CSV/JSON Schema Normalizer

Normalize every row from CSV, JSON, or an Apify dataset into stable, typed, automation-ready records.

What this Actor does

• Accepts inline JSON records, CSV text, JSON text, an Apify dataset ID, and public or user-authorized CSV/JSON URLs.
• Follows JSON next links, page tokens, and offset pagination to the source's natural end.
• Infers one stable schema from the complete input, or applies a buyer-supplied schema.
• Emits exactly one default-dataset item for every delivered source row, including duplicate rows.
• Saves the schema in SCHEMA and counts in RUN-SUMMARY in the default key-value store.
• Charges result-delivered once per production row through Apify's charged dataset push.

There is no maxItems, maxRows, per-source, charge-derived, or overall result ceiling in this Actor. A buyer's Apify spending limit can stop paid delivery, as required by the platform.

Inputs

• records: Inline array of JSON objects.
• csvText: Inline RFC-style CSV text, including quoted delimiters and line breaks.
• jsonText: Inline JSON array or an object containing items, records, results, rows, or data.
• datasetId: Existing Apify dataset iterated to its natural end.
• startUrls: Public or user-authorized CSV/JSON URLs. The production default reads the public Apify Store API and follows all available pages.
• schema: Optional field map. For example, {"amount":{"type":"number","required":true}}.
• csvHasHeader, delimiter, trimStrings, and emptyStringsAsNull: Parsing and normalization controls.
• includeRaw: Include the original row beside the normalized row.
• requestTimeoutSecs: Per-request network timeout; it is not a result-count limit.

sourceMode selects one input surface, so the public default URL is not mixed into an inline-record or dataset run. Rows within the selected source are not deduplicated, so a 1,200-row input produces 1,200 result records when the buyer's spending limit permits 1,200 charged deliveries.

Output

Each source row produces one item with:

• recordId: Stable SHA-256 identity derived from source, row number, and source content.
• rowNumber and sourceRowNumber: Global and source-local row positions.
• sourceType and sourceReference: Input provenance.
• normalized: Complete row normalized against the stable schema.
• valid and validationErrors: Required-field and type-conversion results.
• retrievedAt: Run timestamp.
• raw: Original row only when requested.

The default dataset never adds summary rows, so its item count equals the number of delivered source rows. The inferred/applied schema and run summary are separate key-value-store records.

Quick non-billable check

{
  "mode": "sample",
  "sourceMode": "sample"
}

Production examples

Normalize an Apify dataset:

{
  "mode": "production",
  "sourceMode": "dataset",
  "datasetId": "YOUR_DATASET_ID"
}

Normalize 1,200 inline records by supplying all 1,200 objects in records; no Actor input field truncates the array.

Pricing

The existing Actor pricing remains unchanged:

• result-delivered: USD 0.01 for each validated buyer-facing row delivered to the default dataset.

Sample mode is non-billable. Production stops when Apify reports that the buyer's spending limit has been reached, preventing unpaid processing after the platform limit.

Access and limitations

• Process only public data or inputs you are authorized to use.
• The Actor does not bypass authentication, paywalls, CAPTCHAs, access controls, or source terms.
• Very large sources still depend on the run's platform memory, network, and time resources; the Actor itself does not impose a record-count ceiling.
• Invalid values are retained as null with a field error instead of being silently invented.

Existing public Actor identity

• Owner: zentrafoundry
• Actor ID: lxv37LGvAeUvcZtnM
• Slug: csv-json-schema-normalizer
• Store URL: https://apify.com/zentrafoundry/csv-json-schema-normalizer

This package updates that exact public Actor in place. It does not create, clone, draft, privatize, or change pricing or visibility.