Runs and reproducibility
Every analysis step is a versioned pipeline run suitable for methods sections and audit trails.
Finding runs
| Location | What you see |
|---|---|
| Navbar Jobs | Recent and active jobs across all studies |
| Cardiology dashboard | Recent runs panel |
| Study overview | Last 10 runs for this study |
| Dataset Runs tab | All runs for one recording |
| Subject hub | Cross-dataset runs for one participant |
Run detail page
Open /runs/{id} for:
- Pipeline name, semver version, status, duration
- Outputs — structured metric summaries
- Artifacts — downloadable JSON files
- Inputs — source dataset ID and upstream run IDs (provenance chain)
- Logs — execution messages and errors
Reproduce run
Re-submit the job with identical parameters from the run detail page. Useful for verifying reproducibility after code or dependency updates.
Compare runs
From a dataset Runs tab, select two completed runs and click Compare selected. Opens /runs/compare?left=…&right=… with side-by-side parameters and outputs.
Provenance chains
Common chains:
compute_hrv → preprocess_ecg → dataset
compute_baroreflex → analyze_bp run + compute_hrv run
compute_rsa → preprocess_ecg run + analyze_respiration run
cohort_summary → many compute_hrv runs across subjects
The Inputs tab on each run shows these links explicitly.
Report export
From the study overview, the Report export panel can generate study bundles and export cohort charts as PNG/PDF from the UI.
Server-rendered report bundles with auto-generated methods sections are partially implemented — see product status docs for current scope.
Failed runs
Failed jobs retain error messages on the run detail page and in logs. Common causes:
- Missing upstream preprocess run
- Invalid or too-short RR series
- Missing Ollama configuration for interpret jobs
- File not found (upload path mismatch)
Fix the underlying issue and re-run from the dataset inspector or reproduce from a prior successful run.