Automatic PDF Processor - Changelog

Fixed: the "ANSI" character set followed the program language

When "ANSI" was selected as the character set for an output file, the program used the ANSI character set of the program language (German or English) instead of the one configured in Windows. On systems with a different Windows character set - for example with Central European, Cyrillic, Greek or Turkish language settings, or with the Windows option "Beta: Use Unicode UTF-8 for worldwide language support" enabled - language-specific characters were therefore written or read incorrectly. This affected the "Use template" task, the processing log and the export of the "Save extractions" task. German and Western European systems were not affected.

Improved: date recognition when the space after the comma is missing

With scanned documents, text recognition frequently drops the space after the comma - the text then reads "September 16,2024" instead of "September 16, 2024". Dates written that way were not recognized. As a result, a file either ended up without a date at all, or a different date from further down the document was used instead. The space after the comma is now optional, and both spellings are recognized. This applies to all date formats - locale, any, and user-defined ones.

Improved: clear message when a date rule does not recognize a date

If an extraction rule of the data type "Date" did not recognize a date in the text it read, the program previously reported "text contains no characters" - even though the text was there and simply no date matching one of the enabled date formats was found. This pointed troubleshooting in the wrong direction, for example when the date in the document is written with a spelled-out month name while the rule only searches for the locale format. In this case, the unambiguous message "no date found in the text" is now displayed. In addition, the rule overview in the profile settings now shows the cause of errors directly.

Fixed: "Replace content" sometimes removed one character too many

When replacing content, the first character of the word immediately following could disappear as well if it sat right next to the defined area - "Invoice No. 4711" became "[REPLACED] o. 4711", for example. The cause was that removal used a slightly larger area than reading the text did. What gets removed is now exactly the text that was found in the area. This also affected subsequent replacements: a second replacement searching for a keyword right next to the first one no longer found it.

Fixed: "Clear error logs" did not affect background processing

After clearing the error logs, the deleted entries reappeared as soon as the next error occurred in background processing: it still held its own state in memory and wrote it back in full. Clearing now takes effect across all parts of the program.

Improved: automatic restart of the processing service

When the Windows service is started via the "External processor" setting, the program now also configures the service recovery options. If the service is terminated unexpectedly, Windows restarts it automatically - previously, processing simply stopped until someone noticed.

Improved: error logging under load

Saving the error log no longer blocks the remaining logging. When several errors occurred at the same time, or while the log file was briefly locked by another program (virus scanner, backup, synchronisation service), this could noticeably slow down processing. An endless loop is no longer possible when the log file is locked either.

New: extracted data as filter criteria

The most frequently requested feature so far: the profile filter now offers the new tab "Extracted data". There, the results of the extraction rules can be used directly as filter criteria - for each rule, you can specify whether the extraction must return a value (or must not return one), and the extracted value can be checked with a text, date, or number comparison depending on its data type (including the familiar filter syntax with lists, placeholders, and AND/OR logic). This makes it possible, for example, to process only invoices above a certain amount or only documents with a recognized invoice number.

New: OCR support for scans without a text layer

New profile category "OCR support" (above the filter): for scanned PDF files without a text layer, a temporary text recognition can be enabled. Text filters, the character count filter, and extraction rules then work with the recognized text, while all file operations continue to use the unmodified original. The text recognition uses the same full OCR settings as the task "Make file searchable" and can be limited to the first pages to speed up processing. The temporary copy is deleted automatically once processing is complete. The rule editor also uses the OCR support (checkbox at the bottom left, preset with the category setting): the rule preview then temporarily recognizes scanned example files - extraction rules can thus be configured directly on the scan and show exactly the values that the processing will determine (including a progress window with page progress). The same applies to all other previews of the profile settings (results preview of the filter, extraction results overview, placeholder previews) - each example file is recognized only once per opened settings window. For quick diagnosis, the list of example files now additionally shows the columns "Character count" and "OCR recommended", and with OCR support disabled the preview in the rule editor points out scans without a text layer.

Changed: option "Working directory" removed

The program option "Working directory" has been removed. Temporary files are now always stored in the local application data folder.

Fixed: filter criterion "Modification date" showed wrong values

In the filter dialog, the PDF criterion "Modification date" loaded the settings of the criterion "Creation date" when opened; as a result, the values of the wrong criterion could be applied when saving.

Fixed: "Determined data area is invalid" for rotated scans

PDF pages that are stored sideways by the scanner and only displayed upright via the page's rotation flag always caused keyword-based data extraction to fail with "Determined data area is invalid": the keyword itself was found, but the data area next to it was searched on the non-rotated page. The preview in the rule editor was affected as well (the frame around the data area was missing and changes to the area extensions had no effect). The same issue affected the task "Replace content" with keyword-based area determination; for such pages, the determined area is now also correctly converted into the coordinate system of the original page so that replacement and redaction are applied at exactly the right position. In addition, a data area that collapses due to overly large negative extensions is now reported as invalid instead of extracting random text or replacing content at the wrong position.

