📖Summary

The Cost Center service is built following the [agent-centric modular architecture](/dev/carolopedia/wiki/architecture) of Carolverse. It leverages agile principles to build or modify software using distinct agent identities, each carrying out a specific activity — here, the financial roll-up that turns every service's sales, costs and profit into one accountable view of the whole economy. It is an early **work-in-progress**: the intent is set and the owner is in place, but most of the ledger is still being built.

📖Functional considerations

As the CFO ledger for the whole economy, this service must eventually guarantee a few things end-to-end:

- **One accountable financial view.** Every service's profit and loss rolls up into a single consolidated picture, not scattered per-service numbers.
- **Faithful aggregation.** Sales, costs and profit are summed from the source records, so the consolidated total always reconciles with the parts.
- **Single source of truth.** The roll-up reads from the live stores, never from hand-copied figures.

Most of this is intended, not yet delivered — the service is `wip`.

📖Solution architecture

The intended pattern is a direct instance of Carolverse's [agent-centric modular architecture](/dev/carolopedia/wiki/architecture): a financial roll-up owned by [Midas](/dev/carolopedia/wiki/agent/agt-021) that reads each service's P&L and consolidates it into one ledger. The service has no blocks defined yet, so the pipeline of steps and the control plane that wires them are still to be designed. See the service's team above for the accountable owner.

📖Technologies

Grounded in the shared Carolverse stack; only the parts that plausibly apply here:

- **Python 3** on **FastAPI** / **Flask** behind **nginx**, the standard for Carol-project services.
- **SQLite (WAL)** for the ledger datastore.
- The **registry** and **design store** as the binding sources of truth the roll-up reads from.
- **cron** / **systemd** to schedule the periodic roll-up, once it exists.

No service-specific datastores, schedulers or AI steps are wired yet.

📖Design principles

- **Single source of truth.** Figures come from the live stores, never hand-copied — the shared principle on the [Carolverse Architecture](/dev/carolopedia/wiki/architecture) page.
- **Agent-centric modular architecture.** Each step, once built, has an accountable agent and a doing droid.
- **Reconciliation by construction.** The consolidated total must always equal the sum of its parts.

📖Success criteria

- Every service's sales, costs and profit roll up into **one consolidated ledger**.
- The consolidated numbers **reconcile** with the underlying per-service records.
- The financial view is **owned and accountable** — traceable to its source data.

These are the target outcomes; the service does not meet them yet.

📖Policies

- **Owner equals requester**, and the requester is an agent id, never a human — the financial view is owned by [Midas](/dev/carolopedia/wiki/agent/agt-021).
- **Figures are derived, never hand-entered** — the ledger reads from the source records.
- **Lifecycle operations route through the app and its droids** once they exist — the datastore is not hand-edited.

📖What it delivers today

The service is `wip` and delivers little today beyond its defined intent and ownership:

- An owned, registered cost-center identity (owner [Midas](/dev/carolopedia/wiki/agent/agt-021)) and a stated purpose: consolidate every service's P&L into one view.

No roll-up job, ledger UI or reconciliation report is live yet.

📖What it will deliver

Most of this service's capability is still ahead:

- A periodic **roll-up** that pulls each service's sales, costs and profit into one ledger.
- A **consolidated financial view** of the whole economy, owned by Midas.
- **Reconciliation checks** that confirm the total matches the per-service parts.
- An **operator/HTTP surface** to read the ledger and per-service breakdown.