Inference OS Studio · Orivael Governance Console

Inference OS Demo

one request · many controls · one signed result

Hit Run to execute the full Inference OS pipeline.

      [0] Intent Kernel → [1] Router → [2] Retrieval → [3] Generation → [4] Governance → [6] Audit

Axiom Container

Text Model always present

HuggingFace model ID

Hardware

Pack Options (SRD flags)

α (alpha) 0 = pure 4-bit (4.5 bpw) · 0.5 = balanced · 1.0 = full residue (13 bpw)

Group size Quantization block size — smaller = finer granularity, higher bpw overhead

Top-k layers (%) Fraction of layers to SRD-pack; others kept FP16

GGUF output format Quantization applied after AXM extraction

+ Drop a module here