Skip to main content

Using the library without code

Everything a program does with the library can be done from the command line: give people aliases, turn an alias back into an id, and redact text. Useful when you work with a model by hand, or from a shell script, without writing a program.

Load your class​

So that redact knows your students' names, the directory has to have met them. Load them from a CSV, such as an export of your LMS's participants:

Name,Email,Institutional ID,service_user_id,Group
Marta Quiroga,marta.q@school.example.org,2024-0031,4821,B
"Ferrándiz, Lucía",lucia.f@school.example.org,,4825,"B, mornings"
Hugo Estrella,,2024-0047,,A
pseudonym remember --csv class.csv --service class-2b > class-with-aliases.csv
  • Columns it reads, by their header, ignoring case, spaces, hyphens and underscores: name (required), email, institutional_id and service_user_id (at least one per row). They're the identifiers the library uses.
  • Other columns (group, grades, anything) are kept as they are.
  • The output is the same CSV with an alias column first. It still holds names: it's for you.
  • The whole file goes in one call, and nothing is remembered if a row is wrong: an error names the line (never its content), so you can fix it and run it again.
  • --csv - reads stdin; without --service it uses the default service.

Pseudonymize an export for a model​

Add --only-aliases and the columns that identify someone are dropped, leaving the aliases and everything else:

pseudonym remember --csv grades.csv --service class-2b --only-aliases > grades-for-the-model.csv
alias,Group,Practice 1,Practice 2
A-9NQWD5,B,8.5,7
A-Q30XC4,"B, mornings",7.5,8
A-TA8D1J,A,6,5.5

That file can go to a model: it says who did what, as aliases, and nobody's name, email or id. Check the other columns yourself: a comment column may still name people (pass it through pseudonym redact if so).

The model's answer about A-9NQWD5 can be read with names using pseudonym reveal.

One person​

pseudonym remember --name "Marta Quiroga" --email marta.q@school.example.org --service-user-id 4821

prints A-9NQWD5. Options: --name (required), --email, --institutional-id, --service-user-id (at least one), --service, --link (one alias across services, see linking), and --json for {"alias": "A-9NQWD5"}.

Remembering someone again gives the same alias, as in the library: a person keeps their first alias even if their email changes.

From an alias back to an id​

A script that gets an alias from a model and needs to act on that person in the LMS:

pseudonym resolve A-9NQWD5 --service class-2b

prints 4821, the service_user_id remembered for that person in that service (linked aliases count). With --json: {"alias": "A-9NQWD5", "service": "class-2b", "id": "4821"}. Exit code 2 if the directory doesn't know the alias, or has no id for it in that service.

In a script​

#!/bin/sh
# Grades to a model as aliases, its answer back with names, for the teacher only.
export PSEUDONYM_SERVICE=class-2b
pseudonym remember --csv grades.csv --only-aliases > for-model.csv
ask-the-model --file for-model.csv "Who needs help, and why?" > answer.txt   # your own tool
pseudonym reveal < answer.txt

remember, redact and reveal print results to stdout and messages to stderr, so they compose with pipes. Run such scripts yourself, never through an agent: reveal and resolve undo the aliases (Keeping agents away from the key).