Features - Added a camera quality prioritization setting with options for speed, quality, or balanced modes. - Added speech recognition setting that listens for “inspect” and triggers an inspection. - Model ID is now included in outbound inspection result JSON. Bug Fixes - Added metadata support for the camera quality prioritization setting. - Model metadata now correctly uses the model name from DWVI JSON instead of inferring from the filename. - Fixed an error when downloading models.

Features - Added pre‑compiled loading of models. - Added two new builtin models: yolo26n-seg-enterprise and yolo26n-enterprise. Bug Fixes - Fixed crash that could occur in visual trigger mode due to memory over‑utilization. - Text is now a valid symbology type that can be disabled when using the Symbology model. - Correctly uses label colors from DWVI, if they exist.

Features - Added wildcard support when writing visual trigger logic. - Added setting to disable inspection preview. Bug Fixes - Renamed "WiFi" to "Network" in the Information view. - Model timestamps now displayed in a more human‑readable format. - Fixed handling of model names containing Unicode characters. - Replaced remaining uses of "Inference" with "Inspection" across the app for clarity.

Features - Added support for square images. - Zoom level now persists across app launches. - Added build number to the information view. - Model task is now included in metadata. Bug Fixes - Fixed incorrect sending of inference data. - Fixed gray screen appearing on startup. - Improved handling of missing previous inference images. - Improved error messaging for model-related actions. - Initial compilation is now skipped when possible to improve startup performance. - Renamed metadata key for clarity.

Improvements - Added a setting to configure the model download timeout. - Improved error messaging and UI feedback for model-related actions.

Improvements - Added a setting to control the maximum number of logs retained. - Reduced unnecessary logging for heartbeat, ping, and pong events. - Improved logging clarity for key system interactions. - Basic and advanced displays can be shown on-screen following an inference result. - MES capabilities to allow remote downloading, deleting, and selecting of AI models. - Added zoom functionality to inference result images. Bug Fixes - Improved messaging for connection statuses. - Resolved crash after multiple consecutive inferences. - Ensured logging persists across app launches. - Improved text sizing in inference results.

Bug Fixes - Camera Flip Issue: Fixed unintended 180° flip when using the front-facing camera during inference. - Stability Improvements: Resolved a crash that could occur after multiple consecutive inferences. - Persistent Logging: Logging now continues reliably across app launches. - Detection Limit Fix: Corrected the 30-detection cap during inference.

Features - Built-in OCR: The Symbology model now includes optical character recognition (OCR) capabilities. - Segmentation Support: Added support for segmentation to enhance image analysis and object detection. - Redesigned Inference View: A fresh new look for the inference view improves usability and clarity. Bug Fixes - Shutter Button Visibility: Fixed an issue where the shutter button was invisible in dark environments. - Text Correction: Updated misleading “upload image to dataset” message for clarity.

Important Updates - Streamlined Integration: Inference output JSON has been improved for smoother interaction with deviceWISE. - Enhanced Logging: Logging has been refined to provide clearer and more useful diagnostics.

Features - Symbology Model: Introducing a new built-in model capable of reading 2D codes for enhanced inspection capabilities. Important Updates - Model Compatibility Notice: This version includes significant changes to model handling, which may affect compatibility with previously saved models.

Features - Custom Overlay Images: You can now use thumbnail images as overlays, with the option to customize them per model. - Swipe Navigation: Easily swipe between previous inferences for quicker comparisons and review. Important Update - Model Compatibility Notice: This version includes changes that may not be compatible with older saved models. Please redownload any models that are from past versions.

Bug Fixes - Improved Mobile Layout: Fixed a layout issue affecting the model selector in iPhone Portrait mode. - Settings Cleanup: Removed the obsolete WebSocket IP address setting. - Preview Duration Control: Added minimum and maximum limits to the preview duration setting. - WebSocket Port Validation: Set a maximum value for the WebSocket port setting to prevent invalid entries.

Features - Capture Only Mode Enhancements: You can now send images to a dataset even when not connected to deviceWISE. - Quick Model Display: A new streamlined model preview is available in the models view. Improvements - Studio Connection Setting: Renamed for better clarity. - Live Model Updates: The model detail view now reflects updates made in deviceWISE AI Studio, no need to redownload a model.

This release introduces advanced AI capabilities to Telit Cinterion VI Assistant: - AI Model Deployment: Download trained image detection models from deviceWISE AI Studio for on-device use. - On-Device Inference: Run AI predictions directly within the app using locally stored models. These new features expand the app’s functionality beyond basic image capture, enabling intelligent computer vision workflows for industrial automation and edge AI applications.

This version includes a small fix to the keywords and subtitle of the app listing. In addition, it also fixes the layout for the iPhone app.