CLI reference

The commands and options below are generated from the installed Typer app on every documentation build. Use jevotron COMMAND --help for help in your terminal. jt is an alias for jevotron and accepts the same commands and options.

See the file format reference for parser defaults, examples, and the options accepted by --format-option. jevotron formats NAME shows the same options in your terminal.

jevotron

Find suspicious fields in files and databases. Preview, scan, and review warnings.

Usage: jevotron [OPTIONS] COMMAND [ARGS]...

Option	Description
--install-completion	Install completion for the current shell.
--show-completion	Show completion for the current shell, to copy it or customize the installation.
--help	Show this message and exit.

jevotron preview

Inspect exact model requests without API calls or credentials.

Usage: jevotron preview [OPTIONS] {file}

Option	Description
file	File to assess; detect databases by header, other formats by extension. [required]
--config <path>	Local Python configuration file.
--format <str>	Override input format; see 'jevotron formats'.
--format-option <str>	Parser option KEY=VALUE; repeat as needed.
--table <str>	Database table or view (optionally schema-qualified); repeat to select several. Default: all user tables.
--guidance <str>	Inline instructions describing what is correct; appended after --guidance-file when both are given.
--guidance-file <path>	Read guidance from a UTF-8 document; may be combined with --guidance.
--exemplars <path>	JSON list of user-selected examples.
--model <str>	Model; defaults to jev-1.13.0 or the config value. gpt-* models (e.g. gpt-6-luna) use OpenAI Decisions with OPENAI_API_KEY.
--batch-size <str>	1 keeps per-entry requests (default); auto chooses economical batches up to 64 entries; N > 1 forces shared guidance with at most N entries. Token limits may reduce batch size. Shared guidance changes the prompt layout and can change scores. [default: 1]
--batch-tokens <int range>	Estimated budget for shared-guidance requests; state plus largest question is limited to half this budget. Per-entry requests are sent whole until Jev rejects their size. [default: 48000; 1<=x<=64000]
-f, --field <str>	JSON Pointer to assess, required on every entry; repeat to select multiple fields.
--optional-field <str>	JSON Pointer to assess only where present; repeat to select multiple fields.
--relaxed	Treat every --field as optional: assess whichever are present.
--id-column <str>	Use this top-level scalar field as the entry ID.
-l, --limit <int range>	Maximum entries to preview. [default: 3; x>=1]
-o, --output <path>	Write results to this file; defaults to stdout.
--help	Show this message and exit.

jevotron scan

Assess fields with Jev, reusing cached results for unchanged entries.

Usage: jevotron scan [OPTIONS] {file}

Option	Description
file	File to assess; detect databases by header, other formats by extension. [required]
--config <path>	Local Python configuration file.
--format <str>	Override input format; see 'jevotron formats'.
--format-option <str>	Parser option KEY=VALUE; repeat as needed.
--table <str>	Database table or view (optionally schema-qualified); repeat to select several. Default: all user tables.
--guidance <str>	Inline instructions describing what is correct; appended after --guidance-file when both are given.
--guidance-file <path>	Read guidance from a UTF-8 document; may be combined with --guidance.
--exemplars <path>	JSON list of user-selected examples.
--model <str>	Model; defaults to jev-1.13.0 or the config value. gpt-* models (e.g. gpt-6-luna) use OpenAI Decisions with OPENAI_API_KEY.
--batch-size <str>	1 keeps per-entry requests (default); auto chooses economical batches up to 64 entries; N > 1 forces shared guidance with at most N entries. Token limits may reduce batch size. Shared guidance changes the prompt layout and can change scores. [default: 1]
--batch-tokens <int range>	Estimated budget for shared-guidance requests; state plus largest question is limited to half this budget. Per-entry requests are sent whole until Jev rejects their size. [default: 48000; 1<=x<=64000]
-f, --field <str>	JSON Pointer to assess, required on every entry; repeat to select multiple fields.
--optional-field <str>	JSON Pointer to assess only where present; repeat to select multiple fields.
--relaxed	Treat every --field as optional: assess whichever are present.
--id-column <str>	Use this top-level scalar field as the entry ID.
-l, --limit <int range>	Maximum entries to assess; default all. [x>=1]
-o, --output <path>	Write results to this file; defaults to stdout.
--cache <path>	SQLite assessment cache. [default: .jevotron/cache.sqlite3]
--no-cache	Disable cache reads and writes.
--refresh	Reassess and replace matching cache entries.
-t, --threshold <float>	Emit only entries with anomaly score at least this value (0 to 1). Also sets the warning threshold; default 0.5 or config value.
-w, --warnings-only	Emit only entries at or above the threshold.
-a, --anomalies-only	Emit only entries with a field classified as ANOMALY (or the configured anomaly label).
--sort-score	Buffer results and emit highest scores first.
-s, --sort-confidence	Buffer results and emit highest confidence first; with --anomalies-only, use the most confident anomalous field, otherwise the highest-scoring field.
-O, --output-format <jsonl&#124;csv&#124;yaml>	Report format. [default: jsonl]
--rows <entries&#124;fields>	Report entries with nested fields, or individual field assessments. [default: entries]
--where <str>	SQL row expression filtering assessment results, e.g. label = 'NORMAL' AND confidence >= 0.9. Combines with other filters using AND.
--order-by <str>	SQL ordering, e.g. confidence DESC, score DESC. Buffers the report; ties retain input order.
--help	Show this message and exit.

jevotron formats

List supported input formats, defaults, and parser options. No API key needed.

Usage: jevotron formats [OPTIONS] [name]

Option	Description
name	Show options for this format.
--help	Show this message and exit.

jevotron tables

List database tables, views, columns, and primary keys as JSONL. No API key needed.

Usage: jevotron tables [OPTIONS] {file}

Option	Description
file	File to assess; detect databases by header, other formats by extension. [required]
--format <str>	Override input format; see 'jevotron formats'.
--help	Show this message and exit.

Environment

TYPESAFE_API_KEY supplies the API credential for uncached assessments. Preview and complete cache replays do not require it. With a gpt-* model such as gpt-6-luna, OPENAI_API_KEY is used instead; see OpenAI Decisions.

Precedence

Explicit CLI flags override project config values. Parser defaults apply when neither provides a setting. See project configuration. An explicit Config(parser=...) owns parsing; set its options in that Python config. Combining a custom parser with --format or --format-option is an error.