Generate CSV Files from GPT-4 Mock Data

Overview

This workflow generates three CSV files containing mock user data using OpenAI's GPT-4 model. Each file includes 10 randomly generated fictional character names, emails, subscription statuses, and subscription dates. The data is created in a single execution and saved locally, making it ideal for testing, seeding databases, or preparing sample datasets for demonstrations.

Node-by-Node Breakdown

1. When clicking "Execute Workflow" (Manual Trigger) — No auth. Starts the workflow manually when the user clicks the execute button in the n8n editor.
2. OpenAI (OpenAI node) — API Key auth. Sends a prompt to GPT-4 requesting a JSON array of 10 random fictional users. The prompt specifies name/email patterns (first and last name starting with same letter), subscription logic, and date constraints. Options: n=3 (generate 3 completions), maxTokens=2500, temperature=1.
3. Split In Batches — No auth. Splits the three OpenAI responses into individual items (batch size = 1) so each response is processed separately.
4. Parse JSON (Set node) — No auth. Uses JSON.parse($json.message.content) to convert the GPT-4 response string into a JavaScript array of user objects.
5. Make JSON Table (Item Lists node) — No auth. Splits the array into separate items, one per user record.
6. Convert to CSV (Spreadsheet File node) — No auth. Converts the structured user data into CSV format, adding a header row. The filename is dynamically generated using the batch index (e.g., funny_names_1.csv).
7. Strip UTF BOM bytes (Move Binary Data node) — No auth. Removes UTF-8 BOM bytes that can cause issues when reading the CSV file later. Encodes as UTF-8 without BOM.
8. Create valid binary (Move Binary Data node) — No auth. Converts the cleaned CSV string back into a binary file with proper MIME type (text/csv), filename, and no BOM. Source JSON is not kept.
9. Save to Disk (Write Binary File node) — No auth. Saves the binary CSV file to the local filesystem at ./.n8n/{{ $binary.data.fileName }}.

Two sticky notes provide inline documentation about the workflow’s purpose and the BOM fix.

Setup Instructions

• OpenAI Account: You need an OpenAI API key. Obtain one from OpenAI Platform. In n8n, create a new credential of type “OpenAI” and paste your API key.
• Local File Access: The workflow saves files to ./.n8n/ relative to the n8n execution directory. Ensure the n8n process has write permissions there. To change the save location, update the fileName parameter in the Save to Disk node.
• No other external services are required; all processing happens within n8n.

Use Cases and Variations

• Seed data generation: Quickly produce realistic CSV files for populating test databases or demo environments.
• Custom data sets: Modify the GPT-4 prompt to generate other types of data (products, orders, events) by changing the JSON schema and example.
• Output delivery: Replace Save to Disk with a node like Send Email, Google Sheets, or HTTP Request to deliver the CSV files elsewhere.
• Scheduling: Add a Schedule Trigger to generate data automatically at intervals.
• Batch size: Adjust the batchSize in Split In Batches to control how many completions are processed in parallel.

{
  "id": "6FSx5OMVxp8Ldg8A",
  "meta": {
    "instanceId": "fb924c73af8f703905bc09c9ee8076f48c17b596ed05b18c0ff86915ef8a7c4a"
  },
  "name": "Prepare CSV files with GPT-4",
  "tags": [],
  "nodes": [
    {
      "id": "5b43e57d-1fe1-4ea6-bf3d-661f7e5fc4b0",
      "name": "When clicking \"Execute Workflow\"",
      "type": "n8n-nodes-base.manualTrigger",
      "position": [
        960,
        240
      ],
      "parameters": {},
      "typeVersion": 1
    },
    {
      "id": "291466e8-1592-4080-a675-5e9f486d0d05",
      "name": "OpenAI",
      "type": "n8n-nodes-base.openAi",
      "position": [
        1160,
        240
      ],
      "parameters": {
        "model": "gpt-4",
        "prompt": {
          "messages": [
            {
              "content": "=please create a list of 10 random users. Return back ONLY a JSON array. Character names of famous fiction characters. Make Names and Surnames start with the same letter. Name and Surname can be from different characters. If subscribed is false then make date_subscribed empty. If date_subscribed is not empty then make it random and no later then 2023-10-01. Make JSON in a single line, avoid line breaks. Here's an example: [{\"user_name\": \"Jack Jones\", \"user_email\":\"[email protected]\",\"subscribed\": true, \"date_subscribed\":\"2023-10-01\" },{\"user_name\": \"Martin Moor\", \"user_email\":\"[email protected]\",\"subscribed\": false, \"date_subscribed\":\"\" }]"
            }
          ]
        },
        "options": {
          "n": 3,
          "maxTokens": 2500,
          "temperature": 1
        },
        "resource": "chat"
      },
      "credentials": {
        "openAiApi": {
          "id": "63",
          "name": "OpenAi account"
        }
      },
      "typeVersion": 1
// ... truncated (copy to see full JSON)