Reducto Tools

Reducto turns documents (PDFs and other files reachable by URL) into structured content. Use these tools in a workflow to parse a document into structured chunks, extract specific fields against a JSON schema, or split a document into labeled sections.

Overview

Operations

Configuration

reducto_parse

Parses a document at a public URL into structured content. Request: POST https://platform.reducto.ai/parse.

reducto_extract

Extracts structured data from a document against a JSON schema. Request: POST https://platform.reducto.ai/extract.

reducto_split

Splits a document into sections based on described categories. Request: POST https://platform.reducto.ai/split.

Outputs

All three tools return the same output shape.

reducto_parse

• data (json) — Parsed document content (chunks, blocks, usage).
• metadata (json) — Job metadata. Contains:
  • jobId (string) — Reducto job identifier.
  • status (number) — HTTP status code returned by Reducto.

reducto_extract

• data (json) — Extracted structured data and citations.
• metadata (json) — Job metadata. Contains:
  • jobId (string) — Reducto job identifier.
  • status (number) — HTTP status code returned by Reducto.

reducto_split

• data (json) — Split result with detected sections and page ranges.
• metadata (json) — Job metadata. Contains:
  • jobId (string) — Reducto job identifier.
  • status (number) — HTTP status code returned by Reducto.

YAML Example

reducto_1:
  type: reducto
  name: "Reducto"
  inputs:
    operation: "reducto_parse"
    documentUrl: "https://example.com/document.pdf"
    apiKey: "{{REDUCTO_API_KEY}}"
  connections:
    outgoing:
      - target: next-block-id

Extract example (operation reducto_extract):

reducto_2:
  type: reducto
  name: "Reducto"
  inputs:
    operation: "reducto_extract"
    documentUrl: "https://example.com/invoice.pdf"
    schema: '{"type":"object","properties":{"total":{"type":"number"}}}'
    systemPrompt: "Be precise and thorough."
    apiKey: "{{REDUCTO_API_KEY}}"
  connections:
    outgoing:
      - target: next-block-id

Split example (operation reducto_split):

reducto_3:
  type: reducto
  name: "Reducto"
  inputs:
    operation: "reducto_split"
    documentUrl: "https://example.com/bundle.pdf"
    splitDescription: '[{"name":"invoice","description":"Pages containing an invoice"}]'
    apiKey: "{{REDUCTO_API_KEY}}"
  connections:
    outgoing:
      - target: next-block-id

Tips

• Auth setup: A single Reducto API key authenticates every operation. It is sent as a Bearer token (Authorization: Bearer <apiKey>), so store it as a secret/environment variable and reference it with {{REDUCTO_API_KEY}} rather than hardcoding it.
• Documents must be reachable by URL: documentUrl has to be publicly accessible — Reducto fetches the file itself. Pass an upstream block's output with {{blockName.field}} if the URL is produced earlier in the workflow.
• schema and splitDescription are JSON: For Extract, provide a valid JSON schema object describing the fields to pull out. For Split, provide a JSON array of {"name", "description"} category objects. Malformed JSON will cause the request to fail.
• Reading results: Every operation returns the same shape — use {{reducto_1.data}} for the result payload and {{reducto_1.metadata.jobId}} / {{reducto_1.metadata.status}} for job tracking and the HTTP status.