ORDNA — Change log
What each version brought, newest first. The entries come straight from the change log that ORDNA ships with.
ORDNA is in public beta; until the first commercial release
the version scheme is 0.MINOR.PATCH — PATCH for a bug fix, MINOR for a new
feature or a change to existing behaviour. The first commercial release
will be 1.0.0.
With 0.49.0 (21.10.2026), number 2 of the privacy policy also names the rules of the intake filter and what ORDNA records about them in the provenance journal.
With 0.50.0 (21.10.2026), number 2 of the privacy policy also names the subfolder “Ausgefiltert”, the bringing-back of filtered-out documents and the new entries in the provenance journal (checked invoice amount, copy of the invoice data of a filtered-out invoice).
With 0.51.0 (24.10.2026), number 2 of the privacy policy names the new way to check for updates yourself: Help → Check for updates instead of Check now in the settings. What is transmitted stays the same.
With 0.53.0 (04.11.2026), number 2 of the privacy policy also covers the backup: what ORDNA copies to the chosen folder, what stays out and that the copy is not encrypted.
0.53.1 2026-11-10
- Fixed: text recognition no longer fails at random with “ocrmypdf rc=15” in its last step, the conversion to the long-term archive format PDF/A. In the installed ORDNA the start of Ghostscript failed at times; the documents concerned were filed without that conversion, documents without text under “Not archived”. Documents filed before this update stay as they are; ORDNA does not convert them again.
0.53.0 2026-11-04
- Added: optional backup (Settings, section Backup). ORDNA copies the archive to a folder on another drive at an interval you choose or on request, and copies only what is new or changed. Documents that would be replaced or are no longer in the archive stay in the target under “Ältere Stände”. A running backup finishes even if you close ORDNA; it does not start while ORDNA is closed. Off by default; the copy is not encrypted.
- Added: a moved archive is found again in one step. Choose the archive folder at its new location in the new window Archive not found or under Settings, Document types and target folders. ORDNA assigns the filing folders itself, shows the assignment with the number of documents first and applies the target folders and the copy of the filing journal together. A whole archive folder is no longer accepted for a single class.
- Added: View changes in the update window opens the change log on zuentec.de in your browser, so you can see what is new before you install.
- Added: Help → About ORDNA shows the versions of Ollama and Ghostscript, and the log names them at the start of every run.
- Changed: privacy policy, section 2. The optional backup is a new storage location of your choice: the archive, the Excel overview, the filing journal, your settings and the recognised details go there, unencrypted; mailbox access data and the search index do not.
- Changed: user manual 2.73. New chapter 5.15 on the backup; the manual explains the moved archive, the versions under About ORDNA and the new result line of a folder run; the PDF document title names the manual and what ORDNA does.
- Fixed: after the archive had been moved, Check archive reported almost every document as having no journal entry and could report unchanged documents as changed. It now recognises the documents at their new location.
- Fixed: a folder run without Include subfolders whose documents lie only in subfolders now names the missing tick instead of pointing to the accepted file types.
- Fixed: when the version with all pages of a document whose blank pages were removed could not be moved to Duplicates (for example because the target folder had disappeared), the whole run stopped with an unexpected error. The run now continues and counts it as a failed filing.
0.52.0 2026-10-25
- Added: a ? at the top right of every card opens the user manual at the chapter that explains the card. Help → User manual (F1) still opens it at the beginning.
- Changed: new installations accept PDF, JPG, JPEG and DOCX by default (until now only PDF), so a photographed receipt or a Word document is filed without a detour through the settings. An existing installation keeps its selection.
- Changed: user manual 2.69 and reference guide 1.18. Chapter 5 of the manual follows the five cards of the main page (5.1 to 5.5 carry the card titles); chapter 3.1 says that older Word documents (DOC) are converted only if Microsoft Word is installed, and what a DOCX file keeps without Word; chapter 5.5 names the date fields of the search. The reference guide explains the contract whose other party is missing, and the label “In short:” under each category heading is gone.
- Fixed: without Microsoft Word, Word documents (DOCX) with more than one paragraph were not converted and stayed unprocessed. The plain PDF now also carries headers and footers and text boxes, in the order of the document, and keeps €, quotation marks and dashes. Images are not carried over; DOC files still need Microsoft Word.
- Fixed: the interface language, the appearance and “Don't show again” are kept after a restart. A reinstallation in another language no longer changes the language you chose.
- Fixed: a contract whose other party the recognition returned only as the sender is now filed under Contracts — when its type names a contract or an agreement.
- Fixed: the three date fields of the search are labelled — Date from, Date to, Follow-up due by.
- Fixed: in the dark appearance the calendar symbol of the date fields, check boxes and the text field of the error report are drawn dark too — the calendar symbol was hardly visible.
- Fixed: in German, French and Spanish the window opens tall enough for six search results again (18 pixels short since 0.51.0).
- Fixed: opening the user manual cleared the result lines of the cards. They now stay.
0.51.0 2026-10-24
- Added: a menu bar in the header — File · View · Tools · Settings · Language · Help, operated as in other Windows programs: a click opens a menu, Alt or F10 reaches it from the keyboard (underlined access keys, arrow keys, Esc), F1 opens the user manual and F5 refreshes the overview. Settings opens the settings window directly, without a menu. The ORDNA logo with its version now sits on the right.
- Changed: the symbol buttons in the header are gone. Every function has exactly one way: Recently filed documents and Refresh are under View, the log under Tools; the interface language has a menu of its own; manual, reference guide, change log, privacy, error report under Help. While the overview is being refreshed, a note next to the logo says so.
- Added: File → Exit. It closes ORDNA like the X — and, unlike the X, asks first while a processing run is in progress (Cancel is preselected).
- Added: Help → About ORDNA with version and manufacturer.
- Changed: the update check moved to Help → Check for updates. The Check now button in the settings is gone; the setting “Check for updates at program start” is unchanged. Every result — up to date, no connection, update found — appears in the update window.
- Changed: maintenance moved to the Tools menu. Check naming, Check archive, Open check report and Delete all duplicates are no longer a section of the settings window. They no longer open the log by themselves: while they run, a note in the header says so, and a message window shows the result (findings, as before, in the search results).
- Changed: the “new” column counts only documents that came in through Search a folder or Search the mailbox (including the daily fetch). Opening a document — for example through a link in the Excel workbook — no longer makes it “new”, and neither does correcting it or moving it to another class. A document that just came in stays new after you correct it.
- Changed: the user manual, the reference guide and the change log open in a window of ORDNA's own instead of the program set up for PDF files, so they no longer open full screen; links in them open your browser.
- Changed: Open Excel workbook is a button as wide as Edit rules …, at the bottom left of the archive card and level with it; the two arrows pointing to the search card now sit between the two cards; the intake filter card is marked (optional). The title row of the settings window with OK stays in place while you scroll.
- Changed: errors and warnings use a rust red (#B94927; in the dark design the lighter #E46444) instead of the former red. Error messages are now clearly readable in the dark design, and Remove from archive darkens when you point at it.
- Changed: user manual 2.65 and reference guide 1.15. Every path now names the menus (the reference guide: Settings, section Naming of invoices); chapter 3.3 explains the menu bar and its keyboard use, chapter 4.3 describes File → Exit; figures 1 to 3 show the new header, the settings window with its ten sections and Remove from archive in the new rust red.
