Implements manufacturer-specific device identification for automatic driver selection using System SKU, Machine Type, and other vendor identifiers instead of relying solely on model name pattern matching.
Adds normalized manufacturer detection to handle vendor name variations consistently across Dell, HP, Lenovo, and Microsoft Surface devices.
Extracts comprehensive system information from WMI including baseboard details, BIOS metadata, and firmware versions to support accurate device identification and troubleshooting.
Refactors system information gathering into reusable functions that separate data collection from display logic, enabling the driver mapping feature to leverage device identifiers.
Enhances logging by capturing vendor-specific identifiers while hiding internal matching fields from user-facing output to reduce confusion.
Fixes minor log message wording for clarity when driver installation encounters expected failures.
Enhances HP driver handling to properly track and display SystemId alongside ProductName throughout the driver workflow.
Parses SystemId from HP PlatformList.xml and creates unique entries per ProductName/SystemId combination to avoid conflicts when multiple models share the same product name but different system identifiers.
Updates driver lookup and metadata preservation to include SystemId, MachineType, and ProductName fields across download tasks and result processing, ensuring this information persists through JSON import/export operations.
Improves display name generation to show "ProductName (SystemId)" format for HP models when both values are available, providing clearer model identification in the UI and configuration files.
Standardizes driver metadata handling by replacing simple Make lookups with comprehensive driver metadata lookups that preserve all relevant fields for Dell, HP, and Lenovo vendors.
Enhances driver mapping functionality to automatically extract and store manufacturer-specific identifiers alongside driver paths.
Implements SystemId lookup for HP devices using PlatformList.xml with caching and normalization for improved matching accuracy. Extracts SystemId from Dell model names and MachineType from Lenovo model names using regex pattern matching.
Updates existing mapping entries when identifiers are discovered and adds them to new entries during driver registration. Improves driver matching capabilities by providing additional metadata beyond manufacturer and model name.
Improves consistency in how driver model names and identifiers are processed for different manufacturers (Microsoft, Dell, HP, Lenovo) throughout the codebase.
Introduces helper functions to extract base names from display names and construct standardized display names with identifiers in parentheses format.
Centralizes the logic for converting driver items to JSON model objects, eliminating code duplication across save and download operations.
Enhances validation during driver import and processing to skip invalid entries with missing required fields like MachineType for Lenovo or empty model names.
Ensures proper parsing of model names containing parenthetical identifiers (e.g., "Product Name (Type)") and extracts components correctly for each vendor's requirements.
Removes obsolete CabRelativePath property from Dell driver handling.
- Implements detailed failure tracking for driver downloads, capturing model names and statuses for failed attempts.
- Updates logging to provide clearer messages for both successful and failed downloads, improving traceability.
- Modifies the user interface to display comprehensive error messages when downloads fail, including a summary of failed models.
- Ensures that all exceptions during download and extraction processes are logged and thrown, preventing silent failures.
- Client OSes will now use CatalogIndexPC.xml to identify which ProductLine_SystemID.xml to use to identify which drivers to download. This is inline with how DCU works.
- In the UI, Dell Model names now show the full product line, model number, and system ID in the model column.
- There are many more models now shown due to breaking each model out by systemID (one model will have many systemIDs).
- Downloads per model should be much smaller as prior code was downloading drivers for models that Dell had reused their model number (e.g. Precision/Inspiron/Latitude/Vostro 3520 would result in a very large driver download)
- Dell driver downloads are best effort based on the data from the XML files. In some cases the Dell support website may show a newer driver than what is downloaded. This is rare, but in testing I've seen one or two drivers per model where the XML doesn't have what's listed on Dell's website. Again, rare, but not unexpected.
Enhances the Edge automation process for retrieving authentication tokens with multiple reliability improvements:
- Implements dynamic port allocation to prevent port conflicts
- Adds retry logic with configurable attempts for token retrieval
- Implements proper WebSocket message polling to handle asynchronous responses
- Creates isolated temporary Edge profiles to avoid state interference
- Adds fallback mechanism to extract tokens from cookies when localStorage is unavailable
- Improves error handling and diagnostic logging throughout the workflow
- Ensures comprehensive cleanup of resources including WebSocket connections, browser processes, and temporary profiles
These changes address intermittent failures when the token is not immediately available in localStorage and improve overall robustness of the headless browser automation.
