CLI
Это содержимое пока не доступно на вашем языке.
The anonymize command works both with and without arguments.
Without arguments — launch the web UI
anonymizeOpens your default browser at http://127.0.0.1:<port>. Same as the web UI flow.
With a file — batch process
anonymize redact path/to/contract.docx -o path/to/contract.anonymized.docxRedacts the input file using P0 defaults and writes to the output path. Exit code 0 on success, non-zero on any I/O or parsing error.
Flags:
| Flag | Purpose |
|---|---|
-o, --output PATH | Output path (default: alongside input with .anonymized.<ext> suffix) |
--categories CAT1,CAT2 | Restrict to specific categories. Default: P0+P1 |
--lang LANG | Force language pack (ru / en). Default: autodetect |
--no-metadata-clear | Don’t strip metadata (debug only) |
--report PATH | Emit a JSON report of detected entities (categories + offsets, no raw values) |
-v, --verbose | Enable info-level logs |
Subcommands
| Subcommand | Purpose |
|---|---|
anonymize redact FILE | Process a single file |
anonymize testkit run --corpus DIR | Run the testkit against a golden corpus (CI mode) |
anonymize feedback export | Export the feedback log as JSON to stdout |
anonymize feedback clear | Wipe the local feedback log |
Examples
Redact all .docx in a folder:
for f in *.docx; do anonymize redact "$f"; doneRedact and emit a report (no raw PII in the report):
anonymize redact contract.docx -o contract.anonymized.docx --report contract.report.jsonThe --report output is safe to share — it contains category counts and offset positions but no entity text.