- Changed: privacy policy, section 2 — the update check is started with Help → Check for updates instead of the former Check now button; what is transmitted is unchanged.
- Fixed: Tools → Open check report stayed greyed out after every restart, although the report of the last check with findings was still there, and after a check without findings it stayed enabled and then only said there was no report. It now opens the latest report even after a restart and is greyed out after a check without findings.
- Fixed: intake filter — the archive question could show outdated numbers. Changing a rule, or unticking “also apply the changed rules to the existing archive”, while the question was shown left the old numbers in place, and Yes, apply applied the changed rule. The question now disappears with every change; OK asks again with fresh numbers.
- Fixed: intake filter — “New rule” clicked twice created two empty rules. While the selected rule is still empty, the button now only moves the cursor to its name.
0.50.0 2026-10-21
- Changed: user manual 2.63 and reference guide 1.14. Figure 1 of the manual shows the three-column main window of 0.49.0; chapter 5.15 describes the subfolder, “Apply rule” and the bringing-back of filtered documents; the ⊘ mark is now drawn in the PDFs (it was an empty box since manual 2.61).
- Changed: filtered documents get their own subfolder. What the intake filter sorts out now lies in Not archived\Ausgefiltert, apart from the unreadable originals in Not archived itself — so the subfolder can be emptied without touching an original. The archive overview shows the number of filtered documents behind the count of Not archived (⊘), and every such search hit carries ⊘ with the rule and the reason. Documents filtered out by 0.49.0 are sorted into the subfolder the next time changed rules are applied to the archive.
- Added: changed rules bring documents back. When rules are applied to the existing archive, ORDNA also re-checks everything it filtered out earlier. A document that no rule matches any more comes back: one taken from the archive returns under its old name to its folder (with its invoice data), one sorted out during processing runs through recognition again. The dialog names all numbers first and asks; nothing is deleted.
- Added: “Apply rule” in the rules window checks the entry and puts the rule, named, into the list before OK saves.
- Fixed: documents moved to Not archived by applying rules to the archive did not appear in the search until the next refresh and were not counted as “new”; the dialog closed before the result could be read.
- Changed: privacy policy, section 2. Filtered documents are stored in the new subfolder; the provenance journal now also keeps the checked invoice amount of a filtered invoice and, for an invoice taken from the archive, a copy of its analysis data, so that it can be restored when the document comes back.
0.49.0 2026-10-21
- Added: the intake filter. A new card between the sources and the archive. Rules name a document class (or all), terms that must appear in the text, and for invoices an amount range — for example “invoices up to 25 €” or “correspondence containing a certain name”. A document that matches goes unchanged to Not archived instead of its subject folder; the log names rule and reason (line marked ⊘), and the result line counts the filtered documents. Nothing is deleted. There is no rule by default; up to ten rules can be defined in the Edit rules window, and changed rules can also be applied to the existing archive — ORDNA shows the number of matches first and asks. An amount only counts when it was recognised on the invoice; an invoice without a recognised amount is archived. Manual 2.61, chapter 5.15.
- Changed: the archive path is the first column header of the archive table and opens the archive folder in Explorer; the arrows from both sources point at the new card, and the three columns of the main window are the same height.
0.48.1 2026-10-20
- Changed: recognition no longer depends on a default of Ollama. ORDNA now sets the repetition setting of the recognition model itself, to the value that current versions of Ollama use. An update of Ollama can therefore no longer change how documents are recognised without anyone noticing. With Ollama 0.40 nothing changes; a version older than 0.32.10 now uses the same setting as the current one.
- Changed: the wording of the update function. The section in the Settings window is called Updates, and the tick box reads Check for updates at program start. The German interface, the log and the documents now say “Version” instead of “Fassung” throughout.
- Changed: the setup's space check counts 2,863 MB for Ollama, as measured with Ollama 0.40.6.
- Changed: user manual 2.60. Chapter 4.3 has a new paragraph Closing ORDNA: the X at the top right ends ORDNA, processing under way stops, and the next run takes up the documents it had started; Ollama keeps running as a separate program. Chapters 2 and 5.7 use the new wording of the update function. The reference guide on document recognition is now at document version 1.12.
0.48.0 2026-10-17
- Added: the setup names an older Windows 10. The System check page now also appears on Windows 10 older than version 22H2 (build 19045): Ollama, which performs document recognition, requires this version, and whether recognition works on older versions has not been tested. As before, Next installs anyway.
- Changed: the publisher's mark. The first page of every document — user manual, reference guide on document recognition, change log — now shows the ZUENTEC raven at the top, the following pages the ORDNA mark. The title bar of the ORDNA window shows the raven and “zuentec”; the program icon remains the ORDNA icon.
- Changed: user manual 2.58. Chapter 4.1 mentions the Windows version on the System check page.
0.47.0 2026-10-16
- Security: Ghostscript 10.08.0. Older versions of Ghostscript have a known security vulnerability (CVE-2026-39919) that a specially crafted PDF can trigger. The setup now downloads version 10.08.0; if an older version is installed, the Additional Programs page reports this and ticks the box for the new version. ORDNA shows a notice at start-up while the Ghostscript it uses is older than 10.08.0, and if several versions are installed it now always uses the newest one — also over a Ghostscript found in the system path.
- Added: a message when the window cannot work. If the Windows component Microsoft Edge WebView2 is missing or too old (on some editions of Windows 10 IoT or LTSC), ORDNA says so at start-up and offers Microsoft's download page, instead of showing a window that does not respond.
- Added: the setup checks the computer. A System check page appears only if something is missing — less than 8 GB of memory installed, or the WebView2 component — and lists only that; Next installs anyway. On Next on the Additional Programs page, the setup also checks per drive whether there is enough free space for the ticked programs (Ollama and the recognition model need about 9 GB in total during installation) and asks before continuing if not.
- Added: a notice when the recognition model runs without a graphics card or only partly fits into graphics memory. ORDNA reads this from Ollama itself at the start of a run; recognition then takes one to several minutes per document.
- Changed: user manual 2.57. Chapter 3.2 states the system requirements (at least 8 GB of memory, 16 GB recommended, a graphics card is required); chapters 2, 4 and 6 cover Ghostscript 10.08.0, the new setup checks and the new notices.
0.46.0 2026-10-16
- Added: a notice when the computer has too little memory (prompted by a report from Blondi_2021 in the deskmodder.de forum). Measured, not estimated: without a dedicated graphics card the recognition model takes about 5.3 GB of memory, ORDNA with text recognition up to 1.4 GB more. If the computer has less than 6.7 GB in total, a notice above the source cards says so at start-up; Hide closes it until the next start. A computer with 4 GB is not enough for the recognition.
- Changed: when Ollama refuses to load the model, the log names Ollama's own reason — for example "model requires more system memory" — instead of only "500 Server Error". The message at the start button now points to that reason and no longer blames the graphics driver across the board.
- Changed: user manual 2.55. Chapter 3.2 states the measured memory figures; chapter 6 explains the start-up notice and how to read Ollama's reason in the log.
0.45.1 2026-10-13
- Changed: an e-invoice is never rewritten. A PDF that carries embedded invoice data (ZUGFeRD, Factur-X or the XRechnung profile) now stays exactly as it arrived — no conversion, no new text recognition, no removal of blank pages. Until now a hybrid invoice without the PDF/A mark was converted, and while the embedded invoice file survived, its marking as an e-invoice was lost.
