Web Search Component

Use the Web Search component to perform a targeted web search and get back a structured list of results. This component is the simplest way for your agent to find information on the internet, with advanced options to filter by time, exclude domains, and even generate an AI-powered summary of the findings.

Why this matters

Giving your agent the ability to search the web is a foundational step in creating a powerful and knowledgeable assistant. It can find real-time information, research topics, answer user questions with up-to-date data, and provide the source material for more complex components like the Web Scrape Component.

What You’ll Configure

• Configure Search Settings
• Provide a Search Query
• Handle the Results
• Best Practices
• Troubleshooting Tips
• What to Try Next

Step 1: Configure Search Settings

Fine-tune the search parameters to get the most relevant results for your task.

Setting	Description
Sources Limit	The maximum number of search results to return. Default is 5.
Search Topic	Choose the category for the search. general provides broad web results, while news focuses on recent articles from news outlets. Default is general.
Include Questions and Answers	Enable to receive an AI-generated answer to your query based on the content of the top search results.
Include Image Results	Enable to include a list of query-related images in the response.
Time Range	Filter results to a specific time range from the current date. Options are day, week, month, or year.
Include Raw Content	Enable to include the cleaned HTML content of each search result page in the output. This is useful if you want to analyze the page content without a separate scrape.
Exclude Domains	Provide a comma-separated list of domains to exclude from the search results (e.g., wikipedia.org,pinterest.com).

Step 2: Provide a Search Query

The component requires a single input to know what to search for.

Input	Description
SearchQuery	The default input field that holds the text query you want to search for.

Step 3: Handle the Results

The component provides several outputs to easily access the different parts of the search results.

Output	Description
Output	Contains the entire set of results as a single JSON object, including search results, images, and the AI-generated answer if those options are enabled.
Results	A convenient output containing just the array of search results (links, titles, snippets).
Images	If Include Image Results is enabled, this output will contain the list of related images.
Answer	If Include Questions and Answers is enabled, this output will contain the AI-generated answer string.

Convenient Outputs

While Output contains everything, using the specific Results, Images, and Answer outputs is often easier and cleaner, as it saves you from having to parse the main JSON object.

Best Practices

• Use for Initial Research: Web Search is excellent for the first step of a research task. Use it to find relevant URLs, then pass those URLs to the Web Scrape Component for deep content extraction.
• Enable AI Answers for Quick Summaries: The Include Questions and Answers option is a very efficient way to get a direct answer to a query without needing a separate LLM call.
• Filter for Relevance: Use Time Range and Exclude Domains to narrow your search and improve the quality of your results. For example, when researching a recent event, set the time range to day or week.
• Handle No Results: Your workflow should be able to handle cases where a search returns no results. Check if the Results output is empty and branch your logic accordingly.

Troubleshooting Tips

If your search isn't working as expected...

• Irrelevant Results: Your SearchQuery might be too broad. Try making it more specific. Also, consider using the Exclude Domains setting to filter out noisy or unhelpful sites.
• No AI Answer Generated: The Include Questions and Answers feature depends on the content of the search results. If the top results do not contain a clear answer to the query, the LLM may not be able to generate one.
• Search Fails: In rare cases, the search provider might be temporarily unavailable. You can build retry logic into your workflow using a Condition Component to check for an empty result and loop back if needed.

What to Try Next

• Take the Results array and pass it to a ForEach Loop. Inside the loop, use the Web Scrape Component on each URL to extract its full content.
• Use the Answer output and feed it directly into an API Output Component to create an agent that can answer questions with sourced information.
• Use the search results to feed a RAG pipeline. Scrape the content of the top results and use RAG Remember to add the new information to your agent's knowledge base.