Auto Reader Reduce API

openapi: 3.2.0
info:
  title: Auto-Reader OCR + Translate Reduce API
  description: Multilingual OCR (Arabic-first, manga-aware), translation, and vision escalation.
  version: 1.0.0
servers:
- url: https://api.auto-reader.com
security:
- bearerAuth: []
tags:
- name: Reduce
paths:
  /v1/reduce:
    post:
      summary: Reduce Sync
      description: 'SYNC size-reducer: a scanned PDF (or images) in -> a much smaller,

        SEARCHABLE PDF out, returned directly. Small docs only (larger -> async

        /v1/book). Billed 1 credit/page.


        output="same" makes the result follow the input: a SINGLE image in gives a

        recompressed image back (JPEG, or PNG when the source needs alpha) with no

        OCR and no text layer — because "shrink this photo" shouldn''t hand back a

        PDF. Multiple images, or any PDF, always produce a PDF; output="pdf" (the

        default) always does.


        grayscale=True (THE DEFAULT) does NOT convert the document to greyscale.

        It drops the chroma planes only where a page — or, on a PDF whose own JPEG

        streams are rewritten, an individual strip — carries no real colour, judged

        by connected coloured AREA rather than an average. A stamp, a signature, a

        logo or a coloured rule keeps its colour; blank margins and black text lose

        planes that held nothing. Worth about -6% on a text scan for nothing

        visible. Send grayscale=false to keep every plane regardless.


        vectorize=True additionally erases the original raster text (LaMa inpaint)

        and redraws it as real vector text at the same position — much smaller for

        text-heavy pages and crisp at any zoom. Because a misread becomes a

        permanent, visible error in that mode (unlike the invisible search layer),

        any line the OCR wasn''t confident about gets one more independent

        verification pass (reverify_lines). A line STILL uncertain after that is

        neither erased nor redrawn -- its original pixels stay -- unless the vision

        model (when the service has it) confirms the reading on a crop; each such

        vision call is billed +VECTORIZE_VISION_CREDITS (1 credit) on top of the

        vectorize surcharge, capped per request (VECTORIZE_VISION_MAX_CALLS, 8) and by the caller''s balance. Response headers

        X-Vision-Calls / X-Raster-Lines report it. Only lines that were redrawn

        yet remain uncertain get a PDF comment.


        quality=sharp (PDF output): for artwork / infographic pages with thin

        light text. The page background is downscaled and compressed hard, and

        every detected text box is pasted back at full resolution (PNG, pixel-

        identical, when that is cheap; 4:4:4 JPEG q90 otherwise). Works with

        ocr=false (detection only, compress-only billing) and ocr=true; cannot be

        combined with vectorize. On the lone-image output=same path, high /

        veryhigh / sharp encode without chroma subsampling (and a WebP source may

        come back as WebP when that is smaller). sharp is never larger than the

        plain presets: on a page where its lossless crops cost more than the same

        page as veryhigh (plain light documents), the veryhigh PDF -- or `high`

        when even that is not below the source -- is returned instead and the

        response says so in X-Sharp-Fallback (veryhigh | high; absent when sharp

        itself was the smallest).


        vectorize fonts: Latin/Cyrillic = Noto Sans, Arabic = Noto Naskh (shaped),

        Hindi = Noto Sans Devanagari (shaped), Japanese/Korean/Chinese = the

        matching Noto Sans CJK face. Arabic lines also carry an /ActualText span

        so copy/search return logical text. Languages with an OCR pack: ar fa ur en

        fr es de it pt tr vi zh ja ko ru hi (the pack for hi is `devanagari`, tr/vi

        use `latin`).'
      operationId: reduce_sync_v1_reduce_post
      parameters:
      - name: x-credits-available
        in: header
        required: false
        schema:
          type: string
          default: ''
          title: X-Credits-Available
      requestBody:
        required: true
        content:
          multipart/form-data:
            schema:
              $ref: '#/components/schemas/Body_reduce_sync_v1_reduce_post'
      responses:
        '200':
          description: Successful Response
          content:
            application/json:
              schema: {}
        '422':
          description: Validation Error
          content:
            application/json:
              schema:
                $ref: '#/components/schemas/HTTPValidationError'
      tags:
      - Reduce
components:
  schemas:
    ValidationError:
      properties:
        loc:
          items:
            anyOf:
            - type: string
            - type: integer
          type: array
          title: Location
        msg:
          type: string
          title: Message
        type:
          type: string
          title: Error Type
        input:
          title: Input
        ctx:
          type: object
          title: Context
      type: object
      required:
      - loc
      - msg
      - type
      title: ValidationError
    HTTPValidationError:
      properties:
        detail:
          items:
            $ref: '#/components/schemas/ValidationError'
          type: array
          title: Detail
      type: object
      title: HTTPValidationError
    Body_reduce_sync_v1_reduce_post:
      properties:
        file:
          items:
            type: string
            contentMediaType: application/octet-stream
          type: array
          title: File
        quality:
          type: string
          title: Quality
          default: email
        grayscale:
          type: boolean
          title: Grayscale
          default: true
        flatten_bg:
          type: boolean
          title: Flatten Bg
          default: false
        lang:
          type: string
          title: Lang
          default: auto
        vectorize:
          type: boolean
          title: Vectorize
          default: false
        ocr:
          type: boolean
          title: Ocr
          default: true
        allow_partial:
          type: boolean
          title: Allow Partial
          default: false
        output:
          type: string
          title: Output
          default: pdf
        mode:
          type: string
          title: Mode
          default: document
      type: object
      required:
      - file
      title: Body_reduce_sync_v1_reduce_post
  securitySchemes:
    bearerAuth:
      type: http
      scheme: bearer
      description: API key from POST /signup (nsk_live_...)