Content, web, SEO and document nodes

Generated on 2026-10-20 from a running server by tools/reference.mjs. Do not edit this page by hand.

The steps that read the web and documents, and the steps that build content.

This page lists 21 nodes. How to read a node reference.

Content

Content creation.

Content Generator

Step · in the catalogue: Content · type content.content_generator

Generates structured DSL content from diverse sources with LLM-powered content generation and deterministic layout.

When to use it. To have an AI write a structured page, such as an article or a landing page, from a description, with an optional brief, SEO data and links that must all appear in the text. It needs an LLM model connected; its page layout in DSL format then goes to DSL to HTML, DSL to Markdown or DSL to Elementor.

Effects. costs.

Fields

Name	Type	Required	Default	What it is
`content_description`	Text	yes		Description of the content to generate (user prompt - what, why, for whom).
`seo_strategy`	Text	no		SEO strategy data (from SEO Advisor or manual).
`links`	Text	no		URLs to include (internal/external/references) - MANDATORY integration with link-first approach.
`content_brief`	Text	no		Detailed content brief with specific requirements.
`tone`	Text	no	`professional`	Writing tone: professional, casual, educational, conversational, etc.
`word_count`	Number	no	`1000`	Target word count (approximate).
`structure_hints`	TextList	no		Structure guidance: sections to include, flow suggestions.

Outputs

Name	Type	What it holds
`dsl_content`	Text	Generated content, as a page layout in DSL format (JSON)
`metadata`	Text	Content metadata: topics, keywords, estimated reading time

Entries. In (Entry point)

Exits. Out (Default output port)

Places for resources

Place	Required	Accepts	How many
LLM Model	yes	`llm.model.*`	1

Content > Layout

Layout manipulation.

Add Block

Step · in the catalogue: Content > Layout · type layout.add_block

Adds one or more blocks to the layout at a specified position. Supports single block object or array of blocks.

When to use it. To add one or more blocks, such as a call to action or an extra section, to a page layout in DSL format, at a place named by its path (such as main). Give the block as JSON, or a JSON list of blocks.

Fields

Name	Type	Required	Default	What it is
`layout`	Text	yes		DSL layout JSON.
`target`	Text	yes		Target path (e.g., ‘main’).
`position`	Choice	no	`append`	Insert position. Choices: `append`, `prepend`, `before`, `after`, `replace`.
`block`	Text	yes		Block(s) to add - single object or array of blocks (DSL JSON).

Outputs

Name	Type	What it holds
`layout`	Text	Updated DSL layout JSON

Entries. In (Entry point)

Exits. Out (Default output port)

Move Block

Step · in the catalogue: Content > Layout · type layout.move_block

Moves a block from source to target position in the layout.

When to use it. To move one block of a page layout in DSL format to another place, both named by their path (such as main.block[1] and main).

Fields

Name	Type	Required	Default	What it is
`layout`	Text	yes		DSL layout JSON.
`source`	Text	yes		Source path (e.g., ‘main.block[1]’).
`target`	Text	yes		Target path (e.g., ‘main’).
`position`	Choice	no	`append`	Insert position. Choices: `append`, `prepend`, `before`, `after`.

Outputs

Name	Type	What it holds
`layout`	Text	Updated DSL layout JSON

Remove Block

Step · in the catalogue: Content > Layout · type layout.remove_block

Removes a block from the layout at the specified path.

When to use it. To take one block out of a page layout in DSL format, named by its path (such as root.main.section[2]).

Fields

Name	Type	Required	Default	What it is
`layout`	Text	yes		DSL layout JSON.
`target`	Text	yes		Target path to remove (e.g., ‘root.main.section[2]’).

Outputs

Name	Type	What it holds
`layout`	Text	Updated DSL layout JSON

Replace Block

Step · in the catalogue: Content > Layout · type layout.replace_block

Replaces a block at the specified path with one or more new blocks.

When to use it. To swap one block of a page layout in DSL format, named by its path, for one or more new blocks given as JSON. To keep the block and add next to it, use Add Block instead.

