Dataset Inspector
Inspect a local tabular data source and return a Markdown schema report.
Workflow
- Confirm the user provided a local data file path.
- Run
scripts/dataset-inspect --source <path>. - Use
--type parquet|csv|jsononly when extension-based detection is wrong. - Use
--output <report.md>when the report should be saved. - Summarize the row count, column count, file size, and notable schema details.
Command
scripts/dataset-inspect --source data/sample.parquet
scripts/dataset-inspect --source data/sample.csv --output reports/schema.md
Requirements
The helper script is self-contained except for duckdb. If it is unavailable,
install it in the active project environment before running the script.