Godmode Runtime

What it does

Godmode routes every prompt through a 3-layer pipeline:

1. L1 Router — classifies intent, resolves required capabilities, picks the cheapest model that can handle it (Ollama local first)
2. L2 Executor — runs the model, streams output live, then scores quality 0–1 with a local judge
3. L3 Governor — if score < 0.55, flags ⚠ NEEDS REVIEW (skill mode) or retries with a cloud model (standalone mode)

Result: You get fast, cheap answers for simple tasks. Hard tasks escalate automatically. Every run is tracked so you can see exactly how much you saved vs Claude Opus.

Example use cases

1. Code review without hallucinations

You have a Kotlin service. You want a security review but the local model keeps suggesting Spring patterns that don't exist in your codebase.