**Cost**: 13 credits per processed page.

Extract a structured, editable design tree out of one or more pages of a PDF. Each page becomes a fully-populated **`VisualElement`** tree — the same node shape as `image_to_design` — augmented with the vector / typography / transform fields that only PDFs carry.

## What this API is for

PDFs are vector documents: every paragraph, every shape, every gradient is stored as math, not pixels. This endpoint preserves that math instead of rasterizing it, so the tree you get back can be re-rendered, re-flowed, re-styled, or translated into another format without quality loss.

Typical use cases:
- **PDF → editable design**: open up an exported brochure / spec sheet / one-pager and turn it back into a Figma-like tree.
- **PDF → PPT / HTML / code**: walk the tree to emit slides, web pages, or component code. (Our own `pdf_to_ppt` builds on this.)
- **Content extraction**: pull every text run with its true font, size, and color, plus every vector shape with its actual path data.
- **Asset salvage**: re-export embedded images and vector logos at original resolution.

Differences vs. `image_to_design`: input is a **PDF URL or PDF file**, output is **one tree per requested page** (not a single tree), and the per-node fields tagged "PDF only" in the shared schema **are** populated here (vector paths, character-level styling, transform matrices, etc.).

## Sending the request

This endpoint supports two input modes:

1. **JSON URL mode**: send `application/json` with `pdf_url` and `page_no` when the PDF is already publicly reachable.
2. **Direct upload mode**: send `multipart/form-data` with a local PDF file in the `pdf_file` field and one or more `page_no` fields.

JSON example:

```bash
curl 'https://openapi.codia.ai/v2/open/pdf_to_design' \
  -H 'Authorization: Bearer {codia_api_key}' \
  -H 'Content-Type: application/json' \
  --data '{ "pdf_url": "https://example.com/file.pdf", "page_no": [0, 2, 5] }'
```

Direct upload example:

```bash
curl 'https://openapi.codia.ai/v2/open/pdf_to_design' \
  -H 'Authorization: Bearer {codia_api_key}' \
  -F 'pdf_file=@./file.pdf' \
  -F 'page_no=0' \
  -F 'page_no=2'
```

- **`pdf_url`** *(JSON mode, required)* — public URL of the PDF. Encrypted / password-protected PDFs are rejected.
- **`pdf_file`** *(upload mode, required, binary)* — the PDF itself. Stream the file directly; do not base64-encode it.
- **`page_no`** *(required)* — **0-indexed** page numbers. JSON mode uses an integer array; multipart mode repeats the `page_no` field once per page. Page count and file-size caps follow the global Codia API limits.

## Response shape

Standard Codia envelope; on success `data` carries:

| Field | What it is | How to use it |
|---|---|---|
| `data.configuration` | Same `Configuration` object as `image_to_design` — `baseWidth`, `measurementUnit` (`px` \| `pt`), `scalingFactor`. | Apply `scalingFactor` to every numeric value when rendering. `baseWidth` is the canvas width the trees were measured against (often page width in points or px, e.g. 1940). |
| `data.size` | Page dimensions in document units: `{ width, height }`. | Use to set up artboard / viewport per page. All pages in a single PDF share this size. |
| `data.pages[]` | One entry per requested page. Each entry is `{ visualElement: <Body root> }`. | Iterate pages in order; recurse into each `visualElement.childElements` to reach every region. |

## Walking a page tree

Each `data.pages[i].visualElement` is a `Body`-typed root with the same `VisualElement` shape used elsewhere in this API (see `image_to_design` for the full field guide). The headline pieces:

- **`elementType`** — `Body` at the root, then `Layer` / `Group` / `Text` / `Image` / `Vector` / `Component` inside. PDFs use `Vector` heavily for shapes, gradients, and logos.
- **`childElements`** — nested nodes; recurse depth-first.
- **`layoutConfig` / `styleConfig` / `processingMeta`** — same semantics as `image_to_design`.
- **`contentData`** — for `Text` reads `textValue`; for `Image` reads `imageSource` and/or the base64 in `imageData`; for `Vector` reads `vectorData` (see below).

### PDF-specific fields you can actually use

Fields the shared schema tags "PDF only" — these are real, populated, and useful on this endpoint:

- **`boundingBox`** *(`[x, y, width, height]`)* — exact node geometry in document coordinates. The most reliable positioning source for PDFs.
- **`displayOrder`** — z-order within the page; lower = drawn first.
- **`styleConfig.transformMatrix` / `extendedMatrix`** — affine transform applied to the node. `transformMatrix` is the 6-element PDF form `[a b c d e f]`; `extendedMatrix` is the CSS-compatible string. Apply this to handle rotated text, sheared images, etc.
- **`styleConfig.characterData`** — per-character styling keyed by character index. Use this when a single text node mixes fonts, sizes, or colors (very common in PDFs).
- **`styleConfig.cornerRadius`** — scalar fallback when the four-corner array isn't applicable.
- **`styleConfig.isRightToLeft`** — direction flag for Arabic / Hebrew text.
- **`contentData.vectorData`** — the actual vector path. Contains `pathItems[]` (each item is `{ pathType: "l" | "c", coordinates[] }` — `l` = line, `c` = cubic curve), plus `fillColor`, `strokeColor`, `strokeWidth`, `fillOpacity`, `strokeOpacity`, `lineCapStyle`, `lineJoinStyle`, `dashPattern`, `evenOddRule`, `isClosedPath`. Emit SVG `<path>` directly, or rebuild in your own renderer.
- **`contentData.imageData`** — base64-encoded image bytes for embedded raster images (alongside or instead of `imageSource`).
- **`contentData.svgTemplate`** — pre-rendered SVG markup template for compound vector content.
- **`contentData.maskingMode` / `blendingType`** — PDF graphics-state knobs (clip / soft-mask, blend mode). Map to CSS `mix-blend-mode` / SVG `<mask>` as appropriate.

