Version 2.4.0 | Sept 4, 2025

Overview

ZBrain Builder 2.4.0 enhances accuracy, usability, and performance across monitoring, guardrails, flows, integrations, and RAG. Key highlights include smarter error handling in Knowledge Base creation, configuration validation, refined consent handling and instant notifications in monitoring, as well as an infinite scroll feature for large datasets. Users can edit/delete Knowledge Graph nodes, review files in full clarity, and ensure monitoring test configurations stay aligned with agent and app settings.

This release strengthens transparency, control, and efficiency across the platform.

ZBrain Builder 2.4.0 release overview

New features

Knowledge Base

Enhanced error handling for knowledge base creation

The knowledge base creation process now features enhanced error handling with clear, step-specific messages instead of a generic "Embedding status: errored." Each error provides actionable guidance (e.g., verify API keys, check permissions) to help users resolve issues quickly. The Execute and Finish step interface follows a consistent Figma-aligned design with proper colors and placement for clarity. Errors are displayed at the relevant step and persist until resolved, ensuring transparency and a smoother troubleshooting process.

Navigation

Knowledge-> New Knowledge Base-> Select Knowledge Source → Data Source Configuration → Data Refinement Tuning -> Execute and Finish

Key outcomes

• Improves usability and transparency during knowledge base setup.

• Reduces user confusion with clear and actionable guidance.

• Maintains consistent and professional system messages across browsers and devices.

• Enables faster issue resolution through descriptive, persistent feedback.

Evaluation framework

Verification of monitoring test configurations against agent/app settings

ZBrain Builder introduces configuration validation in monitoring tests to ensure greater accuracy and consistency. When running a test from the Test Evaluation Settings panel, the system now verifies that the model and temperature values match the exact configuration defined in the associated agent or app. Previously, Response Relevancy and Faithfulness evaluations always defaulted to GPT-4o, regardless of the configured settings.

With this update, these metrics now dynamically align with the actual model selected in the entity, making evaluation results more realistic and aligned with runtime behavior.

Navigation

Monitor → Configure-> Event Monitoring Settings → Test

Key outcomes

• Mirrors real-world execution by using the same model and temperature in test evaluations.

• Adapts Response Relevancy and Faithfulness to the configured model instead of being locked to GPT-4o.

• Ensures monitoring outputs reflect actual system configurations, so users can confidently rely on them.

• Eliminates mismatches caused by static evaluation defaults, leading to more accurate insights.

‘Send Notification’ toggle for monitoring test results

ZBrain Builder introduces a Send Notification toggle in the Test Evaluation Settings panel of the monitoring module. This enhancement empowers users to choose whether they would like to receive an email notification containing the evaluation report right after executing a test.

The toggle appears on the right-side Test Evaluation Settings panel under Event Settings → Test. If no flow is selected, it still appears but shows a validation message: ‘Please select a flow before enabling notifications.’ Hovering over the (i) icon displays the message: ‘If enabled, you’ll receive a notification with the evaluation report,’ ensuring clarity for the user. The toggle is currently available for Apps and Reasoning.

Navigation

Monitor → Configure-> Event Monitoring Settings → Test

Key outcomes

• Delivers evaluation results to users instantly after testing instead of waiting for CRON execution.

• Provides clear feedback messages to ensure correct configuration before enabling notifications.

• Works only where relevant (Apps/Reasoning), avoiding unnecessary noise for Agent entities.

RAG

Edit or delete nodes within a knowledge graph

Users can now edit and delete nodes within a knowledge graph, enabling updates to node details or removal of nodes and their direct relations while keeping the rest of the graph intact. Node properties such as name and description can now be updated via a pop-up editor, with changes reflected immediately. Deleting a node requires typing DELETE to confirm, preventing accidental removals. Only the selected node and its direct connections are removed. Even when nodes are removed from visualization, the underlying chunk data remains accessible, ensuring that app and agent responses are not disrupted.

Navigation

Knowledge Base -> Select Knowledge Source → Review Document → Knowledge Graph → Node Details (Popup)

Key outcomes

• Empowers users with direct control over Knowledge Graph refinement.

• Prevents accidental data loss with a confirmation for deletions.

• Improves usability and accuracy by allowing immediate updates to node details.

• Ensures reliable retrieval and continuity of app/agent responses, as underlying chunk data remains accessible even after node edits or deletions.

Improvements

RAG

Improved file review experience

When a user opens a file using the Review File option in the File Content Menu (which displays all chunks), the document is shown at 100% resolution, maintaining its original clarity and dimensions. This eliminates scaling issues, ensures sharp readability across all file types and sizes, and provides a consistent user experience across desktop, tablet, and mobile devices. The fix also preserves smooth performance and stable UI layout during file rendering.

Navigation

Knowledge->Select the Knowledge Source -> File Content-> Review File

Key outcomes

• Displays files opened via View File in their original 100% resolution, eliminating blurriness from scaling.

• Maintains readability across different file types, sizes, and devices (desktop, tablet, mobile).

• Enables users to review files in their original quality without distortion, making analysis more accurate and efficient.

• Ensures smooth rendering with no lag or layout disruptions during file loading.

• Aligns file display with expected UI standards, improving overall user trust and experience.

Evaluation Framework

Progressive data loading with infinite scroll for Apps, Flows and Knowledge Base

Previously, the Apps, Flows, and Knowledge Base (KB) data were loaded in fixed batches of 25 records at a time. This approach created performance bottlenecks and degraded user experience when handling large datasets. To address this, ZBrain Builder introduces infinite scrolling with dynamic data fetching across all modules, including Apps, Flows, and Knowledge Base selection panels. This feature ensures optimized initial load and extended navigation, ensuring smooth loading and performance for large datasets. API calls are triggered automatically when the user scrolls near the bottom of the list. Lists remain accurate and complete, handle empty and end-of-list states precisely, and maintain efficiency even across very large datasets.

Key outcomes

• Delivers a faster and more responsive UI for large datasets.

• Reduces waiting times and improves efficiency in navigating lists of Apps, Flows, and Knowledge Base sources.

• Streamlines workflows with uninterrupted navigation and automatic data handling.

• Supports clear visual feedback to ensure transparency during data fetch operations.