Fields

Name	Type	Required	What it is
`layout`	Text	yes	DSL layout JSON.
`target`	Text	yes	Target path to replace (e.g., ‘root.main.block[1]’).
`block`	Text	yes	New block(s) - single object or array of blocks (DSL JSON).

Outputs

Name	Type	What it holds
`layout`	Text	Updated DSL layout JSON

Content > Layout Adapters

Layout format conversion nodes.

DSL to Elementor

Step · in the catalogue: Content > Layout Adapters · type content.layoutadapters.dsl_to_elementor

Converts DSL content to Elementor-compatible format for WordPress

When to use it. To turn a page layout in DSL format, such as the one Content Generator gives, into an Elementor page for Publish to WordPress. To show images already uploaded, give a JSON list linking each image to its WordPress media ID.

Fields

Name	Type	Required	What it is
`dsl_content`	Text	yes	Page layout in DSL format (JSON), as Content Generator gives it.
`title`	Text	no	Optional page/post title.
`image_mappings`	Text	no	JSON array mapping image IDs to WordPress media IDs: [{image_id, media_id, media_url}].

Outputs

Name	Type	What it holds
`elementor_data`	Text	Elementor-compatible JSON data
`metadata`	Text	Content metadata

Entries. In (Entry point)

Exits. Out (Default output port)

DSL to Email HTML

Step · in the catalogue: Content > Layout Adapters · type content.layoutadapters.dsl_to_email_html

Converts DSL content to email-compatible HTML with table layout and inline styles

When to use it. To turn a page layout in DSL format into the HTML of an e-mail, such as a newsletter, that displays well in mail apps, in your brand colors. For a web page, use DSL to HTML instead.

Fields

Name	Type	Required	Default	What it is
`dsl_content`	Text	yes		Page layout in DSL format (JSON), as Content Generator gives it.
`email_width`	Number	no	`600`	Email width in pixels (default: 600).
`brand_color`	Text	no	`#2563EB`	Primary brand color in hex (default: #2563EB).
`background_color`	Text	no	`#F3F4F6`	Background color in hex (default: #F3F4F6).
`text_color`	Text	no	`#1F2937`	Text color in hex (default: #1F2937).
`font_family`	Text	no	`Arial, Helvetica, sans-serif`	Font family (default: Arial, Helvetica, sans-serif).
`include_responsive`	Boolean	no	`true`	Include media queries for mobile responsive (default: true).

Outputs

Name	Type	What it holds
`email_html`	Text	Generated email-compatible HTML
`preview_text`	Text	Preview text (first 100 characters of content)
`metadata`	Text	Content metadata

DSL to HTML

Step · in the catalogue: Content > Layout Adapters · type content.layoutadapters.dsl_to_html

Converts DSL content to semantic HTML with optional CSS framework support

When to use it. To turn a page layout in DSL format into HTML for a web page, plain or styled for Tailwind or Bootstrap. For an e-mail, use DSL to Email HTML instead.

Fields

Name	Type	Required	Default	What it is
`dsl_content`	Text	yes		Page layout in DSL format (JSON), as Content Generator gives it.
`css_framework`	Text	no	`none`	CSS framework: ‘tailwind’, ‘bootstrap’, or ‘none’ (default: ‘none’).
`include_inline_styles`	Boolean	no	`false`	Include basic inline styles for spacing (default: false).
`wrapper_class`	Text	no		CSS class for the root wrapper div.
`generate_toc_anchors`	Boolean	no	`true`	Generate ID attributes for headings (default: true).

Outputs

Name	Type	What it holds
`html`	Text	Generated HTML content
`metadata`	Text	Content metadata (heading count, word count, etc.)

DSL to Markdown

Step · in the catalogue: Content > Layout Adapters · type content.layoutadapters.dsl_to_markdown

Converts DSL content to Markdown format

When to use it. To turn a page layout in DSL format into Markdown, for a documentation page, a README or any tool that takes Markdown. For a web page, use DSL to HTML instead.