Fixed: file appearing twice in the result when merging via file group on network drives

If the monitored folder was located on a mapped network drive (e.g. L:\...), merging via a file group could include the first file twice in the target file: the folder monitoring tracked the file under the resolved network path (\\Server\...), while the file group scan found the same file again under the drive letter - both entries were treated as different files. The file group scan now uses the same path resolution as the folder monitoring; in addition, the merge task now detects and skips such duplicates in general.

Fixed: existing files were not skipped for new profiles on network drives

The safeguard that skips files already present at the time a profile is created or modified (preventing e.g. mass printing or mass mailing of the existing stock) did not take effect if the monitored folder was located on a mapped network drive. The cause was the same path form mismatch (drive letter vs. network path) as in the merge issue.

Fixed: dynamic email attachments from network drive folders

For dynamic attachments of the "Send file" task, the configured source folder was not resolved via the drive mapping. In service mode, collecting attachments from a folder on a mapped network drive therefore failed; in addition, the main file could incorrectly count as an additional match and trigger an unjustified error with the "Report multiple matches" setting.

Fixed: duplicate pages when merging PDF files

For the "Merge files" task with the option "Additionally append all PDF files of the parent folder" enabled, individual documents could end up twice in the target file: statically selected files located in the monitored folder itself (e.g. a cover sheet) as well as files of a file group were also picked up by the parent folder scan and therefore appended a second time. Files that have already been included are now reliably skipped.

Fixed: German labels in the English program version

For the tasks "Compress file", "Convert to PDF/A", and "Automatically rotate pages", the settings labels were displayed in German even in the English program version. All three tasks are now fully localized.

Improved: "Save extract" task without selected values

When adding a new "Save extract" task, all existing data extraction rules are now preselected as values to save. In addition, the program now warns when saving if not a single value is checked - such a task would produce an empty output file.

Fixed: data extraction rules blocked text recognition for scanned files

If a profile contained data extraction rules and a "Make file searchable" (text recognition) task, scanned PDF files without a text layer could never be processed: data extraction was already applied to the scanned file before text recognition, failed due to the missing text, and thereby also aborted the text recognition task - the very task that would have created the required text. A data extraction failure now only aborts a task if that task actually uses extracted values (e.g. as placeholders in the target folder or file name). A text recognition task without such placeholders therefore runs through; subsequent tasks then extract their values from the searchable file - also in combination with the option "Perform data extraction only once for each file".

New: "Overwrite original" also for printing as PDF file

When printing via a PDF printer, the storage location now directly offers the "Overwrite original" option - the printed PDF file then replaces the source file. Previously, this was only possible via the workaround of manually entering the source folder and file name as the target.

New: hint when combining a document text filter with text recognition

If a profile contains the "Make file searchable" task and at the same time filters by document text, the program now points out when saving that for scanned files the document text only becomes available after they have been made searchable - and recommends splitting the setup into two profiles (make searchable first, then filter).

Improved: unexpected task errors now always appear in the log

If an unexpected internal error occurred in a task, the task could previously be missing from the processing log and the report without comment in rare cases. Such errors are now reliably reported as errors of the respective task.

Improved: more reliable task chains with "Print as PDF file"

When printing via a PDF printer (e.g. "Microsoft Print to PDF"), the program now waits until the printer driver has completely written the target file before the next task starts. If the original file is overwritten in the process, subsequent tasks (e.g. text recognition followed by moving based on extracted data) are now guaranteed to work with the new file content.

Fixed: long-running processing operations could be aborted after 10 minutes

If processing a single file took longer than 10 minutes - for example, text recognition (OCR) of a large document - the internal crash protection incorrectly classified the processing process as hung: the operation was terminated with the message "Processing aborted (process terminated)" and the file was subsequently skipped. Long-running operations (OCR, printing, merging, splitting, stamps/watermarks, redacting, replacing content, and others) now report their progress page by page, so the crash protection only intervenes on genuine hangs - without requiring any settings to be adjusted. Should such an abort still occur, the cause is now stated precisely in the processing log and in the error report.

Fixed: text remnants could remain visible below the redaction bar on scanned documents

For scanned files made searchable via text recognition, the black bar of the "Redact text" task did not always fully cover the descenders of the printed text (e.g. in "g", "p", or "y"). Scanned pages are now detected automatically and redacted with full height plus a safety margin - without requiring any settings to be adjusted. For purely digital PDFs, the previous behavior, which protects neighboring lines, remains unchanged.

Improved: a profile deactivated during a running processing run is now skipped immediately

Previously, deactivating a profile only took effect on the next check run. Now, a profile that is deactivated or deleted during a run immediately stops processing further files; the remaining profiles continue normally. Activating or adding a profile still only takes effect on the next run.

