See llms.txt for all machine-readable content.

Back to Templates

Auto-document and backup workflows with GPT-4 and Airtable

Created by

Created by: Guillaume Duvernay || duv
Guillaume Duvernay

Last update

Last update 11 days ago

Categories

Share

Quick overview

Automatically backs up every new version of your n8n workflows to Airtable on a schedule, then uses AI to generate a plain-English summary of each workflow and a one-sentence changelog of what changed since the last snapshot.

How it works

  1. A Schedule Trigger runs the workflow at the frequency you set (for example, daily), starting with the Get all n8n workflows node, which fetches every workflow from your n8n instance via the n8n API.
  2. In parallel, Search all existing snapshots reads all previously stored version IDs from the Snapshots table in Airtable, and Match snapshot IDs (Merge, keepNonMatches) compares them against the current version ID of each workflow to isolate only the workflows with a new, unsaved version. Only keep new snapshots filters out any leftover items without a version ID before looping through the results with Loop Over Items.
  3. For each new version, Create or update workflow upserts a record in the Workflows table (matching on Workflow ID), setting the workflow name and creation time. Search all snapshots then retrieves the past snapshot count for that workflow, and the Check workflow status Code node decides whether the AI summary needs to be regenerated (on the 1st, 2nd, and every 5th snapshot).
  4. If a previous snapshot exists, Download previous snapshot and Extract from File1 pull the last stored .json backup from Airtable, and Document workflow differences (an OpenAI-backed LangChain chain, via OpenAI Chat Model1) compares it to the current workflow JSON to produce a one-sentence changelog. If it's the very first snapshot, State that this is the first version sets a fixed N/A - First identified version note instead.
  5. When Needs new workflow summary? is true, Re-summarise workflow calls the same AI model to generate a fresh, one-sentence description of the whole workflow's purpose, which Store new workflow summary saves back to the Workflows table's AI summary field.
  6. Store new snapshot* creates the new Snapshots record (version ID, update time, changelog, linked Workflow), then Get full workflow JSON, Move Binary Data, and Extract from File convert the full workflow JSON into a binary file. Store workflow file into Airtable (HTTP Request to the Airtable content API) uploads that file as an attachment to the snapshot's Workflow_file field, and a Wait node pauses 3 seconds before the loop continues, to stay under Airtable's API rate limits.

Setup

  1. Duplicate the Airtable base template (linked in the workflow's sticky note) into your own Airtable account. It contains the two required tables: Workflows (workflow name, AI summary, linked snapshots) and Snapshots (n8n version ID, update time, AI changelog, Workflow_file attachment).
  2. Add your n8n API credentials to the Get all n8n workflows, Get full workflow JSON, and Create or update workflow (n8n node type) nodes so the workflow can list and read your instance's workflows.
  3. Add your Airtable credentials to all Airtable nodes (Search all existing snapshots, Search all snapshots, Create or update workflow, Download previous snapshot, Store new workflow summary, Store new snapshot) and remap each node's base and table to the base you just duplicated.
  4. Add an OpenAI (or compatible) credential to the OpenAI Chat Model1 node, which powers both the Document workflow differences and Re-summarise workflow AI chains. It defaults to gpt-4.1-nano; swap the model if you need higher quality summaries.
  5. In the Store workflow file into Airtable HTTP Request node, replace <AIRTABLE-BASE-ID> in the URL with your duplicated base's ID (starts with "app...", found in the base's Airtable URL). This node also needs an Airtable Token API credential (nodeCredentialType airtableTokenApi) with attachment upload permission.
  6. Set your desired frequency on the Schedule Trigger node (for example, once daily) and activate the workflow.

Requirements

  • An n8n instance reachable via the n8n API (with an API key credential).
  • An Airtable account with a base duplicated from the provided template, plus an Airtable API/token credential.
  • An OpenAI (or compatible) API credential for the two AI summary/changelog steps.

Customization

  • Add a Slack or Email node at the end of the run to notify your team which workflows were backed up.
  • Swap Airtable for Google Drive, Dropbox, Google Sheets, or Notion by adapting the storage and metadata nodes to those services.
  • Edit the Check workflow status Code node's logic to change how often the full AI summary is regenerated (currently: 1st, 2nd, and every 5th snapshot), to control AI token usage.

Additional info

Duplicate the reference Airtable base before running this workflow: https://airtable.com/appPFFj6CUUhZyDPT/shrorM8k6HsUqBACB. The workflow uses Airtable's content upload API endpoint to store the workflow JSON as a file attachment directly from base64 data, avoiding a separate Drive/Dropbox upload step.