26 source-backed candidates across 11 task types. Documentation is checked separately from task execution. Select a task to compare its exact input and output requirements.
Extract text, fixed-column tables and exact labelled fields locally using PDF.js.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Document fields / 文档字段提取
No OCR. Scans, handwriting, merged cells and arbitrary layouts are unsupported. Field extraction matches explicit labels; it is not semantic invoice understanding. No XLSX export.
Documentation: 2026-10-22 · noneExtract PDF or image text and table structure with OCR.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR
JSON contains extracted content; it is not an Excel workbook. OCR accuracy is not measured by this profile.
Documentation: 2026-10-22 · api_keyExtract text, layout and tables with a configured Azure resource.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR
Endpoint is account-specific. Free tiers have limits; local residency and arbitrary language accuracy are not established here.
Documentation: 2026-10-22 · api_keyUse OCR, Layout Parser or Form Parser processors for document analysis.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR · Document fields / 文档字段提取
Processor types have different input, quota and region limits. OCR, layout and form parsing are distinct configurations.
Documentation: 2026-10-22 · google_cloud_iamExtract printed text, tables and form fields through AWS Textract.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR · Document fields / 文档字段提取
AWS SDK signing is required. Table/form outputs need interpretation; no general XLSX export is promised.
Documentation: 2026-10-22 · aws_iamExtract text, structure and table renditions from native or scanned PDFs.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR
PDF input only in this profile. CSV/XLSX are table renditions, not a perfect reconstruction of the entire document. Quotas apply.
Documentation: 2026-10-22 · adobe_client_credentialsParse PDF and office documents into text and structured document elements.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR · Office document text / Office 文档
Parsing modes differ in price and accuracy. Schema-driven LlamaExtract is a separate product and is not implied by this profile.
Documentation: 2026-10-22 · api_keyPartition documents locally into typed text and table elements.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR · Office document text / Office 文档
Model downloads and compute cost are separate. Partitioning results vary by strategy; a cloud API or remote enrichment would change the privacy boundary.
Documentation: 2026-10-22 · noneConvert documents locally with text, layout, OCR and table processing.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR · Office document text / Office 文档
OCR and table quality depend on the pipeline and document. Remote VLM configurations are outside this local profile.
Documentation: 2026-10-22 · noneConvert documents locally to Markdown, JSON or HTML with OCR support.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格 · Scanned document OCR / 扫描件 OCR · Office document text / Office 文档
Code and model weights have different terms. Commercial eligibility is not established here; this is not an unconditional free commercial service.
Documentation: 2026-10-22 · noneExtract text and metadata from PDF and office files locally.
PDF text / PDF 文本提取 · Office document text / Office 文档
This profile covers text and metadata. Structured table reconstruction and OCR require separate configuration and are not claimed.
Documentation: 2026-10-22 · noneInspect PDF text positions and extract tables from machine-generated PDFs.
PDF text / PDF 文本提取 · Text PDF tables / 文字型 PDF 表格
No OCR. Table extraction needs layout-specific checks. Nested arrays are not automatically CSV or XLSX.
Documentation: 2026-10-22 · noneFull machine-readable catalog · Refresh and deduplication status