Fields

Name	Type	Required	Default	What it is
`dsl_content`	Text	yes		Page layout in DSL format (JSON), as Content Generator gives it.
`generate_toc_anchors`	Boolean	no	`true`	Generate anchor links for TOC (default: true).
`heading_prefix`	Text	no		Optional prefix for all headings (e.g., project name).

Outputs

Name	Type	What it holds
`markdown`	Text	Generated Markdown content
`metadata`	Text	Content metadata (heading count, word count, etc.)

Web

Web scraping and content extraction.

HTTP Request

Step · in the catalogue: Web · type http.request

Calls a web API: GET, POST, PUT, PATCH, DELETE or HEAD, with parameters, headers and a body, and gives the status and the answer (as JSON when it is JSON).

When to use it. To use a service that has a web API but no step of its own in the catalog: create or read a record, send data to another tool, call a webhook. Only public addresses can be called, never the local machine or a private network. To read a web page as text, use Scrape Web Page instead.

Example. method: POST, url: https://api.example.com/orders, body: {“customer”: “form.email”}

Effects. reads outside, changes something outside.

Fields

Name	Type	Required	Default	What it is
`method`	Choice	no	`GET`	The HTTP method. Choices: `GET`, `POST`, `PUT`, `PATCH`, `DELETE`, `HEAD`.
`url`	Text	yes		The address to call, in full (https://…).
`query`	Structured	no		Parameters added to the address, as a JSON object: {“page”: “2”}.
`headers`	Structured	no		Headers sent with the call, as a JSON object: {“Accept”: “application/json”}. Put a key in an HTTP API connection rather than here.
`body`	Text	no		What is sent with POST, PUT, PATCH or DELETE — usually JSON.
`content_type`	Choice	no		How the body is written (guessed from the body when left empty). Choices: `json`, `text`, `form`.
`credentials`	Credentials	no		The HTTP API connection whose key signs the call, when the API needs one. Connection of type `http_api`.
`timeout_seconds`	Number	no	`30`	How long to wait for the answer, in seconds.
`max_response_chars`	Number	no	`100000`	The most characters of the answer kept.
`rate_limit_profile`	Choice	no	`medium`	How long to wait and try again when the service is busy (readings only: a call that changes something is never sent twice). Choices: `none`, `fast`, `medium`, `slow`.

Outputs

Name	Type	What it holds
`status`	Number	The HTTP status the service answered (empty when it could not be called)
`body`	Text	The answer, as text
`json`	Structured	The answer read as JSON, when it is JSON
`headers`	Structured	The headers of the answer, as a JSON object
`url`	Text	The address that answered (after any redirect)
`error`	Text	Why the call failed, when it did

Entries. In (Entry point)

Exits. Out (The service answered 2xx), Error (The service answered with an error, or could not be called — see ‘error’)

Scrape Web Page

Step · in the catalogue: Web · type web.scrape

Reads a web page as Markdown, ready for an LLM prompt. Uses the connected web scraper, or the free Direct scraper when none is connected.

When to use it. To read the content of one web page from its link, for example a link a person gave, so an AI step can work on it. Not for finding pages on a subject: use a web search instead.

Effects. reads outside.

Fields

Name	Type	Required	Default	What it is
`url`	Text	yes		The page to read (http or https).
`timeout_seconds`	Number	no	`30`	How long to wait for the page.
`max_content_length`	Number	no	`50000`	Longest content kept, in characters.

Outputs

Name	Type	What it holds
`markdown`	Text	The page content as Markdown
`text`	Text	The same content as plain text
`url`	Text	The URL that was read
`metadata`	Structured	Page details (title, description…) as a JSON object, when the scraper provides them
`structured_data`	Structured	The data the page embeds for machines (JSON-LD, page state, meta tags, microdata), compacted — often holds exact values (prices, ratings, dates) that the readable content only shows in passing
`cleaning`	Structured	What cleaning took out of the page: sizes before and after, and what was removed
`error`	Text	Why reading failed, when it did

Entries. In (Entry point)

Exits. Scraped (The page was read successfully), Failed (The page could not be read — see ‘error’)

Places for resources

Place	Required	Accepts	How many
Web Scraper	no	`web.scraper.*`	1

Web > Scrappers

Web scraping resource nodes and providers.

