feat(commentary): refine rafters-critic persona and improve commentary prompt for humor

feat(media): introduce MediaRouter and stubs for image and speech generation

fix: Update GPU configuration for NVIDIA models to use A100 instead of H200

Refactor modal service and logging setup

feat: Add FP8 quantization support for model serving with environment overrides

feat: Implement well-known typed fields for verdicts in output models

fix(modal): shorten cascade slug to fit DNS limit; add endpoint health tooling

feat: refine structured output handling with guided decoding and mode adjustments

feat(logging): implement structured JSON logging for vLLM and enhance observability

feat: enhance agent personas and improve model performance tuning

fix: update Google Gemma model configuration to use instruction-tuned version

feat: add Nemotron-Cascade-14B-Thinking model configuration to catalogue and README

feat: unify model catalogue and self-hosted routing