generated: '2026-08-29'
method: derived
source: >-
  openapi/_original/abbyy-vantage-processing-openapi.json ($ref graph and id-reference fields), enriched
  from https://docs.abbyy.com/vantage/developer/output/json/json-schema (fetched 2026-08-29)
provider: ABBYY
providerId: abbyy
api: ABBYY Vantage Processing REST API
summary: >-
  Two loosely coupled halves. The PROCESSING half is a strict containment tree — Transaction owns
  Documents, Documents own Files, and everything hangs off one transactionId — governed by a Skill chosen
  at creation time. The CATALOG half is a flat reference-data store (Catalog, Column, Record) that skills
  read from during extraction but that the transaction graph never links to. The extraction RESULT is a
  third model that the OpenAPI barely describes: TransactionDto returns result FILES, and the actual
  extracted-data shape (ExtractedData, DocumentDefinition, RootObject, ExtractedObject, confidence,
  RawValue) lives only in ABBYY's separately published JSON Schema documentation.
entities:
  - name: Transaction
    schema: TransactionDto
    primary_key: transactionId
    description: A processing job. Created against exactly one Skill, then started, polled and downloaded from.
    lifecycle: created -> (files added) -> started -> in-progress -> completed | failed -> deleted
    status_enum: TransactionStatus
    relationships:
      - type: has_many
        target: Document
        via: documents
      - type: has_many
        target: File
        via: sourceFiles
      - type: has_many
        target: RegistrationParameter
        via: registrationParameters
      - type: has_many
        target: SkillParameter
        via: skillParameters
      - type: belongs_to
        target: Skill
        via: skillId
  - name: Document
    schema: ExtractedDocumentDto / TransactionDocumentResult
    primary_key: documentId
    description: A logical document inside a transaction; the unit that classification and extraction act on.
    relationships:
      - type: belongs_to
        target: Transaction
        via: transactionId
      - type: has_many
        target: File
        via: sourceFiles
      - type: has_many
        target: ResultFile
        via: resultFiles
      - type: has_one
        target: Classification
        via: classification
      - type: has_many
        target: BusinessRuleError
        via: businessRulesErrors
  - name: File
    schema: FileDto / TransactionFileMetadataDto
    primary_key: fileId
    description: An uploaded source image or PDF. Names must be unique within a transaction (409 otherwise).
    relationships:
      - type: belongs_to
        target: Transaction
        via: transactionId
      - type: has_one
        target: ImageProcessingOptions
        via: imageProcessingOptions
      - type: has_many
        target: FileRegistrationParameter
        via: registrationParameters
  - name: ResultFile
    schema: ResultFileDto
    primary_key: fileId
    description: A generated output artifact (JSON, XML, searchable PDF and so on) keyed by ResultFileType.
    relationships:
      - type: belongs_to
        target: Document
        via: documentId
  - name: Skill
    schema: SkillDto / SkillListItemDto
    primary_key: skillId
    description: >-
      The trained AI model that drives a transaction. Types are Classification, Document, OCR and Process.
      A Process skill composes other skills.
    type_enum: SkillType
    relationships:
      - type: has_many
        target: Field
        via: fields
      - type: has_many
        target: SkillClass
        via: classes
      - type: has_many
        target: Skill
        via: skills
        note: self-referential — InnerSkillDto, how a Process skill chains child skills
  - name: Field
    schema: FieldDto
    description: An extractable field definition. Self-referential via children, so tables and nested records recurse.
    type_enum: FieldType
    relationships:
      - type: has_many
        target: Field
        via: children
      - type: has_one
        target: TextFieldFormat
        via: textFormat
  - name: Classification
    schema: ClassificationDto
    description: The class assigned to a document plus per-class confidence scores.
    relationships:
      - type: has_one
        target: SkillClass
        via: resultClass
      - type: has_many
        target: ClassificationConfidence
        via: classConfidences
      - type: belongs_to
        target: SkillClass
        via: classId
  - name: BusinessRuleError
    schema: DocumentBusinessRulesErrorDto
    description: >-
      A validation-rule violation surfaced INSIDE a successful result, not as an HTTP error.
      errorType is Normalization, Condition or Unknown.
    relationships:
      - type: belongs_to
        target: Field
        via: currentFieldId
      - type: belongs_to
        target: Rule
        via: ruleId
        note: ruleId is opaque — no operation resolves it to a rule definition.
  - name: Catalog
    schema: CatalogDto / CatalogListItemDto
    primary_key: catalogId
    description: A reference-data table skills look up against during extraction (vendors, GL codes, and so on).
    relationships:
      - type: has_many
        target: Column
        via: columns
      - type: has_many
        target: Record
        via: records
  - name: Column
    schema: ColumnDto
    type_enum: ColumnType
    relationships:
      - type: belongs_to
        target: Catalog
        via: catalogId
  - name: Record
    schema: RecordDto / RecordModel
    primary_key: id
    description: >-
      One catalog row. The id is CALLER-SUPPLIED, which makes AddRecords upsert-like and DeleteRecords
      exactly reversible — the only place in the API a client controls a primary key.
    relationships:
      - type: belongs_to
        target: Catalog
        via: catalogId
  - name: ReindexOperation
    schema: ReindexOperationResult
    primary_key: reindexOperationId
    description: >-
      Async job that republishes a catalog to the search index. Skills do not see catalog changes until it
      completes. State enum ReindexState.
    relationships:
      - type: belongs_to
        target: Catalog
        via: catalogId
  - name: RegistrationParameter
    schema: TransactionRegistrationParameter / TransactionFileRegistrationParameter
    description: Caller-supplied key-value metadata attached to a transaction or a file, readable back and reportable.
  - name: SkillParameter
    schema: TransactionSkillParameter / TransactionSkillParameterDto
    type_enum: SkillParameterType
    description: Per-transaction overrides for the chosen skill.
  - name: Invite
    schema: CreateAdminInviteData
    description: Administrative user invitation. Write-only — no read, list or revoke operation is exposed.
result_payload_model:
  described_in_openapi: false
  documented_at: https://docs.abbyy.com/vantage/developer/output/json/json-schema
  entities:
    - ExtractedDataTransaction
    - ExtractedDataDocument
    - ExtractedData
    - DocumentDefinition
    - RootObject
    - ExtractedObject
    - FileModel
    - Page
  note: >-
    The single most valuable data structure ABBYY produces — the extracted field values with confidence
    scores, RawValue and source coordinates — is NOT in the OpenAPI. The API hands back a downloadable
    result file and the shape is specified in prose documentation only. An agent generating a client from
    the spec gets no type for the answer it came for.
id_conventions:
  format: opaque string
  prefixes: none
  note: >-
    No typed id prefixes (no txn_, doc_, skl_). Ids are bare strings, so a misrouted identifier cannot be
    caught client-side before the call.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com