- Fixed: the account detail in the names of bank documents. A telephone number, a VAT identification number or a bank sort code is no longer taken as the account. If the document contains exactly one card number, ORDNA takes its last four digits (credit card statements were named after the last digits of the card service's phone number); otherwise the detail stays empty and the document is not filed under a wrong name.
- Changed: the setup installs Ghostscript first. Its administrator question and wizard now come right after you click Install and the 65 MB download; Ollama and the recognition model follow without any further question. Until now Windows asked only after the Ollama download and installation — whoever had left the computer found the installation waiting.
- Changed: an older Ghostscript is no longer "nothing to do". If an older working version is installed, the Additional Programs page recommends version 10.07.1, which closes known security vulnerabilities; tick the box to install it alongside — ORDNA then uses it. The box is not ticked for you.
- Changed: the text recognition ships only the libraries it uses. Tesseract now comes with 26 instead of 51 libraries (46 MB less); the others were only needed by training tools ORDNA does not include. The licence texts of all of them are in
THIRD-PARTY-LICENSES.txt, and the source code of the two copyleft libraries among them (JBIG-KIT and GNU libiconv) is supplied inDocumentation\Source. - Changed: user manual 2.54. The links to the reference guide and to the revision history now point to zuentec.de, so they also work in a manual you downloaded; both documents are still supplied locally. The manual also covers the new Ghostscript status and the account rule.
0.45.0 2026-10-10
- Added: a follow-up date per document (suggested by Blondi_2021 in the deskmodder.de forum). In the ✏️ window every archived document — whatever its class — takes a date on which you want to see it again, for instance before a notice period of a contract, an insurance policy or a mobile phone plan ends; Set follow-up for all sets one date for a whole result list. ORDNA reminds you inside the program: at start-up a notice says how many documents are due and how many become due within the lead time (Settings → Follow-up, default 30 days), Show lists them with ✏️ and the open link, and the search has a filter Follow-up due by; due and soon-due documents are highlighted in the result list — with a done link that ticks the follow-up off in one click — and every sheet of the Excel overview ends with a Follow-up column (a real date, so you can sort by it). The date is stored in the filing journal next to the archive like the location of the original — never in the file name, the document itself stays unchanged, and it survives a reinstallation. There is no background service, no autostart, no e-mail and no calendar service: ORDNA reminds you only when you start it. Going back to 0.44.x loses nothing — the older version simply does not show the date.
- Changed: the recognition model gets ten minutes per request instead of three, and a computer that is too slow is named as such. On a machine without a graphics card the model answered after the old limit, so the same file failed in every run with "LLM disturbance". The log now says "model answers too slowly (over 600 s) — this computer is too slow for the recognition model", and the file stays in the source for the next run.
- Changed: direct requests to the local service are checked once, at the entrance. A request body that is not a JSON object, a field of the wrong type or a number out of range used to end in an internal server error at up to 23 routes, and the text "false" switched settings ON. Such requests are now rejected with a message; no way of the user interface was affected.
- Added: a notice when ORDNA no longer responds. If the status cannot be fetched three times in a row, the window says so and asks you to close it and start ORDNA again; a run in progress resumes at the next start.
- Smaller things: in the French and Spanish setup the introduction of the page Additional programs no longer touches the Ghostscript line; the OCR language list names Italian (old prints) and Georgian (old script) instead of showing their codes; French guillemets no longer break at a line end in the manual and the reference, and the Spanish reference writes them without inner spaces, as is usual.
0.44.0 2026-10-08
- Added: two custom attributes per invoice. Under Settings → Naming of invoices you give an attribute a heading of your choice, such as "Contract" or "Cost centre", and a word list (entries separated by
;, at least one entry): if exactly one of the listed words appears in the text of an invoice, ORDNA enters it when filing and adds it to existing invoices on the next refresh; with several matches the field stays empty. You can also enter or correct the value per invoice in the ✏️ window, or for a whole result list with Set attribute for all; what you entered by hand is kept. The attributes appear as columns in the Excel overview, in the search and in the result rows. Headings, list entries and values accept letters, digits, spaces and- . & + ( ); other characters are rejected with a message because they would interfere with the file name. - Added: file names of invoices from building blocks in the order you choose. The date always comes first; issuer, invoice number and the two attributes follow in the order you set — click a row and move it with the arrows. The tick box decides whether a block goes into the file name; an attribute without a value is left out, so two invoices of the same issuer on the same day are named
…_Cable.pdfand…_Internet.pdfinstead of…_2.pdf. As before, ORDNA shows how many files a change would rename and renames them only after Rename and apply. - Changed: the three LGPL libraries (fpdf2, img2pdf, pi_heif) are shipped as separate files in the program folder instead of inside the program archive, so you can replace them with your own versions as the licence allows; the licence page explains how. The complete list of all bundled packages with their licence texts is in
Documentation\THIRD-PARTY-LICENSES.txt. - Note for going back to 0.43.x: choose the naming "Date_Issuer" (no invoice number, no attributes) and apply it before installing the older version — it does not know the new building blocks and would read such file names incorrectly. Attribute values stay in the filing journal and return when you install 0.44.0 again.
0.43.1 2026-10-06
- Fixed: other websites can no longer give ORDNA instructions. The ORDNA service is only reachable from your own computer, but any website open in your browser could still send it requests while ORDNA was running, for example to have the mailbox test send your saved password to a different server. ORDNA now refuses requests that do not come from its own window. A saved password is also only used for the server, port and SSL setting it was entered for: if you change one of them, enter the password again. A request with a malformed mailbox list can no longer delete all mailboxes.
- Fixed: ORDNA never signs in to a mailbox unencrypted, and it checks the server's certificate. If a server offered no encryption (STARTTLS), ORDNA used to send the password in plain text, and the certificate of an encrypted connection was not checked. ORDNA now refuses to sign in without encryption or to a server without a valid certificate, and the message says why. A mail bridge on your own computer (127.0.0.1) is the only exception. If your server is entered as an IP address or uses a self-issued certificate, enter the provider's server name.
- Fixed: date and issuer always stay complete in file names. When a name had to be shortened because the archive lies deep in the folder structure, ORDNA could cut into the issuer and even the date. Now only the descriptive part (subject, kind, branch) is shortened. A target folder may therefore be at most 180 characters long; ORDNA rejects longer ones and names the folder before a run starts.
- Fixed: a damaged settings or mailbox file no longer stops ORDNA or costs you data. If the settings file could not be read, ORDNA did not start at all. It now keeps a copy of the damaged file, restores the target folders and the invoice naming from its filing journal, uses the defaults for everything else and says so above the source cards. A damaged file with mailbox credentials is no longer overwritten when other settings are saved; a note on the mail card asks you to enter the mailboxes again. Mailbox credentials are now written so that a crash cannot leave a half-written file.
- Fixed: failed filings are counted and shown. If a target folder was write-protected or the drive full, the result line still said "0 errors". Every failed filing now counts, the line turns red, and the log names the file and the reason. "Search a folder" no longer leaves working copies behind that the next run would process again.
- Fixed: double clicks and simultaneous actions. Saving settings in two places at once could end with an error, two quick start requests could start two runs, a double click on "Install" in the update window started two downloads and two setups, and a double click on "Report a problem" opened two mail windows. All of this is now prevented.
