askEdgi

New & Improved

Consistent Result Formatting Using Data Stories

An enhancement has been introduced in askEdgi to ensure consistent formatting of analytical results using a designated Data Story. Previously, results could appear with varying numeric precision or formatting, leading to inconsistencies across responses.

With this update, administrators can define organization-wide presentation guidelines as free-text instructions in a specific Data Story. During analysis, askEdgi retrieves these instructions and injects them into the analytic agent’s prompt, ensuring that outputs adhere to the defined formatting standards. If no Data Story is configured, the system continues with default behavior. This improvement delivers consistent, standardized results across askEdgi responses.

Prompt Guidance for Next-Best Question Suggestions

A new Prompt Guidance feature has been introduced in askEdgi to help users determine what to ask next during exploratory or analytical workflows. Previously, users often faced uncertainty after submitting an initial prompt, which could slow down analysis and require manual refinement of the prompt.

With this enhancement, a lightweight guidance agent analyzes the user’s current prompt, context, and response to suggest relevant follow-up questions that deepen or refine the analysis. These suggestions are presented proactively but do not trigger full processing unless the user selects one. The guidance operates independently of the core discovery and execution flow, ensuring minimal system overhead. This improvement helps users move naturally from high-level questions to more detailed insights, reduces the need for prompt engineering, and improves the overall conversational experience.

UI Improvements for RAG Suggestions and Object Navigation

Usability enhancements have been introduced in askEdgi to improve the visual hierarchy and navigation on the RAG Suggestion Table and Object Summary pages. Previously, all action buttons were styled as primary, creating visual clutter and making it difficult to identify the most important action. Additionally, there was no direct way to open objects in the Data Catalog from contextual views.

With this update, action buttons in the RAG Suggestion Table are styled as secondary, with the primary style reserved for critical actions. The Object Summary page also includes a new “Open in Data Catalog” icon with a tooltip, which opens the corresponding object directly in the Data Catalog in a new tab. These changes provide a cleaner interface, clearer action prioritization, and faster navigation across modules.

Enable Tool Calls in Focused Mode for Business Glossary Formulas

An enhancement has been implemented in askEdgi to allow tool calls while operating in Focused mode. Previously, tool execution was restricted in this mode, preventing the system from applying Business Glossary formulas and related logic when generating responses.

With this update, askEdgi can now invoke the necessary tools even in Focused mode, enabling accurate processing of Business Glossary formulas and ensuring consistent results across interaction modes.

Update Custom Fields

A new capability has been introduced in askEdgi to update metadata through the Take Action > Update Metadata Fields workflow. Previously, this action did not support updating custom metadata attributes via askEdgi.

With this update, users can now modify Custom Fields directly for all objects using the Take Action feature. This implementation of custom fields applies to all object types.

Take Action: Support for Updating Governance Roles

A new enhancement enables users to update governance roles directly from askEdgi analysis results through the Take Action framework. Previously, askEdgi could recommend potential Owners, Stewards, or Custodians based on metadata usage. Still, users had to manually navigate to governance screens to apply these changes, resulting in lost context and additional effort.

With this update, authorized users can now execute the Update Governance Roles action in askEdgi to assign roles to selected cataloged assets. The action supports mapping suggested users to governance roles, such as Owner, Steward, Custodian, and other configured roles, and validates mandatory fields and asset eligibility. Only governable assets are selectable, and the system enforces object-level permissions and role assignment privileges.

Role updates require explicit user approval and are processed as background jobs, with status tracking and detailed audit logging of changes, including previous and new values. This enhancement converts AI-driven governance recommendations into actionable outcomes, reduces manual steps, preserves analytical context, and ensures secure, compliant role management.

Configure Actions | Pluralized Action Labels

An update has been made to the askEdgi Configure Actions to improve label clarity and consistency. Previously, action names were displayed in singular form (for example, Assign Tag, Create Term, Post Question, Create Data Quality Service Request), even when applied to multiple selected records.

