Pulse Tools

Pulse is a document-understanding (OCR) provider that extracts text, structure, figures, and structured data from PDFs, Word documents, PowerPoints, spreadsheets, and images. Use this tool in a workflow when you need to turn an uploaded file or a document URL into clean markdown, HTML, chunked text, or structured output for downstream blocks.

Overview

Operations

Configuration

pulse_parser

Parse documents (PDF, images, Office docs) using the Pulse OCR API. Provide either a file upload or a document URL/path — at least one file input is required.

Notes:

• A valid apiKey is always required; the request is rejected if it is missing or empty.
• You must supply a file input: either upload a file (fileUpload/file) or provide a filePath. If neither is present the request fails with a "Missing file input" error.
• filePath values that are not absolute paths must be valid HTTP or HTTPS URLs.

Outputs

pulse_parser

• markdown (string) — Extracted content in markdown format.
• page_count (number) — Number of pages in the document.
• job_id (string) — Unique job identifier.
• plan-info (json) — Plan usage information (e.g. pages_used, tier).
• bounding_boxes (json, optional) — Bounding box layout information.
• extraction_url (string, optional) — URL for extraction results (for large documents).
• html (string, optional) — HTML content if requested.
• structured_output (json, optional) — Structured output if a schema was provided.
• chunks (json, optional) — Chunked content if chunking was enabled.
• figures (json, optional) — Extracted figures if figure extraction was enabled.

YAML Example

pulse_1:
  type: pulse
  name: "Pulse"
  inputs:
    operation: "pulse_parser"
    filePath: "https://example.com/contract.pdf"
    pages: "1-3,5"
    chunking: "semantic"
    chunkSize: 512
    apiKey: "{{PULSE_API_KEY}}"
  connections:
    outgoing:
      - target: next-block-id

Tips

• Set your Pulse API key once as an environment variable (for example PULSE_API_KEY) and reference it as {{PULSE_API_KEY}} rather than pasting the raw key into the block.
• Provide exactly one file source per run: a fileUpload from the file-upload control or a filePath URL/path. If both are set, the uploaded file takes precedence.
• Reference the parsed output in downstream blocks with double braces, e.g. {{pulse_1.markdown}} for the markdown text or {{pulse_1.chunks}} when chunking is enabled. For very large documents Pulse may return an extraction_url instead of inline content, so handle that case when wiring follow-on steps.