Changes the default Windows version from 24H2 to 25H2 across the FFU development tool to reflect the latest Windows 11 release.
Adds dynamic products.cab download functionality for Windows 11 using Windows Update service API instead of static MCT links. This is due to a change in how the MCT pulls the products.cab file.
Increments internal preview version for build and deployment scripts to reflect next preview cycle, ensuring logs and artifacts show the correct release identifier for traceability.
- Sorts top-level config keys before serialization for deterministic files and cleaner diffs.
- Increases JSON depth to 10 to retain nested settings.
- Writes JSON as UTF-8 via Set-Content for consistent encoding.
- Applies across config export and UI save flows.
- Enables selecting multiple existing FFU images to include on the deployment USB for easier distribution and testing.
- Adds a UI option with selectable, sortable list from the capture folder, refresh support, and persisted selections.
- Validates that selections exist when the option is enabled to prevent empty runs.
- Supports unattended/CLI flows by prompting early or accepting a preselected list for USB creation; deduplicates and logs chosen files.
- Always includes the just-built (or latest available) FFU as a base.
- Improves no-FFU handling and streamlines multi-FFU selection workflow.
Adds per-app control for additional accepted exit codes and ignoring non‑zero exit codes to improve handling of installers with nonstandard returns.
Exposes editable fields in the app list UI, persists them across search defaults, import/export, and pre-download save, and applies overrides during app resolution to honor configured behavior.
Applies name sanitization when persisting the app list and when building/checking Win32 and Store download directories.
Prevents invalid characters in folder names, aligns persisted names with on-disk structure, and improves detection of existing content to avoid redundant downloads and errors.
Adds persistence of AdditionalExitCodes and IgnoreNonZeroExitCodes when exporting the UI list to prevent losing custom exit handling settings and maintain parity with the primary save routine.
Introduces a new common function, `ConvertTo-SafeName`, to sanitize strings by removing characters that are invalid in Windows file paths.
This function is now used consistently when creating directory and file names for drivers (Dell, HP, Lenovo, Microsoft) and applications to prevent path-related errors. It replaces several ad-hoc sanitization methods with a single, more robust implementation.
Extracts the logic for importing supplemental assets (Winget, BYO, Drivers) into a new reusable function. This function is now called by both the manual and automatic configuration loaders, reducing code duplication.
Enhances the manual configuration loading process with more robust error handling. It now provides specific user-facing error messages for file read failures, empty files, and invalid JSON, improving the user experience when loading a malformed configuration.
When loading a configuration, if optional supplemental files like AppList.json are referenced but not found, an informational message is now displayed to the user instead of failing silently.
Implements a deferred cleanup mechanism for driver source folders when they are compressed to a WIM and also used for WinPE.
When drivers are compressed, the original source folders are now preserved if they are also needed for WinPE driver injection. A marker file is created in these preserved folders.
A new cleanup step is added after the WinPE media creation to remove these preserved folders, ensuring they are available when needed but not left behind permanently.
Enhances the model name normalization function to better handle variations in hardware model strings. This change introduces specific rules to canonicalize "All-in-One" and screen size variants (e.g., "-in" or "inch") for more reliable matching against driver mapping rules.
Additionally, optimizes performance by normalizing the system model once before the comparison loop. Logging is also added to show the original and normalized model strings for easier debugging.
Introduces a new feature, `UseDriversAsPEDrivers`, that allows WinPE drivers to be sourced directly from the main driver repository.
When enabled, the script scans all available drivers, parses their INF files, and copies only the essential driver types (e.g., storage, mouse, keyboard, touchpad, system devices) needed for WinPE. This eliminates the need to maintain a separate, manually curated `PEDrivers` folder.
The UI is updated with a new checkbox that becomes visible when "Copy PE Drivers" is selected, making this a sub-option. Parameter validation is also adjusted to support this new workflow.
Introduces a "Restore Defaults" feature in the UI to reset the environment. This action removes generated configuration files, ISOs, downloaded apps, updates, drivers, and FFUs.
The post-build cleanup logic is refactored from the main build script into a new common function. This new function is used by both the standard build process and the new restore defaults feature, promoting code reuse and simplifying maintenance.