### A minimal traversal

```ts
for (const page of response.data.pages) {
  console.log('page', page.visualElement.elementName, response.data.size);
  walk(page.visualElement);
}

function walk(node, depth = 0) {
  const pad = '  '.repeat(depth);
  if (node.elementType === 'Text') console.log(pad, 'TXT', node.contentData?.textValue);
  if (node.elementType === 'Image') console.log(pad, 'IMG', node.contentData?.imageSource);
  if (node.elementType === 'Vector') console.log(pad, 'VEC', node.contentData?.vectorData?.pathItems?.length, 'segs');
  node.childElements?.forEach(c => walk(c, depth + 1));
}
```

## Inputs, limits, billing

- **Input** — either JSON with `pdf_url` or `multipart/form-data` with `pdf_file`. Encrypted / password-protected PDFs are rejected.
- **Page selection** — `page_no` uses **0-based** indices. JSON mode sends an integer array; multipart mode repeats the `page_no` field once per page. Out-of-range indices are silently dropped from the response.
- **Billing** — consumes the unified Codia Open API balance (`GET /v2/open/credits`). 13 credits per processed page.
- **Latency** — synchronous. Throughput is per-page; a long document may take many seconds. For very long PDFs, batch the request by `page_no` ranges and parallelize.
- **Determinism** — same PDF + same `page_no` yields the same structural tree, but `elementId`s are regenerated each call; don't rely on cross-call ID stability.

## Errors

Uses the standard envelope with non-zero `code` and a descriptive `message`. HTTP status codes:
- `400` — bad input (missing `pdf_url` / `pdf_file`, malformed `page_no`, corrupt / encrypted PDF).
- `401` / `403` — missing / wrong / unauthorized API key.
- `402` — the subscription is inactive or credits are exhausted. See the global Errors section for recovery steps.
- `429` — rate-limited; back off and retry.
- `500` — internal error; safe to retry once.

### `measurementUnit`

`pdf_to_design` returns whichever unit the source PDF used. Most editorial / business PDFs report `pt` (PostScript points: 72 = 1 inch). PDFs exported from design tools at a specific pixel target may report `px`. Read this field per response and convert if your downstream renderer assumes one unit.

> **Sync vs. async**
> This synchronous endpoint returns the result inline — fine for short PDFs. For documents with more than ~5 pages, prefer `POST /v2/open/tasks` with `operation=pdf_to_design` to avoid HTTP client timeouts, get webhook delivery, and support cancellation. Inputs and credit cost are identical to this v2 endpoint.

사용자가 지정하는 선택적 키(최대 255자). 설정하면 성공한(2xx) 응답이 7일 동안 캐시됩니다. 같은 키와 동일한 요청 본문으로 재시도하면 캐시된 응답이 반환되며 크레딧이 추가로 차감되지 않습니다. 같은 키로 다른 본문을 보내면 409가 반환됩니다. 네트워크 장애 시 POST를 안전하게 재시도할 수 있도록 하려면 이 키를 사용합니다. multipart/form-data 업로드는 폼 필드와 파일 내용으로 요청을 비교하므로, 클라이언트가 새 multipart boundary를 선택하더라도 재시도한 업로드에도 캐시된 응답이 반환됩니다.

Idempotency-Key

처리할 PDF의 공개 URL. Codia가 직접 가져올 수 있어야 합니다.

pdf_url

처리할 페이지 번호(0부터 시작). 예: [0, 2, 5].

page_no

처리할 PDF 파일. multipart 바이너리 파트로 전송합니다.

pdf_file

처리할 페이지(0부터 시작)마다 multipart page_no 필드를 한 번씩 반복합니다.

변환에 성공했습니다. `data.pages`는 배열이며, 처리된 페이지마다 항목이 하나씩 있습니다. 각 항목에는 `image_to_design`과 동일한 `VisualElement` 구조의 `visualElement` 트리가 포함됩니다(작업 설명에 기재된 PDF 전용 필드 추가). `data.size`에는 페이지 크기가, `data.configuration`에는 캔버스 스케일과 단위가 담깁니다.

code

message

디자인 출력의 전역 렌더링 설정

configuration

pages

size

data

요청 본문의 형식이 잘못되었거나 필수 매개변수가 누락되었습니다

API 키가 없거나, 유효하지 않거나, 폐기되었습니다

계정 잔액이 부족하거나 구독이 만료되었습니다

계정에 이 엔드포인트에 대한 액세스 권한이 없습니다

Idempotency-Key가 다른 요청 본문과 함께 재사용되었습니다.

속도 제한에 도달했습니다. `Retry-After` 헤더에 표시된 초만큼 기다린 후 재시도하십시오.

서버에서 예기치 않은 오류가 발생했습니다

Vector

VELORA SKINCARE · SELL SHEET 2026

48 units

$49.00

MSRP

$17.50

WHOLESALE

24 months unopened

Shelf life

24 units · 3.1 kg

Case pack

Amber glass bottle, glass dropper

Packaging

15% vitamin C · hyaluronic acid · niacin

15% vitamin C · hyaluronic acid · niacinamide

Key actives

VLR-GS-030

Brightening vitamin C serum with hyaluro

Brightening vitamin C serum with hyaluronic acid. Fragrance-free,
vegan, dermatologist tested.

Glow Serum 30 ml

V E L O R A · W H O L E S A L E

Image

PDF 페이지를 래스터화하지 않고 원본 벡터, 글꼴, 색상을 유지한 편집 가능한 디자인 트리로 변환합니다.

PDF를 디자인으로

입력

출력

관련 API