Slovakia RUZ Filing Changes & Corrections Metadata
Pricing
from $0.00005 / actor start
Slovakia RUZ Filing Changes & Corrections Metadata
Read Slovak RÚZ changed accounting units, statements, financial reports and annual-report metadata. Export official IDs, filing dates, attachment hashes and bounded continuation receipts. Current observations, not a complete revision history.
Pricing
from $0.00005 / actor start
Rating
0.0
(0)
Developer
Michal Pšenčík
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
3 days ago
Last modified
Categories
Share
Read the four anonymous Slovak RÚZ modified-ID feeds: accounting units, accounting statements, financial reports and annual reports. Get current filing metadata, official parent IDs, attachment hashes and an explicit coverage receipt. Old accounting periods can change today: the source's statement 2222095 covers 2014, was filed in 2015 and was modified in October 2026.
This is a bounded metadata-feed experiment. It does not establish a complete revision history, prove a legally filed correction, or distinguish new records from corrections.
Input
{"changedSince": "2026-10-16","entityTypes": ["statement"],"pageSize": 1,"maxPagesPerFeed": 1,"maxRows": 1,"requestIntervalMs": 2000}
Omitted changedSince resolves once to yesterday UTC; only calendar dates from 2000 onward are accepted. Defaults select statement, report, annual_report, unit in that order, 10 IDs/page, one page/feed and 20 aggregate observations. Bounds are 1–25 IDs/page, 1–4 pages/feed and 1–100 rows/run. maxRows must be at least pageSize. Feeds rotate by whole page. The request interval accepts 1100–5000 ms; this build always waits at least 2000 ms after completion.
Pass previousVersions with at most 1000 keys like "ruz:statement:2222095": {"metadataHash":"<64 lowercase hex>","sourceModifiedDate":"2026-10-16","sourceStatus":null}. Hashes must use previousNormalizationVersion "1". The Actor compares only entities it encounters; it has no persistent watch database. A later run can supply the prior row's hash/date/status.
Output and resuming
Each paid dataset row has the complete version-1 schema: sourceKey/sourceId/entityType/sourceUrl, source date/status/deleted, metadata, metadataHash/versionId, comparison fields, fetchedAt, apiVersion and warnings. Metadata contains official IDs, periods, submission/preparation/approval dates, accessibility, classification codes, child IDs and attachment ID/MIME/bytes/pages/language/hash. Missing scalars are null and unexposed arrays empty. Empty arrays do not prove source-wide absence. Names, addresses, IČO, DIČ, cover-sheet identity, filenames, financial tables and PDFs are excluded. Nonpublic unit metadata is suppressed entirely; report visibility gates content and attachments.
changeType is observed, metadata_changed, source_date_changed, unchanged or source_deleted. Only exact source ZMAZANÉ is deletion; neither absence nor a failed HTTP request implies deletion. versionId identifies a canonical observed version; A→B→A reuses A's ID. It is a downstream dedup key, not a unique billing receipt.
Free default KV records SUMMARY and OUTPUT report coverage, errors and a portable continuation. SUMMARY includes per-feed pages/IDs/stored rows/afterId/complete, attempts, response bytes, emitted version IDs and resumeAdvice. RUN_STATE is internal run-local migration bookkeeping. Reserve pageSize free row slots before starting a page. Leftover capacity may be unused. A page cursor advances only after every listed row is verified in dataset storage. A failed or interrupted page keeps its old cursor; replayed prefix observations may be delivered and billed again.
For ordinary row/page caps only, autoResumeAllowed is true. Pass the exact continuation, same changedSince/entityTypes and same pageSize; explicitly reducing pageSize is allowed for recovery. Never invent a large cursor. Receipts expire after seven days. Operational interruptions disable automatic resume: inspect delivery, resolve the error, raise the platform allowance, or reduce pageSize as SUMMARY directs. A one-ID page that still fails needs investigation. After completion, start a fresh sweep using nextChangedSince (original sweep UTC start day minus one), with fresh cursors. A complete sweep means observed feed ends, not an immutable register snapshot. Serialize runs sharing one consumer cursor and update downstream baselines only from stored rows.
Pricing
Configured reference prices: $0.00005 per automatic start unit and $0.001 per useful dataset observation stored. Unchanged checks, overlapping/replayed checks and explicit tombstones cost the same as changed observations. Parent enrichment has no separate item. Errors, empty results and KV coverage have no item fee; every run can still incur its platform start fee. Ten delivered rows plus one start total $0.01005 in reference event fees. Actual platform allocation, usage mode and pricing configuration determine the invoice; local logs do not verify prices.
Limits and source
HTTP only, fixed HTTPS origin, one request in flight, at least two seconds idle after every attempt, two bounded retries for transient errors. All redirects must retain the allowlisted route/query. 403, challenge/HTML and 429 stop the entire source without bypass. Retry-After appears in SUMMARY. 400/404 are errors, never fabricated empty feeds or tombstones.
Ceilings: 100 HTTP attempts including parents/retries/redirects, 20 seconds/attempt, 2 MiB decoded response, 32 MiB conservative external traffic accounting, 300 seconds source work, 8 MiB internal serialized storage traffic, 128 KiB/row, and 500 operations per storage category. Budget reservations can stop earlier: before every push the Actor reserves the push's readback and receipt writes, and checks the remaining spending allowance. It uses the smaller of the SDK allowance and the run's total charge cap minus verified stored-row costs and the start event. The durable delivery ledger also counts committed appends whose responses were lost, and is checked against dataset readback on restart. Storage API operations use no SDK retries and a twenty-second timeout. Each append is attempted once; after a lost response, exact dataset count and row readback decide whether it was stored. An unconfirmed append is reported as delivery_uncertain with that row's versionId and is never replayed automatically. Migration resumes only reconciled stored rows, with the halt reason, Retry-After and request pacing intact; an interrupted HTTP operation fails closed because its final attempt/traffic state is uncertain. Final receipts on a hard kill are best effort; use the last durable checkpoint.
Source: RÚZ Open API, published under CC0. The free native API is an alternative. Current metadata is not historical versions; child changes need not update parent dates and records may disappear without tombstones. UTC overlap is a cautious polling recommendation, not proof of source timezone or completeness. No cloud throughput guarantee.
Official IDs can relate to natural persons. The Actor user is the controller for downstream personal-data processing; use an appropriate purpose and retention policy. Do not enrich or infer personal details. Raw development fixtures contain published source data outside the output whitelist.
Apify-IP access, real platform billing/spend behavior, allocated memory and invoices must be verified before publication. Quiet/empty-scan economics and buyer demand remain unverified.