Questions answered from another workflow’s data

Questions answered from another workflow’s data

An n8n workflow that answers a question from data that a second workflow returns. You type a question, an OpenAI model reads
whatever the second workflow hands back and answers from it. The pattern suits any source n8n can reach, such as a sheet, a
database or an API, because the second workflow does the fetching and this one does the asking.

What is inside

• workflow.json: the workflow as n8n exports it, five nodes. Four are the question flow and the fifth is the retriever that
  points at your second workflow. Two sticky notes sit beside them. It holds no credential.
• README.md: these notes, with a small sample second workflow you can paste in to try it.

What we changed from the original copy

• Workflow Retriever pointed at a workflow on the author's own n8n, which was not part of the file. Its ID is
  YOUR_SUBWORKFLOW_ID now.
• The file had no second workflow, so the example question had nothing to read. The sample below fills that gap. It is ours, not
  the author's.
• The credential link on OpenAI Chat Model is gone, and the file is named for what it does.
• Nothing else. The example question is as the file had it.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An OpenAI API key.
• A second workflow that starts with an Execute Workflow Trigger node and ends with a node that returns your data.

A sample second workflow

Copy this JSON, open a new workflow in n8n and paste it onto the canvas. Name it and save it. It returns three made-up customer
records with notes, so the example question has something to find. The addresses use example.com, which is reserved for
examples.

{
  "name": "Customer notes (sample)",
  "nodes": [
    {
      "name": "Execute Workflow Trigger",
      "type": "n8n-nodes-base.executeWorkflowTrigger",
      "typeVersion": 1,
      "position": [0, 0],
      "parameters": {}
    },
    {
      "name": "Customer notes",
      "type": "n8n-nodes-base.code",
      "typeVersion": 2,
      "position": [220, 0],
      "parameters": {
        "jsCode": "return [\n  { json: { name: 'Jay Gatsby', email: 'jay.gatsby@example.com', notes: 'Hosts large parties. Prefers evening calls.' } },\n  { json: { name: 'Daisy Buchanan', email: 'daisy.buchanan@example.com', notes: 'Asked for a quote in March. Waiting on the budget.' } },\n  { json: { name: 'Nick Carraway', email: 'nick.carraway@example.com', notes: 'Referred by Jay Gatsby. Wants a demo.' } }\n];"
      }
    }
  ],
  "connections": {
    "Execute Workflow Trigger": {
      "main": [[{ "node": "Customer notes", "type": "main", "index": 0 }]]
    }
  }
}

The sample ignores the question and returns every record. That works for three records. For real data, read the question in the
Code node with $input.first().json.query and return only the matching rows.

Set it up

1. Paste the sample above into a new workflow and save it. Look at the address bar: the part after /workflow/ is its ID.
2. In n8n, create another workflow, open its menu and choose Import from File. Pick workflow.json.
3. In Workflow Retriever, replace YOUR_SUBWORKFLOW_ID with that ID.
4. Add an OpenAI credential and select it in OpenAI Chat Model.
5. Press Execute Workflow and open Retrieval QA Chain2 to read the answer. It should give the notes and the email for Jay Gatsby.
6. To ask your own question, change the text in Example Prompt.

How a run goes

1. When clicking "Execute Workflow" starts the run by hand.
2. Example Prompt sets one field, input, to What notes can you find for Jay Gatsby and what is his email address?
3. Retrieval QA Chain2 takes that input as its question. At node version 1 it reads {{ $json.input }}.
4. Workflow Retriever runs your second workflow and passes it the question as a field named query. Each item the last node
   returns becomes one numbered passage of context.
5. OpenAI Chat Model writes the answer from that context, and the chain returns it.

Change the look

• The question: the input value in Example Prompt. To take questions from a chat or a form, swap the trigger and set input
  from it.
• The data: your second workflow. Put any nodes you like between its trigger and its last node.
• The model: OpenAI Chat Model.
• Which workflow is read: the ID in Workflow Retriever.

Good to know

• The retriever does no search of its own. Everything the last node of your second workflow returns is given to the model as
  context. A big result costs more and can overflow the model's limit, so filter inside the second workflow.
• The second workflow receives the question as query and nothing else. Only the final run's output is used, not the steps
  before it.
• OpenAI Chat Model sets no model, so n8n uses its default for that node, gpt-5-mini. OpenAI's deprecations page, read on 12
  November 2026, gives 11 December 2026 for its shutdown, with gpt-5.6-terra as the replacement. Pick another model before then.
• Where the data goes: the question and the whole result of your second workflow go to OpenAI on every run.
• Billing: OpenAI charges for each run, in proportion to the size of the data you hand over.