generated: '2026-07-19'
method: derived
source: openapi/kita-underwriter-openapi.yml, openapi/kita-capture-openapi.yml
also_searched: https://www.kita.ai/documentation
summary: 'Two loosely-coupled domains. The Underwriter domain is rooted on Application — a
  loan file that owns documents, a credit picture, a memo and a borrower conversation. The
  Capture domain is rooted on Document — a single processed file that may belong to a batch or
  a folder and may be extracted through a custom schema. The two are related conceptually
  (Capture is the extraction engine beneath Underwriter) but expose separate identifiers and
  hosts, and no cross-domain identifier is documented.'
domains:
- name: underwriter
  base_url: https://underwriter.kita.ai/api/v1
- name: capture
  base_url: https://portal.usekita.com
entities:
- name: Application
  domain: underwriter
  description: A loan application / credit file.
  identifier:
    primary: id (UUID)
    alternate: app_id (human, e.g. APP-1234) — both are accepted in the path
    client_key: external_ref (caller's own loan or record ID; drives idempotent intake)
  status_field: status
  states: [submitted, awaiting_docs, underwriting, question_raised, memo_drafted, approved, declined]
  fields: [business_name, borrower_email, borrower_phone, loan_type, loan_amount,
    application_context, send_outreach, file_completeness, document_count]
  operations: [intakeApplication, createApplication, listApplications, getApplication,
    updateApplication, deleteApplication]
- name: Document
  domain: underwriter
  description: A borrower document attached to an application and run through extraction.
  identifier:
    primary: docId
  status_field: status
  states: [awaiting, processing, verified, low_confidence, missing]
  fields: [kita_raw, recommendations, inconsistencies, download_url]
  operations: [uploadDocuments, listDocuments, getDocument, deleteDocument]
- name: CreditPicture
  domain: underwriter
  description: The deterministic assessment computed by a versioned calculation engine. Three
    layers, each null until computed.
  identifier:
    primary: none — singleton per application
  layers:
  - name: spread
    description: Adjusted EBITDA and net income, DSCR, margins, debt-to-worth, current ratio,
      reserve months, LTV, a sensitivity table and per-metric provenance.
  - name: adjustments
    description: Normalization line items such as owner-comp add-backs and non-recurring
      items, each tagged ai or lo.
  - name: decision
    description: Policy-engine routing recommendation, auto-route flag, and per-rule pass/fail
      with observed vs expected.
  operations: [getCreditPicture]
- name: Memo
  domain: underwriter
  description: The cited credit memo, synthesized from documents and conversation.
  identifier:
    primary: none — singleton per application
  fields: [last_synthesis_at, sections]
  child_shape:
    name: MemoSection
    fields: [section_number, title, body, locked_by_lo, updated_at]
  freshness_fields: [synthesis_in_progress, is_stale, stale_reasons, new_doc_count,
    new_message_count, last_run_status, last_synthesis_at]
  operations: [synthesizeMemo, getMemo, getMemoStatus]
- name: ConversationMessage
  domain: underwriter
  description: A message in the borrower thread.
  ordering_field: display_order
  operations: [getConversation]
- name: Transcript
  domain: underwriter
  description: Plain-text transcript of a document's pages.
  fields: [pages, block_count, char_count, has_text, text]
  operations: [getDocumentTranscript, listApplicationTranscripts]
- name: CaptureDocument
  domain: capture
  description: A single file submitted for extraction.
  identifier:
    primary: documentId (integer)
  status_field: status
  states: [pending, processing, completed, failed]
  fields: [document_type, filename, processing_time_seconds, uploaded_at, metadata,
    extracted_data, fraud_detection, total_cost_usd, cost_report]
  operations: [processDocumentAsync, getDocumentResult, listDocuments, getDocumentSummary,
    exportDocument, customExportDocument, getDocumentJob]
- name: Batch
  domain: capture
  description: Up to 100 documents submitted from URLs in one request.
  identifier:
    primary: batch_id (string, prefix batch_)
  fields: [status, total_documents, completed, failed, progress_percent, status_url,
    total_cost_usd]
  operations: [createBatch, getBatch, getBatchResults]
- name: Folder
  domain: capture
  description: A container for grouping documents, e.g. by applicant.
  operations: [createFolder, listFolders, listFolderDocuments]
- name: ExtractionSchema
  domain: capture
  description: An enterprise-defined custom extraction shape that overrides the default
    vocabulary output.
  operations: [listSchemas, createSchema, getSchema, updateSchema, deleteSchema]
- name: Webhook
  domain: capture
  description: A registered outbound delivery endpoint with a signing secret.
  fields: [url, events, active]
  operations: [createWebhook, listWebhooks, getWebhook, updateWebhook, deleteWebhook,
    testWebhook, rotateWebhookSecret, getWebhookSecret, listWebhookDeliveries,
    getWebhookStats, getWebhookDeadLetterQueue]
- name: VerificationResult
  domain: capture
  description: Cross-document verification across 2–50 already-processed documents.
  fields: [document_count, documents, document_summary, cross_doc_score, signals,
    field_verifiability, corroboration]
  operations: [verifyDocuments, verifySingleDocument]
relationships:
- from: Application
  to: Document
  type: has_many
  via: path /applications/{id}/documents
- from: Document
  to: Application
  type: belongs_to
  via: path /applications/{id}/documents/{docId}
- from: Application
  to: CreditPicture
  type: has_one
  via: path /applications/{id}/credit
- from: Application
  to: Memo
  type: has_one
  via: path /applications/{id}/memo
- from: Memo
  to: MemoSection
  type: has_many
  via: sections
- from: Application
  to: ConversationMessage
  type: has_many
  via: path /applications/{id}/conversation
- from: Document
  to: Transcript
  type: has_one
  via: path /applications/{id}/documents/{docId}/transcript
- from: Batch
  to: CaptureDocument
  type: has_many
  via: documents[].document_id
- from: CaptureDocument
  to: Batch
  type: belongs_to
  via: batch_id
- from: Folder
  to: CaptureDocument
  type: has_many
  via: path /api/v1/folders/{id}/documents
- from: CaptureDocument
  to: ExtractionSchema
  type: belongs_to
  via: organization-configured schema override
- from: VerificationResult
  to: CaptureDocument
  type: has_many
  via: document_ids[]
- from: Webhook
  to: CaptureDocument
  type: observes
  via: job-completion delivery
- from: CaptureDocument
  to: FraudDetection
  type: has_one
  via: fraud_detection
id_prefixes:
- entity: Application
  prefix: APP-
  example: APP-1234
  note: human-readable app_id; the primary key is a UUID
- entity: Batch
  prefix: batch_
  example: batch_abc123
- entity: Capture API key
  prefix: kita_prod_
- entity: Underwriter API key
  prefix: kita_uw_
render: none — no subway/ diagram in this repo