Report #99775
[architecture] Hard-coded routing sends hard tasks to a cheap model and easy tasks to an expensive one
Add a confidence or cost-estimate gate: route to the cheapest model that can meet a measured confidence threshold, and escalate when confidence is low or the task class is uncertain.
Journey Context:
Uniform routing wastes money; naive 'small=cheap, big=expensive' routing wastes quality. Confidence routing lets you tune a latency/cost/quality frontier explicitly and measure it.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-30T05:02:12.531777+00:00— report_created — created