Improved: the order of items in the "Replace content" task no longer matters

If multiple replacements were configured where a shorter value is contained in a longer one (e.g. "Faust" and "Dr. Faust"), the longer value could no longer be found after the shorter one had already been replaced, depending on the configured order. Replacements are now automatically processed with the longest search text first - the configured order does not need to be considered. The "Redact text" task was never order-dependent.

Fixed: a PDF file could be attached twice when sending emails

If a dynamic attachment or a file group matched the same file that was already being sent as the main attachment, the PDF file could end up in the email twice with an identical name - especially with the option "Replace special characters in attachment names" enabled or with monitored folders on network drives (drive letter vs. UNC path). The attachment check now recognizes the same file regardless of how the path is written.

New: the processing log shows the unmet filter criteria for "No match" entries

If a file is not processed by any profile, the details of the log entry now show for each rejecting profile which filter criteria were not met. This makes it much easier to find the cause of unexpected "No match" results.

New: working directory check at program startup

If the working directory configured in the options is not available or not writable - for example after switching the Windows user account - the program now points this out at startup and offers to reset it to the default directory. Previously, temp file operations failed in this case without any visible indication.

Improved: protection against duplicate processing with multiple Windows user accounts

If the program accidentally runs under two Windows user accounts of the same computer at the same time (e.g. during an account migration), each instance previously processed the files separately - with duplicate results such as emails being sent twice. The lock against parallel processing now also works across user accounts. Please note: across different computers, such detection is technically not possible - a monitored folder should always be processed by only one installation (see the note in the program help, chapter "Monitored Folders").

New: the "Save extractions" task now also supports XML and JSON

Extracted values - for example from PDF form fields or text areas - can now be output not only as CSV but also as an XML or JSON file, ideal for further processing by other programs and interfaces. For XML, you can choose whether the rule names are written as an attribute or as the element name; the suitability of the rule names is checked when saving the task. Each processed PDF file produces its own output file.

New: output multi-line values as a list (XML/JSON)

For selected rules, multi-line values - e.g., invoice line items - can be split at line breaks and written as a list (XML: sub-elements with a freely selectable name, JSON: array). Checked rules are always output as a list so that the structure of the output file remains consistent for downstream programs.

Improved: program help revised

The program help available as a PDF now includes all illustrations. In addition, the example for setting up the PDF processor as a Windows service via "sc create" has been corrected: paths containing spaces require escaped inner quotation marks, otherwise the service will not start.

Fixed: catch-up processing incorrectly reported “no match” on network drives

If the monitored folders were located on a mapped network drive (e.g. “H:\”), catch-up processing could reject all files with “no match” since version 2.5.0, even though the profiles were unchanged and regular monitoring processed the same files correctly. The cause was the network drive letter resolution introduced in version 2.5.0 for running as a Windows service, which was not yet applied during catch-up. Catch-up processing now uses the same path resolution as regular processing.

Improved: “Treat recurring files as new” now also covers manually moved files

The profile option “Treat recurring files as new” previously only applied to files that the program itself had moved, renamed, or deleted. If an already processed file was moved out of the monitored folder by the user and later placed back there, it remained unprocessed. The program now also detects such manually removed files and processes them again when they reappear, provided the option is enabled. Nothing changes for profiles without this option.

Fixed: text recognition could fail to start at all on some systems

In rare cases, a required OCR language file was not placed in the expected location during initial setup, causing text recognition to fail consistently with an initialization message. The program now checks for this file at every start and, if needed, automatically restores it from the data already present – with no action required. You can also trigger the restore at any time via the menu “Other → Repair OCR data”.

Fixed: text recognition could fail for an entire batch in rare cases

On some systems, text recognition (OCR) could abort with a message that the engine could not be initialized – usually because real-time antivirus briefly locked one of the required OCR files; as a result, every other file in the same run failed as well. When this happens, text recognition now waits briefly and retries automatically, so processing completes reliably. Tip: add the program folder and the “OCRData” folder to your antivirus exclusions to avoid such delays entirely.

Fixed: delay when shutting down Windows

In rare cases, Windows displayed a message during shutdown stating that the background processing could not be closed. Processing now responds to the shutdown immediately and exits reliably, so this message no longer appears.

Improved: OCR data is checked at startup and restored automatically

In rare cases, the data folder for text recognition (OCR) was incomplete, causing text recognition to fail with an error message. The program now checks this folder at every startup and automatically restores any missing components. In addition, you can trigger the restore manually at any time via the menu “Other → Repair OCR data”. Any language packs you have already downloaded are preserved.

Fixed: skewed text layer with automatic page orientation correction

When the “Automatically correct page orientation” option was enabled in a text recognition task, the recognized text layer could be placed at an angle for certain documents scanned in landscape format. As a result, search, data extraction and renaming based on recognized text failed. The correction now works reliably, so that the page content and the recognized text line up exactly again.

