficient; both are adjustable to balance between agent convergence time and decision quality.