Version 2.3.8 | Aug 7, 2025

Overview

ZBrain Builder 2.3.8 enhances monitoring, app response visualization, retrieval accuracy, and agent control. Key highlights include interactive ECharts rendering in app responses, customizable flow selection for monitoring, and a new latency-based performance metric. Guardrails now detect hallucinations using the app’s Knowledge Base, while the Agentic Retrieval toggle aligns KB behavior with real-time agent reasoning. Users can also manage execution more efficiently by deleting individual agent tasks.

ZBrain Builder 2.3.8 release overview

New features

Flows & Pieces

Interactive chart rendering with Apache ECharts

ZBrain Builder now supports native rendering of Apache ECharts directly within the app interface, enabling users can visualize data outputs as rich, interactive charts. Whenever an app or API response includes an ECharts script marked by the ''' echart identifier, the platform automatically detects, parses, and renders the chart using the embedded configuration. Rendered charts include standard interactive features, such as tooltips, zoom, legends, and dynamic responsiveness, across both desktop and mobile devices. The supported formats are- line, bar, pie, scatter, and other core ECharts visualizations.

Key outcomes:

• Charts are displayed inline within the message flow, ensuring a seamless viewing experience that doesn't interrupt other content.

• If the configuration contains errors or is unsupported, the system displays a user-friendly error message instead of breaking the interface.

• This feature enhances data comprehension and decision-making within ZBrain apps by allowing users to interact directly with real-time visual outputs, eliminating the need for external visualization tools or manual rendering.

Evaluation framework

Custom & ZBrain template Flows for monitoring configuration

ZBrain Builder now supports a dual-tabbed interface, featuring Default and Custom tabs in the side panel, enabling users to select from both custom-built Flows and predefined ZBrain template Flows.

• Custom Flows: Displays all flows previously created by the user.

• ZBrain Flows: Displays a curated list of prebuilt ZBrain templates.

This flexibility allows teams to either reuse existing workflows or rapidly configure standard monitoring notifications, such as Gmail notifications, without needing to build from scratch.

Navigation: Login-> Monitor-> Event Settings-> Select Flow

Key outcomes

• Prebuilt templates simplify and accelerate the monitoring setup process.

• Prevents duplication by restricting the re-adding of already used template flows.

Response Latency Metric under Performance category

To improve operational visibility, ZBrain Builder introduces a new evaluation category- Performance, which includes the Response Latency metric. This metric enables users to monitor and enforce execution time thresholds for AI Agents, Flows, or Apps. Users can input latency thresholds in either seconds or minutes. When a threshold is breached or satisfaction is achieved, the system triggers Success or Failure evaluations and sends relevant notifications.

Key outcomes

• Enabling tracking and enforcement of latency SLAs

• Optimizing execution speed and ensuring responsive user experiences across intelligent workflows.

• These enhancements significantly streamline the monitoring setup process and make performance evaluation more granular and actionable, reinforcing ZBrain Builder’s enterprise-grade observability and automation capabilities.

Guardrails

Hallucination detection for Apps

ZBrain Builder introduces hallucination detection for apps as a new guardrail parameter, designed to ensure response accuracy and alignment with enterprise knowledge. This feature automatically validates AI-generated outputs against the linked Knowledge Base to detect factual inconsistencies or unsupported claims, helping maintain trustworthy and grounded responses. Each app generates a YAML file for hallucination detection based on its Knowledge Base. Responses are evaluated using LLM-as-a-judge to detect mismatches or unsupported claims. If hallucination is detected, the app returns a fallback message, ensuring that only accurate, trusted responses are delivered.

Navigation path: Apps → Configure your bot → Add guardrails → Hallucination Detection

Key outcomes

• Prevents hallucinated or unsupported outputs by leveraging LLM-as-a-judge evaluation.

• Delivers only trusted, knowledge-backed responses by blocking those that fail validation.

• Enhances enterprise compliance and trust in AI-generated content.

RAG

Agentic Retrieval toggle in knowledge base set up

ZBrain Builder now supports the creation of Agentic Retrieval during knowledge base creation and editing. Users can enable a toggle to agent-style, RAG-based retrieval logic, allowing for smarter, LLM-driven sub-query planning and more accurate, context-rich results.

Navigation: Login → Knowledge → Data Refinement Tuning → Agentic Retrieval Toggle

Key outcomes:

• Allows alignment of retrieval logic with agent runtime behavior.

• Improves response accuracy by enabling LLM-based query decomposition.

• Increases transparency and control over how knowledge is retrieved during live agent execution.

• Helps achieve a better understanding of data segmentation and text preprocessing.

Improvements

Agents

Agent task deletion from the Agent Activity screen

ZBrain Builder now allows users to remove individual agent tasks directly from the Agent Activity screen. A delete icon for each task opens a confirmation modal. Upon confirmation, the task is removed from the queue; if canceled, no changes are made. This streamlines agent management by allowing users to clear stuck, failed, or unnecessary tasks.

Navigation Path: Login → Agents → Select an Agent-> Agent Dashboard-> Agent Activity → Running Task → Delete Icon

Key outcomes

• Gives users greater control over agent execution.

• Maintains a clean, actionable task queue.

• Simplifies management of failed or redundant agent tasks.

• Enhances execution monitoring and operational hygiene.

Flows & Pieces

Webhook-based agent triggers for Salesforce & Jira

ZBrain Builder now provides native support for agent auto-triggers through custom webhooks for Salesforce and Jira. Users can now configure real-time agent invocation directly within the ZBrain Builder UI when new records or issues are created or updated.

Navigation path:

Login → Agents → Select Agent → Agent Settings → Create Queue → Salesforce / Jira

Key outcomes

• Real-time automation without external dependencies as triggers is powered by custom internal webhooks, which reduce latency and improve maintainability.

• These webhook-based triggers enable seamless, real-time agent automation across business functions, fully integrated within the ZBrain Builder platform for reliable, reduced latency.