Everything starts from an empty r-io: no use case is loaded. The project is created from the Crisis domain and kept free of any reference base, so that every node of the situation is built during this walkthrough.

1. In r-ioga, open the Projects workspace, tab Create project, and click the Crisis card.
2. Replace the default name My project with Fire at Albi Cathedral. Choose the name now: it identifies the knowledge space and cannot be changed afterwards.
3. In the Reference projects card, click Clear all references.
4. In IA & RAG Knowledge documents, keep the default AI provider, Mistral, then click Create project. If its key turns out to be rejected (an assistant fails with an invalid API key), switch the project to OpenAI: Project List, Edit project, IA Provider, Save changes.

ResultProject List shows Fire at Albi Cathedral with the Collaborative and Crisis domains, an empty References column and twelve launchers in Applications. Every application opened from this row works in the context of the project.

The AI Creator Assistant of r-ioda creates instances of the metamodel from a short description. For places it geocodes the address and brings back the outline and a photo.

1. Launch r-ioda from the Applications column and open the Context And Partners tab.
2. Click the robot icon AI Creator Assistant, then Territory.
3. Enter Albi, type City, and click Find territories.
4. Keep the proposed place Albi, France, click Select, then Push.

ResultThe Albi territory is created in the Active space, with its city outline and a photo.

The building is created the same way. A Fixed_Onto relation then says where it stands.

1. Open the AI Creator Assistant again, choose Building, enter Albi Cathedral and click Find buildings.
2. Select Sainte-Cecile Cathedral of Albi and push it.
3. Right-click the cathedral: the Connexions disponibles menu lists the relations it can carry. Click Fixed_Onto.
4. Drag the free end of the new arrow onto Albi.

ResultThe cathedral is fixed onto Albi: its geometry will later place the sensor and the alert on the map.

The two places created by the assistant are real geographic shapes, not just icons on a diagram. r-ioplay, the map of the project, shows them over the satellite view.

1. From Project List, launch r-ioplay: the map opens on the territory of the project.
2. Zoom on the old town, by the Tarn, to find the cathedral.

ResultAlbi is the outline of the commune and the cathedral the footprint of the building, both brought back by the geocoding. The sensor fixed onto the cathedral and the fire alert will be drawn at this very place. The assistant gives each area a random greyish tint that gets stronger as the area gets smaller: the city very pale and very transparent, a site a little stronger, a building stronger still, so that the building stays readable over the city.

A role groups the functions an actor can perform. The assistant asks the configured LLM which functions a firefighter provides: the exact list changes from one run to another.

1. Open the AI Creator Assistant, choose Role, keep the type Person and enter Firefighter, then click Find.
2. Click Push. r-ioda recommends storing roles and functions in the Freeze space, the knowledge the deduction draws on: answer Yes.
3. Check only the Freeze status filter to see them.

ResultThe Firefighter role provides about six functions, among them Extinguish fires, all frozen.

Before fighting a fire, a firefighter must confirm it. This function is added by hand, with an output variable, and Extinguish fires needs it: the fire is only fought once it is confirmed.

1. Right-click Firefighter and choose Provides → Function. Double-click the new function and name it Confirm Fire.
2. In DATA, Output variables declaration, declare a variable Fire Confirmed of type Radio with the possible values Yes and No, then Save.
3. Right-click Extinguish fires, choose Needs and drop the arrow on Confirm Fire.
4. Double-click the Needs edge. In Condition Evaluation, type @: the editor proposes @Fire Confirmed[Confirm Fire]. Complete it into @Fire Confirmed[Confirm Fire] == "Yes" and Save.

ResultExtinguish fires Needs Confirm Fire: whenever the deduction retains the extinction, it adds the confirmation before it. The condition of the edge, a single boolean expression, @Fire Confirmed[Confirm Fire] == "Yes", guards Extinguish fires: the variable is read on the task deduced from Confirm Fire when the process runs. In the diagram a CE badge (Condition Evaluation) marks the edge, and hovering it shows the code.

