CSV Reconciliation Audit

Compare two UTF-8 CSV exports by exact keys and identify amount differences before importing data into another workflow. Built for agent pipelines, automation developers and operations teams.

What you receive

One JSON report in the default dataset, including:

• Matched records within the amount tolerance you specify.
• Amount discrepancies, with the signed difference (right minus left).
• Records found only in the left or right file.
• Summary counts, declared currency and matching policy.

Example

{
  "leftCsv": "id;amount\n001;1.234,56\n002;100,00",
  "rightCsv": "id;amount\n001;1.234,57\n002;90,00",
  "delimiter": ";",
  "keys": ["id"],
  "amountColumn": "amount",
  "decimalSeparator": ",",
  "groupSeparator": ".",
  "tolerance": "0.01",
  "currency": "CLP"
}

Record 001 matches within tolerance. Record 002 has a difference of -10.00. Identifiers remain strings, so leading zeros are preserved. Calculations use exact decimal arithmetic.

Input requirements

Supply CSV text directly, not file URLs. Declare the delimiter, the normalized key column names, one amount column and its separators. Header normalization lowercases names, replaces non-word runs with underscores, and assigns unique suffixes when necessary. Inspect column names locally if your source has complex or duplicate headings.

The key can have multiple columns, but each combined key must be unique and nonempty in each file. The amount column cannot be part of the key. Both CSVs must use the same separators and currency. Currency is your declaration; the Actor does not verify currency against individual rows. Empty amounts, malformed numbers, inconsistent column counts or duplicate keys cause the run to fail without a custom report charge.

Pricing unit

One completed report is one reconciliation-report event, regardless of matching or unmatched row count. The current configured event price must be checked in the Pricing tab before running. An automatic Actor start event may also apply as shown there. Invalid inputs do not trigger our custom report event; platform start or usage charges may still apply. No separate custom per-row charge is requested.

Limits

• Total serialized input: 1,000,000 UTF-8 bytes.
• Up to 10,000 nonblank records and 200 columns per CSV.
• Maximum serialized report: 8,000,000 bytes; larger reports are rejected before a custom charge is attempted.
• One-to-one exact matching; no fuzzy matching, many-to-one aggregation or currency conversion.
• No Excel, PDF, image parsing, external URLs, bank connection or ERP writes.

The report is a data comparison, not confirmation that funds moved or accounting records are correct. A tolerance match does not prove that two transactions represent the same real-world event. Check key selection and results before acting.

Privacy

This hosted version receives input through Apify and stores the output in the run's dataset. Apify also stores run inputs and platform metadata according to its settings and policies. Do not assume zero retention. Only send files you are authorized to process. The application does not call a language model, fetch external URLs, execute CSV formulas, or log file contents. Use your platform storage settings to manage access and retention.

Automation

After publication, invoke this Actor with the Apify API or a compatible Apify MCP client using the input above. Publication, pricing and access settings are controlled by the publisher. This source package itself is not a live paid service.