Limits and Permissions
3 min read
This page covers what Oracle can see, what it won’t answer, and the daily limits that apply.
What It Can See
Section titled “What It Can See”Never more than you can. Oracle follows the same permissions as each screen. If your role can’t open Defects, you can’t ask about them.
That includes conditions. Test cases that have an open defect against them returns test cases, but it has to read defects to find them. A role without defect access can’t ask it, and so can’t learn which test cases have defects. The same goes for readiness columns: they are worked out from defects, so a role that can’t read defects can’t add them as a column or sort by them.
Records you can’t read aren’t hidden from the answer afterwards. Oracle never knows about them, so the question can’t be asked in the first place.
What Oracle Sees
Section titled “What Oracle Sees”It never sees what is written inside a record. No test case title, no defect description, no comment, no step text.
It is shown your project’s vocabulary: what the fields are called and the values they hold, such as your folder names, label names, release titles and team members, together with your question. That is how it turns “in the Checkout folder” into a real condition. See What Oracle Is Shown.
What It Won’t Answer
Section titled “What It Won’t Answer”| Question | Answer |
|---|---|
Which tester is fastest? | Can’t be answered. Nothing records it. |
How many defects will we have next month? | Can’t be answered. Oracle doesn’t predict. |
Show me the important ones | One question back, with answers to pick from |
Defects assigned to Sarah, when nobody is called Sarah | A message saying there is no Sarah, not an empty table |
The last one matters most. An empty table looks like good news. A message about the name tells you it was a typo.
Questions It Can’t Express
Section titled “Questions It Can’t Express”- No sums or averages. Oracle counts records; use Observatory for sums and averages. It can list records, count them, or count them per group, such as “failures per requirement” or “defects by severity”.
- Conditions combine with “and”. There is no “or” and no brackets. A question that needs two answers is two questions.
- One sort column.
- No paging. You get a hundred rows by default, and up to five hundred if you ask for more. For more than that, narrow the question.
- Rich text isn’t searched. Conditions can’t match the text inside test case steps or shared steps.
Daily Limits
Section titled “Daily Limits”Each workspace can ask up to 1,000 questions, draw 500 charts and get 100 AI readiness explanations a day:
| Action | Limit |
|---|---|
| Asking a question | 1,000 a day |
| Drawing a chart | 500 a day |
| AI readings of a readiness explanation | 100 a day |
The limits are shared by everyone in the workspace, not set per person. They reset each day.
They are the same on every plan. A bigger plan doesn’t raise them and a smaller one doesn’t lower them. What your plan sets is your monthly AI credit allowance. These daily limits are far above what a team reaches in normal work, and we may adjust them.
Corrections don’t count. Removing a condition, sorting, loading more rows, adding readiness columns or redrawing a chart doesn’t use AI, so none of them count. Answering one of Oracle’s follow-up questions does count, because Oracle works the question out again. See How to Ask.