Datamolino capture API: returned data, review and integration checks
The API accepts invoices and receipts and returns captured document data for your application. Datamolino’s API service can include human review where automation needs help, with the fields and handling agreed for the project. Your application still needs to match the result to its own records and handle exceptions. Invoice-coding controls are not part of the current API scope.
Does human review mean our application needs no checks?
No. Provider review and customer checks serve different purposes. Your application must still confirm the right source, entity and supplier, handle duplicates and validate the data needed for its accounting or operational decisions.
Agree the output before building the integration
Start with the fields your application actually uses. The public API overview describes supplier and invoice identifiers, dates, currency, net, tax and total amounts, plus line items where available. Extra fields and custom processing need an agreed scope.
For each field, decide how your application handles a missing value, an unexpected format or a value that does not match its master data. A captured supplier name is not automatically the internal supplier identifier in your system.
| Captured information | Application check |
|---|---|
| Invoice number and source reference | Does this result belong to the intended file, and has the invoice already been imported? |
| Supplier name or tax identifier | Which existing supplier should receive the transaction? |
| Currency, net, tax and total | Are the values complete and consistent with the supported document and the application’s checks? |
| Invoice lines | Are the required descriptions, quantities and amounts present? Which missing values need review? |
Define what happens when automation needs help
Agree which documents or fields receive human review and how exceptions are handled. The original 2022 article described operator intervention, but the service agreed for your integration determines the current scope.
Confirm which documents or fields can receive human review, what happens when the source is unreadable and how completion time is measured. Do not assume that every document is manually checked or that a percentage accuracy claim applies to your document mix.
Capture technology and accounting coding are different. Recognition reads the document. Datamolino’s web-app coding uses saved choices and deterministic rules. The API use case here is captured data, not control of the web app’s coding or approval configuration.
Build around a document lifecycle
- Submit. Keep your own reference to the source and the upload result. An accepted upload is not a completed capture.
- Wait for completion. Use the agreed status and notification mechanism. Handle a delayed result without blindly uploading the same file again.
- Retrieve. Store the returned data with the source reference and processing outcome.
- Validate. Run master-data matching, amount and duplicate checks required by your application.
- Resolve exceptions. Assign incomplete or disputed results to a defined review queue.
The published status guide describes processing, ready, duplicate and failure-related outcomes. It is an older reference, so use the current endpoint documentation supplied for your project as the implementation contract.
A multi-invoice PDF needs another check: associate each resulting invoice with the original upload. One file can produce several document results. Preserve that relationship rather than treating a parent batch as one invoice.
Run a trial with exceptions, not only clean PDFs
Include a digital invoice, a detailed invoice, a new supplier layout, a receipt photo and an ordered PDF containing several invoices. Add a duplicate and a file with missing or unreadable information. Use documents you are authorised to process.
For each, record required fields, returned values, completion time, corrections and the final disposition. Count invoice-level failures as well as individual field errors. A result that reads most fields correctly may still be unusable if the missing field determines the supplier or amount.
Treat capture as an asynchronous process. An accepted upload is not a completed result. Use the agreed completion webhook or status checks before retrieving and validating the fields.
Agree these points before the production handoff
- Supported document types, languages, field names and output formats.
- Authentication, permissions and how each entity’s data stays in the intended scope.
- Completion notification, retry behaviour and duplicate handling.
- Human-review coverage, exception handling and service levels.
- Retention, access and the source-to-result relationship.
- Commercial terms for the actual volume and custom work.
Use the developer overview to discuss API access and a representative sample. The useful deliverable from that discussion is an agreed data contract and trial result, not an assumption that every web-app feature is available through the API.
Can a custom agent upload documents?
An authorised application or agent can use the agreed API to upload selected documents and retrieve captured data. Mailbox access, file selection and scheduling belong to that integration. This is not a built-in Codex or ChatGPT connector.