Direct Scraper

Resource · in the catalogue: Web > Scrappers · type web.scraper.direct

Scrape web pages using direct HTTP requests (free, simple, limited features).

When to use it. To read web pages for free, with no account, from a step that reads pages such as Scrape Web Page. For pages that only show their content once scripts run in the browser, use Firecrawl Scraper instead.

Effects. reads outside.

Plugs into a place that accepts web.scraper.*.

Firecrawl Scraper

Resource · in the catalogue: Web > Scrappers · type web.scraper.firecrawl

Scrape web pages using Firecrawl API (premium, JS rendering, clean markdown).

When to use it. To read pages the free scrapers cannot, such as pages built by scripts in the browser, with your own Firecrawl account (an API key). Connect it to a step that reads pages, such as Scrape Web Page.

Effects. reads outside.

Fields

Name	Type	Required	Default	What it is
`credentials`	Credentials	yes		Firecrawl API credentials. Connection of type `firecrawl`.

Plugs into a place that accepts web.scraper.*.

Jina Reader Scraper

Resource · in the catalogue: Web > Scrappers · type web.scraper.jina

Scrape web pages using Jina AI Reader (free, clean markdown output).

When to use it. To read web pages as clean text for free through Jina AI Reader, with no account, from a step that reads pages such as Scrape Web Page. Useful when the Direct Scraper keeps too much clutter.

Effects. reads outside.

Plugs into a place that accepts web.scraper.*.

Web > Search

Web search nodes backed by external search providers.

Web Search (SearXNG)

Step · in the catalogue: Web > Search · type web.search.searxng

Searches the web through a SearXNG instance and returns normalized organic results.

When to use it. To search the web in a workflow through a SearXNG server you run or trust, given by its address, with no paid account; the results go to the next steps. For an agent that searches by itself, use AI Agent Web Search Tool (SearXNG).

Effects. reads outside, costs.

Fields

Name	Type	Required	Default	What it is
`credentials`	Credentials	yes		SearXNG credentials. Connection of type `searxng`.
`query`	Text	yes		Search query.
`top_k`	Number	no	`5`	Maximum number of results.
`language`	Text	no		Optional language hint.
`site`	Text	no		Optional site/domain filter.
`rate_limit_profile`	Choice	no	`medium`	Rate limiting and retry profile. Choices: `none`, `fast`, `medium`, `slow`.

Outputs

Name	Type	What it holds
`provider`	Text	Search provider identifier
`query`	Text	Original search query
`effective_query`	Text	Executed search query including filters
`result_count`	Number	Number of search results returned
`results_json`	Text	The results (rank, title, URL, snippet) as JSON
`urls`	TextList	Returned result URLs
`error`	Text	Why the search failed, when it did

Entries. In (Entry point)

Exits. Found (The search ran — its results are in the outputs (there can be none)), Failed (The search could not run — see ‘error’)

Web Search (SerpApi)

Step · in the catalogue: Web > Search · type web.search.serpapi

Searches the web through SerpApi and returns normalized organic results.

When to use it. To search the web in a workflow through your own SerpApi account (an API key); the results go to the next steps. For an agent that searches by itself, use AI Agent Web Search Tool (SerpApi).

Effects. reads outside, costs.

Fields

Name	Type	Required	Default	What it is
`credentials`	Credentials	yes		SerpApi credentials. Connection of type `serpapi`.
`query`	Text	yes		Search query.
`top_k`	Number	no	`5`	Maximum number of results.
`language`	Text	no		Optional language hint.
`site`	Text	no		Optional site/domain filter.
`rate_limit_profile`	Choice	no	`medium`	Rate limiting and retry profile. Choices: `none`, `fast`, `medium`, `slow`.