New: profile export now includes the lists and counters in use

When you export profiles, the query lists and named sequential counters they use – including each counter’s current value – are now saved along with them. After an import – for example on a new computer or following a reinstallation – they are available again without having to recreate them by hand. Existing lists and counters with the same name are kept unchanged. If a rule still refers to a missing list or a list file that cannot be found after the import, a note points this out.

Fixed: clear message for an invalid email address

If a recipient address contained an error, a technical message that was hard to understand used to appear. The program now clearly states the affected address, making the problem easier to find and fix.

Improved: clearer error log when network problems occur

A monitored network folder that stayed unreachable used to fill the error log with hundreds of identical entries, burying important messages. Such recurring environment messages are now consolidated, keeping the error report sent to support concise and meaningful.

New: warning when a profile blocks other profiles

If a profile has the "Stop processing after this profile" option enabled and at the same time has no filter criteria, it matches every file in its monitored folders and thus ends processing. Because profiles are applied in alphabetical order of their name, later profiles monitoring the same folders no longer get their turn. When saving, the program now points out this situation and names the affected profiles, making the mistake easy to avoid (e.g. an unfiltered catch-all profile "All receipts" that overrides the alphabetically later profile "Accounting").

Improved: network drives now fully usable in the Windows service

If a profile monitors a network drive or saves files to it (e.g. drive H:), the program now automatically detects the associated network path. This means files are reliably found and saved even when processing is handled by the Windows service, which otherwise cannot see individual drive letters. The paths in the profiles remain unchanged.

Fixed: protection against duplicate processing during simultaneous operation

If the Windows service and the program window accidentally ran at the same time, files could be processed twice. The program now ensures that only one instance processes at a time, avoiding duplicate access and conflicts.

New: notice and automatic start of the processing service

When you enable the „External processor“ option, the program now checks whether the associated Windows service is running. If it exists but is stopped, it can be started directly and set to start automatically. If no service exists, a notice points this out. This avoids the common misunderstanding that the checkbox alone would already take over processing.

New: backup for support

Via the “More” menu you can now create a special backup that you can safely send to support. License files are excluded, and all passwords (PDF passwords from options and profile settings as well as the account passwords for sending e-mail) are removed automatically; the e-mail account user name and sender address are replaced with a placeholder. Settings, profiles and error logs are kept for troubleshooting. An enclosed info file transparently lists what the backup contains and what it does not. The backup also includes a status snapshot (SystemSnapshot.txt) with versions, license and profile status, the reachability of the monitored folders, and the errors of the last seven days – making it easier to pinpoint problems.

New: continuous diagnostic log

The background processor now keeps a permanent, day-by-day rotating log (folder “Health”, up to 14 days) with regular heartbeats and a few key events: start and stop, a monitored folder that is not reachable (e.g. a disconnected network drive), file-processing timeouts, and restarts after a crash. It also adds a regular scan summary (how many files were seen, newly detected, or why they were skipped) and a warning if processing makes no progress for an extended period despite being allowed to run. This keeps even older disruptions traceable – without having to enable a detailed diagnostic log beforehand. The log is included in the support backup.

When such a stall is detected, the program can also notify you actively: if the “Report general errors” option is enabled, a notification e-mail is sent to the configured recipients (and an all-clear once processing resumes). Independently of that, the incident is written to the Windows event log – even in service mode without an open program window.

Fixed: temporarily unreadable files are no longer lost

If a PDF file was briefly unavailable while being read – for example because it was still open in another program, a network drive dropped out for a moment, or access was delayed – it was previously skipped permanently and never retried. Such temporary read failures are now retried automatically for up to four hours (without a flood of error e-mails), so the file is processed as soon as it becomes available again. Only after that is it treated as permanently faulty. Genuine read errors (e.g. a corrupted file) are still reported as an error immediately.

During catch-up, such temporarily unprocessable files are now listed in the summary at the end – together with locked files and including the reason – and are automatically carried over to the next catch-up run instead of being lost.

The former log tab “Locked files” has been renamed to “Currently unprocessable” and now shows, in addition to locked files, those that are currently being retried – with a status column (e.g. “Locked” or “Retrying: read error”). This makes it easy to see at a glance which files are currently stuck.

Fixed: files skipped due to the time limit now appear in the error log

When a file exceeded the configured processing time limit and was skipped, this previously showed up only as a counter but not as a log entry – so the affected file remained invisible. Such files are now listed correctly under „Errors“, including the profile and a message stating that the time limit was exceeded.

Also fixed: if the processing of a file ended unexpectedly (time limit, abort or program shutdown), the corresponding entry in the processing details could show an endlessly increasing duration. Such interrupted operations are now properly closed as „aborted“.

