Six questions. One picture each.
Each mental model answers one question about how Sofrito works. They build on each other, so they are best read in order.
-
Responsibility and data access
Who does what, and who sees what?
Two layers build the platform. Two layers use it.
Four layers, the three lines between them, and who answers for each.
-
The flow of data
What happens to data between arrival and report?
Keep a full copy once. Rebuild everything else from it.
Five boxes from source to published report, and what each one holds, changes and keeps.
-
Classification
How is the data governed?
A class is a label. Behind it is a configuration.
Classes that RHC decides, and the five settings Sofrito applies for each one.
-
From question to report
How does a question become a report, and how official is the result?
Requests go down. Reports move up.
Where a request is served, and the four steps a report can stand on.
-
The role of AI
What is AI allowed to do, and who answers for it?
Agents belong to people. LLMs are built in.
Agents and LLMs on both sides of the key line, and what goes into and comes out of each.
-
The system
What is Sofrito made of?
Separate jobs. Shared data.
The deployments around the stored data, and how code reaches each one.