From an interactive activity to an ASTRA/MySTRA report
Prepare an interactive activity with your harness, experiment with it in the browser, then let Studium turn what you chose to keep into a local report you can check before sharing anything. The viewer contains no agent. The five templates ship with synthetic examples; a technical example is not a validated teaching experience.
Prepare the activity
In a Studium session already started, ask the harness:
Use the studium-activity-templates skill. Get the public context of my current scene, choose a template suited to its question and prepare it in a new folder. Show its sources and limits. Wait for my prediction.
The harness calls get_operator_context and saves the JSON reply as context.json. It can then run,
from the repository and its Python environment:
.venv/bin/python tools/experience_templates.py list
.venv/bin/python tools/experience_templates.py prepare \
--template parameter-comparison --context context.json --out activity
activity/task.md guides the adaptation and activity/index.html is the viewer to open. The
templates linked-representations, evidence-claim, diagnose-analysis and
conclusion-robustness use the same command. The default examples are not derived from your scene's
sources: a shared identifier does not prove a scientific link. If the needed inputs are missing, the
harness must say so and present the sample as a separate illustration.
| Option | Effect |
|---|---|
--language fr |
Start the viewer in French (default en) |
--translations FILE |
Local translation dictionaries; untranslated content stays in its source language, flagged |
--supports FILE |
A local manifest of supporting media; see Language and supports |
--config FILE |
Your own inputs, following the template contract |
No translation API is called. Preparing does not link new data to a publication: keep the sources, reference values and controls actually performed. A new objective or scientific conclusion is for the teacher to review.
Experiment and choose what to keep
In the viewer, choose whether a local trace should be prepared for a possible export. Write your prediction before showing the results, vary the control or choose a representation, then write your observation and conclusion. Optional fields, confidence included, may stay empty. The path without a trace also completes the activity, without any download.
Download the trace only if you want to. A partial export is possible and does not claim the activity
is finished. Save the chosen file as actions.json. The download holds what you submitted in the
viewer; the report then applies its own choice about including your texts.
If the harness also uses the private experiment journal, the prediction must be recorded there before the results are revealed. A later download must not be re-entered as a prediction the engine would have recorded earlier.
Produce a copy to review
Ask the harness:
Use studium-learner-report to prepare a local report from this actions.json file and the definition of my activity. Exclude my personal texts and show me the result before any sharing.
.venv/bin/python tools/learner_trace.py validate actions.json
.venv/bin/python tools/learner_report.py export report \
--actions actions.json --activity activity/activity.json
.venv/bin/python tools/learner_report.py verify report
Open report/index.html. The report also holds ASTRA, the ARA trace, the MySTRA source and the
structured files that can be verified. Actions keep their declared origins and sequence; the
activity definition makes choice identifiers and parameters readable.
To include your original words on purpose, generate a new copy with --include-original-text. That
choice applies to this report only; it does not follow from consenting to journal storage. A journal
can also be selected with --journal; the browser and engine timelines stay separate, and matching
scene identifiers do not prove two files come from the same person. None of the commands above sends
the report to a teacher or an external service.
What the teacher can examine
The question, recorded actions, parameters, evidence and any texts the student chose to include. A missing help trace means "unknown". ASTRA checks and file fingerprints verify structure and technical consistency, not learning nor authorship. The learner report guide details the limits and how to open the report in the shared MySTRA viewer.
Two more activities to ask for
To diagnose an analysis:
Use
diagnose-analysisfor the public steps and results of this scene. Prepare several plausible repairs, with their explicitly provided results and how to check them. Let me pick the suspect step and explain my diagnosis before discussing the repair. If a result is missing, say so; never present an invented value as an executed computation.
To stress-test a conclusion:
Use
conclusion-robustness. Compare a reference hypothesis with defensible alternatives, their results and an explicit criterion with units. Let me predict the consequences, then examine where the conclusion changes. State what remains unknown outside these scenarios.
A teacher can adapt these formats to another discipline by replacing sources, steps, hypotheses, units and values in the configuration; the viewer code stays the same. A new scientific run is the harness's job; the viewer then shows the captured results and traces local choices.