GPT Action History and AI Memory Keys in Workflows

By Sparkforge Support Team ·Last reviewed May 3, 2026

GPT actions in Sparkforge workflows can now retain history across five scopes — sub-account, workflow, execution, step, or custom — for more relevant, context-aware responses per contact.

GPT actions in Sparkforge workflows now retain conversation history. Each action can pull prior context into the next prompt, so responses stay relevant to the contact and the situation.

History scopes

  • This Sub Account: remembers GPT conversation across every workflow in the sub-account that uses this scope.
  • This Workflow: retains history within a single workflow.
  • Per Execution: retains history within one execution of a workflow.
  • This Step: retains history for a specific action across multiple executions.
  • Custom: a named history bucket you can reuse across workflows or steps.

Enable history on a GPT action

  1. Enable the option in Labs.
  2. Add the GPT Powered By OpenAI action to your workflow.
  3. Toggle Enable history on.
  4. Choose the History for scope from the dropdown.
  5. Add system instructions if needed.

Advanced options

  • System Instructions: persistent rules every GPT action follows in this workflow.
  • Exclude instructions from history: keeps system instructions out of the stored conversation.
  • Exclude responses from history: useful when responses are sentiment scores or analytics outputs you don't want feeding back into future prompts.

Things to know

  • Enabling history defaults the action to GPT 4 models.
  • History only works for the Custom action type.
  • History is independent for each contact — one contact's conversation never bleeds into another's.

Frequently Asked Questions

Which history scope should I pick?
Use Per Execution for one-off, in-flight reasoning. Use This Workflow or This Sub Account for ongoing personalization. Use Custom when multiple workflows need to share context.
Does history mix across contacts?
No. History is scoped per contact, so personalization stays accurate and conversations don't leak between people.
Can I keep system instructions out of the stored history?
Yes. Toggle Exclude instructions from history to keep your prompt rules private and out of the conversation log.
Why does enabling history switch me to GPT 4?
GPT 4 models handle longer context windows reliably, so history-enabled actions default to GPT 4 to keep responses coherent.

Was this article helpful?

Related Articles