Outputs

Name	Type	What it holds
`provider`	Text	Search provider identifier
`query`	Text	Original search query
`effective_query`	Text	Executed search query including filters
`result_count`	Number	Number of search results returned
`results_json`	Text	The results (rank, title, URL, snippet) as JSON
`urls`	TextList	Returned result URLs
`error`	Text	Why the search failed, when it did

Entries. In (Entry point)

Exits. Found (The search ran — its results are in the outputs (there can be none)), Failed (The search could not run — see ‘error’)

Web Search (Tavily)

Step · in the catalogue: Web > Search · type web.search.tavily

Searches the web through Tavily and returns normalized organic results.

When to use it. To search the web in a workflow through your own Tavily account (an API key); the results go to the next steps. For an agent that searches by itself, use AI Agent Web Search Tool (Tavily).

Effects. reads outside, costs.

Fields

Name	Type	Required	Default	What it is
`credentials`	Credentials	yes		Tavily API credentials. Connection of type `tavily`.
`query`	Text	yes		Search query.
`top_k`	Number	no	`5`	Maximum number of results.
`language`	Text	no		Optional language hint.
`site`	Text	no		Optional site/domain filter.
`rate_limit_profile`	Choice	no	`medium`	Rate limiting and retry profile. Choices: `none`, `fast`, `medium`, `slow`.

Outputs

Name	Type	What it holds
`provider`	Text	Search provider identifier
`query`	Text	Original search query
`effective_query`	Text	Executed search query including filters
`result_count`	Number	Number of search results returned
`results_json`	Text	The results (rank, title, URL, snippet) as JSON
`urls`	TextList	Returned result URLs
`error`	Text	Why the search failed, when it did

Entries. In (Entry point)

Exits. Found (The search ran — its results are in the outputs (there can be none)), Failed (The search could not run — see ‘error’)

SEO

SEO analysis and optimization.

Internal Link Suggester

Step · in the catalogue: SEO · type seo.links.suggest

Suggests relevant internal links using structure-first selection with semantic scoring.

When to use it. To choose which pages of your own site a new article should link to, with link texts written by an AI, from the site structure Sitemap Structure gives. Not for links to other sites: it only suggests pages of your site.

Effects. costs.

Fields

Name	Type	Required	Default	What it is
`article_topic`	Text	yes		Article topic or title.
`article_summary`	Text	no		Brief article summary/outline (optional).
`site_structure`	Text	yes		Site structure (JSON), as the Sitemap Structure step gives it.
`max_links`	Number	no	`10`	Maximum links to suggest.
`parent_pillar_url`	Text	no		Parent pillar URL (required for best results).
`quota_parent_pillar`	Number	no	`1`	Links to parent pillar.
`quota_sibling_pillars`	Number	no	`2`	Links to sibling pillars.
`quota_same_pillar_leaves`	Number	no	`4`	Links to leaves in same pillar.
`quota_other_pillars`	Number	no	`2`	Links to other relevant pillars.
`quota_cross_pillar_leaves`	Number	no	`1`	Links to leaves in other pillars.

Outputs

Name	Type	What it holds
`suggested_links`	Text	List of suggested links (JSON array)
`total_suggestions`	Number	Number of links suggested

Entries. In (Entry point)

Exits. Out (Default output port)

Places for resources

Place	Required	Accepts	How many
LLM Model	yes	`llm.model.*`	1
Embeddings	no	`ai.embeddings.*`	1

Sitemap Analyzer

Step · in the catalogue: SEO · type seo.sitemap.analyze

Parses XML sitemap and extracts page information for SEO analysis.

When to use it. To list the pages of a website from its XML sitemap (such as https://example.com/sitemap.xml), optionally with each page’s title and description, as the start of an SEO workflow. Give its pages to Sitemap Structure.

Effects. reads outside.