With this change, action labels now appear in plural form to accurately reflect bulk operations. Updated labels include Assign Tags, Create Terms, Post Questions, and Create Data Quality Service Requests.

Organization Recipes | Intent-Based Automatic Recipe Execution

In the askEdgi module, the system now evaluates existing Organizational Recipes before generating analysis dynamically, enabling better reuse of predefined workflows. The functionality is now enhanced to analyze user intent, match it with Organizational Recipes, and execute the relevant recipe automatically in the background within the same thread.

The following improvements have been implemented:

  • Organizational Recipes are identified using intent and metadata such as name, description, and tags.

  • Matching recipes are executed automatically instead of generating new commands.

  • Execution occurs within the same thread without impacting user experience.

  • Results from the executed recipe are displayed directly in the conversation.

Take Action | Configure Action UX Improvements

In the askEdgi module, when an action was already configured, the Configure Action flow displayed the list of actions instead of the mapped template and included inconsistent UI behavior. The functionality is now enhanced to directly display the mapped template when Configure Action is clicked, remove the Back button, replace the Save button with Update, and improve visibility of the action name.

Service Desk | Legacy Data Quality Status

In the Service Desk module, there was no structured mechanism to capture and process legacy data quality context for Data Quality Issue requests. The functionality is now enhanced with a new template property, Legacy Data Quality Status, which enables capturing context when the status is moved to “Legacy Data Quality Issue” and triggers an internal fulfillment job to store the context against the object.

The following changes have been implemented:

  • Introduced template property Legacy Data Quality Status with default behavior for Data Quality templates.

  • Enabled new action “Mark as Legacy Issue” with status “Legacy Data Quality Issue.”

  • Context provided in Corrective Action is stored against the object when the status is updated.

AI Prompt and Context Visibility

In the askEdgi module, users could not view the AI prompt and context used for generating columns or results, limiting transparency and validation. The functionality is now enhanced to provide read-only visibility of the prompt and context through the AI icon or a dedicated action.

The following improvements have been implemented:

  • Prompt and context are visible for AI-generated outputs.

  • Additional details such as function, input columns, output column, and timestamp are available.

  • Access is restricted to read-only for validation and explanation purposes.

Take Action | Create Project Tasks

In the askEdgi module, analytical results could not be converted into actionable and trackable work within Projects. The functionality is now enhanced to allow creation of Project Tasks directly from askEdgi results, supporting parent-child hierarchy, object and non-object tasks, and assignment handling.

Take Action | Create Data Quality Rule

In the askEdgi module, analysis results could not be directly converted into Data Quality rules and required manual intervention. The functionality is now enhanced to allow creation of Data Quality rules from askEdgi results through a governed action framework with mandatory mapping, explicit approval, and asynchronous execution.

Workspace Download Control (Role-Based)

In the askEdgi module, users could not download data from workspace objects, and the download behavior was inconsistent across different object types. The functionality is now enhanced to support download for all workspace objects with role-based access control, ensuring consistent behavior across workspace, analysis, and chat views.

The following improvements have been implemented:

  • Download is enabled for catalog and non-catalog workspace objects.

  • Role-based configuration controls access to the download functionality.

  • Behavior is consistent across workspace, analysis, and chat views.

Prompt Guidance for Next-Best Question Suggestions

In the askEdgi module, users lacked guidance on what to ask next after submitting a prompt during exploratory workflows. The functionality is now enhanced with a Prompt Guidance agent that analyzes the current prompt, context, and response to suggest relevant next-best prompts under “You can ask Edgi.”

Thread-Level Prompt History and Summary

In the askEdgi module, users could not easily navigate, search, or recall previous prompts within long chat threads. The functionality is now enhanced with a Prompt History feature that provides a slider view with search and navigation capabilities to access prompts and their corresponding results.