- Fixed: Settings window. A class can no longer be given the folder of another class or a folder inside it. If saving fails, the reason now appears next to OK and the section concerned opens. Ticking a class no longer discards a folder path you typed in another row but had not saved yet.
- Fixed: smaller things. Spanish stays a fixed OCR language, as the window shows. Reading the schedule no longer fails when PowerShell is missing from the system path. If the update server answers with an error, ORDNA says so instead of reporting no internet connection. The licence page of the setup now names all bundled components correctly; the full licence texts of every bundled package are in Documentation\THIRD-PARTY-LICENSES.txt.
- Changed: the setup asks you to stay until Ghostscript is installed. Windows asks for administrator rights and Ghostscript's own window needs "Finish"; only after that does the installation continue without you. Reported by Holgi from the deskmodder.de forum.
0.43.0 2026-10-05
- Added: copy the log with personal details hidden. Above the log (📋) there are four check boxes, all ticked by default: account numbers (every number with five or more digits keeps only its last four, e.g. xxxx9304), user name in folder paths, email addresses, and issuer names taken from the file names in your archive (replaced by Aussteller-A, Aussteller-B and so on). They apply only to the copy button; the display and Ctrl+C stay unchanged, and if the details cannot be hidden, nothing is copied. Useful whenever you pass on the log, for example in a forum. Suggested by Holgi from the deskmodder.de forum.
- Fixed: log lines no longer appear several times. When ORDNA was busy, for example while the recognition model was working on a slow computer, the window asked for news faster than it got answers, and the same line could appear many times in a row. It now waits for each answer before asking again.
- Fixed: "Search a folder" never reads your archive. Until now only the Invoices folder was left out. If the archive lay inside the source folder, for example when reading Documents with "Include subfolders" while the archive is in Documents\ORDNA\Archiv, ORDNA read its own archive again: every run parked another copy in "Duplikate", and with "Delete source file after archiving" ticked it removed archived documents from their folder. ORDNA now skips every target folder, wherever it lies, together with its "Duplikate" folder, and the log names the skipped folders. A target folder, a folder inside it and the archive root folder can no longer be chosen as source folder. Reported by Holgi from the deskmodder.de forum.
- Fixed: all statements of one account now carry the same name. For bank documents, the account part of the file name is now always the last four digits of the account number or IBAN, failing that of the customer or card number. Until now ORDNA took over what the recognition delivered, so one account could appear with four, eight or ten digits, and when asked for the last four digits the recognition often picked the first four. It now reads the complete number, and ORDNA shortens it itself. A detail without any digit counts as missing. Names already in the archive stay as they are; ✏️ adjusts an older name the next time you click Apply. Reported by Holgi from the deskmodder.de forum.
0.42.0 2026-10-04
- Added: the setup can download the recognition model right away. The Additional Programs page has a third check box, "Download the recognition model now". It is ticked while the model is missing and runs as the very last step, after everything that asks a question, so you can leave the computer alone: a console window shows the download (about 4.6 GB) and closes by itself, and ORDNA is ready as soon as the setup ends. Without the tick, ORDNA downloads the model on the first run as before. Thanks again to Holgi from the deskmodder.de forum.
- Fixed: text in the ORDNA window can be selected and copied again. Since version 0.20.0 the window blocked any selection, so the log could not be copied, neither with the mouse nor with Ctrl+A. Text can now be selected everywhere except on buttons. The log window has a new button with two overlapping rectangles that copies the whole log with its times; Ctrl+A inside the log selects only the log. Reported by Holgi from the deskmodder.de forum.
- Fixed: answering "Yes" when uninstalling now removes downloaded OCR languages as well. Languages added with "Add language" are stored next to the program. Until now they stayed behind after every uninstallation, and with them the program folder, even after answering Yes to removing settings and data. They are now removed with Yes and kept with No, like the settings, and the question at the end of the uninstallation names them. Ollama with the recognition model and Ghostscript are separate programs and are uninstalled separately.
- Fixed: no log file in the program folder. When the invoice step had to run text recognition itself, it wrote a log file (pdfa_check_log.csv) into the program folder. It now goes to the log folder, and uninstalling removes a copy left by an earlier version.
- Changed: the setup's information page describes the uninstallation completely. By default only the program is removed; at the end ORDNA asks whether settings and data should be removed as well (default No); your document archive is never touched.
- Added: "Privacy" in the menu. It opens the privacy policy for ORDNA on zuentec.de in your browser, in the language of the interface. ORDNA itself transmits nothing in doing so.
- Changed: the error report no longer contains the Python version. "Report a problem" now lists exactly what the manual and the privacy policy name: version, Windows version, interface language, recognition model, OCR languages and the number of documents and active classes, plus your own text.
- Improved: the log no longer calls a document in "Not archived" archived. After such a document it now says that the original lies unchanged in "Not archived", and the summary of a folder or mailbox run states how many of the filed documents went there.
- Fixed: the log stopped after 800 lines. After 800 lines in one session, no further lines appeared in the log window until ORDNA was restarted. It now keeps showing new lines.
- Improved: the log says why a document was not filed in its folder. When ORDNA recognises a document type but does not file it there, the log line now names the missing detail (for example the account of a bank statement) or says that the recognition was not sure. Such documents still go to Other documents rather than into the wrong folder.
- Improved: leftovers of an interrupted run are named. If a run was interrupted, the next run first processes the copies left behind in ORDNA's working folder. The log now lists each of them by name and says that they do not come from the chosen folder or mailbox.
- Fixed: duplicates are parked in today's archive. If your archive had moved, a document filed before the move and imported again was parked in the Duplicates folder at the old location. It now goes to the Duplicates folder of the archive you use today.
- Fixed: licence text. The link to the licence of the language model llama3.1 led to a page that Meta has removed; it now points to the licence in Meta's GitHub repository. The licence text also still recommended Ghostscript 10.05.1; it now names 10.07.1, like the setup, and lists every use of the internet.
- Changed: wording. The Updates section no longer calls the update check the only feature that uses the internet: it is the only contact with the manufacturer, and the other uses are listed in the manual, chapter 2. The German edition says "Version" instead of "Fassung" throughout.
0.41.0 2026-10-03
- Improved: the setup tells you what happens next. The "Ready to Install" page now lists the chosen archive folder and, for each ticked additional program, what the next click will download and what you will be asked to do (Ollama: a download of more than 1 GB followed by a silent installation of several minutes; Ghostscript: an administrator prompt and a wizard to finish). While an additional program installs, the setup shows a progress page with a moving bar instead of standing still. After Ollama has been installed, the last page explains that Ollama's own "Welcome to Ollama" window may be closed. The check boxes on the Additional Programs page are no longer clipped on displays scaled above 100 %. Thanks to Holgi from the deskmodder.de forum for the detailed report.
- Improved: the first run waits for the recognition model instead of turning you away. While the model is still downloading (once, about 4.6 GB), Process folder and Fetch mailbox now start anyway: the progress bar under the button shows the download in percent, and the run continues by itself as soon as the model is there. The red message "start again afterwards" and the second click are gone.
- Improved: a folder without processable files says why. If the chosen folder contains no file of an accepted type, the result line lists what is there instead, by file extension, and points to Settings → Accepted file types.
0.40.0 2026-10-02
- New: pick a location of the original from a list. The three fields for the location of the paper original (on the Search a folder card, in the pencil window and under Set location for all) now open a list of every location already used in the archive, the same values as in the last column of the Excel workbook. Typing narrows the list, and a new location can still be entered by hand. Picking instead of typing saves work and keeps one binder from turning up under two spellings.
