Azure Content Understanding — Layout Analyzer

Extract structured content from documents using Azure's prebuilt-layout analyzer. Outputs Markdown and structured JSON with text, tables, figures, and document hierarchy.

Setup

Set environment variables:

export AZURE_CU_ENDPOINT="https://YOUR_RESOURCE.services.ai.azure.com/"
export AZURE_CU_API_KEY="YOUR_KEY_HERE"

Optional: set API version (defaults to 2025-05-01-preview):

export AZURE_CU_API_VERSION="2025-11-01"

Quick Usage

Analyze a URL and print Markdown

node scripts/analyze.mjs --url "https://example.com/document.pdf"

Analyze a local file (pipe via stdin)

cat invoice.pdf | node scripts/analyze.mjs --stdin --markdown output.md --output result.json

Save both Markdown and full JSON

node scripts/analyze.mjs --url "https://example.com/report.pdf" \
  --markdown report.md \
  --output report.json

Direct API Call

When the script isn't available, use curl:

# Submit analysis (preview API)
curl -s -X POST "$AZURE_CU_ENDPOINT/contentunderstanding/analyzers/prebuilt-layout:analyze?api-version=2025-05-01-preview" \
  -H "Ocp-Apim-Subscription-Key: $AZURE_CU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://example.com/doc.pdf"}'

# Response includes Operation-Location header — poll that URL for results

For GA API (2025-11-01), the body format changes:

{"inputs": [{"url": "https://example.com/doc.pdf"}]}

Output

Markdown

The analyzer produces GitHub Flavored Markdown preserving:

• Headings (h1–h6)
• Tables (as HTML blocks)
  Selection marks (☒ checked, ☐ unchecked)
  Figures (with references)
  Paragraphs with reading order

  Structured JSON

  The full result includes detailed per-element data:

  • pages — dimensions, word/line counts per page
  • paragraphs — text blocks with bounding regions and semantic roles
  • tables — cells with row/column spans
  • figures — detected images/charts with bounding regions
  • sections — hierarchical document structure

  Supported Formats

  PDF, JPEG, PNG, BMP, TIFF, HEIF, DOCX, XLSX, PPTX, HTML

  Best Practices

  • Async operation — the API returns 202; poll Operation-Location for results
  • Poll interval — 3 seconds is reasonable; results typically arrive in 5–60 seconds
  • Large documents — up to 2,000 pages supported; processing time scales linearly
  • File upload — use Content-Type: application/octet-stream with binary body
  • Tables — rendered as HTML in markdown for complex layouts (merged cells, etc.)

  API Reference

  See references/api.md for full request/response details.

  版本历史

  共 1 个版本

  • v1.3.0 当前
    2026-03-30 15:45 安全 安全

  安全检测

  腾讯云安全 (Keen)

  安全，无风险

  查看报告

  腾讯云安全 (Sanbu)

  安全，无风险

  查看报告

  🔗 相关推荐

  design-media

  FLUX.2-pro Image Generation

  zwcih

  通过 Azure AI Foundry 使用 Black Forest Labs FLUX.2-pro 生成图像。适用于请求创建、生成图片、插图、照片或艺术作品时。

  ★ 0 📥 854

  office-efficiency

  Excel / XLSX

  ivangdavila

  创建、检查和编辑 Microsoft Excel 工作簿及 XLSX 文件，支持可靠的公式、日期、类型、格式、重算及模板保留功能。

  ★ 384 📥 146,277

  office-efficiency

  Gog

  steipete

  Google Workspace 命令行工具，支持 Gmail、日历、云端硬盘、通讯录、表格和文档。

  ★ 931 📥 187,164

  Skill工具集 © 2026