Improved: a file causing a one-off crash is no longer skipped permanently right away

If a file caused the processing to crash, it was previously skipped permanently straight away to avoid an endless loop. A single, temporary crash could therefore block an otherwise processable file forever. Such files are now retried a few times and only skipped once they repeatedly cause a crash – while the protection against endless loops is retained.

Improved: greater stability with hanging or faulty files

The actual file processing now runs in a separate, supervised worker process. If a single file hangs (for example a defective PDF during text recognition) or causes a crash, only that worker process is affected: processing ends it deliberately, skips the file in question, and continues normally with the remaining files. Previously, such a file could freeze the entire processing or bring it down.

An additional safety timer detects a process that stops making progress for a longer period (e.g. a hang already while scanning the files) and restarts it automatically. It automatically aligns with the per-file processing time limit so that long but correctly running files are not aborted. When the program is closed, this process is also shut down cleanly.

Fixed: catch-up continues after a restart of the processor

If the background processor is restarted or terminated during a catch-up run, the progress window now detects this, shows a note, and automatically continues the remaining files after the restart – instead of stalling. The files still pending remain in the queue, so nothing is lost.

Fixed: a file was incorrectly reported as an error when restarting during processing

If the program (or background service) was closed or restarted at the exact moment a file was being processed, that file could afterwards be misclassified as having “caused a crash” and skipped permanently – even though no crash had occurred. A file interrupted by a shutdown or restart is now simply processed again on the next run – without an error message and without being skipped. Only if a file actually causes the processing to crash is it skipped, to avoid an endless loop; this crash detection remains fully effective.

Fixed: freeze when saving profiles repeatedly within a session

If several profiles were opened and saved one after another within a single program session, the program could stop responding from the second save onward. The cause was an internal reference that was not fully released when the profile settings were closed and that accumulated over the session. It is now reliably cleaned up, so any number of profiles can be saved in sequence without freezing. This mainly affected setups with many profiles, such as invoice processing.

New: named counters shared across profiles

The sequential number can now be set up as a named counter that can be shared across multiple rules and even across all profiles – creating continuous, unique numbering without duplicates. The new „Manage counters“ window lets you create and name counters and define their start value, increment, format, and an optional reset interval. In a rule's „General“ tab you simply pick the desired counter from a list. Existing profiles are migrated automatically.

New: preview of the resulting columns when saving extractions

For the „Save extractions“ task, a new preview shows directly which columns the generated file will contain and in which order. This makes the column selection easy to grasp at a glance, without having to scroll within the selection area.

Fixed: a few missing English translations

Some individual texts were still displayed in German in the English interface and are now translated correctly.

New: custom output format for date values

For the „Date“ data type, you can now define your own output format on the „Formatting“ tab (e.g., „yyyy-MM-dd“ for 2026-06-15). The „Insert date component“ button and the accompanying legend make it easy to assemble year, month, and day; a live preview shows the result immediately. If the field is left empty, the system's default date format is used as before. The formatted date is also available at the top of the placeholder menu, so it can be inserted directly into file names, e-mails, or stamps.

New: format number values

For the „Number“ data type, the output can now be tailored: the number of decimal places, leading zeros, the thousands separator, and the decimal separator can be set using simple input fields. A live preview shows the result immediately. Formatting is optional and disabled by default, so nothing changes for existing profiles.

New: „Last change“ column and clearer „Start date“

The profile list now shows an additional „Last change“ column that indicates when a profile was last edited. The former „Created/Modified on“ column is now called „Start date“ and shows more clearly from which point in time a profile processes files regularly.

New: clean up pending operations when changing or deactivating a profile

When a profile is changed in a way that affects which files it selects (filter, monitored folders, or delayed processing), or when a profile is deactivated, the program now asks how to handle that profile's still-pending operations – that is, files queued for delayed processing and pending retries. This keeps outdated operations from running under changed settings.

Improved: retries of inactive profiles are paused

Pending retries (for example after a previously failed print) of a deactivated profile are now paused instead of executed, and only continue once the profile is reactivated.

Fixed: delayed processing across time zones

An issue in delayed processing has been fixed where queued files could be discarded incorrectly depending on the time zone.

Fixed: missing English labels in the rule editor

Missing English labels were added in the rule editor for the „Text of page(s)“ data determination.

Bug fix: time zone deviation in PDF dates

If the creating program (e.g., a scanner) stores the PDF creation or PDF modification date in universal time (UTC), the time was previously not converted to local time. As a result, for example, the option „Set modification date to: PDF modification date“ could set a file date that deviated by several hours. PDF dates with time zone information are now correctly converted to local time. The correction takes effect everywhere PDF dates are used (setting the file date, placeholders, filters, catch-up processing).

Catch-up processing: live progress already while determining files

The progress window now reports progress already while the files to be processed are being determined. Previously, for large file sets the display appeared frozen at „0“ during this phase.

