Skip to main content
The Playground is where you design a document workflow before shipping it in code. It gives you the same core primitives as the API: Extract, Schema, Tables, Split, async jobs, saved presets, and generated SDK calls.

​
Layout

After a run completes, the Playground keeps the original document and the extracted output side by side so you can validate citations, tables, and structured data without leaving the workflow. Pulse Playground showing a source PDF beside extracted markdown and table output
AreaWhat it is for
New ExtractionUpload files, configure a pipeline, and run it.
Pipeline stepsAdd Extract, Tables, Split, and Schema steps.
Document viewerCompare extracted content against the original PDF or image.
Output tabsReview Markdown, Tables, Split assignments, Schema output, citations, chunks, and figures.
Pipeline LibraryReopen saved extractions, organize folders, share outputs, and rerun steps.
Developer sectionManage API keys and usage.

​
Pipeline Builder

Every new run starts with Extract. From there, add the step that matches the output you need. Pipeline step menu with Split, Tables, and Schema options Use the right-side configuration panel to tune each step before running the pipeline. Pulse pipeline configuration screen with Extract, Split, and Schema steps

​
Extract Step

Use Extract to parse the document into reusable content. Extract step configuration panel with model, page range, figure, HTML, bounding box, spreadsheet, and chunking options
SettingUse when
Page rangeYou only need known pages, such as 1-5 or 3,7,12.
Figure descriptionsCharts, diagrams, or screenshots need textual descriptions.
Show imagesYou want image URLs or inline visual assets in the output.
Return HTMLYour downstream renderer prefers HTML over markdown.
ChunkingYou are preparing content for search, RAG, or retrieval.
Spreadsheet optionsHidden rows, columns, or sheets matter for Excel files.

​
Schema Step

Use Schema to extract named fields into JSON. The editor supports generated schemas, manual editing, field descriptions, schema prompts, saved presets, and reruns. Use it for invoices, forms, contracts, policies, statements, and any document where a known output shape is more important than preserving full table layout. Schema Builder step with visual and JSON modes, field controls, schema prompt, effort, and page range settings Use the AI helper when you want Pulse to draft a schema from a short prompt, the attached document, or an existing schema. AI Schema Helper modal with document, schema, and prompt context options

​
Tables Step

Use Tables when the table is the product. Enable merge for cross-page tables and chart-to-table conversion when charts contain values you want in rows and columns. Tables step configuration panel with merge tables and charts-to-tables options

​
Split Step

Use Split when a long document contains sections that need different treatment. Split topics should be clear and mutually distinct. After Split, you can apply per-topic schemas or table extraction. Split step configuration panel with split input name and description fields Use the AI helper to draft split topics from a prompt, the attached document, or existing split inputs. AI Split Helper modal with document, split input, and prompt context options

​
Presets

Presets let you reuse stable configurations:
PresetStores
Extract presetExtraction settings like page range, chunking, figures, and spreadsheet options.
Split presetTopic names and descriptions.
Schema presetJSON Schema, descriptions, and schema prompt.
Tables presetMerge, table format, chart conversion, and related table options.
Use inline configs while exploring. Save presets once the workflow is stable.

​
Show Code

Show Code generates Python, TypeScript, or cURL from the current pipeline. If a step uses a saved preset, generated code references the config ID instead of inlining the full config. Show Code modal with generated Python for a Pulse extraction pipeline This is the recommended bridge from Platform to production:
  1. Build the workflow visually.
  2. Test it on representative documents.
  3. Save stable presets.
  4. Generate code.
  5. Put the generated chain behind your own app, queue, or data pipeline.

Build Your First Pipeline

A hands-on Platform tutorial.

Choose Your Workflow

Decide which steps to add.

Step Preset Library

Save and reuse configs.

Moving to Production

Turn Playground work into SDK calls.