Tarbes Crisis is a project skeleton: its context model is empty, but its project carries a RAG document that describes the town hall of Tarbes. Everything starts from r-ioga.

1. In r-ioga, Projects workspace, click the card Tarbes Crisis.
2. In Use case loading options, click Load without changing anything.
3. Close Use-case successfully loaded !!! with OK: r-ioga opens Project List.

ResultThe project row shows twelve launchers. While loading, the RAG document of the use case is ingested: the server log reports Finished storing 5 text segments into the embedding store.

The context will be extracted by an LLM from the documents of the project. Both are set in the project itself.

1. In Project List, click Edit project (pencil icon in the Actions column).
2. Scroll down to IA & RAG Knowledge documents.
3. Leave with Cancel: nothing needs to change.

ResultIA Provider Openai, IA Model Gpt-5.5; the knowledge base holds Tarbes_Crisis_Case_Study_Context_Form.pdf (506.9 KB). The note under it links the blank form, Template_Case_Study_Context_Form.pdf.

Anyone can describe a crisis environment without knowing r-io: fill in the blank three-page form, then attach it to the project as a RAG document. Tarbes Crisis ships with a filled copy.

1. Download the blank form from the link of Edit project: city, monitored sites, monitored goods, sensors, specific roles, additional context, crisis management contacts.
2. Read the filled form of Tarbes: the city, its town hall site and main building with their GeoJSON outlines, the smoke sensor SS_MB_1 fixed onto the main building, and two contacts with their role.

ResultThe form lists what the agents must discover: 1 city, 1 site, 1 building (plus a parking lot), 1 sensor, 2 people and their 2 roles.

r-ioda notices that the project has an active RAG and offers to search it for the context and the partners.

1. Launch r-ioda from Project List and open the Context And Partners tab: the model is empty.
2. Click the AI Creator Assistant button of the editor, then the link here of the sentence Your RAG is activated.
3. In Find Context and Partners, keep the concepts checked by default (Sensor, Territory, Person, Role, Building) and click Find.

ResultThe LLM Reasoning window opens as soon as the search starts.

The search is a graph of agents. Each one extracts a kind of instance from the document; the last one checks the whole model against the RAG.

1. Watch the LLM Reasoning window: each agent turns from PENDING to DONE with its duration.
2. When the badge reads Completed, read the final summary on the right, then Close.

ResultEleven agents, all DONE, in about 30 seconds. Role finds the roles; RolePurposeDiscovery runs only when a role has no function yet and invents its functions; Territory, Person, Building and Sensor extract the instances; ModelController answers COMPLETE when the model matches the document.

The preview shows the model that the agents built. Pushing it writes it into the context of the project.

1. Read the preview graph of Find Context and Partners (zoom out with its editor toolbar).
2. Keep the status ACTIVE and click Push.
3. To the recommendation to move roles and functions to the Freeze space, answer Yes.
4. Look at the Active space, then at the Freeze space with the status filters.

ResultThe Active space shows 6 nodes: TARBES, the site Town Hall, the building Main Building fixed onto it, the sensor SS_MB_1 fixed onto the building, Alice Johnson (Role: Security agent) and Emma Thomson (Role: Town hall secretary), both acting on Town Hall. The Freeze space holds Town hall secretary and its five functions; Security agent already exists in the reference project Reference Crisis Commons Roles.

The sensor found in the document is ready to play: its CEP rule was generated with it, and r-iosemit can plan the actuality it raises.

1. Launch r-iosemit from Project List: there is no game scenario yet.
2. Click Create a scenario, name it Smoke in the Town Hall and click Save.
3. In TIMELINE, click Add an actuality or a message…, set the date to 13:00, select Fire detected at Main Building and click Save, then save the scenario.
4. Open the DATA tab.

ResultThe only sensor actuality offered is Fire detected at Main Building (📡 SS_MB_1, rule R1 · false → true). The timeline holds 1 date and 2 DATA steps; DATA shows SS_MB_1, DEDICATED - SMOKE, Connected, with the dataset smoke_in_the_town_hall_ss_mb_1.

Playing the date sends the smoke value: the CEP rule of the sensor raises the actuality on the building.

1. In r-iosemit, click Next step.
2. Launch r-iota from Project List, then click OPEN DETAILS and Cancel.

ResultThe sensor starts and receives its two values. r-iota shows 1 SURE ALERTS, a red status and the fire on Main Building. The details list Fire detected at Main Building (Fire_Alert) with the objective Treats Fire detected: both are cloned from the reference project Reference Crisis Commons Problems.