DuckDuckGo Search integration for AI Flowi Workflow

The DuckDuckGo Search API query is covered here through AI Flowi Workflow’s native DuckDuckGo Search component. It takes required search text in input_value, offers two optional limit controls and returns a Table. The catalogue does not describe a standalone public JSON endpoint that you can call outside the workflow.

This component is documented in the catalogue; not production-tested by Ai Flowi. Its field reference comes from the AI Flowi Workflow 1.9.2 export dated 9 October 2026. The page explains how to inspect a small search result without inventing a successful response, access guarantee or provider quota.

By Ai Flowi · Published 10 October 2026

One provider-specific search entry

On a phone, focus and scroll the DuckDuckGo tables to read the query contract beside its limits.

The identity is duckduckgo.DuckDuckGoSearchComponent. The full catalogue lists it separately from Web Search, whose generic mode selection is documented on its own page. DuckDuckGo Search has no exported Web, News or RSS selector; use the fields actually present on this entry.

Required query and two different limits

A query connection and a typed string value describe different ways to supply the same field. For the first inspection, enter a deliberate phrase directly. Once that behavior is understood, connect a Message-producing input if the task needs changing search text. Validate the task’s text before a search; a required flag alone does not describe whitespace or blank-query rejection.

The SerpAPI limit discussion explains the shared distinction between result count and snippet context.

Related: Wikipedia and Wikidata integration for AI Flowi Workflow.

Table output requires field inspection

The port name is dataframe, the display label is Table and the exported type is Table. The entry does not name result columns, ordering fields, source-address fields or error rows. A Table connection therefore establishes a compatible container for the next node, not a complete schema for selecting URLs or writing summaries.

Inspect the result before choosing any field-based operation. Identify the actual source references, snippet representation and empty-result behavior you observe. If a source cannot be resolved from the returned data, keep it out of a citation-dependent output until you can inspect it. Do not replace a missing source address with a guessed URL based on a title.

No exported key field: what that means

The complete input list contains no API key, account selector or saved-connection identifier. That is a statement about this component’s exported settings. It does not establish unrestricted provider access, an official hosted JSON API, a service-level agreement or permission to redistribute retrieved material.

Use documented connection mechanisms if your deployment requires access configuration outside this DuckDuckGo node.

A query-to-display inspection

example configuration, not a recorded test

1. Add DuckDuckGo Search with identity duckduckgo.DuckDuckGoSearchComponent. Enter the illustrative query workflow data validation documentation in input_value.
2. Set max_results to 3 and max_snippet_length to 240. These numbers are proposed inspection settings; accepted ranges and enforcement need checking in your own run.
3. Add Chat Output and connect dataframe to its required input_value. Table appears in the receiving field’s accepted types.
4. Inspect the displayed data and node result. Record the field names and source references actually available, with the query and both limit settings.
5. Change only one limit for a second trial. Compare the returned representation before adding a formatter or a language-model step.

Use the Chat Output reference when configuring how the DuckDuckGo candidate set is displayed.

Separate discovery, source review and failure

For documentation discovery, resolve the component identity before accepting a reference. A page about a similarly named node can look relevant while describing an incompatible input contract. The query exercise below shows how to narrow that identity question.

If a run returns no usable candidates, retain that condition as “no usable result for this trial.” If the node reports an access or provider error, retain the error instead. These observations require different next steps. The catalogue does not define an automatic branch for either condition, so verify their actual representations before designing unattended processing.

When do I need a retrieval interface other than DuckDuckGo?

No engine options, page offset, language selector, location field, timeout or retry control is exported here. The component description establishes web search with customizable limits; it does not document news-specific retrieval, full-page loading, a ranking guarantee or an exact freshness window.

Choose a documented URL-reading path when you already know the page to inspect, and use the generic Web Search reference if its exported mode controls are the requirement. This page does not turn the DuckDuckGo component into a consumer browser tutorial or a general HTTP API specification. Test the native search contract before depending on it in a larger flow.

Related: Exa Search integration for AI Flowi Workflow.

A documentation-discovery query exercise

Suppose you need the validation instructions for a named workflow component. Prepare a first phrase containing the component name and the behavior you need to understand, rather than an entire chat conversation. A long conversation may contain several unrelated requests, while this node has one required search-text field. Decide which request is the discovery question before putting text into input_value. This does not assume support for a particular search operator or query grammar.

In the illustrative workflow data validation documentation trial, classify each inspectable candidate by its purpose: a component reference, a tutorial showing one configuration, a general article or an unrelated match. A tutorial can help explain a configuration without being the authoritative definition of every input. A reference can define an input without demonstrating a successful run. This classification answers the documentation question more directly than accepting a page because its title contains all the query words.

For a follow-up phrase, replace the broad behavior with the exact component name you already know. Hold the trial limits at their previous values. Ask whether the candidate now addresses the input that prompted the search. If it discusses a different component with a similar label, revise the entity wording before enlarging the search. The choice to narrow a phrase concerns the identity being sought; it is separate from the shared limit trade-off linked above.

Before replacing typed text with a Message connection, inspect a proposed incoming message containing two questions. Determine which question would be sent to this search node and how the other should be handled by your workflow design. The catalogue supplies no query-extraction field here. Connecting a conversational message does not itself select a discovery question from that message, so establish the intended text upstream before relying on dynamic input. Do not paste credentials or private workspace details into a query simply because they appeared in the conversation.

For this documentation task, an empty result means the component-specific question still lacks a candidate reference. Broaden the wording only if the exact-name phrase was too narrow; a provider error instead calls for resolving the failed execution. Avoid turning either condition into a fabricated documentation link. If a colleague already supplied the authoritative page address, searching for it again adds an unnecessary discovery step; read that known address through the URL component.

Search inspection usage

Plans: Go from USD8/month billed yearly (USD10 monthly), Plus from USD16 (USD20 monthly), Pro 100 from USD80 (USD100 monthly), Pro 200 from USD160 (USD200 monthly) and Pro 500 from USD400 (USD500 monthly), with 5,000 to 250,000 Flowi Tokens/month and top-ups; 1 token per node run; third-party API and model usage is not included.

Other native retrieval categories are indexed in the integrations hub.