- Improved: the user manual is easier to start with. It now opens with a one-page quick start, a "What do you want to do?" table and a clickable table of contents, and it carries three figures of the program window. The PDF uses a larger type. The content is otherwise unchanged; the chapter numbers stay the same.
0.39.0 2026-10-01
- New: location of the paper original. For documents whose paper original you keep, ORDNA can now note where it is, for example in which binder and on which shelf. Enter it on the Search a folder card before a run; it applies to every document of that run and clears itself afterwards. With Include subfolders, ORDNA can append the subfolder names, so one run over several binders keeps them apart. You can add or change the location of any document with the pencil, and set it for a whole result list at once with Set location for all. Searching for the location finds every document filed there, the result list shows it, and the Excel workbook lists it in a new last column on every sheet. The location is never part of the file name, and the document itself stays unchanged.
- Fixed: a date at the start of a file name no longer makes an invoice. Scan programs such as Epson Document Capture Pro name each scan with the date first. ORDNA took such a file for an invoice that was already named and filed it in Invoices unchecked: a delivery note could end up there, and a real invoice kept the scan name instead of its invoice date and issuer. ORDNA now reads every file. If a file arrives with a name that already starts with a date and its issuer appears in the document, ORDNA keeps that name.
- Fixed: German Fraktur can be added as an OCR language again. The entry was labelled frk and could no longer be downloaded, because the model has been renamed. It is now listed as Fraktur (German) and downloads the current model.
0.38.1 2026-09-30
- Fixed: a silent installation no longer waits for a click. When Ghostscript or Ollama was missing, the setup showed a note at the end even when it was started silently, for example during an automatic update. The note now appears only in a normal installation; a silent one writes it to the setup log. The same applies to the error messages shown while downloading or installing these two programs.
0.38.0 2026-09-27
- New: several mailboxes. Under Settings → Mailboxes (IMAP) you can now set up more than one mailbox. A list shows each of them with a tick for "include when fetching" and a dot for the result of its last fetch — green, red or grey; clicking a row opens that mailbox for editing, and + Add mailbox and Remove mailbox do what they say. Fetch mailbox and the daily automatic fetch collect all ticked mailboxes one after the other. A mailbox that cannot be reached no longer holds up the others; the result line names it in red. The period and the trash option apply to all mailboxes. Your existing mailbox becomes the first entry — there is nothing to enter again.
- New: STRATO in the provider list. Choosing it fills in the server and port; the hint below the password field says which password to use.
- Changed: Test sign-in checks what is typed in the fields. It used to test the saved settings, so a new address had to be saved before it could be tested.
- Fixed: error results looked like successes. When a mail or folder run ended with an unexpected error, its card showed the message in the same grey as a normal result. It is now red.
- Fixed: handled mails are remembered in order. Once the list of handled mails reached its limit of 5,000 entries, ORDNA dropped random entries instead of the oldest ones, and a recent mail could be processed a second time.
0.37.2 2026-09-27
- Fixed: the reference guide on document recognition. It said that an invoice whose date or issuer is uncertain gets the ending
_Xin its file name. ORDNA has not done that for a long time; the file name stays clean. The guide now describes what actually happens: if the date or the issuer is missing, ORDNA puts0000-00-00orUnbekanntin its place. All four languages. - Fixed: password-protected PDFs in the reference guide and in the user manual. Both still said that a keyword in the file name of a locked PDF could send it to a specialised folder. ORDNA has not done that since 0.33.0: every password-protected PDF goes to Encrypted, whatever it is called, and keeps its name. The reference guide already said so in its chapter on Encrypted, but contradicted itself in three other places (all four languages). In the user manual, the troubleshooting chapter of the English, French and Spanish editions still advised giving such a file a telling name; it now has its own row for Encrypted, as the German edition already had.
0.37.1 2026-09-25
- Changed: the setup page "Important information". It used to explain, on 84 lines, how to download and install Ghostscript and Ollama by hand — while the setup offers to do exactly that itself a few windows later, with the boxes already ticked. The page now says so up front and points to chapter 4.2 of the user manual for the manual route; the download addresses still appear at the end of the installation whenever something is actually missing.
- Changed: the user manual. Chapter 4.2 now gives the command
ollama pull llama3.1for fetching the recognition model in advance. It used to appear only in that setup window.
0.37.0 2026-09-09
- New: choose how invoices are named. Under Settings → Naming of invoices you pick what follows the invoice date: the issuer only (as before), issuer and invoice number, or invoice number and issuer. The number is the one from the analysis, shown in the Excel overview and correctable in the ✏️ window; where ORDNA does not know it, the name carries the placeholder XXX. Slashes and underscores in a number become hyphens. Switching renames all existing invoices in one go — after a preview that tells you how many files are affected and shows examples, and only after you confirm. Choosing the previous form renames them back. Other document classes are not affected.
- Changed: the ✏️ window. When the chosen naming carries the invoice number, correcting the number also renames the file; the preview shows the new name as you type.
0.36.0 2026-08-20
- New: report a problem straight from the help menu. The menu has a new entry, "Report a problem". It opens a window where you describe what happened, shows you the complete text that would go to the manufacturer — version, system, language, model and document counts, and nothing else — and then opens your usual e-mail program with that text addressed to report@zuentec.de. ORDNA never sends anything itself: the message leaves your computer only when you click Send in your own e-mail program. If no e-mail program is set up, "Copy text" puts the report on the clipboard.
- New: file names are now checked for length and for what Windows allows. Very long entries are shortened automatically when ORDNA files a document, so that neither the file name nor the full path exceeds what Windows and the text recognition tools can handle. Only the last entry of a name is shortened — the date and the issuer always stay complete.
- Changed: correcting details by hand no longer accepts a name that cannot be created. If your entries would produce a name that is too long or otherwise not allowed, the preview shows it in red, explains underneath exactly what is wrong and by how much, and "Apply" stays greyed out until you shorten the entry. Nothing is shortened behind your back here — you decide what to cut.
- Changed: "Check naming" also reports names that are too long or not allowed. Existing documents are only listed, never renamed automatically; you correct them with ✏️ as before.
0.35.1 2026-08-19
- Fixed: with several Ghostscript versions installed, ORDNA could pick the oldest one. Versions were compared as text, where "9" counts as higher than "1" — so a leftover Ghostscript 9 beat a current 10. Versions are now compared as numbers. The same correction applies to the setup program, which showed that same version on its "Additional programs" page, and to its warning about the harmful 10.6 versions.
- Fixed: a leftover folder from an incomplete Ghostscript removal is no longer mistaken for a working installation. ORDNA now looks for the program itself, not just for the folder it used to live in.
- Fixed: a missing target folder no longer reappears silently. When ORDNA set a duplicate aside, it created the
Duplikatefolder together with everything above it — so a target folder you had renamed, or one on a drive that was temporarily disconnected, came back as an empty second archive. ORDNA now stops and asks, which is what it already did everywhere else.
0.35.0 2026-08-19
- Fixed: the list of findings survives your corrections — for good. Applying a correction started a refresh of the overview, and that refresh in turn re-ran the search, which cleared the list you were working through. Apply now keeps the ✏️ window open and moves on to the next document of the list; only after the last entry does the window close and the overview refresh. Take out of archive behaves the same way.