Updates the visibility of UI panels for Winget and drivers when a previous environment is automatically loaded.
This ensures that if Winget apps or driver models are present, their corresponding UI sections are made visible. Additionally, it updates the "select all" checkbox state for Winget results and attempts to pre-select the hardware make for loaded drivers.
Implements a new feature to automatically load the previously saved environment when the UI is launched.
This improves user experience by restoring the last saved configuration, including selected applications and drivers, eliminating the need to manually reload them on each run.
The process loads the main `FFUConfig.json` and then proceeds to load associated Winget, BYO App, and Driver lists if they are defined. UI elements and checkboxes are updated accordingly to reflect the loaded state.
- Changed the default disk size parameter from 30GB to 50GB in BuildFFUVM.ps1 and FFUUI.Core.psm1 to accommodate larger virtual machines.
- Updated tooltip and default value in the UI XAML file to reflect the new disk size.
Comments out the logic that forces app installation when building from a downloaded ESD file. This workaround was implemented to prevent an OOBE reboot loop but is no longer required. This should speed up scenarios where you want to download the ESD media, install the latest CU and .NET CU, and capture the FFU.
Renames `Remove-DisabledUpdates` to `Remove-DisabledArtifacts` to better reflect its expanded scope.
This function now also removes Office installation scripts and downloaded content if the Office installation is disabled via the `$InstallOffice` flag.
The function call is moved to run before app installations to ensure artifacts are removed prior to the installation phase.
Introduces a new function to remove residual artifacts for updates that are disabled via script flags.
If updates for Defender, MSRT, OneDrive, or Edge are turned off, this change ensures that any related files are deleted from the build environment. This prevents unnecessary files from being included in the final image.
Clarifies the distinction between the application architecture to be downloaded and the target Windows architecture for installer pruning.
Renames the `SelectedWindowsArch` parameter to `WindowsArch` and introduces a new `ApplicationArch` parameter. This makes the download and subsequent installer selection logic more explicit and easier to understand, especially for MS Store apps where multiple installers might be available.
Refines the post-download cleanup process for applications with multiple installers.
When multiple installers are downloaded for an application, this change introduces logic to select the most appropriate one based on the target Windows architecture. It prioritizes universal installers first, then architecture-specific installers, before falling back to the previous method of selecting the newest file by signature date.
This ensures a more accurate installer is chosen, especially when Winget provides multiple architecture options for a single package.
Improves the reliability of WinGet app processing by making several key changes.
The build process now deletes the `WinGetWin32Apps.json` file before each run to ensure it is always freshly generated.
The UI no longer relies on `WinGetWin32Apps.json` to detect previously downloaded content. Instead, it checks directly for the application's content on disk, preventing unnecessary re-downloads.
This change also introduces a feature allowing users to override the default silent install commands for Win32 apps. By specifying `CommandLine` or `Arguments` properties in `AppList.json`, these values will be used to update the corresponding entries in `WinGetWin32Apps.json` during the build process.
Eliminates the read-only field and derives the features list directly from the checked items, producing a sorted semicolon string when collecting config. Avoids duplicated state, prevents desynchronization between UI elements, and yields deterministic ordering for persistence.
- Added logic to check for per-user Appx packages that are not provisioned for all users, which could block the Sysprep process.
- Introduced a hash set to store provisioned package families and collect current user Appx packages while excluding frameworks and non-removable packages.
- Implemented removal of non-provisioned Appx packages with error handling and re-checking to ensure all blockers are resolved before proceeding with Sysprep.
- Creates a new parameter [bool]InjectUnattend
- This will take the FFUDevelopment\Unattend\unattend_[arch].xml file and copy it to the FFUDevelopment\Apps\Unattend and rename the file to unattend.xml.
- This is useful for situations where you don't use the USB drive for deploying the FFU but still have Unattend-related customizations that you want to apply.
- Improved detection of installer candidates by reading YAML configuration for NestedInstallerFiles.
- Added logic to resolve silent install switches from YAML, prioritizing block-level settings.
- Enhanced fallback mechanisms for selecting installers when multiple candidates are found.
- Updated command construction to accommodate new installer resolution logic.
rbalsleyMSFT