RAG Retrieval | Metadata-Aware Retrieval and Scoring Model

In the Data Catalog module, asset retrieval now uses a metadata-aware RAG retrieval architecture to improve how datasets are identified and ranked for analysis. The system evaluates both relevance and governance signals to ensure that returned assets align with the user query and meet defined data governance standards.

The following improvements have been implemented:

  • Retrieval combines vector similarity and keyword-based search to identify candidate assets

  • Candidate assets are re-ranked using metadata signals and governance attributes

  • StructuralScore evaluates glossary hierarchy, tag hierarchy, and custom field relevance

  • GovernanceScore evaluates CDE, data quality, certification status, and authoritative flag

  • Final ranking is calculated using a weighted scoring formula

  • Only top-ranked assets are selected for analysis and response generation

  • Weight configuration is supported to align with organizational metadata and governance practices

Workspace | Enable data download for all workspace objects

In the askEdgi Workspace, data download is now supported across all workspace objects to ensure consistent access and usability. This enhancement enables users to download data from catalog tables, non-catalog tables, AI-generated tables, and uploaded files across workspace, analysis, and chat views.

The following improvements have been implemented:

  • Download is available for all workspace objects regardless of source

  • Download access is controlled using role-based configuration

  • Admins can configure allowed roles using the defined system setting (askEdgi.workspace.download.allowed.roles)

  • The download option is available in table view and analysis/chat outputs

  • Permission is enforced consistently across workspace, analysis, chat, and recipe outputs

  • The download option is hidden or disabled for users without the required roles

Prompt Discovery and Recipe Creation

In askEdgi, prompt discovery and reuse capabilities have been enhanced to improve usability and reduce manual effort in recipe creation. This enhancement enables users to discover prompts across chats and convert them into reusable recipes.

The following improvements have been implemented:

  • Prompts are displayed as a flat list across all chats, sorted by execution date

  • Search is supported based on the prompt text

  • Filters allow selection by source chat and time

  • Users can select multiple prompts across chats and dates

  • Selected prompts can be converted into recipe steps

  • Recipe steps are created in execution order with auto-extracted ingredients

  • Only valid analytical prompts are included for selection

Data Quality | Structured Data Quality Context Extraction

In askEdgi, data quality handling has been enhanced to support structured context extraction for legacy data issues. This enhancement enables the system to classify issues and apply appropriate remediation through governed workflows.

The following improvements have been implemented:

  • Mandatory classification of legacy data quality issues during approval

  • Conditional capture of Data Cleanup Context for remediation

  • The cleanup context is stored and linked to the respective object

  • Workspace objects are highlighted when a cleanup context exists

  • Users can review the cleanup context before execution

  • Backend cleanup generates transformed datasets without modifying source data

  • Post-cleanup indicators and notes are displayed for transparency

  • Users can revert to the original data by removing the cleaned dataset

AI Generative Description | AI Instructions configuration

In the AI Prompt Questions module, AI description generation has been enhanced to support standardized and controlled output. This enhancement introduces a new configuration layer for managing reusable instruction templates.

The following improvements have been implemented:

  • A new AI Instructions tab has been added for managing instruction templates

  • Administrators can create, edit, activate, and deactivate instructions

  • Instructions are scoped by Object Type and Description Type

  • Only active instructions are available in the AI Description Generator

  • Instruction text is displayed in truncated format with full view on hover

  • Standardized prompts ensure clear, concise, and consistent descriptions.

Audit Trails | Take Action Logging and Source Tracking

In the Audit Trails module, actions triggered through askEdgi Take Action are not consistently tracked, resulting in limited visibility into who performed actions and what changes were made across Service Desk, Data Quality, Business Glossary, and Data Catalog. The functionality is now enhanced to ensure all Take Action activities are logged with complete execution details and source tracking across modules.

