tags: curated discovery keywords (remove competitor names, dedupe, add search terms)

layout: quantized ONNX builds moved to panther-read-intent-classifier-edge (prior revisions remain accessible)

docs: edge split (variants -> -edge repo, onnx_fp16 paths) + Patronus Ark note

layout: add onnx/onnx_fp16 FP16 export (uniform live layout)

docs: fix LICENSE reference in license section

license: add Apache-2.0 license text

license: Apache-2.0 + upstream (MIT) attribution

docs: full model card per Patronus template (correct file links + real metrics)

Sync current L2 export for unified-v3-routing

docs: link Patronus Protect (patronus.studio)

Replace unified-multitask-model-augmented-v3

Replace unified-multitask-augmented-v3-routing

initial commit