- Fixed: Check archive now watches documents you have corrected by hand. Renaming a document, moving it to another class or taking it out of the archive left no fingerprint in ORDNA's records, so those files were counted as unknown and a later change to them would never have been noticed. They are covered again: on this archive the share under watch rose from 85.0 % to 99.3 %, and files listed as having no record fell from 303 to 15.
0.34.0 2026-08-19
- Added: the class Encrypted can now be chosen when correcting a document. Password-protected files that were filed before that class existed still sit in subject folders; the ✏️ window now lets you move them across. The file keeps its original name — no details are needed — and ORDNA checks that it really does ask for a password, so that folder stays the complete list of what needs unlocking.
- Fixed: the findings of Check naming stay on screen while you work through them. Correcting one of the listed documents used to clear the whole list, and you had to start the check again after every single fix. Now only the line you have dealt with disappears.
- Changed: Check archive shows what it found. Instead of a bare count and a pointer to a log file you could not open, the findings now appear in the ordinary hit list — with the reason under each line and the ✏️ button to act on them, which is also the way to take an empty or unreadable file out of the archive. A new button opens the full log in a spreadsheet program.
0.33.0 2026-08-18
- Added: a folder of its own for password-protected documents. ORDNA cannot read what is locked, so such a file used to be filed by what its name happened to say. It now goes to the new Encrypted folder instead — always, whatever it is called —, unchanged and under its original name. The folder sits between Other documents and Not archived in the overview. Documents already filed stay where they are.
- Added: ORDNA now asks about each missing filing folder separately. When a folder cannot be found — after a fresh installation, or because you moved or renamed one — you are offered the choice folder by folder: look for it, create it, or skip it. It used to ask about all of them at once and could only look for the first. If nothing has ever been filed, creating them is offered right away; if the drive is merely disconnected, creating stays blocked, as before.
- Added: Settings → Maintenance can now check the naming. The new button lists every document whose file name does not follow the pattern of its type — a missing date, a date in the wrong form, a missing entry. The findings appear in the ordinary hit list, so the ✏️ button is right there to put them straight. Encrypted and Not archived are left out: there the original name is intended.
- Added: Settings → Maintenance can now clear the duplicate folders. The new button moves everything in the Duplikate subfolders to the Windows recycle bin, after telling you how many files and how much space that is. They remain restorable from the recycle bin until you empty it. ORDNA checks beforehand that the drive really recycles rather than deletes; if it does not, nothing is touched.
0.32.0 2026-08-17
- Added: the Excel overview now names the currency of each invoice. Where an invoice states its own currency —
CHFon a Swiss bill, for example — ORDNA reads it off the document and labels that amount accordingly, instead of using the symbol you chose. The new Currency column names it; an empty cell means the currency could not be read off the document, and the amount then carries your symbol as before. ORDNA reads the currency from the document itself and never guesses it, so nothing changes for the invoices where it is absent. Still no conversion between currencies, and recorded amounts are untouched.
0.31.0 2026-08-16
- Added: you can now choose the currency symbol. Amounts used to be labelled with a euro sign wherever they appeared — in the search fields, in the ✏️ window and in the Excel overview — which was simply wrong outside the euro area. Under Currency in the settings you can now enter any symbol, up to five characters; ORDNA starts out with the symbol of your Windows region setting. It is a label and nothing more: the amounts ORDNA has recorded do not change, documents already filed are untouched, and no currency is converted.
- Changed: the revision history of the user manual now sits in a document of its own. It is linked from the last chapter of the manual and, unlike the manual, is kept in English only.
0.30.0 2026-08-16
- Added: ORDNA can now tell you when a new version is out. In the settings under Updates it will check zuentec.de — and, if you say so, download and install the new version by itself, confirming afterwards that it worked. This check is switched off by default, and it is the only feature in ORDNA that uses an internet connection at all: nothing about you or your documents is ever transmitted. Check now looks once without switching anything on. No update is carried out while a folder or mailbox run is working.
- Changed: the program icon now carries on a dark taskbar. It had been hard to make out there: the icon sat on a dark blue tile that all but merged with the background. The tile is now white, with the mark in the brand blue on it — clearly set off on a dark taskbar and unchanged in effect on a light one.
- Changed: the settings window opens as an overview. Instead of one long scroll it now shows only the headings of its seven sections; a click expands the one you need, and a second click closes it again. Several may stand open at once.
- Changed: the "new" column no longer speaks green. The column heading, the figures, the arrow beneath them and the new tick of the search row now use the brand colour and are set in bold type, which is what tells them apart from the Count column beside them. Green stays reserved for what it is meant for: confirming that something worked.
0.29.0 2026-08-16
- Changed: the main window now follows the path of your documents. The two source cards sit on the left, one above the other — Search a folder on top, the word or below it, then Search the mailbox — and an arrow leads right to the Archive overview, which used to be on the left. The word and the arrow say what is expected where: a choice between the two sources, a direction from source to archive. The Search & edit documents card is unchanged.
- Added: the archive overview now leads into the search from both of its number columns. Clicking a figure under Count shows all documents of that class; the green figure under new shows just the new arrivals, as before. Two arrows below the table say where the numbers lead, and each click clears the other search filters first so that nothing quietly narrows the result.
- Added: a link to the Excel workbook. Open Excel workbook at the top right of the Archive card opens
Document Overview.xlsxin whichever program is set for Excel files on your computer. - Changed: the only new tick is now called new, sits in green at the right-hand end of the search row and lines up with the green arrow and the new column it belongs to.
- Corrected note (nothing in the program changed): the ↺ button in the new column heading is permanently visible. The 0.24.0 note above says it appears only once there is something to count — that was never the case, and the manual said the same. Both are now put right.
- Changed: the program icon is easier to see. It used to be the dark blue mark on a transparent background, which all but disappeared in the dark Windows taskbar. It is now the white mark on a filled brand-blue tile — legible on the taskbar, in the title bar, in Explorer and on the setup file alike.
0.28.0 2026-08-15
- Fixed: the daily automatic mailbox fetch now takes effect when you set it. Its tick sits on the mail card, but only the OK button of the ⚙️ window ever wrote it — so setting the tick and closing the program left you without any automation, and nothing said so. Tick and time are now saved immediately, confirmed by a green ✓, and the manual no longer asks you to click a button that did not exist.
- Added: Undo in the ✏️ editing window. The new button resets every field to the state the document was opened with — including a document class picked by mistake — while the window stays open. It is greyed out as long as nothing has changed.
- Changed: the button in the window header is now called Close and asks before discarding unapplied changes (apply and close, discard and close, stay here). Until now the window carried two identically labelled buttons, and closing dropped pending edits without a word.
0.27.0 2026-08-15
- Added: the ✏️ editing window can now step through the search results. Two arrow buttons in its header jump to the previous or next document of the result list, with the position shown between them ("3 / 17"); at the start and end of the list the respective arrow is greyed out. If you changed details without applying them, a prompt asks first whether to apply and continue, discard and continue, or stay here — nothing is lost by accident.
- Changed: an open search now follows the archive. Whenever the archive overview changes — new documents arrive, a run files something — the result list refreshes by itself; previously it kept the old state until a filter was touched.
- Changed: resetting the new counter (↺) now also clears the search, exactly like its ✕ button. Before, a search for "only new" documents kept showing results for a count that no longer existed.