Actors are persons holding a role. The assistant creates one with a face and positions it in the chosen territory; the person is then renamed with a fixed name so that the login is known in advance, and linked to an application user who will do the person’s tasks in r-io4U, the mobile application.

1. Check the Active filter again (keep Freeze) and open the AI Creator Assistant, Person.
2. Set Number of persons in group to 1, Roles to Firefighter, Located in to Albi, click Find, then Push.
3. Open the person and rename it Owen Hartley when the generated person is a man (firstName Owen, lastName Hartley), Olivia Hartley when it is a woman: the login is the same. In REQUIRED INFORMATION, ACTIONS, click Generate corresponding user.
4. The Register … as user dialog proposes the login ohartley (first letter of the first name followed by the last name) and the default password shown at the top of the dialog: click Save, then Save the person.

ResultOne active person holds the Firefighter role and owns an r-io4U account: the tasks deduced for the firefighters will be assigned to him and appear on his phone.

A crisis is anticipated before it happens: a frozen alert describes the fire that could break out, and a frozen objective describes how it is treated. Both stay frozen: each time a sensor detects the fire, an active copy of them is created, and the frozen pair remains for the next time.

1. Open the Objectives tab and check only the Freeze filter.
2. In the header Palette menu, click Fire_Alert. Double-click the new node, name it Fire in Albi Cathedral and Save.
3. Double-click it again and, in ACTIONS, click Generate associated objective.
4. Double-click the objective Treats Fire in Albi Cathedral. In Broken relations (target [me] <- source), table Satisfies, click Add, choose the source instance Extinguish fires and save each dialog.

ResultTwo frozen nodes: the alert Fire in Albi Cathedral and its objective Treats Fire in Albi Cathedral, which Extinguish fires satisfies: the objective is treated by the action that puts the fire out, and the confirmation comes with it through Needs. The satisfaction is a broken edge because the function lives in another view.

The last piece is the sensor that watches the frozen alert. A dedicated sensor sends a boolean; its rule raises an active alert near the frozen alert it references when the value becomes true.

1. Back in Context And Partners, with the Active filter checked, open the AI Creator Assistant, Sensor.
2. Choose the type Dedicated (nature Smoke), the event type Fire In Albi Cathedral, the mode Fixed Onto Asset and the asset Sainte-Cecile Cathedral of Albi, then Find and Push.
3. Double-click the sensor and rename it Smoke sensor of Albi Cathedral.

ResultThe sensor carries a MAX threshold that references Fire in Albi Cathedral (FREEZE, Fire_Alert). Saving it deploys its CEP rule.

The situation is described, but nothing has happened.

1. Launch r-iota from Project List.

Resultr-iota shows No alert detected and a green status. The frozen alert is knowledge, not an incident.

r-iosemit replays a generated dataset: no smoke on the first three steps, smoke from the fourth one.

1. Launch r-iosemit and click Start sensor on the card of Smoke sensor of Albi Cathedral.
2. Click Play selected dataset step by step (the swipe_right_alt icon) four times.

ResultAt the fourth step the value switches from false to true. The CEP rule of the sensor fires. The frozen alert impacts no good under the sensor: it is an anticipated crisis, not the incident itself. The rule therefore creates a new active alert, Fire in Albi Cathedral at Sainte-Cecile Cathedral of Albi, which impacts the cathedral and is Near the frozen alert; its objective is an active copy of the frozen one, Near it too, through which the deduction finds Extinguish fires. The frozen alert and objective stay untouched. (A frozen alert that impacts a good the sensor is fixed onto would be activated in place instead.)

The fire is a sure alert. The deduction looks for functions that satisfy its objective and for actors able to perform them, then deploys and supervises the process.

1. In r-iota, check the sure alert and click FIND SOLUTION(S).
2. Keep the preselected strategy, leave Deduce and run on YES and click Find.
3. Wait for Deduction, deployment and supervision completed, then open the second tab of the dashboard, Execution graph.
4. Hover the CE badges of the two flows leaving Confirm Fire to read the condition each one carries.