The following improvements have been implemented:

  • Take Action executions are logged under Others > askEdgi > Take Action with Action Name, Job ID, execution details, Updated By, and Date

  • Creation of Data Quality Service Desk tickets via Take Action is logged with Object Type, Summary, Description, Updated By, and Date

  • Tags assigned through Take Action include a new Source column with value ‘askEdgi’ in Audit Trails

  • Data Quality Rules created via Take Action include Source as ‘askEdgi’ in Audit Trails and are marked in Audit History

  • Terms created through Take Action include a new Source column with value ‘askEdgi’ in Governance Catalog

  • Logging is applied consistently across Service Desk, Data Quality, Business Glossary, and related modules to ensure traceability of actions.

Fixed

Recipes Marketplace | Publishing Error Resolved

In Recipes | Marketplace, an issue in which publishing certain recipes displayed an “[object object]” error message has been resolved. The error occurred during publication, even though the recipe was created and executed successfully, while other recipes were published without issues. This issue prevented specific recipes from being added to the marketplace.

The publication process has been updated to handle these recipes correctly, enabling them to be published to the marketplace without any errors.

Fixed Governance Role Cascading Across Hierarchy

In askEdgi, an issue where governance roles did not cascade to related hierarchy levels when the “Cascade to Hierarchy” option was enabled has been resolved. Even when cascading was set to true, updating governance roles for selected objects, such as tables, did not apply the changes to associated hierarchical objects as expected.

The update process has been corrected to ensure governance roles properly cascade across the hierarchy when the option is enabled.

Business Glossary Formulas Now Apply Without Pinning

In askEdgi, an issue where the Business Glossary Formula was applied in Analysis Mode only when the related data object was pinned has been resolved. When the object was not pinned, the configured formula did not run, even though it was correctly defined for the dataset. As a result, responses did not reflect the intended glossary logic.

The formula application logic has been updated to ensure Business Glossary Formulas are applied in Analysis Mode regardless of whether the object is pinned.

Improved askEdgi Response When Data Is Missing

In askEdgi (RAG), an issue where clicking a suggested follow-up question did not display a meaningful response when the required data was unavailable has been resolved. When a dataset lacked necessary fields, such as a timestamp for trend analysis, the system displayed “Unable to Process Request” and did not clearly present the selected question or explain the limitation.

Response handling has been improved to provide clear, informative messages when data are insufficient and to suggest alternative valid questions based on the available dataset.

Workspace Project Selection Update

In the Data Catalog Summary Page, when “My Workspace” was set as the default project, the Add to Workspace pop-up showed all projects, including those where the object was already added. This made it difficult to understand the correct selection. This issue has been resolved. The pop-up now displays only the relevant project details.

Action Names Display

In askEdgi Configure Actions, some action names were not shown in the correct format. A few actions followed the expected naming, while others did not. This issue has been resolved. All action names are now shown in a consistent format, such as Assign Tags, Create Terms, Create Data Quality Service Requests, Create Data Quality Rules, Post Questions, Update Governance Roles, and Update Custom Fields.

Data Products Access Based on Edition

In askEdgi Data Products, adding objects to askEdgi was earlier controlled by the system host type setting instead of the edition setting. This was due to incorrect mapping in the implementation. This issue has been resolved. The control now works based on the “askedgi.edition” system setting. When the value is Data Analytics, adding objects from Data Products is allowed. When the value is Meta Data Analytics, this option is not available.

Deleted Tables Removed from Analysis

In askEdgi Workspace, when a table was pinned and later deleted, it was still used in analysis. This continued even after refreshing the workspace and confirming that the table no longer existed. This issue has been resolved. Deleted tables are now removed from the system context immediately. Only currently available and selected tables are used in the analysis.

3D Chart Issue

In askEdgi, when an analysis required a 3D chart, the system was not able to generate it. This limited the ability to view multi-level data clearly. This issue has been resolved. askEdgi can now generate 3D charts as expected for such requests.

Recipe Suggestions Display

In askEdgi Recipes, while executing a recipe, the suggestions section showed incorrect values. This happened when the connection name contained the “|” character, which affected how the table was displayed. As a result, object names appeared in place of other details.

