Curated

(5/5)

Host Your Own AI Deep Research Agent with n8n, Apify and OpenAI o3

2. Ask Clarifying Questions [Read more about form nodes](https://docs.n8n.io/integrations/builtin/core-nodes/n8n-nodes-base.form/) To handle the clarification questions generated by the LLM, I used the same technique found in my "AI Interviewer" template ([link](https://n8n.io/workflows/2566-conversational-interviews-with-ai-agents-and-n8n-forms/)). This involves a looping of dynamically generated forms to collect answers from the user.

Integrations Used (9)

Outputparserstructured

Workflow Structure (66 nodes)

100%

Form Trigger

On form submission

Form

Research Request

Set

Set Variables

Langchain.lm Chat Open Ai

OpenAI Chat Model2

Langchain.chain Llm

Clarifying Questions

Notion

Get Existing Row

Langchain.output Parser Structured

Structured Output Parser1

Notion

Set In-Progress

Split Out

Feedback to Items

Set

Set Initial Query

Split In Batches

For Each Question...

Execute Workflow

Generate Learnings

Form

Ask Clarity Questions

Set

Get Initial Query

Set

Accumulate Results

Is Depth Reached?

Langchain.lm Chat Open Ai

OpenAI Chat Model4

Langchain.chain Llm

Report Page Generator

Langchain.output Parser Structured

Structured Output Parser4

Set

Get Research Results

Notion

Create Row

Execute Workflow

Generate Report

Execute Workflow

Initiate DeepResearch

Form

Confirmation

Form

End Form

Execute Workflow Trigger

DeepResearch Subworkflow

Execution Data

Switch

JobType Router

Langchain.chain Llm

Generate SERP Queries

Langchain.lm Chat Open Ai

OpenAI Chat Model3

Notion

Get Existing Row1

Langchain.lm Chat Open Ai

OpenAI Chat Model1

Langchain.output Parser Structured

Structured Output Parser2

Langchain.chain Llm

DeepResearch Report

Split Out

SERP to Items

Split In Batches

For Each Query...

Markdown

Convert to HTML

No Op

Item Ref

Aggregate

Combine & Send back to Loop

Set

HTML to Array

Http Request

Web Search

Split Out

Tags to Items

Set

Top 5 Organic Results

Langchain.lm Chat Google Gemini

Google Gemini Chat Model

Langchain.chain Llm

Notion Block Generator

Set

DeepResearch Results

Has Results?

Split Out

Results to Items

Split Out

URLs to Items

Set

Parse JSON blocks

Set

Set Next Queries

Http Request

Page Contents

Filter

Valid Pages

Filter

Valid Blocks

Code

URL Sources to Lists

Has Content?

Merge

Append Blocks

Set

Empty Response

Markdown

Convert to Markdown

Split In Batches

For Each Block...

Langchain.lm Chat Open Ai

OpenAI Chat Model

Langchain.chain Llm

DeepResearch Learnings

Langchain.output Parser Structured

Structured Output Parser

Notion

Set Done

Http Request

Upload to Notion Page

Set

Research Goal + Learnings

Node Types:

default

transform

trigger

logic

http

action

code

Requires credentials

Prerequisites

Required Credentials (5)

Google Palm

Used by: Google Gemini Chat Model

Docs

HTTP Header Auth

Used by: Web Search

Docs

Http Query Auth

Used by: Web Search, Page Contents

Docs

Notion

Used by: Create Row, Get Existing Row +4 more

Docs

OpenAI

Used by: OpenAI Chat Model, OpenAI Chat Model1 +3 more

Docs

Multiple services required

This workflow needs 5 different credentials. Make sure you have accounts for all services before importing.

Complexity

Advanced

Expert level. Custom code, complex logic, many integrations.

Setup time: ~1 hour

Score: 100/100

Custom Code

Conditional Logic

Loops

11 Integrations

When to Use This Workflow

AI-Powered Content Creation

Automate content generation, summarization, or analysis using AI capabilities.

Save hours of manual work while maintaining quality and consistency.

Real-time Response

React instantly when events happen - new orders, form submissions, or API calls.

Zero delay between trigger and action for time-sensitive workflows.

Error Reduction

Eliminate human error from routine processes with consistent automation.

Improve accuracy and reliability across your workflows.

workflow.json

{
  "meta": {
    "instanceId": "408f9fb9940c3cb18ffdef0e0150fe342d6e655c3a9fac21f0f644e8bedabcd9",
    "templateCredsSetupCompleted": true
  },
  "nodes": [
    {
      "id": "645ae2b1-799e-49be-8bdf-12cd1bb739e6",
      "name": "Structured Output Parser",
      "type": "@n8n/n8n-nodes-langchain.outputParserStructured",
      "position": [
        1680,
        1140
      ],
      "parameters": {
        "schemaType": "manual",
        "inputSchema": "{\n \"type\": \"object\",\n \"properties\": {\n \"learnings\": {\n \"type\": \"array\",\n \"description\": \"List of learnings, max of 3.\",\n \"items\": { \"type\": \"string\" }\n },\n \"followUpQuestions\": {\n \"type\": \"array\",\n \"items\": {\n \"type\": \"string\",\n \"description\": \"List of follow-up questions to research the topic further, max of 3.\"\n }\n }\n }\n}"
      },
      "typeVersion": 1.2
    },
...