AI Research Report Generator using Jina and Gemini - n8n Workflow

Researchers and Academics: Automating the literature review and report drafting process, ensuring all sources are properly cited.
Content Marketing Teams: Generating long-form, authoritative content quickly based on up-to-date web data.
Automation Specialists: Users looking for advanced examples of chaining multiple AI agents and services within an n8n workflow.
Technical Analysts: Individuals needing structured data aggregation and transformation using custom n8n node scripting.

This sophisticated automation represents a powerful application of the n8n workflow system, solving the common challenge of synthesizing disparate online information into a single, cohesive, and cited report. It leverages Jina AI's ability to efficiently search the web and read content from various URLs, then passes this raw data to a multi-stage chain of Google Gemini 2.5 Flash models.

The core value provided by this n8n template lies in its high degree of structure and quality control. It doesn't just generate text; it employs a dedicated 'Generator Agent' with strict citation rules and then routes the output through an 'Evaluator Chain' n8n node that verifies citations, checks factual consistency, and optimizes markdown formatting. This ensures that the final output delivered by the n8n workflow is polished, authoritative, and ready for use.

This comprehensive n8n workflow operates through a structured process, triggered by a user's query:

1. Initiation (n8n trigger): The process begins with the 'When chat message received' n8n node, which acts as the primary n8n trigger, capturing the user's research topic or question.


2. Web Search: The user input feeds into the 'Search web' Jina AI n8n node, which performs a targeted web search and returns a list of relevant URLs.


3. URL Preparation: A 'Code' n8n node (Python script) extracts and cleans up the URLs, preparing them for iteration.


4. Content Loop: The 'Loop Over Items' n8n node initiates a batch processing loop, handling each URL individually.


5. Reading and Summarization: Inside the loop, the 'Read URL content' Jina AI n8n node fetches the full text content of the URL. This content is immediately routed to the 'Summarizer Agent' n8n node (powered by Gemini 2.5 Flash), which produces a structured summary including the source URL.


6. Rate Limit Management: A 'Wait' n8n node is strategically placed after summarization to introduce a 1-second pause, ensuring the n8n workflow adheres to API rate limits during heavy looping.


7. Data Aggregation: After the loop completes, the 'Transform' n8n node (another Python script) consolidates all individual, structured summaries into a single JSON object. This is a critical step before report generation.


8. Report Generation: The aggregated summaries are fed into the sophisticated 'Generator Agent' n8n node, which uses detailed system instructions and the Gemini 2.5 Flash model to synthesize the information and draft the full research report, complete with inline citations.


9. Evaluation and Refinement: Finally, the raw report enters the 'Evaluator Chain' n8n node. This final Gemini model ensures citation accuracy, verifies factual support, and formats the output into a professional, publication-ready markdown structure, delivering the high-quality output of this specific n8n workflow.

To deploy this powerful n8n workflow, follow these steps:

1. Import the n8n template: Copy the provided JSON data and navigate to your n8n instance. Click 'New' > 'Import from JSON' and paste the code.


2. Jina AI Credentials: Locate the 'Search web' and 'Read URL content' n8n node instances. You must set up or select your Jina AI API credential. This allows the n8n workflow to access web content.


3. Google Gemini Credentials: Locate the 'Summarizer Model', 'Generator Model', and 'Evaluator Model' n8n node instances. Configure a credential for the Google Gemini API (using the Google Palm/Gemini connector) to enable AI processing.


4. Set the n8n trigger: Activate the 'When chat message received' n8n trigger node to ensure the workflow is listening for input. You may need to configure a chat platform connection (like Telegram or Slack) or use the built-in webhook ID if deploying a self-hosted instance.


5. Test Execution: Run a test query through the chat trigger to ensure all connections and the entire n8n workflow execute successfully.

When chat message received (n8n trigger): The starting point of this n8n workflow. It listens for incoming chat messages, using the user's input as the search query for the research task.
Search web (Jina AI n8n node): Performs the initial web search using the user's chatInput. Key configuration: operation: search, with the query dynamically mapped from the incoming message.
Code (n8n node - Python): Acts as a data transformer, iterating over the raw Jina AI results to extract only the necessary url property into a simplified list structure required for the subsequent loop.
Loop Over Items (n8n node): Essential for handling multiple search results. This splits the list of URLs into individual items, allowing the n8n workflow to process and summarize each source sequentially.
Read URL content (Jina AI n8n node): Fetches the full, clean content for each URL passed from the loop. This content is crucial raw material for the AI summarization step.
Summarizer Agent & Summarizer Model (LangChain Agent n8n node): An AI agent using Gemini 2.5 Flash to summarize the fetched web content based on the original user query, utilizing a 'Structured Output' n8n node parser to ensure JSON consistency.
Wait (n8n node): Configured to wait for 1 second. This core n8n node manages API load and prevents rate limiting during the loop execution.
Transform (n8n node - Python): Aggregates the structured summaries from the loop into a single, cohesive input list (output) for the final Generator Agent.
Generator Agent & Generator Model (LangChain Agent n8n node): This highly configured n8n node utilizes the Gemini 2.5 Flash model and an extensive prompt (including source analysis, structure requirements, and detailed citation protocols) to draft the full research report from the summarized data.
Evaluator Chain & Evaluator Model (LangChain Chain n8n node): The final quality control stage. This n8n node refines the generated report, focusing specifically on citation verification, markdown optimization, and ensuring professional structure before the final output.