Reports¶

DatasetLint reports are available as console output, JSON, Markdown, or static HTML.

datasetlint examples/minimal_dataset --format console
datasetlint examples/minimal_dataset --format json
datasetlint examples/minimal_dataset --format markdown
datasetlint examples/minimal_dataset --format html

Use the dedicated report command when you want a report artifact:

datasetlint report examples/minimal_dataset --out datasetlint-report.json
datasetlint report examples/minimal_dataset --out datasetlint-report.md
datasetlint report examples/minimal_dataset --out datasetlint-report.html

Formatted validation output still prints to stdout. Redirect output when you want those representations as files:

datasetlint examples/minimal_dataset --format json > datasetlint-report.json
datasetlint examples/minimal_dataset --format markdown > datasetlint-report.md
datasetlint examples/bad_dataset --format html > report.html
datasetlint report examples/bad_dataset --out report.html

The HTML report is plain static HTML with no frontend build step. It includes the dataset path, pass/fail state, issue counts by severity, checks run, adapter name and mode, dataset fingerprint, config summary, dataset summary, stats when available, and an issue table with severity, check, file, row, message, and metadata. HTML output intentionally omits generated_at so repeated runs over the same inputs produce stable HTML. JSON and Markdown reports still include generated_at.

Committed sample reports:

• examples/reports/minimal_report.json
• examples/reports/bad_report.json
• examples/reports/bad_report.md

Validation Report Fields¶

Python validation returns LintReport:

• dataset_path
• generated_at
• datasetlint_version
• dataset_fingerprint from a lightweight file-manifest hash
• dataset_summary
• checks_run
• adapter
• config
• issues
• findings
• stats
• passed

Static HTML renders the same report data except for generated_at, which is omitted for deterministic output.

Each issue includes:

• check_name
• severity
• message
• file
• row
• metadata

Rows are 1-based CSV rows. The header is row 1 and the first data row is row 2. Native lint issues include metadata.suggestion when DatasetLint has a concrete fix hint for the rule.

The committed JSON schema is schemas/report.schema.json.

Adapter Manifest And Validation Fields¶

Adapter inspection returns DatasetManifest:

• dataset_name
• adapter_name
• dataset_root
• version
• sequences
• frames
• sensors
• annotations
• calibration
• splits
• metadata
• limitations
• provenance

Adapter validation returns AdapterValidationReport:

• adapter_name
• dataset_root
• detected
• valid
• validation_mode
• checked
• not_checked
• limitations
• errors
• warnings
• coverage
• stats

Use datasetlint export-manifest DATASET --adapter NAME --output manifest.json when you need a JSON artifact for adapter output.

Severity Levels¶

Severity	Meaning
error	Dataset is invalid for the current rule
warning	Dataset has suspicious or risky data
info	Informational finding

Validation passed is true when there are no error issues.

Exit Codes¶

Code	Meaning
0	Command completed and did not meet the configured failure threshold
1	Validation failed the --fail-on threshold, adapter validation failed, or diff regressions were found with --fail-on-regression
2	Invalid usage, unknown check group, bad adapter, or invalid config

CI Behavior¶

Default validation fails only on errors:

datasetlint examples/minimal_dataset

Fail on warnings too:

datasetlint examples/minimal_dataset --fail-on warning

Fail on any finding:

datasetlint examples/minimal_dataset --fail-on info