Model Config
Domain: Models & Providers
Motivation
Model config is the per-model and per-route configuration that shapes each call: sampling (temperature, maxTokens), reasoningEffort, and image budgets. The adapter validates an explicit or configured reasoning effort against the exact model before any provider I/O and materializes an adapter-configured output cap when a request omits one.
Concrete Example
On dsh-llm-deepseek-api-key: reasoningEffort: high (off | low | high | max), maxTokens: 256000, and image caps like maxImagesPerRequest: 600.
Analogy
The dial settings on a camera — exposure, focus, and frame size for each shot.