Skip to content

Structural demo
For Anyone evaluating TumorQuantAI or learning its output layout
Hands-on steps Clone, run one command, inspect a local report
Prerequisites Linux and Python 3; no GPU, Docker, model weights, account, or internet after cloning
Storage A small synthetic result directory; no WSI download or model cache
Writes to tumorquantai-demo/ by default

Not a biological prediction

The demo is a structural software fixture. Its samples, counts, classes, failures, and images have no biological or validation meaning.

Run it

git clone https://github.com/cfarkas/tumorquantai.git
cd tumorquantai
tumorquantai demo

Expected final output includes an absolute path ending in:

TumorQuantAI structural demo complete.
No HistoPLUS inference ran; values have no biological meaning.
Open first: /your/checkout/tumorquantai-demo/START_HERE.html

The demo does not contact HistoPLUS, Hugging Face, or Zenodo. It uses bundled fixtures and a stub worker to exercise the same structural stages used for slide discovery, sample scattering, failure auditing, aggregation, status, and report.

Inspect the result

Open tumorquantai-demo/START_HERE.html. Then compare:

  • aggregated_celltypes/sample_aggregation_audit.csv: successful and failed fixture samples remain explicit;
  • aggregated_celltypes/celltype_counts_by_sample.csv: only completed fixture samples have numeric columns; and
  • tumorquantai_report.json: machine-readable structural summary.

Run the readers directly:

tumorquantai status tumorquantai-demo
tumorquantai status tumorquantai-demo --json
tumorquantai report tumorquantai-demo

The status output must distinguish a fixture zero from a fixture failure. Do not interpret either as biology.

Stop, repeat, and clean up

Press Ctrl+C if the demo is interrupted, then repeat tumorquantai demo. It is safe to rerun. To remove only this demo, first confirm the exact directory and then delete it:

test "$PWD/tumorquantai-demo" != "/" &&
  rm -r -- "$PWD/tumorquantai-demo"

Do not run a recursive deletion against a variable you have not printed and checked.

Next: prepare a public one-slide quickstart, or inspect your own slides without inference.