Execute Question
What this screen is for
This is where one question from the Intelligence Hub is run: you supply its inputs, the console performs the reads the question declares, and the answer is shown. Every question is a read, so nothing on this page changes your tenant; the invariants that guarantee it, and the rules that decide which questions you may run, are on the Hub page:
A deep link, a bookmark, a scenario card, a Recent entry and Execute Again all arrive here, and the page decides the same way for each of them.
What must exist first
The question must be offered to you here and now. The page resolves it and renders exactly one of three surfaces:
- Offered: the question's summary, then Input Parameters.
- Blocked: the summary, then the reason in place of the form. Nothing is requested from the server. A question that only needs a tenant shows the Requires entering a tenant section instead: it enters the tenant chosen in the scope, or, with none chosen, tells you to choose one in the hub's scope selector. A platform administrator sees the scope the answer is asked in above the form.
- Not applicable, or an identifier that does not exist: This question is not available, with a button back to the Hub. The two cases look identical on purpose, so that a link cannot be used to learn which questions exist.
A question about one tenant runs inside a tenant. On the platform it runs only when it has a platform equivalent for the Scope chosen on the Hub, which the address carries to this page; this page has no selector of its own. A platform-wide question runs from the platform, under Whole platform.
Decisions you make here
Supplying the inputs
The summary states how many reads the question makes, how many inputs it takes, its output type and who it is written for. Input Parameters then asks for each input: text, a number, a date or date range, one option from a list, or several values separated by commas. An identifier such as a key ID, a user or an exchange context is offered from the tenant's own list as you type; you can still type a value that is not in the list. A question with no inputs offers Execute Question directly.
Reading the answer
The answer is computed from the reads performed when you pressed Execute Question, so it reflects the tenant or the platform at that moment. The result card shows when it was executed, how long the reads took, how many records the answer holds and the inputs you gave.
Below it, depending on the question: a table or a chart, each beside Raw JSON, a timeline, key metrics, or several sections. When a table holds one page of a larger read, it says which records it shows and offers Previous page and Next page; a table chosen from that page, such as the ten slowest operations, says instead how many of the matching records it is based on.
Export CSV downloads the answer's table, chart points, metrics or timeline entries; for an answer of several sections, the section whose tab is open. Export PDF downloads a report of the whole answer, every section included, with the question, the tenant or the whole platform it covers, when it ran (in UTC) and the inputs it was given, in the console's language.
When every read the question makes is refused or fails, the error is shown under the form and no result is produced: a refused read says you lack the permission, a missing key or resource says so and names it. When a question makes several reads and only some fail, the answer is built from the ones that succeeded.
Running it again
Execute Again clears the result and shows the form. The console keeps each answer for five minutes, keyed by tenant, or on the platform by the chosen scope, and by question and inputs: the same question with the same inputs inside that window shows the kept copy, with its original execution time, and performs no read. Change an input, or wait, for a fresh read.
What is recorded
Each execution that performs a read adds one entry to your history in this browser: the question, the time, the record count and the duration, and neither your inputs nor the answer. A cached answer adds none. The audit service separately records every audit read a question makes, as it does for any reader of the trail.
How it relates to other screens
Back to Hub returns to the catalogue, or Back to the scenario you came from:
Past executions, and Execute Again for each:
The screens whose data the questions read in full: