Get a Bank Statement
GET/api/v8/partner/bank-statements/:document_id
Veryfi's Get a Bank Statement endpoint allows you to retrieve a previously processed Bank Statement.
Request
Path Parameters
The unique identifier of the document.
Query Parameters
A field used to determine whether or not to return bounding_box and bounding_region for extracted fields in the Document response.
A field used to determine whether or not to return the score and ocr_score fields in the Document response.
Responses
- 200
Returns a processed Bank Statement.
- application/json
- Schema
- Example (from schema)
Schema
- API_V8_PARTNER_BANKSTATEMENTS_BANKSTATEMENT
- API_V8_PARTNER_BANKSTATEMENTS_DETAILEDBANKSTATEMENT
- Array [
- ]
- Array [
- ]
- Array [
- ]
- Array [
- _FraudulentPDFSignalReason
- _FraudulentPDFMetadataReason
- _FraudulentPDFSplicedPageReason
- ]
- Array [
- _FraudulentImageSignalReason
- _FraudulentImageEditorReason
- _FraudulentImageGeneratorReason
- ]
- Array [
- ]
- Array [
- ]
- Array [
- BankStatementFraudPage
- Array [
- ]
- Array [
- ]
- Array [
- ]
- ]
- Array [
- Array [
- ]
- Array [
- ]
- ]
- Array [
- ]
- Array [
- ]
- Array [
- ]
- Array [
- ]
- Array [
- _FraudulentPDFSignalReason
- _FraudulentPDFMetadataReason
- _FraudulentPDFSplicedPageReason
- ]
- Array [
- _FraudulentImageSignalReason
- _FraudulentImageEditorReason
- _FraudulentImageGeneratorReason
- ]
- Array [
- ]
- Array [
- ]
- Array [
- BankStatementFraudPage
- Array [
- ]
- Array [
- ]
- Array [
- ]
- ]
- Array [
- Array [
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- ]
- Array [
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- Array [
- MOD1
- MOD2
- ]
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- Array [
- MOD1
- MOD2
- ]
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- Array [
- Array [
- MOD1
- MOD2
- ]
- Array [
- MOD1
- MOD2
- ]
- ]
Possible values: non-empty
A custom identification value. Use this if you would like to assign your own ID to documents. This parameter is useful when mapping this document to a service or resource outside Veryfi.
meta object
Possible values: non-empty
A custom identification value. Use this if you would like to assign your own ID to documents. This parameter is useful when mapping this document to a service or resource outside Veryfi.
pages object[]
Possible values: <= 1
The average OCR score of the page.
The width of the page.
The height of the page.
Possible values: non-empty
Default value: ``
Tags associated with the document.
Possible values: <= 1
The average OCR score of the whole document.
Possible values: non-empty
The version of the model used to process the document.
Possible values: non-empty
The original file name of the uploaded document.
Possible values: [api.email, api.web, api, browser_extension, lens.bill, lens.invoice, lens.long_receipt, lens.other, lens.receipt, lens.web, lens]
Default value: api
The source of the document's submission for processing.
duplicates object[]
An array of duplicate documents found in the system.
The id of the duplicate document.
Possible values: non-empty and <= 2083 characters
The url of the duplicate document.
fraud BankStatementsFraud
An object that contains additional information to identify fraud.
Possible values: <= 1
Confidence of Fraud Detector in it's prediction
Possible values: [green, yellow, red]
Color from Fraud Detector: green means legitimate, yellow means review needed and red means fraud
Possible values: [LCD photo, not a document, duplicate, digital background, screenshot, digital tampering, fraudulent pdf, fraudulent image, high velocity, critical velocity]
List of attributions which marked the document as fraud
details BankStatementFraudDetails
Strictly typed descriptions and reasons keyed by fraud type.
lcd_photo _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
not_a_document _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
duplicate _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
digital_background _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
screenshot _ScreenshotDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
reasons object[]
Possible values: [mobile_screenshot, other_screenshot]
Possible values: non-empty
digital_tampering _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
fraudulent_pdf _FraudulentPDFDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
reasons object[]
Possible values: [text_overlay, font_mismatch, modified, date_inversion, producer_content_mismatch, writer_mismatch, null_metadata, xmp_info_mismatch, future_metadata_date, raster_profile_mismatch, uncompressed_raster, double_jpeg, scrubbed_provenance]
Possible values: non-empty
Possible values: non-empty
Possible values: [creator, producer]
Possible values: non-empty
Possible values: non-empty
Possible values: > 0
1-based number of the first page written after the pages surrounding it. Later pages may belong to the same block.
fraudulent_image _FraudulentImageDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
reasons object[]
Possible values: [double_jpeg, scrubbed_provenance, unprofiled_png, unprofiled_webp, synthetic_raster]
Possible values: non-empty
Possible values: non-empty
Possible values: [icc_profile, exif:Software, exif:ProcessingSoftware, xmp:CreatorTool, png:Software, jpeg:comment]
Possible values: non-empty
Possible values: [generator_signature, generator_writer_profile]
Possible values: non-empty
Possible values: non-empty
high_velocity _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
critical_velocity _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
submissions object
The amount of submissions from specific device id, used in velocity fraud detection.
Possible values: [text_similarity, field_matching]
Method used to detect duplicates for fraud
duplicates object[]
An array of duplicate documents found for purposes of fraud detection
The id of the duplicate document.
Possible values: non-empty and <= 2083 characters
The url of the duplicate document.
fraudulent_pdf FraudulentPDFResult
Results of the pdf analysis.
Possible values: non-empty
What incremental PDF revisions changed.
incremental_edits object[]
Which word an incremental PDF revision replaced, and on which page.
Possible values: > 0
1-based number of the page the edit landed on.
Possible values: non-empty
The replaced word, or null when the word was inserted.
Possible values: non-empty
The replacing word, or null when the word was deleted.
Possible values: non-empty
Possible values: > 0
1-based number of the first page written after the pages surrounding it. Later pages may belong to the same block.
Score for a PDF an automated browser rendered from a desktop operating system. It is reported here but scored against the generated document signal, so it does not move the fraudulent pdf score.
Possible values: non-empty
Operating system named by the headless browser that rendered the PDF. Reported for every headless render, including the server platforms that do not score.
Score for a raster that was JPEG-compressed more than once. An original capture is compressed exactly once, so a second compression means the file is not the first-generation scan it presents itself as. Disabled by default, and reported as 0.0 until it is enabled for your account.
Possible values: > 0
JPEG quality of the first compression, recovered from the coefficient histograms. Evidence for a reviewer rather than a score, and absent when nothing coherent was recovered.
Score for a raster whose encoder signature carries no provenance: grayscale content in a subsampled colour JPEG, no EXIF, a placeholder JFIF density and a stock quantization table. Reaches the same conclusion as double_jpeg for a re-encode too coarse to leave a recoverable first compression. Disabled by default, and reported as 0.0 until it is enabled for your account.
fraudulent_image FraudulentImageResult
Results of the image forensics analysis, for a document that arrived as a picture rather than a PDF. Absent when the signal is switched off for your account.
Score for a document image that was JPEG-compressed more than once. An original capture is compressed exactly once, so a second compression means the file is not the first-generation photo it presents itself as. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a JPEG-family raster can answer it, so a PNG reports null. Disabled by default.
Score for a document image whose encoder signature carries no provenance: grayscale content in a subsampled colour JPEG, no EXIF, a placeholder JFIF density and a stock quantization table. Reaches the same conclusion as double_jpeg for a re-encode too coarse to leave a recoverable first compression. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Disabled by default.
Possible values: > 0
JPEG quality of the first compression, recovered from the coefficient histograms. Evidence for a reviewer rather than a score, and absent when nothing coherent was recovered.
Score for an image that names an image editor in its own metadata. Read as a reason to review rather than to reject: an editor naming itself on a photograph may be an honest user cropping a receipt before uploading it, which is why it scores lower here than the same tool named on a PDF. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Formats that carry neither an ICC profile nor EXIF, which is BMP and GIF, can never answer it.
Possible values: [icc_profile, exif:Software, exif:ProcessingSoftware, xmp:CreatorTool, png:Software, jpeg:comment]
Which metadata surface named the editor. Exactly one is ever reported, the first that matched.
Possible values: non-empty
The literal string the file declared, which is what a reviewer acts on rather than the score.
Score for an image a server-side rendering library declares it wrote, such as libgd naming itself in a JPEG comment segment. Scores far above the editor signature because the finding is different: an editor names itself on a photograph somebody cropped, while a rendering library names itself on a page nothing ever photographed. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a format that can carry metadata can answer it.
Possible values: [icc_profile, exif:Software, exif:ProcessingSoftware, xmp:CreatorTool, png:Software, jpeg:comment]
Which metadata surface named the library. The same surfaces the editor signature reads, and as there exactly one is ever reported.
Possible values: non-empty
The literal string the file declared, which is what a reviewer acts on.
The same finding on a format where nothing is declared: a PNG whose chunk sizing, compression level and absence of provenance are libpng's own defaults, which is how the same libraries emit PNG. Reported rather than inferred from the file naming itself, because on this format it never does. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a PNG can answer it.
Possible values: non-empty
The writer profile that matched, naming the evidence rather than restating the score. Deliberately has no _field companion, unlike the two signatures above: the profile is read off the container itself rather than off one named surface.
Score for a raster too flat to have come off a sensor: an indexed palette, very few distinct colours, or no chroma variation at all. Ships at 0.0 and so contributes nothing to score — it is reported for the rate while a screenshot of a genuine receipt is still indistinguishable from a drawn one. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire.
Possible values: non-empty
Which of those measurements matched, sorted. Evidence for a reviewer rather than a score, and the set grows as the detector learns to name more of them, so read it as a list of strings rather than a fixed vocabulary.
Score for a PNG carrying no colour or provenance chunk at all, as a rendering library writes and a camera pipeline does not. Ships at 0.0 for the same reason as synthetic_raster: screenshot tools and our own rasterizer write the identical bare file. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a PNG can answer it.
The same check on WebP, and ships at 0.0 for a different reason than the two above: too little genuine WebP has been seen to say what the rate would be. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a WebP can answer it.
Possible values: [original, render]
Which bytes were read. original is the upload as submitted and is the only value for which the compression scores above can be reported. render is a page render taken after cropping, rotation and format conversion, which destroys compression evidence and can manufacture it, so those two report null against it. Null means nothing was selected to read at all; a file that was selected but could not be fetched or opened reports a source together with a non-zero unreadable_source.
Possible values: non-empty
Formats of the files that were read, as the image library names them. Note HEIF rather than HEIC, and MPO for a .jpg holding multiple frames, as a phone HDR capture does.
How many selected files could not be opened. Can exceed one on a multi-file upload. A non-zero value here alongside a set source is the expected shape when an original has already been deleted, rather than a fault.
Highest of the scores above, and what the fraudulent image signal contributes. Follows the same rule they do: null when not one of them was in a position to answer, and 0.0 when at least one was and nothing fired. So a page render scores a real 0.0, the compression pair having declined but the editor signature having run. The checks that ship at 0.0 are in the max like any other, which is why they can report a hit without moving it.
pages object[]
An array containing fraud info about each extracted page
is_lcd ClassBoolField
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
The extracted value.
flags object[]
List of flags which marked the document as fraud
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: non-empty
ai_generated ClassBoolField
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
The extracted value.
handwriting object[]
An array containing handwriting info about each extracted page
Possible values: >= 8, <= 8
Bounding region of the artifact in [x1,y1,x2,y2,x3,y3,x4,y4] format
Possible values: non-empty
Type of the artifact
digital_tampering object[]
An array containing digital tampering info about each extracted page
Possible values: >= 8, <= 8
Bounding region of the artifact in [x1,y1,x2,y2,x3,y3,x4,y4] format
Possible values: non-empty
Type of the artifact
Possible values: non-empty and <= 2083 characters
A signed URL to access the auto-generated PDF created from the submitted document. This URL expires 15 minutes after the response object is returned and is resigned during every GET request.
The unique number created to identify the document.
Possible values: non-empty and <= 2083 characters
A signed URL to access the auto-generated thumbnail created for the submitted document. This URL expires 15 minutes after the response object is returned and is resigned during every GET request.
accounts object[]required
summaries object[]required
Summary information
Possible values: non-empty
The title or label that captures the main information
The overall amount calculated from all transactions
transactions object[]required
A list of transactions associated with the account.
The value indicating the position of where the transaction appears on the bank statement.
custom_fields CustomFields
A user-defined dictionary that contains all the custom fields generated by applying specific rules and regular expressions to the extracted data.
The balance after any credit or debits have been applied from this transaction.
Possible values: non-empty
A credit card number associated with this transaction.
The amount credited from this transaction.
The date of the transaction in ISO 8601 format.
The date the transaction was posted in ISO 8601 format.
The amount debited from this transaction.
Possible values: non-empty
The description of the transaction.
Possible values: non-empty
The category of the transaction.
Possible values: non-empty
The vendor of the transaction.
Possible values: non-empty
The unique identifier of the transaction.
Possible values: non-empty
The OCR text extracted from the transaction.
Possible values: non-empty
The account number associated with the bank statement.
The balance at the beginning of the statement period.
The balance at the end of the statement period.
Possible values: non-empty
The International Bank Account Number
Possible values: non-empty
The address of the account holder.
Possible values: non-empty
The name of the account holder.
Possible values: non-empty
The account number associated with the bank statement.
Possible values: non-empty
The type of account associated with the bank statement.
Possible values: non-empty
The unique identifier for businesses used for tax purposes
Possible values: non-empty
The address of the bank.
Possible values: non-empty
The name of the bank.
Possible values: non-empty
The unique identifier assigned to a bank for tax purposes
Possible values: non-empty
The URL for the website of the bank.
The balance at the beginning of the statement period.
Possible values: non-empty
The currency code associated with the bank statement.
The date the payment is due in ISO 8601 format.
The balance at the end of the statement period.
Possible values: non-empty
The International Bank Account Number
The minimum amount due for the statement period.
The end date of the statement period in ISO 8601 format.
The start date of the statement period in ISO 8601 format.
Possible values: non-empty
The routing number associated with the bank statement.
The date of the statement in ISO 8601 format.
Possible values: non-empty
The unique identifier associated with the bank statement.
Possible values: non-empty
The unique identifier for a bank used in international transactions
summaries object[]requireddeprecated
Deprecated. Please use accounts.summaries instead. Summary information
Possible values: non-empty
The title or label that captures the main information
The overall amount calculated from all transactions
Possible values: non-empty
The account number associated with the bank statement.
Possible values: non-empty
The routing number associated with the bank statement.
Possible values: non-empty
A custom identification value. Use this if you would like to assign your own ID to documents. This parameter is useful when mapping this document to a service or resource outside Veryfi.
meta object
Possible values: non-empty
A custom identification value. Use this if you would like to assign your own ID to documents. This parameter is useful when mapping this document to a service or resource outside Veryfi.
pages object[]
Possible values: <= 1
The average OCR score of the page.
The width of the page.
The height of the page.
Possible values: non-empty
Default value: ``
Tags associated with the document.
Possible values: <= 1
The average OCR score of the whole document.
Possible values: non-empty
The version of the model used to process the document.
Possible values: non-empty
The original file name of the uploaded document.
Possible values: [api.email, api.web, api, browser_extension, lens.bill, lens.invoice, lens.long_receipt, lens.other, lens.receipt, lens.web, lens]
Default value: api
The source of the document's submission for processing.
duplicates object[]
An array of duplicate documents found in the system.
The id of the duplicate document.
Possible values: non-empty and <= 2083 characters
The url of the duplicate document.
Possible values: <= 1
How close is the match
fraud BankStatementsFraud
An object that contains additional information to identify fraud.
Possible values: <= 1
Confidence of Fraud Detector in it's prediction
Possible values: [green, yellow, red]
Color from Fraud Detector: green means legitimate, yellow means review needed and red means fraud
Possible values: [LCD photo, not a document, duplicate, digital background, screenshot, digital tampering, fraudulent pdf, fraudulent image, high velocity, critical velocity]
List of attributions which marked the document as fraud
details BankStatementFraudDetails
Strictly typed descriptions and reasons keyed by fraud type.
lcd_photo _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
not_a_document _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
duplicate _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
digital_background _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
screenshot _ScreenshotDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
reasons object[]
Possible values: [mobile_screenshot, other_screenshot]
Possible values: non-empty
digital_tampering _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
fraudulent_pdf _FraudulentPDFDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
reasons object[]
Possible values: [text_overlay, font_mismatch, modified, date_inversion, producer_content_mismatch, writer_mismatch, null_metadata, xmp_info_mismatch, future_metadata_date, raster_profile_mismatch, uncompressed_raster, double_jpeg, scrubbed_provenance]
Possible values: non-empty
Possible values: non-empty
Possible values: [creator, producer]
Possible values: non-empty
Possible values: non-empty
Possible values: > 0
1-based number of the first page written after the pages surrounding it. Later pages may belong to the same block.
fraudulent_image _FraudulentImageDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
reasons object[]
Possible values: [double_jpeg, scrubbed_provenance, unprofiled_png, unprofiled_webp, synthetic_raster]
Possible values: non-empty
Possible values: non-empty
Possible values: [icc_profile, exif:Software, exif:ProcessingSoftware, xmp:CreatorTool, png:Software, jpeg:comment]
Possible values: non-empty
Possible values: [generator_signature, generator_writer_profile]
Possible values: non-empty
Possible values: non-empty
high_velocity _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
critical_velocity _FraudTypeDetails
Possible values: non-empty
Human-readable explanation of the fired fraud type.
submissions object
The amount of submissions from specific device id, used in velocity fraud detection.
Possible values: [text_similarity, field_matching]
Method used to detect duplicates for fraud
duplicates object[]
An array of duplicate documents found for purposes of fraud detection
The id of the duplicate document.
Possible values: non-empty and <= 2083 characters
The url of the duplicate document.
fraudulent_pdf FraudulentPDFResult
Results of the pdf analysis.
Possible values: non-empty
What incremental PDF revisions changed.
incremental_edits object[]
Which word an incremental PDF revision replaced, and on which page.
Possible values: > 0
1-based number of the page the edit landed on.
Possible values: non-empty
The replaced word, or null when the word was inserted.
Possible values: non-empty
The replacing word, or null when the word was deleted.
Possible values: non-empty
Possible values: > 0
1-based number of the first page written after the pages surrounding it. Later pages may belong to the same block.
Score for a PDF an automated browser rendered from a desktop operating system. It is reported here but scored against the generated document signal, so it does not move the fraudulent pdf score.
Possible values: non-empty
Operating system named by the headless browser that rendered the PDF. Reported for every headless render, including the server platforms that do not score.
Score for a raster that was JPEG-compressed more than once. An original capture is compressed exactly once, so a second compression means the file is not the first-generation scan it presents itself as. Disabled by default, and reported as 0.0 until it is enabled for your account.
Possible values: > 0
JPEG quality of the first compression, recovered from the coefficient histograms. Evidence for a reviewer rather than a score, and absent when nothing coherent was recovered.
Score for a raster whose encoder signature carries no provenance: grayscale content in a subsampled colour JPEG, no EXIF, a placeholder JFIF density and a stock quantization table. Reaches the same conclusion as double_jpeg for a re-encode too coarse to leave a recoverable first compression. Disabled by default, and reported as 0.0 until it is enabled for your account.
fraudulent_image FraudulentImageResult
Results of the image forensics analysis, for a document that arrived as a picture rather than a PDF. Absent when the signal is switched off for your account.
Score for a document image that was JPEG-compressed more than once. An original capture is compressed exactly once, so a second compression means the file is not the first-generation photo it presents itself as. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a JPEG-family raster can answer it, so a PNG reports null. Disabled by default.
Score for a document image whose encoder signature carries no provenance: grayscale content in a subsampled colour JPEG, no EXIF, a placeholder JFIF density and a stock quantization table. Reaches the same conclusion as double_jpeg for a re-encode too coarse to leave a recoverable first compression. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Disabled by default.
Possible values: > 0
JPEG quality of the first compression, recovered from the coefficient histograms. Evidence for a reviewer rather than a score, and absent when nothing coherent was recovered.
Score for an image that names an image editor in its own metadata. Read as a reason to review rather than to reject: an editor naming itself on a photograph may be an honest user cropping a receipt before uploading it, which is why it scores lower here than the same tool named on a PDF. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Formats that carry neither an ICC profile nor EXIF, which is BMP and GIF, can never answer it.
Possible values: [icc_profile, exif:Software, exif:ProcessingSoftware, xmp:CreatorTool, png:Software, jpeg:comment]
Which metadata surface named the editor. Exactly one is ever reported, the first that matched.
Possible values: non-empty
The literal string the file declared, which is what a reviewer acts on rather than the score.
Score for an image a server-side rendering library declares it wrote, such as libgd naming itself in a JPEG comment segment. Scores far above the editor signature because the finding is different: an editor names itself on a photograph somebody cropped, while a rendering library names itself on a page nothing ever photographed. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a format that can carry metadata can answer it.
Possible values: [icc_profile, exif:Software, exif:ProcessingSoftware, xmp:CreatorTool, png:Software, jpeg:comment]
Which metadata surface named the library. The same surfaces the editor signature reads, and as there exactly one is ever reported.
Possible values: non-empty
The literal string the file declared, which is what a reviewer acts on.
The same finding on a format where nothing is declared: a PNG whose chunk sizing, compression level and absence of provenance are libpng's own defaults, which is how the same libraries emit PNG. Reported rather than inferred from the file naming itself, because on this format it never does. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a PNG can answer it.
Possible values: non-empty
The writer profile that matched, naming the evidence rather than restating the score. Deliberately has no _field companion, unlike the two signatures above: the profile is read off the container itself rather than off one named surface.
Score for a raster too flat to have come off a sensor: an indexed palette, very few distinct colours, or no chroma variation at all. Ships at 0.0 and so contributes nothing to score — it is reported for the rate while a screenshot of a genuine receipt is still indistinguishable from a drawn one. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire.
Possible values: non-empty
Which of those measurements matched, sorted. Evidence for a reviewer rather than a score, and the set grows as the detector learns to name more of them, so read it as a list of strings rather than a fixed vocabulary.
Score for a PNG carrying no colour or provenance chunk at all, as a rendering library writes and a camera pipeline does not. Ships at 0.0 for the same reason as synthetic_raster: screenshot tools and our own rasterizer write the identical bare file. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a PNG can answer it.
The same check on WebP, and ships at 0.0 for a different reason than the two above: too little genuine WebP has been seen to say what the rate would be. Null when no file was in a position to answer, which is not the same as a 0.0 saying one was examined and this did not fire. Only a WebP can answer it.
Possible values: [original, render]
Which bytes were read. original is the upload as submitted and is the only value for which the compression scores above can be reported. render is a page render taken after cropping, rotation and format conversion, which destroys compression evidence and can manufacture it, so those two report null against it. Null means nothing was selected to read at all; a file that was selected but could not be fetched or opened reports a source together with a non-zero unreadable_source.
Possible values: non-empty
Formats of the files that were read, as the image library names them. Note HEIF rather than HEIC, and MPO for a .jpg holding multiple frames, as a phone HDR capture does.
How many selected files could not be opened. Can exceed one on a multi-file upload. A non-zero value here alongside a set source is the expected shape when an original has already been deleted, rather than a fault.
Highest of the scores above, and what the fraudulent image signal contributes. Follows the same rule they do: null when not one of them was in a position to answer, and 0.0 when at least one was and nothing fired. So a page render scores a real 0.0, the compression pair having declined but the editor signature having run. The checks that ship at 0.0 are in the max like any other, which is why they can report a hit without moving it.
pages object[]
An array containing fraud info about each extracted page
is_lcd ClassBoolField
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
The extracted value.
flags object[]
List of flags which marked the document as fraud
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: non-empty
ai_generated ClassBoolField
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
The extracted value.
handwriting object[]
An array containing handwriting info about each extracted page
Possible values: >= 8, <= 8
Bounding region of the artifact in [x1,y1,x2,y2,x3,y3,x4,y4] format
Possible values: non-empty
Type of the artifact
digital_tampering object[]
An array containing digital tampering info about each extracted page
Possible values: >= 8, <= 8
Bounding region of the artifact in [x1,y1,x2,y2,x3,y3,x4,y4] format
Possible values: non-empty
Type of the artifact
Possible values: non-empty and <= 2083 characters
A signed URL to access the auto-generated PDF created from the submitted document. This URL expires 15 minutes after the response object is returned and is resigned during every GET request.