0.26.1 2026-08-14
- Fixed: old dot-matrix payslips are recognised far more reliably. Payslips from the nineties and early two-thousands carry different headings — "Verdienstübersicht", the pre-printed note "Gilt als Verdienstbescheinigung", the column heading "Lohnart" or "Tarifgehalt". These wordings now count as payslip evidence and override a wrong "invoice" or "bank statement" guess, exactly like the known headings. Measured against the whole archive first: none of the new wordings appears on a genuine document of another category.
- Fixed: four German plain-text date forms are now read. Dates written out as "26. August 2004", dotted dates with a two-digit year ("05.11.03"), months in plain text ("DEZEMBER 2000") and the statement field "12.03/1" found on old payslip forms no longer cost a document its required detail. Measured against the recorded model answers first: seventeen dates gained, none misread; ambiguous forms (slash dates, English month names) deliberately stay out.
- Fixed: when a payslip had been mistaken for a bank statement, the correction could adopt the bank of the salary account as the employer name. The bank field is no longer used as an employer source — if no employer is evidenced, the document goes to the catch-all stage instead of receiving a wrong name.
0.26.0 2026-08-14
- Added: after a mailbox fetch or a folder run, the document overview and the search now update by themselves — exactly as if you had clicked 🔄. Newly filed documents therefore show up right away when you click a green number in the archive overview; previously the search only knew them after a manual refresh. The update starts only when the run actually filed at least one document, so an empty daily fetch leaves overview and search untouched.
- Fixed: the note in the ✏️ editing window still asked for a click on 🔄 after applying a correction — that refresh has been starting on its own since 0.22.0. The note now says so, in all four languages.
- Changed: the counter of new documents no longer restarts at midnight. The green new column counts from your last reset (↺) — across midnight as well; until you reset it for the first time, it counts from the start of the day. The card description now also mentions that clicking a green number shows exactly those documents in the search.
- Changed: the archive overview shows the filing path on its own, without a label in front of it.
0.25.0 2026-08-13
- Added: clicking a green number in the archive overview now shows exactly those documents. The search opens with the document class as filter and the new only new tick set; all other filters are cleared. The tick can also be set by hand — it restricts any search to the documents counted in the new column.
- Changed: the search card is now called Search & edit documents — the previous name "Document search" did not reveal that details can be corrected right there via the ✏️ button.
- Changed: the search runs by itself — while you type and on every filter change. The Search button is gone, and the only new tick sits at the far left of the search row.
- Changed: the filing line of the archive overview now shows the root folder of the archive (instead of the invoices folder) in the same font size as the path fields — as plain text, so it does not look editable.
- Added: every document class now carries its symbol from the reference guide — 🧮 tax, 💶 payslip, 🏦 bank, 🧾 invoice and so on. It appears in front of the class name throughout the application (archive overview, search filters, recently filed documents, editing window) and in the class table of the user manual, so a class looks the same wherever you meet it.
- Changed: the archive overview and all class lists now show the document classes in the order in which ORDNA checks them — the same order the reference guide uses to explain the categories. Previously the lists followed the historical order in which the classes were built. Nothing changes in your archive or in the recognition itself; only the display order is new.
0.24.0 2026-08-13
- Added: you can now set the counter of new documents back to zero. The archive overview counts in its right-hand column what has been added since midnight; the ↺ button in that column heading starts the count over from this moment, which is handy when you read in several folders one after another. Nothing in your archive changes — only the display starts afresh. The button appears as soon as there is something to count, and the counter returns to counting the whole day by itself at the next midnight.
- Changed: that column is now labelled new instead of "new today", because after a reset it no longer counts the whole day.
- Changed: the help symbol in the header is no longer red. Red is reserved for errors, so it now carries the same blue as the other controls.
- Added: a reference guide that explains how ORDNA sorts your documents. "How ORDNA recognises and sorts your documents" replaces the earlier appendix: one chapter per category, each built the same way — typical examples, what ORDNA looks for, which details it needs, how the file is named, what belongs there and what does not, and a borderline case. A table of more than thirty everyday cases answers the question "why is this document in that folder?". It is written for readers without technical knowledge and comes in all four interface languages. You open it from the menu, right below the user manual — previously the appendix existed in German and English only, and the French and Spanish manuals pointed at the English one.
0.23.0 2026-08-12
- Changed: the subfolder that holds parked copies — duplicates, the complete version of a document whose blank pages were removed, and anything you take out of the archive by hand — is now called
Duplikateinstead of_Duplikate. The leading underscore is gone because some cloud storage services treat such names specially and may skip them while synchronising. Existing folders are not renamed for you: ORDNA writes into the new one from now on, and ⚙️ → Maintenance → Check archive reports any leftover_Duplikatefolder so you can move its contents over yourself. Nothing is ever deleted, exactly as before.
0.22.0 2026-08-12
- Added: ORDNA now tells you when a filing folder has gone missing. If you rename, move or delete one, a window offers to find it or to create it anew. Until then nothing is archived — and above all, no folder is created behind your back: previously ORDNA silently made a new empty one and filed into it, leaving you with two archives without a word. If all folders are missing at once, the drive is usually just disconnected, so creating is blocked in that case.
- Changed: the window now opens at the size the interface needs — the four cards and room for six search results below them, centred on your screen. Before, it filled the whole screen, which left a lot of empty space on wide monitors. You can still resize it freely.
- Changed: after you correct a document, the refresh of overview and search starts on its own. You no longer have to remember to click the refresh symbol.
- Changed: when you move a document into a different class, details that exist in both classes are kept — date and issuer stay as they are instead of being emptied.
- Fixed: a document moved into a different class now appears in the "new today" column of its new folder. It was missing there before.
- Added: you can now correct details that were recognised incorrectly. Find the document with the search, click ✏️ on the hit and put the details right — a misread issuer, a swapped date, a wrong amount. Fields that form part of the file name rename the file for you; you see the new name before you apply. Everything else stays as it is: the document class, the folder and the document itself.
- Added: corrections you make by hand are kept. ORDNA does not overwrite them again with a result of the automatic recognition, not even on a later refresh.
- Added: you can move a document into a different class. If an invoice ended up under "Other documents", choose the right class in the correction window: ORDNA moves the file into that folder, asks for the details it needs there and renames it accordingly. This is where most mistakes sit — "Other documents" and "Not archived".
- Added: the document is shown next to the fields while you correct them, so you no longer work blind. A button opens it in your usual PDF program as well.
- Added: category, receipt kind and document class are now chosen from a list instead of typed.
- Added: you can take a document out of the archive. It is not deleted: it moves into the
_Duplikatesubfolder, from where you can bring it back at any time. - Fixed: in a German, French or Spanish installation the three drop-down lists of the document search stayed English until you switched the language by hand.
0.21.0 2026-08-11
- Removed: continuous folder watching is gone. ORDNA no longer watches a folder on its own; you read a folder in when you click Process folder. The feature did not work reliably enough to be useful. Everything else about reading in folders is unchanged — you can keep using the same folder as a permanent drop-off point for your scanner or your phone uploads and read it in whenever it suits you. Documents already archived are recognised as duplicates and are not filed twice.
- Removed: ORDNA no longer starts when you sign in to Windows, and it no longer places an icon next to the clock. Both existed only to keep the folder watch running. Start ORDNA from its desktop shortcut when you need it. An autostart entry from an earlier version is removed when you update.
