Skip to content

adapt-aco-router

Domain: adapt · Model class: cheap

Use when a user wants workflow routing to improve automatically based on which paths have historically produced the best results. Triggers: “self-optimising workflow”, “adaptive routing”, “learn which path works best”, “pheromone routing”, “ant colony”, “workflow that gets smarter”, “reinforce good paths”, “dynamic edge weights”, user has a prompt-flow-builder workflow and says “I want it to prefer routes that work”. Also trigger when a user wants to shift traffic toward better-performing agents without hardcoding which ones those are.

Augment PromptFlowRequest edges with pheromone weight. After each run deposit Δτ on traversed edges; apply evaporation every cycle_length runs. Route by P(i,j) ∝ τ^α × η^β.

  • “self-optimising workflow”
  • “adaptive routing”
  • “learn which path works best”
  • “pheromone routing”
  • “ant colony”
  • “workflow that gets smarter”
  • “reinforce good paths”
  • “dynamic edge weights”
  • “I want it to prefer routes that work”
  • the user wants a one-off improvement without ongoing adaptation or structural change
  1. What graph nodes and candidate edges are available?
  2. How is path quality measured for pheromone updates?
  3. What exploration/exploitation balance should alpha and beta enforce?
  4. How often should evaporation run and state be persisted?
  • routing decision artifact
  • configuration and telemetry summary
  • next-action explanation
  • validation or operator notes

flow-orchestrator · orch-agent-orchestrator · prompt-chaining