• While determining the files, a running animation and a live counter are shown („X checked, Y of them within the period“).
• During the subsequent processing, the progress bar now shows the actual progress.

Bug fix: truncated catch-up dialog

In the „Apply profiles subsequently“ dialog, the bottommost options could be cut off and unreachable depending on font size or display scaling. The dialog now automatically adjusts its size to the content.

Bug fix: blank pages during text recognition (OCR)

For certain PDF files with a defective internal structure, the „Make file searchable“ task could produce an output with inserted blank pages (error message „The page count of the OCR output differs from the original file“). Such files are now automatically processed via a workaround that bypasses the faulty structure—the output thus contains exactly the same pages as the original.

New: Deferred processing

Profiles can now process files with a time delay: a matching file is queued when detected and only processed after a configurable waiting period. Ideal for cleanup profiles following the pattern „delete files a certain time after they arrive“.

• Enabled per profile under Advanced → Deferred processing with a freely selectable delay (seconds to years).
• The waiting period is counted from the moment the file was first detected. Files already present before activation remain untouched — only files arriving afterwards are queued.
• Queued files are not read again until they are due and do not clutter the status display; once the waiting period has elapsed, they are processed automatically.
• The number of pending deferrals is shown in the profile list (column „Deferred“).
• During catch-up processing, active deferrals can optionally be ignored (process files immediately) or the affected profiles can be skipped.

Renamed: „Pause before processing“

The former option „Delay before processing“ (a short wait before a profile starts processing) has been renamed to „Pause before processing“ to clearly distinguish it from the new feature. Its behavior is unchanged.

Profile settings: new „Advanced“ page

The profile „General“ page has been streamlined: less frequently needed options are now grouped on a dedicated „Advanced“ page — making it easier to get started without losing any functionality.

• The „Advanced“ page contains, among others: extract values only once, wait for file groups, profile-specific processing times, pause before processing, re-check the filter before each task, the retry options, and deferred processing.
• If one of these options deviates from its default, the page label shows a counter (for example „Advanced (2)“) — so you can tell at a glance whether a profile uses advanced settings.

Catch-up processing: faster filtering of large file sets

When catch-up processing is limited to a period, files outside that period are now skipped without reading them first. For large folders containing several thousand files, this makes the check considerably faster.

• This applies to the date types „File modification date“ and „File creation date“, as these values can be read directly from the file system.
• For the PDF-internal date types („PDF modified date“ and „PDF creation date“) each file still has to be opened to read the date stored inside the document. If you limit a very large set to a short period, it is therefore best to choose one of the two file date types.

Files that are already named correctly are no longer re-detected on every check

When a task renames a file within the same folder and the file is already named correctly (source and destination are identical, for example a file synced back from the cloud), nothing happens — and the file is no longer mistakenly treated as “recurring”.

• Previously such files could be re-detected on every check cycle (“File X of Y”) even though they remained unchanged and were skipped anyway — this inflated the display and re-read the files unnecessarily.
• In addition, recurring files are no longer re-checked at all when no profile would reprocess them anyway. Existing entries of this kind disappear automatically — no action is required.

Some optimizations

Regression fixed - faster processing of large folders when using a file name filter

When processing many files — for example several thousand PDFs on a network drive (NAS) — the program now again checks the file name first, before a file is actually opened and read.

• Files that do not match the file name, path or size filter are skipped without their content having to be read from disk.
• Especially on slow network drives this considerably shortens the overall time when only a subset of the files needs to be processed.
• Profiles that filter by PDF content, metadata or date still open the files in question as before — only files that already cannot match based on name, path or size are skipped.

More information during processing

While processing, the status display now shows the progress as “File X of Y” together with the name and folder of the file currently being processed.

• This makes it easy to see at a glance how many of the files matching the time period and filter criteria have already been processed, and which file is currently being handled.

Text redaction: no longer removes characters from the line above

When redacting text (the “Redact text” task), very tight line spacing could cause part of the line immediately above to be removed as well.

Significantly smaller internal processing database

The internal database that stores the processing log could accumulate empty placeholder entries over time which were no longer cleaned up automatically. Under heavy use this could grow to several hundred MB.

• On the first program start after the update the database is cleaned up and compacted once.
• Expected size reduction: typically from several hundred MB down to a few MB.
• Entries with real processing information (success, error, retries) are fully preserved.

Filter “Page count” extended with “equals” comparison

• The filters for page count, attachment count and character count now also offer the “equals” comparison. This makes it easy to filter for documents with an exact count.

Filter “page count is between” corrected (inclusive)

The “is between” comparison was strictly exclusive (open interval). To match the value 2, you had to enter 1 and 3. That was unintuitive and did not match common conventions (Excel, SQL).