This issue has been resolved. The “|” character is now properly handled before generating the output, ensuring suggestions display the correct and relevant details during recipe execution.

Long Chat Issue

In askEdgi, after several prompts in the same chat, results stopped appearing. Even questions that worked earlier in the same chat did not return results when asked again. This issue has been resolved. The system now handles longer chats properly and continues to return results without stopping.

Recipe Screen Issues

In askEdgi My Workspace, while saving chats as recipes, the ingredients section was not displayed. Also, the page could not be scrolled down, which blocked further actions. This issue has been resolved. Ingredients are now shown correctly, and the page scroll works without any issues.

Recipe Hover Text

In askEdgi Recipes, the hover text did not appear when it contained hyperlinks. It only worked when the content had plain text. This issue has been resolved. Hover text now appears correctly for both plain text and content with links.

Tooltip Text

In askEdgi, the tooltip shown for “Add to Workspace” displayed incorrect or unclear information. This issue has been resolved. The tooltip now shows clear and correct information about the action.

Recipe Suggestions After Object Removal

In askEdgi, when an object was removed from the workspace and a recipe was executed, the removed object still appeared in suggestions. This issue has been resolved. Only objects that are currently available in the workspace are shown in suggestions.

Data Quality Rule Creation – Data Format Check (Regex)

In askEdgi, within the Take Action feature, Data Quality Rules were not created for the function Data Format Check (Regex), even when recommended through the Data Quality Rule Recommendation recipe. This resulted in missing rule creation for applicable scenarios.

This issue has been resolved. Data Quality Rules are now created successfully for the Data Format Check (Regex) function.

Data Quality Rule Creation – Missing Input and Success Criteria Values

In askEdgi, within the Take Action feature, Data Quality Rules were created without input values, success criteria, and validation details, even when these were provided during rule creation. This resulted in an incomplete rule configuration. This issue has been resolved. Data Quality Rules are now created with input values, success criteria, and validation details as expected.

Data Quality Rule Creation – Irrelevant Fields and Incorrect Labels

In askEdgi, within the Take Action feature, while creating Data Quality Rules, irrelevant fields and incorrect labels were displayed in the configuration form. This created confusion and impacted clarity during rule creation. This issue has been resolved. Only relevant fields are now displayed with correct labels in the Data Quality Rule configuration form.

Update Governance Roles – Invalid User Handling

In askEdgi, within the Take Action feature, when a Viewer license type user or an invalid user was assigned to a governance role, the existing governance roles on the object were removed, and the value was set to empty. In addition, job status showed success while the operation failed.

This issue has been resolved. Existing governance roles are now retained when invalid users are provided, and job logs correctly display failure with the appropriate message “Failed to update governance role due to invalid user ID.”

Recipes – Ingredient Limit and Validation Handling

In askEdgi Recipes, when the number of selected ingredients exceeded 9, the recipe could not be saved, and no validation message was displayed. In addition, existing ingredients were not displayed during recipe editing.

This issue has been resolved. Validation messages are now displayed when the ingredient limit is exceeded, and existing ingredients are shown correctly during edit.

Workspace Logs

In the askEdgi module, workspace logs were displayed in chronological order, requiring users to scroll to view recent entries. The issue has been resolved, and now logs are displayed in reverse chronological order, with the latest entries appearing at the top.

Configure Action | Unassign Action

In the askEdgi module, users could not remove or delete a configured action from the Configure Action step after it was assigned. The issue has been resolved, and now users can unassign an action using the Unassign Action option, where a confirmation pop-up is displayed to confirm or cancel the action.

Workspace | Copy option for objects and data

In the askEdgi Workspace, the issue where users could not copy object names or data values, which led to manual effort and inconsistent behavior. The issue has been resolved, and copy functionality is now available for workspace objects and table data, aligning with Data Catalog behavior.


Copyright © 2026, OvalEdge LLC, Peachtree Corners, GA, USA.

Last updated

Was this helpful?