Flexible live insights query

Runs a custom performance query against Meta, Google Ads or TikTok and returns the platform's rows as is, for reports the fixed endpoints don't cover.

The account's platform picks the contract:

Meta (facebook/instagram): forwards caller-chosen fields, breakdowns and filtering to any Meta insights node and returns Meta's rows verbatim. objectId (required) selects the node; level sets row granularity. Semantic validation is Meta's: an unknown field or invalid breakdown combination returns a 400 carrying Meta's message. For long ranges or agency-scale accounts prefer the async variant (POST /v1/ads/insights/reports).

Google Ads (googleads): raw GAQL passthrough. Send any read-only GAQL SELECT via query (campaign/keyword/search-term/geo/demographic/asset/shopping resources, change_event, any segments.*) and rows come back verbatim (camelCase, counters as strings). Results are paged at a fixed 10,000 rows; follow paging.nextPageToken with pageToken. adAccountId (alias customerId) is only needed when the connection has several Google Ads accounts. Semantic validation is Google's: an invalid query returns a 400 carrying Google's message (note: selecting segments.date requires a finite date filter).

Queries run against Google Ads API v25, so write GAQL against the v25 field reference. One exception is translated for backward compatibility: the legacy campaign.start_date / campaign.end_date (removed by Google in v23) are rewritten to campaign.start_date_time / campaign.end_date_time, and rows still carry campaign.startDate / campaign.endDate as YYYY-MM-DD. In WHERE, =, <, <=, >, >=, BETWEEN and IS [NOT] NULL against a 'YYYY-MM-DD' literal are translated; any other form returns Google's 400. New code should select the _date_time fields directly.

TikTok (tiktok/tiktokads): passthrough of TikTok's synchronous report (/report/integrated/get/). Send adAccountId, dataLevel, dimensions, metrics, fromDate/toDate (TikTok caps the span at 365 days) and optionally filtering in the same [{"field", "operator", "value"}] shape as Meta (operator is TikTok's filter_type, e.g. IN; array values are JSON-encoded for TikTok). Rows come back verbatim as { dimensions, metrics } with page-number paging.

De-duplicated reach across a set of campaigns, ad groups or ads: filter the set and group by country_code instead of the entity id. TikTok then counts each person once across the whole set and range, per country. Example, two campaigns for a month: dataLevel=AUCTION_CAMPAIGN&dimensions=country_code&metrics=reach,impressions,frequency&fromDate=2026-09-01&toDate=2026-09-30&filtering=[{"field":"campaign_ids","operator":"IN","value":["1876574798182050","1876575504573650"]}]. For a set of ads use dataLevel=AUCTION_AD with ad_ids. A filter on a finer entity than dataLevel (e.g. ad_ids at AUCTION_CAMPAIGN) returns 400: TikTok would otherwise widen it to every parent entity containing those ids. When the set delivers in several countries, reach is per country; summing the rows counts a person reached in two countries twice. AUCTION_ADVERTISER rejects entity filters.

import Zernio from '@zernio/node';const zernio = new Zernio({ apiKey: process.env.ZERNIO_API_KEY });const { data } = await zernio.adinsights.queryAdInsights({  query: {    accountId: 'account_abc123',  },});console.log(data);

{  "objectId": "string",  "adAccountId": "string",  "customerId": "string",  "fieldMask": "string",  "data": [    {}  ],  "paging": {    "after": "string",    "nextPageToken": "string",    "page": 0,    "pageSize": 0,    "totalRows": 0,    "totalPages": 0  }}