Nv Reason Cxr

NV-Reason-CXR

Purpose

• Used for command-shape or live NV-Reason-CXR chest X-ray reasoning smoke tests. Not for diagnosis or clinical reporting.
• Use the wrapper exactly as documented; do not replace the upstream entrypoint with a handwritten implementation.
• Manifest I/O: inputs are chest_xray_image_or_fixture; outputs are result_json.

Instructions

• Read skill_manifest.yaml before changing arguments, side effects, or validation gates.
• Run scripts/run_nv_reason_cxr.py through the documented command below; pass --out-dir only for generated fixtures or harness-managed artifact directories.
• If a host agent exposes run_script, use run_script("scripts/run_nv_reason_cxr.py", args=[...]); otherwise run the Bash/Python command shown below.
• Check the emitted JSON and paired verifier guidance before treating the run as evidence.
• When reporting a completed run, return the full wrapper JSON or at minimum the complete output.response_text exactly as emitted, including any model-generated <think>...</think> and <answer>...</answer> sections. Do not collapse the result to labels unless the user explicitly asks for a summary.

Available Scripts

Prerequisites

• Local backend requirements: GPU/CUDA when declared by the manifest; Python packages listed in runtime.side_effects.pip_packages.
• API backend requirements: public network access to the Hugging Face Space; no local PyTorch, Transformers, CUDA, model cache, or Hugging Face token.
• Side effects: emits result JSON on stdout; may write generated fixture artifacts under the caller's --out-dir; may cache model assets under ~/.cache/huggingface/ for local inference; and may contact https://huggingface.co, https://github.com, or https://*.hf.space outside --mock mode.
• Run commands from the repository root unless an existing section below says otherwise.

Limitations

• This is a thin wrapper. Image preprocessing, model inference, and decoding are delegated to Hugging Face Transformers and the NV-Reason-CXR-3B model.
• Output is not a diagnosis, clinical report, treatment recommendation, or triage decision. It is engineering evidence and must be reviewed by a qualified professional before any medical use.
• The model may hallucinate findings, miss subtle abnormalities, misread support devices, or produce overconfident prose.
• The committed fixture uses a generated synthetic PNG and deterministic mock response so CI can verify wrapper behavior without downloading model weights. Mock mode is not a substitute for model inference.
• The hf-space-api backend depends on public Hugging Face Space availability and API compatibility.
• Not for clinical deployment, clinical interpretation, autonomous diagnosis, treatment decisions.

Troubleshooting

Runs NVIDIA-Medtech NV-Reason-CXR-3B for chest X-ray image interpretation through either the documented local Hugging Face Transformers inference path or the public Hugging Face Space API. The wrapper does not reimplement the model, image preprocessing, or decoding.

Exact Runnable Surface

For command-shape smoke tests and JSON fixtures, use this repo-root wrapper path exactly:

python skills/nv-reason-cxr/scripts/run_nv_reason_cxr.py PATH_TO_CXR_OR_FIXTURE --mock --out-dir OUT_DIR

For local live image inference, omit --mock only when the user asks for live model inference. Local is the default backend:

python skills/nv-reason-cxr/scripts/run_nv_reason_cxr.py PATH_TO_CXR_OR_FIXTURE \
  --prompt "Find abnormalities and support devices." \
  --backend local

For public API inference without local model packages, use:

python skills/nv-reason-cxr/scripts/run_nv_reason_cxr.py PATH_TO_CXR_OR_FIXTURE \
  --prompt "Find abnormalities and support devices." \
  --backend hf-space-api

Do not invent Medical AI Skills run, eval_engine/run.py, infer.py, or python -m nv_reason_cxr commands for ordinary user runs.

Preconditions

For --backend local, install the inference dependencies in the environment that will run the skill:

pip install torch==2.7.1 torchvision==0.22.1 transformers==4.56.1 Pillow

The model weights and remote model code are loaded from nvidia/NV-Reason-CXR-3B revision 056bd0383b35226554da9dc5866e095df174ae19 through Transformers. They may download to the Hugging Face cache on first use. Set TRANSFORMERS_OFFLINE=1 or pass --local-files-only only after the weights are already cached.

CUDA is expected for practical inference. CPU execution may work for small tests but is slow and must be requested explicitly.

For --backend hf-space-api, no local PyTorch, Transformers, CUDA, model cache, or Hugging Face token is required. The backend sends the image and prompt to the public nvidia/nv-reason-cxr Hugging Face Space.

Check the local environment before downloading weights or running inference:

python skills/nv-reason-cxr/scripts/run_nv_reason_cxr.py --check-setup

The setup report checks importable dependencies, CUDA visibility, Hugging Face cache state, and the recommended next step.

Operational environment variables:

Prompt Routing

Choose both the model prompt and the user-facing output mode before running the wrapper. Routing order matters: exact model-prompt requests use pass-through/raw-only mode first; otherwise report-generation requests take precedence over general analysis and specific-question routing.

Use pass-through/raw-only mode only when the user explicitly asks to send an exact prompt to the model, such as "call the model with this prompt exactly: ...". Pass only that exact model prompt as --prompt.

Use abnormality-analysis mode when the user asks to analyze, examine, or find abnormalities in a chest X-ray. Treat local image paths, uploaded filenames, backend choices such as "use API" or "use local", output delivery instructions, and other agent orchestration text as wrapper instructions, not model prompt content. Do not include local filesystem paths, backend names, or "use API" in --prompt unless the user explicitly asks to send that exact text to the model. For ordinary abnormality-finding requests, use the documented prompt, usually --prompt "Find abnormalities and support devices.", with the requested backend.

Use report-generation/two-call mode if the user asks to write, create, or generate a structured report, chest X-ray report, radiology report, or report. If sufficient raw model context for the same image is already available, especially output from Find abnormalities and support devices., skip the context-gathering call. Otherwise first run the wrapper with --prompt "Examine the chest X-ray." to gather context, but do not show that first call. Then run the wrapper again with a multi-turn transcript prompt:

User: Find abnormalities and support devices.

Assistant:
<raw model context>

User: Write a structured report.

Treat the second call as the completed run.

Use default-prompt/context-answer mode when the user asks a specific question about a finding, such as presence, count, location, or characterization, or mixes general analysis with specific questions. Run the wrapper with --prompt "Find abnormalities and support devices." before answering the original question in plain text prefixed exactly with Answer:. Base the answer only on the raw model output context and the image.