The unique number created to identify the document.
Possible values: non-empty and <= 2083 characters
A signed URL to access the auto-generated thumbnail created for the submitted document. This URL expires 15 minutes after the response object is returned and is resigned during every GET request.
accounts object[]required
summaries object[]required
Summary information
name object
The title or label that captures the main information
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
total object
The overall amount calculated from all transactions
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
transactions object[]required
A list of transactions associated with the account.
The value indicating the position of where the transaction appears on the bank statement.
custom_fields CustomFields
A user-defined dictionary that contains all the custom fields generated by applying specific rules and regular expressions to the extracted data.
balance DetailedFloatField
The balance after any credit or debits have been applied from this transaction.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
card_number object
A credit card number associated with this transaction.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
credit_amount EnrichedDetailedNonNegativeFloatField
The amount credited from this transaction.
Whether the value is enriched.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
date object
The date of the transaction in ISO 8601 format.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
posted_date DetailedDateField
The date the transaction was posted in ISO 8601 format.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
debit_amount EnrichedDetailedFloatField
The amount debited from this transaction.
Whether the value is enriched.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
description object
The description of the transaction.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
category object
The category of the transaction.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
vendor object
The vendor of the transaction.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
transaction_id object
The unique identifier of the transaction.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
text object
The OCR text extracted from the transaction.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
number objectrequired
The account number associated with the bank statement.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
beginning_balance DetailedFloatFieldrequired
The balance at the beginning of the statement period.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
ending_balance DetailedFloatFieldrequired
The balance at the end of the statement period.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
iban_number object
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
account_holder_address objectrequired
The address of the account holder.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
account_holder_name objectrequired
The name of the account holder.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
account_number objectrequired
The account number associated with the bank statement.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
account_numbers object[]required
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
account_type objectrequired
The type of account associated with the bank statement.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
account_vat_number objectrequired
The unique identifier for businesses used for tax purposes
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
bank_address objectrequired
The address of the bank.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
bank_name objectrequired
The name of the bank.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
bank_vat_number objectrequired
The unique identifier assigned to a bank for tax purposes
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
bank_website objectrequired
The URL for the website of the bank.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
beginning_balance objectrequired
The balance at the beginning of the statement period.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
currency_code objectrequired
The currency code associated with the bank statement.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
due_date objectrequired
The date the payment is due in ISO 8601 format.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
ending_balance objectrequired
The balance at the end of the statement period.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
iban_number object
The International Bank Account Number
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
minimum_due objectrequired
The minimum amount due for the statement period.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
period_end_date objectrequired
The end date of the statement period in ISO 8601 format.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
period_start_date objectrequired
The start date of the statement period in ISO 8601 format.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
routing_number objectrequired
The routing number associated with the bank statement.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
routing_numbers object[]required
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
statement_date objectrequired
The date of the statement in ISO 8601 format.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
statement_number objectrequired
The unique identifier associated with the bank statement.
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
swift objectrequired
The unique identifier for a bank used in international transactions
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
summaries object[]requireddeprecated
Deprecated. Please use accounts.summaries instead. Summary information
name object
The title or label that captures the main information
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
Possible values: non-empty
The extracted value. Nullable: null means nothing was extracted for this field. A parent object listed as required (for example document_type) only guarantees the object is present — value can still be null.
total object
The overall amount calculated from all transactions
Possible values: <= 1
The score shows how confident the model is that the predicted value belongs to the field. See confidence scores explained for more information.
Possible values: >= 8, <= 8
An array containing (x,y) coordinates in the format [x1,y1,x2,y2,x3,y3,x4,y4]` for skewed images and handwritten fields. The bounding region is more precise than bounding box, otherwise it's the same.
bounding_box object[]
Possible values: >= 5, <= 5
An array containing relative coordinates in the format [page_number,x1,y1,x2,y2] for the extracted field from img_url before any rotation.
number
integer
Possible values: <= 1
The score which shows how confident the model in recognizing value symbols. See confidence scores explained for more information.
Possible values: [0, 90, 180, 270]
The angle of rotation of the document in degrees.
{}