Provider Routing: Switch Models Without Rewriting the App

This build lab focuses on the provider router that lets Hermes-style systems choose models by job instead of by brand loyalty. The goal is not to copy a private machine setup. The goal is to learn the architecture pattern well enough to build a small, classroom-safe version.

The app calls one router interface; the router decides whether to use local Hermes, a hosted open model, or a frontier API based on privacy, difficulty, latency, and budget.

Build the small version

1. 1Draw or write a routing table with task labels, preferred providers, fallback providers, and escalation rules.
2. 2Mark which parts are user-facing, which parts are internal, and which parts require approval.
3. 3Choose one low-risk workflow and implement only that workflow first.
4. 4Add one failure case before adding a second feature.
5. 5Write a short operator note: what the agent may do, what it must ask about, and what it must never do.

route(request):
  if request.contains_private_data:
    return local_hermes

  if request.needs_long_reasoning:
    return frontier_reasoning_model

  if request.is_batch_or_low_stakes:
    return cheap_open_model

  return default_chat_model

The big idea: router is not decoration. It is part of the product architecture students need before an agent becomes safe enough to use with real people.