• From now on the inclusive meaning applies: “between 2 and 5” matches 2, 3, 4, 5.
• Existing profiles are migrated automatically on first start – the resulting match set stays the same.

Example files reorderable (profile settings)

• In profile settings → example files, entries can now be moved up and down using two new arrow buttons.
• For preview calculations (rename, storage location, e-mail recipients, …) the first example file that actually exists on disk is now preferred. If the first file in the list is no longer available, the next existing one is used automatically instead of showing an error.

Missing example files are clearly indicated

• The example files page now shows an “Exists” column (Yes/No) as the first column, indicating whether each configured file is still present on disk; missing files are highlighted in red. The display refreshes automatically — copy a file back and it reappears as present shortly afterwards.
• In the example-file selection lists (e.g. in the rule editor), files that no longer exist are hidden.
• In the “values to determine” preview, a missing example file now shows the clear note “example file no longer exists” instead of a technical error message — and stale preview results are no longer left behind.
• When opening the rule editor, we now point out if none of the configured example files exist anymore.

Processing log: “Size/pages” display

• The processing log now shows the file size together with the page count, e.g. “139 KB / 3 pages”.

Cleaner response to Windows shutdown

• The application now exits reliably when Windows shuts down or the user logs off. The message “This application is preventing shutdown” should no longer appear.

Diagnostic log for support cases

In Program Options → Other, the diagnostic log can now be enabled:

• Writes detailed information about the processing flow to a log file in the AppData directory
• Disabled by default – can be enabled when needed and turned back off after diagnostics

Data extraction: “Text of page(s)” with “No determination required”

• When Text of page(s) was selected as the data source together with No determination required as the page determination, the preview incorrectly reported “Keyword not found”
• Fixed: the preview now reliably delivers the text of all pages and shows the extracted data

Additional

• When exporting a single profile from the context menu, the profile name is now used as the suggested filename
• Some optimizations

New stamp styles

The Add Stamp/Watermark task now offers three style variants for text stamps:

• With frame (classic) - the existing stamp look with frame, ideal for status markers like “PAID” or “APPROVED”
• Text only - plain text without frame or background, suitable for subtle labels such as processor initials
• Text with background (highlighter) - text with a colored background rectangle in highlighter look, optimal for compact sorting markers on archived emails and documents

More design freedom

• Font freely selectable - all installed system fonts are available
• Font size and bold style configurable separately
• Background color and padding for the highlighter style fully customizable (default: yellow marker)

Existing profiles

• Profiles from earlier versions keep their previous stamp appearance. The new variants can be enabled per profile without any migration step

• Some optimizations

• Some optimizations
• New option in the "Processing" category - Allow empty input text for query rules with a fallback value

Applies to data determination rules of the "Query" or "Query (with list)" type. If enabled, an empty input text (e.g., after failed OCR) will NOT result in the error "Text contains no characters," provided that either the option "Query result may be empty" is selected or a fallback value ("Value if not found" or "If no entry found") has been configured.

Example: Scanned invoices without recognizable OCR text will still be assigned the fallback name (e.g., "not available") and processed further, rather than being terminated with an error message.

• Save attachments + attachment name filter: factur-x.xml / ZUGFeRD attachments were not found in some third-party system PDFs
• Unnecessary retrieval of barcode data when opening profile settings

• Date recognition: Custom formats without a day (e.g., `MMMM yyyy`) were not recognized
• Some optimizations

• Some optimizations and corrections

• Fixed a problem with the filter

• In the options, some settings were cut off at certain resolutions

• Options -> Processing Log allows you to create CSV logs of processing activities in a folder
• Some optimizations

• In certain cases, the report period was incorrect when an Excel report was sent automatically

• New option: Processing -> Background Processor is managed externally. This option should be enabled when running the program as a service or when starting it regularly via the Windows Task Scheduler, to prevent the program from creating a duplicate background process.
• Some optimizations and corrections

• "Save as Draft" is now also supported for sending via SMTP and Online Exchange.
  If you want to review messages before they are actually sent, enable the "Save as draft" option in the sending options of the "Send file" task. Messages will not be sent immediately but stored in the Drafts window, where you can review and adjust recipients, subject, message text and attachments before manually triggering the send.
  The button for opening the Drafts window appears in the top bar of the main window if there are any drafts available.
• File Size Stability Check: New option in program settings that ensures a new file's size has not changed for a configurable duration before processing begins. Prevents potential errors for files on network drives that are still being copied.
• Configurable File Availability Timeout: The maximum wait time for locked files is now configurable in program options (default: 10 seconds, previously hardcoded to 5 seconds).
• Overwrite Original for Merge: The "Overwrite original file" option is now available for the "Merge PDFs" task.
• Extracted text is automatically trimmed: Leading and trailing empty lines are automatically removed during data extraction (data area and text of pages).
• "Remove Pages" task -> new options: QR code value contains/does not contain + Remove pages without a QR code

• For older barcodes, the character encoding may be incorrect. If necessary, it will now be read with the correct encoding.

