Scenario gallery

A scenario is a JSON file: a list of actions, in order. It is Diwall’s unit of work — you will write some, reuse some, and put them under version control.

These four ship with the tool. They work on a fresh installation, with no prior configuration, except the last one — which needs an encrypted directory.

1 · Read without capturing

The lightest one. It waits for the page, then reads its title — no screenshot, no image analysis.

{"url": "https://example.com",
 "actions": [
   {"type": "attendre_selecteur_present", "selecteur": "body"},
   {"type": "evaluer", "script": "document.title"}]}

diwall-rpa --scenario /opt/diwall/scenarios/exemples/sondage_fast.json --guide-version 1.3

{"succes": true, "http_status": 200,
 "evaluations": [{"script": "document.title", "valeur": "Example Domain"}],
 "respect": {"actions_executees": 2, "duree_totale_ms": 1174}}

Two actions, just over a second. This is the shape to reach for when you only need to know what a page says.

2 · See, then click by number

Same page, but with --som: every interactive element is numbered, and the click names a number rather than a selector.

{"actions": [
   {"type": "attendre_selecteur_present", "selecteur": "body"},
   {"type": "cliquer_som", "id": 1}]}

diwall-rpa --scenario /opt/diwall/scenarios/exemples/navigation_som.json --som --guide-version 1.3

{"succes": true, "http_status": 200,
 "respect": {"actions_executees": 2, "duree_totale_ms": 1160}}

How the numbering works →

3 · Diagnose a page that resists

Same skeleton, different purpose: point it at a local interface that misbehaves and read erreurs_js and erreurs_console in the answer. A blank page with a silent HTTP 500 behind it looks exactly like an expired session — the console is what tells them apart.

diwall-rpa --scenario /opt/diwall/scenarios/exemples/depannage_local.json --guide-version 1.3

{"succes": true, "http_status": 200, "erreurs_js": [], "erreurs_console": [],
 "evaluations": [{"script": "document.title", "valeur": "Example Domain"}]}

Empty lists here, because example.com has nothing to report. Substitute your own URL and they fill up.

4 · Authenticate from the encrypted directory

The most demonstrative, and the one that closes the page rather than opening it: it needs a mounted, configured encrypted directory.

{"auth_indicator": ".user-menu",
 "actions": [
   {"type": "evaluer", "script": "document.title", "contient": "Example"},
   {"type": "remplir", "selecteur": "input[name=\"username\"]",
    "valeur": "depuis_secrets", "secret_cle": "username"},
   {"type": "remplir", "selecteur": "input[name=\"password\"]",
    "valeur": "depuis_secrets", "secret_cle": "password"},
   {"type": "cliquer", "selecteur": "button[type=\"submit\"]"},
   {"type": "attendre_selecteur_present", "selecteur": ".user-menu"}]}

No output is shown here, and that is deliberate: without that directory this scenario does not run, and printing the error it raises would prove nothing about the feature. Set one up first, then run it against a target of your own.

Two details carry the whole security model. "valeur": "depuis_secrets" means the password is resolved inside the browser process, never in your shell. And the first action is an assertion, not a step: if the page is not the one expected, the scenario stops before typing anything anywhere.

The gallery grows from real use

Four scenarios is a deliberately modest starting point. journal.py --exporter-skill turns a successful real run into a reusable scenario — the gallery grows at the pace of actual usage, never with examples invented to fill the page.