Version 4.8

  • Published on Apr 14, 2026
  • 3 minute(s) read
  • eA
 Prev Next

Version 4.8

14 April, 2026

AI Agents

  • HTTP Node Response Header:
    The HTTP node now supports capturing response header data in addition to the response body. A new optional "Response Header Content" field has been added under Output settings, allowing flow designers to store header values in custom context variables and use them within subsequent flow steps. This capability is essential for integrations where critical information, such as authentication tokens, is returned via response headers.

  • Knowledge Base Document Download:
    Users can now download previously uploaded documents directly from the Knowledge Base Data Sources page. The download action is available for documents in Draft status, providing convenient access to original source files without requiring external storage lookups.

  • AI Testing Export-Import Function:
    AI Testing scenarios can now be exported and imported across projects, enabling reusable test configurations and streamlined environment setup. Users can selectively export up to 10 scenarios at a time as a ZIP file and import them into another project with duplicate name detection and validation. Scenario definitions and configurations are transferred, while historical simulation results and evaluation data are excluded.

  • AI Testing Schedule Run:
    A new scheduling capability has been introduced for AI Testing scenarios, allowing users to configure one-time or recurring test runs. Recurring options include Daily, Weekly, and Monthly frequencies with configurable end dates. When a test is triggered while another is already running, the new test is automatically queued and executed in FIFO order, supporting up to 50 queued runs.

  • Partial Reveal for Masking Entities:
    Masking configuration now supports revealing a specified number of characters at the start or end of detected PII values. For each entity, users can choose between Reveal Start, Off, and Reveal End modes with a configurable character. This enables use cases such as displaying the last four digits of a credit card number while keeping the rest masked.

  • HTTP Node Mock HTTP Response:
    A new Mock HTTP Response toggle has been added to the HTTP node, enabling flow designers to define a static JSON response for testing purposes. When enabled, the node returns the mock data instead of making a real HTTP call, removing external service dependencies during flow development. This feature is available exclusively in test mode.

  • Data Flow Voice Gateway Integration:
    The platform now supports integration with Data Flow as a voice gateway, enabling advanced speech processing and VAD capabilities through the Sync API. Audio files managed by Data Flow remain in their origin MinIO storage as the single source of truth, eliminating redundant copies. A new storedExternally parameter in the attachments payload controls whether audio files are saved to the local MinIO or referenced by URL, while word-by-word speech recognition output is delivered via the SrWordDetails channel data parameter for dynamic audio masking.

Improvements

AI Agents

  • Conversation History API Scope Extension:
    The Conversation History API can now be called using the dataservice_external_session_metrics scope, broadening access for external integration scenarios.

  • Knowledge Base Reporting – Knowledge Retrieval Tool Support:
    Queries triggered through the Knowledge Retrieval Tool are now listed on the Knowledge Base reporting page alongside other query types. Tool-sourced queries display a dash in the Response column, default to a "found" status, and do not show Stand Alone Question, Response, or cached answer details in the detail panel.

  • TTFT (Time To First Token) Metrics on AI Agent Dashboard:
    The existing response time widgets on the AI Agent dashboard have been replaced with Time To First Token (TTFT) metrics, providing streaming-aware performance measurement. The updated widgets display average TTFT along with 95th and 99th percentile values, offering more precise insight into initial response latency.

  • Minor UI Improvements:
    Several minor improvements have been made to the user interface to enhance user experience and usability.

  • Minor Bug Fixes:
    Various minor bugs have been addressed to improve overall system stability and performance.