Selector survival
Predict which UI locators will break in the next release before they break your tests — what to send, what the answers mean, and how to act on them.
Task: selector_survival · MCP tool: ostra_check_selector · Price: $0.25 per 1M locators ($0.125 in a batch)
Selector survival answers will this locator survive the next release? Run it on the locators your tests depend on before a release, or on the locators a pull request adds, and replace the ones likely to break before they cost you a red build.
What to send
input.locators is a list of 1 to 500 locators. For each:
| Field | What it is | Where to get it |
|---|---|---|
locator_id | Your name for the locator, echoed back. | Page-object property names work well: checkout.apply-coupon. |
strategy | How it finds its element: css, xpath, test_id, text, role, none. | Playwright getByTestId → test_id, getByRole → role, getByText → text, locator('//…') → xpath, other locator() → css. Cypress cy.get('[data-cy=…]') → test_id. Selenium By.xpath → xpath, By.cssSelector → css. |
dom_locator_found | Whether it matches an element in the current build's DOM. | Open the page in the build under test and check the locator (await page.locator(sel).count() > 0). |
history | Up to 100 { release, resolve_rate } entries, oldest first. resolve_rate (0–1) is the share of runs in that release where the locator resolved. | Record it per release from your test runs, or from the check above (1.0 found, 0.0 not found). |
A locator with no history and no dom_locator_found carries no evidence and gets defer (insufficient_evidence).
Example
curl https://ostra.testrelic.ai/api/v1/predict \
-H "Authorization: Bearer $OSTRA_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "ostra-lite",
"task": "selector_survival",
"input": {
"locators": [
{
"locator_id": "checkout.apply-coupon",
"strategy": "css",
"dom_locator_found": true,
"history": [
{ "release": "2026.10.1", "resolve_rate": 1.0 },
{ "release": "2026.10.2", "resolve_rate": 0.92 },
{ "release": "2026.11.1", "resolve_rate": 0.71 }
]
},
{
"locator_id": "nav.account-menu",
"strategy": "role",
"dom_locator_found": true,
"history": [
{ "release": "2026.10.2", "resolve_rate": 1.0 },
{ "release": "2026.11.1", "resolve_rate": 1.0 }
]
}
]
}
}'output (trimmed to the key fields):
{
"decisions": [
{ "locator_id": "checkout.apply-coupon", "decision": "breaks", "p": 0.78, "probabilities": { "survives": 0.22, "breaks": 0.78 } },
{ "locator_id": "nav.account-menu", "decision": "survives", "p": 0.96, "probabilities": { "survives": 0.96, "breaks": 0.04 } }
]
}Decisions come back in input order. This request is billed for 2 locators.
Answers and what to do
decision | Means | Next step |
|---|---|---|
survives | The locator is expected to keep working. | Nothing. |
breaks | The locator is likely to stop matching in the next release. | Replace it with a sturdier one before release: prefer getByRole with an accessible name, or a dedicated test id. |
defer | Not enough evidence or not confident. | Add history or a DOM check, or review it by hand. |
A falling resolve_rate across releases (like checkout.apply-coupon above) is the clearest warning sign; CSS and XPath locators tied to layout break more often than role and test-id locators.
Getting better answers
- Keep history. Record each locator's resolve rate per release; even three releases help a lot.
- Check the current DOM.
dom_locator_foundis cheap to compute in a preview environment. - Use stable ids. Name
locator_idafter the page object so results map straight back to code. - Batch. One request takes up to 500 locators;
/batchhalves the price.
Check selectors before release has a complete script and CI job.
Reference
- REST:
POST /api/v1/predictwithtask: "selector_survival". - MCP:
ostra_check_selector.
Test impact
Give Ostra a code change and the tests that could be affected; get back the tests to run first, ranked. What to send, how selection and fallback work, and how to use the answer in CI.
Pro
TestRelic's testing model for writing and fixing tests, analysing failures and logs, reading screenshots and video, and planning coverage — limits, thinking, tools, images and video, streaming, sessions, cold starts, compatibility and pricing.