ResultThe status turns blue and the pulsar blue: the fire is handled. The deduction started from Extinguish fires, which satisfies the objective, and placed its prerequisite Confirm Fire before it: the process chains the two, both assigned to the firefighters. It copied the condition of the Needs edge onto the Sequence_Flow Confirm Fire → Extinguish fires, and added a second flow, otherwise, from Confirm Fire to the end, carrying its negation: exactly one of the two is followed. The schedule shows Late because the generated functions have no duration.

The tasks of the process are human tasks assigned to the firefighter: he receives them in r-io4U, the mobile application, and completes them there, typing the output the process expects.

1. Open r-io4U and log in as ohartley (the web build can do it on its own with ?autologin=ohartley and a local web/autologin_ohartley.json, see the r-io4U README). The map shows Owen Hartley; menu → Tasks opens his to-do list, empty before the deduction. Leave it open.
2. Once the solution is supervised in r-iota, the tasks appear in the to-do list on their own: Confirm Fire is started (orange), Extinguish fires expected (grey).
3. Tick Confirm Fire: a form titled with the task name asks for its output, Fire Confirmed, and recalls what to do. Choose Yes and touch Done.
4. Tick Extinguish fires once it has started (it has no output): r-io4U announces that the process is over. Then look at r-iota.

ResultThe output typed in r-io4U is sent to the engine with the completed task: Yes makes the condition true and Extinguish fires starts as soon as Confirm Fire is done. When both are done, r-iota shows No alert detected with a green status: the process is over. The flames stay on the cathedral, though: the process only says that the tasks were done, not that the fire is out. With No, the otherwise flow would have ended the process without sending anyone to fight the fire. The same tasks are also visible, for the administrator, in the r-ioga to-do list.

The fire zone is the active objective and its alert, drawn with the FIRE effect of Fire_Alert. During the process the objective is cloned as EXPECTED; the clone disappears at the end, the active objective stays. Only the expert, once the field has confirmed that the fire is out, closes the crisis.

1. In the r-iota toolbar, click Delete positive deviations (the plus icon next to Refresh).
2. The Positive Deviation Analysis dialog lists the objective Treats Fire in Albi Cathedral and the actuality Fire in Albi Cathedral, both marked YES. Keep them and click Delete.
3. Click Refresh.

ResultThe active objective and alert are deleted: the flames disappear and the cathedral is drawn as a building again. The frozen alert and objective remain in the Freeze space. The supervision of the solution ends with this validation: the process is over and no active objective is left, so a new detection will be a new alert.

The anticipated crisis is still frozen, so the sensor can raise the fire again. This time the firefighter finds no fire: the process takes the otherwise flow.

1. In r-iosemit, click Stop selected dataset, then play the dataset step by step again: at the fourth step the smoke comes back.
2. In r-iota, a new sure alert shows the cathedral on fire. Click FIND SOLUTION(S) and Find as before: the same process is deduced.
3. This time answer as the administrator, from the r-ioga to-do list, which lists the tasks of every actor: tick the new Confirm Fire, choose No for Fire Confirmed, Next, then Save. (The same task also appears, live, in Owen Hartley’s r-io4U.)
4. In r-iota, validate the end of this false alarm with Delete positive deviations, Delete, then Refresh.

ResultNo makes the condition false: the engine follows the otherwise flow to the end and Extinguish fires never runs. Because that task satisfies the objective, its expected twin would otherwise never be released: the otherwise flow lists the objectives of the tasks it skips (set at deduction), and the engine releases them when it follows the flow. r-iota then shows the same end state as after a completed process, and the expert closes the incident.

The project now holds a complete, reusable crisis model. It can leave this server as a single archive.

1. In r-ioga, Project List, click Export project (.zip) in the Actions column.
2. Choose whether to include the time series data and confirm with Yes: the browser downloads Fire at Albi Cathedral.zip.
3. On another server, open the Projects workspace, tab Import ZIP, and select the archive.

ResultThe archive holds the models (context, objectives, collaborative process, in the active and frozen spaces), the CEP rule of the sensor, the game scenario and its dataset, and the photos of the nodes. The import restores the project as it was when exported: export it before playing the sensor to hand over a crisis that is still to come, or at the end to hand over the resolved situation.