fields say what a check compares. Each one is a path into the API response. A check keeps only those fields, in the shape of the response, and reports a change when they differ from the previous check.
fields, the whole response is compared, less its requestMetadata. For Web Scraping that is the page as markdown, or the data your aiExtractRules extracted; it takes no fields.
A field that is missing from the response is kept as null, so a field appearing or disappearing is a change too.
Lists
Many responses carry lists: Zillow’sproperties, Amazon’s product.variants, a SERP’s organicResults. A list is tracked with list, which says how its items are read:
keys: the item fields that tell items apart. With keys, a check reports items that were added, removed, and changed (an item with the same key whose tracked fields differ). Order never matters.fields: the item fields to compare. Empty, the whole item is kept, but only the keys are hashed: a listing whose price moved is not a change, only a listing that appeared or disappeared is.
Two entries for one list merge into one rule:
properties[id].price and properties[id].status as separate entries become fields: ["price", "status"]. An entry with fields: [] absorbs the others. Two different keys for one list, or a path tracked both with and without list, are refused with 400.
In the dashboard the same rules are written as lines:
properties[id].price, product.images[], localResults.places[title,address].rating.What a Snapshot Looks Like
With{ "path": "properties", "list": { "keys": ["id"], "fields": ["price"] } } and { "path": "searchInformation.totalResults" }, a Zillow check keeps:
GET /monitors/{id}/resources/{resourceId}/snapshot returns the latest one.
What a Diff Looks Like
A changed result carriesdiff:
changes lists value changes by path; lists has one entry per tracked list that gained, lost or changed items. A result also carries patch, a unified diff of the snapshot text, for display.
Without fields the diff is { "kind": "object", "changes": [ … ] }, or { "kind": "text", "changes": [{ "type": "added" | "removed", "value": "…" }] } for a page as markdown.
Suggested and Default Fields
GET /monitors/endpoints returns, for each endpoint, fields.suggested (paths worth tracking) and fields.default (what the dashboard picks for a new monitor). The API applies no defaults: a monitor created without fields compares the whole response.
Changing Fields
PUT /monitors/{id} with new fields rebuilds the baseline from the last response the API gave for each resource: no request is made and no credits are spent. History stays; the next check compares against the new baseline.