CLI Reference
Complete reference for every FineType command — flags, defaults, and examples.
FineType ships as a single binary with six commands. Together they form a pipeline from raw text to typed, validated, query-ready tables.
Commands
| Command | Description |
|---|---|
infer | Classify a single value or batch of values into semantic types |
profile | Profile a CSV file — detect the semantic type of every column |
schema | Export the JSON Schema for a type (or set of types) |
load | Generate a runnable DuckDB CREATE TABLE AS statement from a profiled file |
validate | Validate data quality against taxonomy schemas |
taxonomy | Browse and filter the type taxonomy |
The pipeline
The commands map to a natural workflow:
- Explore — use
inferto classify individual values and build intuition about your data. - Discover — use
profileto scan an entire CSV and see what FineType detects in every column. - Capture — use
schemato export the JSON Schema for any detected type, giving you a machine-readable contract. - Gate — use
validateto check data quality against those schemas before loading. - Cast — use
loadto generate a DuckDB SQL statement that casts every column to its detected type.
You don't have to use every step. profile alone gives you a quick answer; load takes you from CSV to typed table in one command. The pipeline is there when you need rigour.
Getting started
If you're new to FineType, start with the Getting Started guide for a hands-on walkthrough covering installation, profiling, and schema export.
Output formats
Every command that produces output supports multiple formats via the -o / --output flag:
| Format | Flag value | Use case |
|---|---|---|
| Plain text | plain | Human-readable terminal output (default) |
| JSON | json | Programmatic consumption, piping to jq |
| CSV | csv | Spreadsheets, further analysis |
| Markdown | markdown | Documentation, reports |
| Arrow | arrow | High-performance IPC, streaming to DuckDB |