Beginner AI Dataset Generator using OpenAI + LangChain in n8n

This n8n workflow dynamically generates a realistic sample dataset based on a single topic you provide. It uses OpenAI (via LangChain) and n8n’s built-in nodes to:

1. Generate structured JSON data for 5 columns with 3–5 values each
2. Flatten that data into a single text blob
3. Infer meaningful column names via a second AI call
4. Pivot, split, merge, and rename columns automatically
5. Output a clean, labeled dataset ready for export or further processing

---

⚙️ Prerequisites

1. OpenAI API Key

   • Visit: https://platform.openai.com/account/api-keys
   • Create a new key
   • In n8n: Credentials → New → OpenAI API, paste key, name it “OpenAi account”

2. LangChain nodes enabled in your n8n instance

🥇 Step 1: Set Up OpenAI Credential

1. Go to OpenAI API Keys
2. Create and copy your key
3. In n8n: Credentials → New → OpenAI API → paste key as “OpenAi account”

🥈 Step 2: Manual Trigger

• Add Manual Trigger to start the workflow

🥉 Step 3: Set Topic

• Add a Set node named Set Topic to Search
• Field: Topic = n8n use cases (or any topic you choose)

✨ Step 4: Generate Structured Data

• LangChain Agent node Generate Random Data
• Connect to OpenAI Chat Model1 and Tool: Inject Creativity1
• System prompt: instruct AI to output 5 columns of realistic values in JSON

🔧 Step 5: Parse AI Output

• Structured Output Parser to validate JSON

🔄 Step 6: Flatten Data

• Code node Outpt all Data to One Field
• Joins all values into a comma-separated string for column naming

🧠 Step 7: Generate Column Names

• LangChain Agent Generate Column Names
• Connect to OpenAI Chat Model2
• Prompt: infer 5 column names from the string

🔢 Step 8: Pivot Names Row

• Code node Pivot Column Names transforms array into { column1: name1, … }

🪓 Step 9: Split Columns

• 5 SplitOut nodes to break each array back into rows per column

🔗 Step 10: Merge Rows

• Merge node Merge Columns together using combineByPosition

🏷️ Step 11: Rename Columns

• Set node Rename Columns assigns the AI-generated names to each column

🔗 Step 12: Final Output

• Merge Append Column Names combines data and header row

---

🏁 Done! You now have a fully AI-driven, labeled dataset generated from a single topic—no external services needed. Easily extend by adding a Google Sheets or HTTP node to export.

📬 Need Help or Want to Customize This?

📧 [email protected]
🔗 LinkedIn