Fields

Name	Type	Required	Default	What it is
`sitemap_url`	Text	yes		Sitemap XML URL (e.g., https://example.com/sitemap.xml).
`fetch_content`	Boolean	no	`false`	Fetch page content (title, H1, meta description, excerpt).
`max_pages`	Number	no	`0`	Maximum pages to analyze (0 = all).

Outputs

Name	Type	What it holds
`pages`	Text	List of sitemap pages (JSON array)
`total_pages`	Number	Total number of pages found

Sitemap Structure

Step · in the catalogue: SEO · type seo.sitemap.structure

Analyzes sitemap pages using LLM to create hierarchical content structure (pillars & clusters).

When to use it. To group a site’s pages into main topics (pillars) and topic clusters with an AI, from the page list Sitemap Analyzer gives. Its result feeds Internal Link Suggester.

Effects. reads outside, costs.

Fields

Name	Type	Required	Default	What it is
`pages`	Text	yes		Sitemap pages (JSON array), as the Sitemap Analyzer step gives it.
`site_name`	Text	no	`Website`	Website name/brand.
`max_pillars`	Number	no	`10`	Maximum content pillars to identify.
`scrape_pages`	Boolean	no	`true`	Scrape pages to analyze content and internal links (slower but more accurate).
`max_pages_to_scrape`	Number	no	`50`	Maximum pages to scrape for analysis (0 = all).
`max_topics_per_page`	Number	no	`8`	Maximum topics to extract per page.
`min_topic_similarity`	Number	no	`0.3`	Minimum similarity threshold for topic edges (0.0-1.0).
`max_topics_after_reduction`	Number	no	`100`	Maximum topics after set-cover reduction.
`cluster_cohesion_threshold`	Number	no	`0.5`	λ₂ threshold for cluster cohesion (0.0-1.0).
`min_cluster_size`	Number	no	`2`	Minimum cluster size.
`assignment_margin`	Number	no	`0.1`	Minimum gap between best/2nd cluster for assignment (0.0-0.3).
`min_child_support`	Number	no	`0.1`	Minimum fraction of cluster pages pillar must link to (0.0-1.0).

Outputs

Name	Type	What it holds
`site_structure`	Text	Site structure with pillars, topic clusters, and orphan pages (JSON)
`total_pillars`	Number	Number of content pillars identified
`total_clusters`	Number	Number of topic clusters identified

Places for resources

Place	Required	Accepts	How many
LLM Model	yes	`llm.model.*`	1
Embeddings	yes	`ai.embeddings.*`	1
Scraper	no	`web.scraper.*`	1

Documents

Read the text of documents (PDF, Word, text).

Extract Document Text

Step · in the catalogue: Documents · type document.extract_text

Reads a PDF, Word (.docx), or plain-text/Markdown file — an uploaded file (file_id) or raw bytes (file_base64) — and returns its content as Markdown, ready for an LLM prompt.

When to use it. To get the text of a PDF, Word (.docx), text or Markdown file a person uploaded, so an AI step can read it. Not for web pages: use Scrape Web Page instead.

Fields

Name	Type	Required	What it is
`file_id`	Text	yes	Reference of an uploaded file (the fileId a SourceInput upload returns) — use this OR file_base64. The file name and format come from the upload.
`file_base64`	Text	yes	The document’s raw bytes, base64-encoded — use this OR file_id.
`file_name`	Text	yes	Original file name, including its extension (e.g. ‘resume.pdf’) — the extension is what selects PDF, Word, or plain-text reading. Required with file_base64; taken from the upload when file_id is given.

Outputs

Name	Type	What it holds
`markdown`	Text	The document’s content as Markdown — headings, lists and emphasis preserved where the source format carries them
`text`	Text	The same content with Markdown syntax stripped, for word counts or search
`format`	Text	Which extractor read the file: ‘pdf’, ‘docx’, or ‘text’
`metadata`	Structured	Extraction details (e.g. page_count, word_count) as a JSON object
`error`	Text	Why extraction failed, when it did

Entries. In (Entry point)

Exits. Extracted (The document’s text was read successfully), Failed (The document could not be read — see ‘error’)