Release Notes 1.5 (Feb-2024)

Updated on February 24, 2024
Updated on February 24, 2024

Release Date: 13-Feb-2024

AppExchange Link: Download GPTfy

New Functionalities

  • COFi Filters for AI Mass Processing: Cross-Object Filter Condition has been introduced to the AI Mass Processing feature. This allows the qualification of the parent record based on conditions applied to its child records during processing.

  • Support for Additional AI Models: GPTfy now supports Azure-based RAG and AWS-based RAG.

  • GPTfy Console on Outlook Salesforce Plugin: Users can now execute the GPTfy prompt and generate responses in Outlook. GPTfy console has been optimized for this experience.

  • GPTfy Console in Experience Cloud: GPTfy is now available on the Salesforce Experience sites. Customers can expose AI prompts to Portal users in a controlled manner. GPTfy console has been optimized for this experience.

Enhancements

  • Prompt Version Reinstate: It allows the user to store the versions of prompts. This helps revamp the older prompts that were working as expected. Whenever a user clicks the activate button, a new version gets created. Once the reinstate button is clicked, the older version is brought in and status is made as a draft.

  • Updated Prompt UI: Prompt user interface is updated to enhance the user experience.

  • Prompts Availability: Prompts can now be conditionally made available on GPTfy User Console. This helps control the number of prompts available on a record based on the record field values.

  • API Data Source: Customers can configure their own API Data source and bring in data that resides outside of Salesforce. This additional data is added it to the JSON request that is sent out to the Gen-AI model.

  • Prompt Deactivate Button: Admins can now deactivate prompts that are not being decommissioned.

  • Only Activated models are available: While creating the Prompt user can only select the Activated AI Model.

  • AI Settings: Enable/Disable prompt versioning and Enable/Disable Exception log have been added.

  • Response Mapping Framework: Response Mapping Framework feature is added to the AI Model. It Establish a JSON node within the connection to map fields returned by the AI to corresponding fields in Security Audit records.

  • Connector class on AI Model: Customers often need to call out their own custom logic to establish a connection. Connector Class is designed to take care of such requirements. Connector Class can be added to the AI Model if named credentials are not required.

  • Delete Card: Delete button is given to the Cockpit cards to delete the unnecessary cards.

  • License Management: License Management is added on the AI Settings page. Users can manage the Pro and Enterprise license keys here.

Bug Fixes

  • Prompt Catalog list view columns are sortable now by Name, Object, etc.

  • Security filter is now applied in a specified order.

  • Retention is only enabled when there is a value in ‘Delete security record after (Days).

  • When a new purpose is created, and a prompt is added, the value in metrics shows the correct count.

  • AI Mass processing can be scheduled only when a prompt scheduler is added.

  • The GPTfy console will only be rendered to the licensed end user.

  • Metadata for cards were not brought in, so the cards for AI models were missing.

  • Metadata for data sources is not updating, so cards for AI data sources are not updating. 

  • When we save metadata for a card, creating duplicate AI models.

  • Only Custom AI Models can be deleted, the system throws an error when a user try to delete the managed package models.

Known Issues

  • Incorrect metrics in Activated AI Models and number shows on metrics.

  • Image processing issue – Docx format is not supported by cognitive service.