DAX prompt: a Power BI measure from your data model

# Role
You are a Power BI modeler who writes DAX measures that return the right number in every visual, slicer and total row, and who explains filter context in plain words.

# Goal
Write a DAX measure that calculates {{measure_wanted}}, for the report that {{audience}} uses, so that the figure in each visual matches the business definition below and the total row can be trusted as much as the rows above it.

# Success criteria
- It uses only the tables, columns and relationships described below, with their exact names.
- It returns the right result everywhere it will be used: {{visuals}}.
- Ratios use DIVIDE, intermediate values are held in VAR, and the measure reads top to bottom.
- Time intelligence (year to date, prior year) runs on the date table, which must be marked as a date table with one row per day and no gaps.

# Constraints
- Do not invent a column, table or relationship. If the measure needs something the model lacks, such as a date table, stop and say what to add.
- Prefer a measure to a calculated column, and a simple column filter inside CALCULATE to FILTER over a whole table.
- Do not change relationships or their filter direction. If the measure needs an inactive relationship, use USERELATIONSHIP and say so.

# Output
1. The measure in a code block, ready to paste, with a suggested format string.
2. Filter context in plain words: what the measure does in one row of the visual, on the total row, when a slicer is set, and when a filter comes from a related table. Mention context transition if the measure iterates. Keep this under 200 words.
3. Test values: a table with the columns Filter, Expected result, Where the expected figure comes from, DAX query to check. Take expected results from the figures I trust below, and write each check as a DAX query (EVALUATE with SUMMARIZECOLUMNS) to run in DAX query view or DAX Studio. For example, one row might read: "Region = North, Month = March | 12,400 | finance report | EVALUATE SUMMARIZECOLUMNS(...)". That example shows the form only.
4. Edge cases: what it returns for blanks, a zero denominator, a period with no rows, and a filter on more than one item.

# Stop rules
If the business definition can be read two ways, for example which date an order counts on when the fact table has both an order date and a ship date, ask before writing the measure. If no trusted figures are given, say which two or three I should look up to check it.

# Context
Business definition of the measure: {{definition}}

Model: tables, key columns with data types, and relationships with their direction:
"""
{{model_description}}
"""

Existing measures I already use:
"""
{{existing_measures}}
"""

Figures I already trust, for checking: {{known_values}}