CSO chart

Turn a question into a correctly filtered CSO table and a simple chart.

Related projects: None in the catalogue.

Calling the tools

Steps name each operation as source/operation (e.g. ppr/ppr_price_stats), using the typed tool name as the operation. On the default surface, run it with ireland_call, for example:

{ "source": "cso", "operation": "cso_search_tables", "args": { "query": "consumer price index", "limit": 10 } }

If an argument shape is unclear, call ireland_describe with { "source", "operation" } first; it returns the schema and an example. search, fetch and nearby are top-level tools. If the server was added with ?toolsets=<ids> or ?toolsets=all, the typed tool (e.g. cso_search_tables) can be called directly with the same args.

Steps

1. Find the table. Call cso/cso_search_tables with { "query": "consumer price index", "limit": 10 }. Pick the table whose title, frequency and date range fit. Tell the user the code you chose.
2. Read the dimensions. Call cso/cso_get_table_metadata with { "table_code": "CPM01" }. Note each dimension code (statistic, time, area, etc.) and the category codes you need.
3. Pull a small slice. Call cso/cso_get_data with table_code and filters mapping dimension codes to arrays of category codes, e.g. { "table_code": "F1001", "filters": { "<dim code>": ["<cat code>"] } }. Always filter: requests over 10,000 cells are rejected. Keep limit modest (≤500).
4. For census population by county, cso/cso_area_profile is a shortcut.
5. Shape the data into rows of { period, series, value, unit }. Keep the unit from the table.
6. Chart. Choose: line for time series, bar for areas/categories. Emit a Vega-Lite spec (see references/vega-lite.md) or, if the client cannot render charts, a Markdown table.

Output format

### <Table title> (<table code>), <series>, <first>–<last period>
<chart spec or table>
Notes: <unit>, <base period if an index>, <provisional flags>.
Sources
- CSO, table <code> "<title>" (CC BY 4.0), last updated <date from metadata>, retrieved <date>: <url>

Rules

• Follow ../_shared/citations.md. Title the chart with the table code and period.
• Never mix tables or units in one series. Never extrapolate or forecast.
• If an index, state the base period (e.g. Dec 2023 = 100).
• If the right table is unclear, show the top 3 candidates and ask.