• In the profile settings, under the "General" category, you can now enable or disable the verification of filter criteria before each task
• When merging, the primary file can be overwritten, just as with other tasks
• Fixed an issue that could cause the program to freeze when saving a profile

• New options for removing pages: QR code value contains/does not contain + Remove pages without a QR code
• Optional delay before applying a profile (Profile Settings -> General)
• New option for sending via SMTP: Ignore SSL/TLS certificate errors

• The deployment of a only partially compatible version of a PDF library could cause crashes with some PDF files.

• New format for log files

• Exchange On-Premise is no longer supported in version 2. When importing options from version 1, this sending method (if enabled) caused the application to crash.
• Some optimizations

• Profile import from Version 1 - Incompatibility fixed
• Optional maximum duration per file processing (Options -> Processing)
• A new processing cycle was triggered after renaming or moving files back to the original folder
• Simple file grouping - improved protection against re-processing
• The catch-up process can be resumed after crashes caused by external libraries

• Fixed issue with running as a service application. If PDFProcessor.exe is started by the Service Control Manager, the parameter -h may no longer be used (see Help).

• For the "Query (with list)" data type, there is now a new option called "Use longest match". This option is enabled by default for new rules, but disabled for existing rules. However, we recommend enabling it.

• Fixed issue with using sequential numbers
• Profiles, options, and licenses can now also be imported from V1 backups

• When previewing data extraction results, the second row of the top toolbar was occasionally only fully visible after changing the category and going back
• When configuring the "Send file" task, a control element was not fully visible for a specific configuration

• Advanced filter options for dynamic attachments
• New option in print settings: Print with default printer
• Add annotation: Placeholder menu and optional size specification added
• Beta status has been ended

• Split by keyword: the option 'If the search term does not appear in the document, return an error' has been split into two options. Additionally, #Return an error if the number of pages remains unchanged'.
• Filter expanded. 'Customer <EntryFromList...> <BeginOfRegex...>' is now also supported, i.e. combinations of text and placeholders between <AND> and <OR>.
• Does not contain + AND error fixed: now correctly 'NOT(A AND B)'
• Insert barcode: optional label at top, bottom, right, or left
• Error corrections

• New data source ‘Sequential number’
• Placeholder menu added to ‘Insert barcode’

Version 2 is a comprehensive update with many new features and improvements.

New Task Types

• Compress PDF files - Reduce file size through image optimization and removal of unnecessary elements
• Create ZUGFeRD/Factur-X e-invoices - Embed user-provided XML invoice data into PDF (ZUGFeRD 2.0/2.1, Factur-X 1.0)
• Convert to PDF/A - For revision-proof long-term archiving (PDF/A-1b, PDF/A-2b, PDF/A-3b)
• Add stamps and watermarks - Text, images, or dynamic content from extracted data
• Insert barcodes - QR codes, Code 128, DataMatrix, Aztec, and other formats
• Anonymize documents - Automatically replace sensitive data with placeholder text
• Remove pages - Delete cover pages, blank pages, or specific page ranges
• Embed attachments - Integrate receipts, delivery notes, or other documents into PDF

Extended Task Management

• 21 task types - Rename, Move, Copy, Delete, Print, Send email, Merge, Split, OCR, Save attachments, Save extractions, Apply template, Change file protection, Compress, Embed attachments, Create ZUGFeRD, Convert to PDF/A, Remove pages, Add stamps, Insert barcodes, Replace content
• Unlimited tasks per profile - No longer limited to 3 tasks
• Freely selectable order - Tasks can be sorted via drag & drop

Processing Times

• Time windows for profiles - Processing only at certain times (e.g., only during office hours)
• Weekday-based scheduling - Different times for different weekdays
• Diagnostics view - Overview of all configured processing times

Batch Printing for Large Documents

• Automatic splitting of large print jobs into smaller batches
• Prevents system instability and long spooling times for documents with many pages
• Configurable thresholds and batch sizes

Command Line Extensions

• Profile selection in headless mode for Windows Task Scheduler
• Different scheduled tasks can execute different profiles
• Example: PDFProcessor.exe --headless --run_once --profile "Invoices"

New Date Placeholders

• Prior year placeholders: For documents referring to the previous year period
• Prior month placeholders: Ideal for monthly billing and archiving
• Prior day placeholders: For daily reports with back-reference

Additional Improvements

• Modernized .NET 8.0 framework for better performance and stability
• Improved barcode and QR code recognition with machine learning
• Extended filter options for odd/even page numbers
• Optimized memory management when processing large PDF files
• Updated OCR engine for improved text recognition
• Automatic cleanup for known files (time-based cleanup)

• Optimizations and bug fixes.