Introducing History for GPT actions..!!
TABLE OF CONTENTS
What's New ?
- We're excited to announce a significant enhancement to our GPT action: the ability to retain and utilize history. The GPT actions will be able to produce more relevant and personlized responses based on the past interactions with GPT actions.
Type of History
There are 5 types of History that can be stored
- This Sub Account - Remembers all GPT conversation across all workflows in the sub account where the History type is "This Sub Account"
- This Workflow - Remembers all GPT conversation within the workflow where the History type is "This Workflow"
- Per Execution - Remembers all GPT conversation within the workflow for a single execution where the History type is "Per Execution"
- This Step - Remembers all GPT conversation for the specific action across multiple executions where the History type is "This Step"
- Custom - Let's the user create a custom History type that can be used within the same workflow or multiple workflows
How to use?
- Enable the option in Labs
- Select the "GPT Powered By OpenAI" action
- Switch on "Enable history"
- Select the "History for" from the dropdown
- Add system instructions
Advanced Options
- System Instructions - User can add Instructions if required to give more information to GPT action to get more specific and desired output. These are the rules that can be added which will be followed by all the GPT actions.
- Exclude instructions from history - This toggle can be used when user wants to exclude the System Instructions from the "history"
- Exclude responses from history - If this is turned on then responses of this action will not be stored in the history. This is useful when the responses are sentiment (positive or negative) or any analytics that were received as output.
Points to be noted
- If Enable History is switched on then by default GPT 4 models will be selected.
- History will only work for "Custom" action type
- History will be independent for each contact
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article