# SFCR notebook generation prompt

Generate a single valid SFCR notebook for the sfcr browser app.

Default to compact YAML. Use expanded runtime JSON only if the user or calling client explicitly asks for JSON.

Before generating the notebook:

1. Read `../notebook-guide.md` for the notebook structure, cell types, naming conventions, and matrix rules.
2. Read `../sfcr-notebook.schema.json` for the machine-readable expanded JSON runtime constraints.
3. For YAML output, start from `../notebook-examples/starter.example.notebook.yaml`.
4. For JSON output, start from `../notebook-examples/starter.example.notebook.json`.
5. Use `../notebook-examples/bmw.example.notebook.yaml` or `.json` as the default reference for sectors and accounting bands.
6. Use `../notebook-examples/gl6-dis-rentier-v2.example.notebook.yaml` or `.json` when the model splits households or needs distributional accounting.

YAML requirements:

- Return only one YAML document.
- Set `format` to `sfcr-notebook-yaml` and `formatVersion` to `1`.
- Set `metadata.version` to `1`.
- Include unique `id`, `title`, and an ordered `cells:` list.
- Use wrapped compact cell entries, where each item has exactly one cell-type key such as `markdown`, `matrix`, `equations`, `solver`, `externals`, `initial-values`, `run`, `chart`, `table`, or `sequence`.
- Do not use older top-level shorthand sections such as `introCell`, `variables`, `equations`, `parameters`, `baselineRun`, `charts`, `tables`, or `cellOrder`.
- Keep cell ids stable and descriptive in kebab-case.
- Use compact array rows for matrices, equations, externals, and initial values when possible.
- Always quote equation and external descriptions inside compact row arrays.
- If you create matrix cells, include both `columns` and `sectors`, and make each row match the column count exactly.
- Include at least one baseline run for executable models.
- Prefer concise markdown explanations before scenarios.
- If the model has stocks with lagged terms (`varName'`, `lag(...)`, or `[-1]`), provide matching initial values unless the default initialization is clearly acceptable.

JSON requirements:

- Return only one JSON object.
- Set `metadata.version` to `1`.
- Include unique `id`, `title`, and `cells` fields.
- Keep cell ids stable and descriptive in kebab-case.
- Keep model ids consistent across `equations`, `solver`, `externals`, `initial-values`, and `run` cells.
- If you create matrix cells, include both `columns` and `sectors`, and make each row `values` array match the column count exactly.
- Use `band` on matrix rows to group accounting lines.
- Include at least one `run` cell for executable models.

Recommended notebook order:

1. Overview / intro markdown cell
2. Balance sheet matrix if applicable
3. Transactions-flow matrix if applicable
4. Sequence cells derived from matrices or dependencies if useful
5. Equations
6. Solver
7. Externals
8. Initial values
9. Baseline run
10. Chart or table
11. Scenario markdown + scenario run + scenario chart/table

Compact YAML row reminders:

- Matrix row: `[band, label, ...values]`
- Equation row: `[name, expression, "description", unit, type, role]`
- External row: `[name, value, "description", unit, type]`
- Initial value row: `[name, value]`
- Use object rows for non-constant external series that need `kind: series` and `valueText`.

Output rules:

- Return raw YAML by default, or raw JSON only when JSON is explicitly requested.
- Do not wrap the output in markdown fences.
- Do not add commentary before or after the notebook.
- Do not use YAML anchors, aliases, merge keys, or duplicate keys.
- Prefer patterns already used in the starter, BMW, and GL6 DIS rentier examples.