Skip to main content

GPTfy 1.12 Release Notes

Release Date: Oct 30, 2025

New Features​

  • Agentic Framework : Build and manage intelligent Salesforce-native bots (Agents) that integrate system prompts, AI models, and RAG sync capabilities. Configure custom permissions and timeouts for each Agent with role-based access control. Create dynamic skills that execute specific business operations including record creation, updates, and data retrieval tasks. Access Agents through multiple channels: GPTfy Console, Utility Bar, and REST API endpoints. Maintain comprehensive audit trails with RAG-based knowledge integration for context-aware responses.
  • RAG Sync : Bidirectionally sync Salesforce Knowledge Base articles with external AI RAG model datastores while maintaining Salesforce KB as the authoritative source. Configure automated scheduled syncs or trigger on-demand synchronization with data transformed into standardized JSON format. Apply flexible filtering and extraction rules to control which specific articles are synchronized to the RAG model.
  • Response Validation Framework : Define include/exclude word or phrase patterns to validate AI responses, or create custom Apex-based validation logic for complex business rules. Display instant warning messages when AI responses contain excluded words/phrases or lack required words/phrases. Compatible with all prompt types (Text, JSON, Canvas) and supported across Console, AI Mass Processing, Test Console, REST API, and Flow.
  • Auto File Processing using Response API : Enable file processing workflows without manual file selection by configuring files as content versions in Data context mapping. Trigger file processing through Salesforce automation or third-party systems via REST API calls for seamless external integrations. Automatically display chat icon after response generation for immediate file-based questioning with full conversation mapping to security audit records.
  • LoopBack : Enable AI response simulation mode using custom JSON response templates without consuming actual LLM credits for various testing scenarios. Speed up exception handling testing and workflow debugging with predictable, controlled responses across all GPTfy features. Maintain full security auditing and compliance tracking even for simulated test responses.

Enhancements​

  • Apex Security Layer Improvements : Real-time validation now throws "Invalid Apex Class Name." error when invalid Apex class names are provided while activating apex security layer records. System prevents activation of prompts containing inactive apex security layers with proper error messaging. Apex security layers now properly included in exports for both individual prompt exports and global security layer exports.
  • GPTfy Canvas Enhancements : Canvas prompts display different UI states based on activation status with read-only datatable for active prompts and full editing capabilities for inactive prompts. New custom UI for adding prompt elements with validation filtering and sequence field accepting values 1-999. Canvas prompts can now be cloned along with all associated Prompt Elements. Added Print icon across all interfaces allowing users to save responses as PDFs with proper styling preservation. Full support in global import/export functionality including all prompt elements and summary prompts.
  • Test Console Enhancements : Added zoom and print icons in test console on successful response rendering. Full support for Canvas-type prompts in GPTfy test console.
  • Permission Set-Based Prompt Visibility : Added "Permission Sets" field to prompts with multi-select functionality working alongside existing profile visibility, enabling flexible access control where prompts are visible to users with matching profiles OR assigned permission sets.
  • Time Saved Coefficient Configuration : Added "Time Saved Coefficient" number field to Prompt allowing customizable reading speed coefficient per prompt while maintaining 8 as default value.
  • Voice Configuration Improvements : Voice configuration details list can now be sorted across all three columns.
  • Default Model Update : GPTfy Connection default model is now gpt-4o-mini.
  • Grok-3 Model Support : Added support for Grok-3 AI model, providing users with additional AI model options for enhanced performance and response capabilities.
  • Enhanced Prompt Library : Expanded the default prompt package with new pre-built prompts featuring advanced functionalities and improved use case coverage.

Issues Fixed​

  • Global Export Fixes : Fixed non-functional 'Reset' button for AI Model, Security Layers, API Data Source, and Catalog Cards tabs. Resolved issue where 'Auto Refresh Duration' field value in Canvas Prompt was not fetched during export.
  • Dynamic Formula Field Fixes : Fixed error handling for dynamic formula fields where users encountered "Duplicate field name" message when saving edits.
  • Prompt Action Fixes : Update action target fields can now be modified after saving without requiring deletion and recreation of the action.
  • Test Console Fixes : Fixed "Attempt to dereference a null object" error in test console for prompts without data context mappings, now properly displays validation message.
  • Clone with Related Fixes : Fixed issue where API data source and Apex Security Layer were not being cloned. Fixed error when DCM includes formula field during duplication and prompt activation.
  • Validation Fixes : Fixed issue where users could save Prompt without selecting File Processing Model. Added proper validation for Auto Refresh Duration, Temperature, Top P, Max Output Tokens, and Track Time saved Coefficient fields. Latest created Voice Configuration details now get picked up when multiple records exist with same profile and prompt for an object.

Known Issues​

  • Apex Security Layer Selection : Dropdown in Data context mapping shows only first-created Apex Security layer when multiple security layers share the same apex class name.
  • Canvas Response Validation : Response Validation framework is not supported on Canvas Prompt's elements.
  • Agentic Chat Hyperlinks : Hyperlinks in Agentic chats are not visible and only appear on hover.
  • Response Validation Editing : Response validation on a prompt remains editable even after the prompt is activated.
  • File Processing Model Import : File Processing AI Models are not getting imported using global import when Prompt is being imported.
  • Dynamic Formula Field Error : Shows incorrect error message on Console when GPTfy user runs prompt with inaccessible Dynamic formula fields.
  • Prompt Action with Validation Errors : Prompt Action Framework executes even when security audit is errored due to response validation.
  • JSON Action Mapping : Create record and open form prompt (JSON) action fails when prompt component created on related object is mapped.
  • Custom Field Reference : Referencing custom field in create record/open form mapping fails during prompt activation.