AI-Powered Multi-Stage Web Search and Research Suite

This n8n workflow provides a robust, AI-powered multi-stage web search and research suite. It leverages a webhook trigger to initiate complex research tasks, dynamically fetches information using HTTP requests, processes data with a Large Language Model (LLM) chain, and structures the output for clear, actionable results. The workflow is designed to handle multiple research queries in batches, incorporating delays to manage API rate limits or ensure data availability.

What it does

This workflow automates the following steps:

1. Receives Research Requests: It starts by listening for incoming data via a webhook, which typically contains a list of research queries.
2. Batches Queries: It splits the incoming research queries into manageable batches, allowing for processing multiple items sequentially or with controlled concurrency.
3. Performs Web Searches (Placeholder): For each item in a batch, it makes an HTTP request. This node is intended to interact with a web search API (e.g., Google Search API, SERP API) to fetch raw search results based on the query.
4. Introduces Delays: After each batch of web searches, it pauses for a configurable duration. This is crucial for respecting API rate limits and preventing service overload.
5. Processes with LLM: The raw search results are then fed into a Basic LLM Chain powered by a Google Gemini Chat Model. This chain is responsible for analyzing the search results, extracting relevant information, synthesizing insights, and performing the core "research" aspect of the workflow.
6. Structures Output: An Auto-fixing Output Parser and a Structured Output Parser are used to ensure the LLM's output is consistently formatted into a predefined structure (e.g., JSON), even if the LLM's initial response is slightly malformed. This ensures downstream systems can reliably consume the research findings.
7. Responds to Trigger: Finally, the workflow responds to the initial webhook trigger with the structured, AI-processed research results.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance (self-hosted or cloud).
• Webhook Integration: An external system or application configured to send POST requests to the n8n webhook URL.
• Google Gemini API Key: Credentials for the Google Gemini Chat Model to power the LLM chain. This requires configuring a credential in n8n.
• Web Search API (External): While the workflow includes an HTTP Request node for web search, you will need to configure it to connect to your preferred web search API (e.g., Google Custom Search API, SerpApi, etc.) and provide the necessary API keys/credentials.

Setup/Usage

1. Import the Workflow:
   • Copy the provided JSON code.
   • In your n8n instance, go to "Workflows".
   • Click "New" or "Import from JSON" and paste the workflow JSON.
2. Configure Credentials:
   • Locate the Google Gemini Chat Model node.
   • Click on the "Credential" field and select an existing Google Gemini credential or create a new one, providing your API key.
   • Locate the HTTP Request node (intended for web search). Configure its URL, headers, and body to interact with your chosen web search API. You may need to add a new credential if your API requires authentication.
3. Activate the Webhook:
   • The Webhook node will display a unique URL once the workflow is saved and activated.
   • Configure your external system to send POST requests to this URL with your research queries (e.g., an array of strings or objects containing a query field).
4. Customize Workflow Logic (Optional):
   • HTTP Request (Web Search): Adjust the HTTP Request node to match the specific API endpoint and parameters of your chosen web search engine. You'll likely need to extract the search query from the incoming webhook data.
   • Loop Over Items: Modify the "Batch Size" and "Delay" in the Loop Over Items (Split in Batches) and Wait nodes to optimize performance and respect API rate limits.
   • Basic LLM Chain: Customize the prompt in the Basic LLM Chain to guide the Google Gemini model on how to process the search results and what kind of output to generate.
   • Structured Output Parser: If your desired output structure changes, update the schema in the Structured Output Parser node.
5. Activate the Workflow: Toggle the workflow to "Active" in the n8n editor.
6. Test: Send a test request to the webhook URL from your external system to ensure the workflow executes as expected.