Modelpublic

thetom-ai/PsyGuard-gpt-oss-20B-Orchestrator-alpha

sourceHugging Faceapache-2.0updated 3mo agoView on Hugging Face

0likes72downloads

Model Card

PsyGuard-gpt-oss-20B-Orchestrator-alpha

Experimental gpt-oss-20B orchestrator fine-tune. Built on openai/gpt-oss-20b.

alpha — experimental. Internal evals only. Dogfood at your own risk.

Format: Q8_0 GGUF (llama.cpp / LM Studio / Ollama), MXFP4 experts. The GGUF is the default orchestrator (spine+boundary+build merged) — run it directly:

bash

llama-server -m PsyGuard-gpt-oss-20B-Orchestrator-alpha-Q8_0.gguf -ngl 99

For the full 3-mode content router, the LoRA adapters are also included: build (default), compute (math), review (vet a sub-agent's result), selected per request by router.py. Merge spine+boundary first; compute/build are alternates on top; review stacks on build. Harmony format, low/medium reasoning effort.