Gradient Biotech

Runs and reproducibility

Every analysis step is a versioned pipeline run suitable for methods sections and audit trails.

Finding runs

LocationWhat you see
Study homeDataset list with ingestion status
Dataset workspacePipeline status and latest results per dataset
Cohort analysisBatch, NBS, and prediction run status
Runs pageFull run history table for the study

Runs page

Open /areas/neurology/studies/{id}/runs for a table of every pipeline run in the study — pipeline type, status, and timestamps — with a delete action per run. /areas/neurology/studies/{id}/report redirects to the shared cross-area report composer, where neurology plots can be added alongside other areas' results into one document.

There is currently no dedicated UI for auto-generated methods text, per-run CSV/manifest export, study annotations, or one-click re-run from a completed record, even though some of that provenance (parameters, pipeline version, artifact paths) is recorded on every run and returned by the job-status API. Use the Interpret panels for a narrative summary of a run, and cite run_id and pipeline_version directly from the run record in your own methods text.

Provenance chains

Common chains:

graph_metrics → connectivity matrix dataset
eeg_features → eeg_preprocess run → EEG dataset
functional_connectivity → fMRI NIfTI dataset
cohort_nbs → batch graph_metrics runs → group-labeled datasets
microstates → eeg_preprocess run → EEG dataset

Artifacts

Pipeline outputs live under data/neuro/artifacts/{run_id}/. Access via:

  • Metadata API artifact endpoints
  • Direct filesystem for local development

Failed runs

Failed jobs retain error messages in the job poll response. Common causes:

  • Invalid or non-square connectivity matrix
  • Missing upstream preprocess run for EEG features
  • Unsupported file format or corrupt upload
  • Missing Ollama configuration for interpret jobs

Fix the underlying issue and re-run the pipeline from the dataset workspace or cohort page.

Versioning

Every run records pipeline_version (currently 0.1.0 for neurology pipelines). Include run IDs and versions in methods sections for reproducibility.