- Changed: closing the window now always quits ORDNA — there is no longer a case in which it keeps running in the background.
0.20.2 2026-08-11
- Changed: ORDNA now opens across the whole working area of your screen — the space next to the taskbar. On a wide screen the three columns then stand side by side and the document search is in view without scrolling; previously the window kept a fixed size and showed only about half the page. You can still resize it however you like.
0.20.1 2026-08-10
- Fixed: the window now fits the screen it opens on. On smaller screens it no longer reaches under the taskbar, and it can be resized down as far as the screen allows. On every screen it opens in the middle instead of off to one side.
0.20.0 2026-08-10
- Changed: ORDNA now runs in a window of its own instead of in your web browser. Nothing about the way you work with it changes — the same page, the same buttons — but it no longer looks like a website, and it no longer appears among your browser tabs. Your documents were always processed on your own computer only; now that is what you see.
- Changed: closing the window quits ORDNA. If a folder watch is running, ORDNA instead keeps working as an icon next to the clock, so that the folder you asked it to watch stays watched. Right-click that icon to reopen the window or to quit for good.
- Changed: starting ORDNA a second time no longer opens a second window; it brings the existing one to the front.
- Changed: the desktop shortcut now starts the program directly. The intermediate launch page is gone, and with it the need to reinstall after moving the program folder.
- Changed: the user manual and the change log under , search hits and the "recently filed" list now open in whichever program you have set for that file type — your usual PDF reader, not a substitute view.
- Note: your colour scheme, language choice and dismissed hints start out empty once after this update, because the window keeps its own settings. The language follows the one chosen during installation.
0.19.0 2026-08-07
- Added: one sender, one spelling. If the target folder already holds documents from the same sender, ORDNA now adopts their spelling, provided the new one differs only in punctuation, capitalisation or spaces — so
Prof. Dr. med. Fabian M, Stubywith a misread comma joins the documents already filed under… M. Stuby, andCALO.SOL GMBHbecomesCALO.SOL GmbHagain. The German transliterations count as the same name as well (Zuendorf=Zündorf). Names that differ by a whole word stay apart: two authorities, suppliers or physicians whose names merely resemble each other are never merged. - Fixed: day and month swapped. Where the date printed in the document was carried over the wrong way round, ORDNA now puts it right — but only when exactly one of the two readings appears in the document. If neither can be verified, the value found is kept: nothing is guessed here either.
0.18.0 2026-08-07
- Added: health records are now filed even when the recognition cannot name the kind of document. If the text carries a customary designation — physician's letter, discharge summary, laboratory report, surgical report, histology report, medication plan, medical certificate and the like — ORDNA adopts it as the description; the classic opening of a physician's letter ("we report on …") counts as evidence too. Until now such a document left its class for lack of that one detail and ended up under "Other documents". Nothing is invented: the designation has to appear in the document, and a description already recognised is never replaced.
- Fixed: a letterhead code is no longer taken for the issuer. Hospital forms carry a ward or case code next to the logo — the recognition read it as the sender, and the file name kept it. Such a code now gives way to the company name found in the document text. Where the text offers no company name, the previous value is kept: an empty issuer would leave the document without a name at all.
0.17.1 2026-08-07
- Fixed (internal): the installation package now carries the time-zone database that one of ORDNA's components may ask for. Nothing changes in daily use — no tested run ever asked for it. It is included so that the package stays complete on a computer with nothing else installed, which is what ORDNA promises.
0.17.0 2026-08-07
- Added: a payslip is now filed even when its employer cannot be read from the document. ORDNA then falls back to an employer name that is already established in your payslip folder and also appears in the document — two independent pieces of evidence. If two different established names would fit, nothing is assumed and the document is left in "Not archived", exactly as before. This closes a gap where the same employer was read from one payslip but not from the next, even though both used the identical form.
- Changed: when blank pages are removed while filing, nothing is lost any more. The complete version with all its pages goes into the
_Duplikate(duplicates) subfolder of the target folder, under the same name prefixed withOrg_. - Fixed: the employer on a payslip is recognised more reliably. Wording taken from the payroll lines themselves (for example "AG-Zuschuss", where "AG" is short for the German word for employer) and the franking imprint on the envelope were occasionally mistaken for the company name; both are now ruled out. Where several spellings of the same company appear, the properly capitalised one wins.
0.16.0 2026-08-06
- Added: ORDNA can now remove blank pages while filing. If your scanner scans both sides, every sheet printed on one side only produces a blank back page — the new tick box Remove blank pages when filing (⚙️ settings) leaves them out. It is switched off by default. A page is only removed when two independent checks agree: it carries no readable text and no visible content. A border strip of 16 millimetres stays out of the check, so that filing holes and folded corners are never mistaken for content. At least one page always remains, single-page documents are never touched, and the original in the "Not archived" folder always keeps all its pages.
0.15.0 2026-08-06
- Changed: for payslips, every version is now kept. A single month often comes with two documents — the payslip and a later correction, which frequently shows only a partial amount. Until now the correction displaced the original one into the
_Duplikatefolder; now both sit side by side, told apart by an appended number (2021-01_Employer.pdfand2021-01_Employer_2.pdf). Genuine duplicates are still recognised as before: the very same file, or the same document saved anew with identical content. - Added: a password-protected PDF is now filed by its file name when that name says clearly what the document is — for example "Patientenkopie", "Hausrat", "Onlinebanking", "Kaufbelege", "Darlehen" or "Baufinanzierung". ORDNA cannot read such a file, so previously every one of them ended up in "Not archived". The file keeps its existing name, because without readable content there is neither a date nor an issuer for the usual scheme.
- Added: payslips are now also recognised by the words "Lohnabrechnung", "Verdienstabrechnung" and "Entgeltabrechnung" in the text. Swiss payslips in particular lack the German heading ORDNA relied on and used to be filed as invoices.
- Fixed: the recognition model sometimes reports a month as
202102or09-2020instead of2021-02. ORDNA now reads those spellings too. Previously the document lost a mandatory entry and dropped into "Not archived" — in one scan run this affected seven of thirteen payslips. - Fixed: the employer on a payslip was read from the wrong line. On a payslip "AG" is short for Arbeitgeber (employer), not for a company form, so lines such as "Zusatzbeitrag AG" were mistaken for the company name. A heading of the form itself ("Steuerdaten") is no longer appended to the employer either — the same series of payslips used to be named inconsistently.
- Fixed: documents whose name does not follow ORDNA's naming scheme no longer produce nonsense columns in the Excel overview. A file called
Patientenkopie_6.pdfappeared there with the date "Patientenkopie" and the issuer "6"; those fields now stay empty. - Fixed: the Stop button of the folder watch had no effect while the watch was running — it could only be switched off through the tick box.
0.14.1 2026-08-05
- Fixed: if you install ORDNA again into an archive folder that already holds an archive, ORDNA now keeps the existing folder names — even when you pick a different setup language than the first time. Without this, a second set of folders would have appeared beside the first, and the documents already archived would have gone missing from the overview.
0.14.0 2026-08-05
- Added: a fresh installation now names your archive folders in the language you chose during setup —
Invoicesinstead ofRechnungenfor an English installation, and likewise in French and Spanish. The names stay as they are afterwards: switching the interface language later never renames a filled archive. Existing installations keep their folder names unchanged. - Added: uninstalling now offers to remove your settings and data as well (mailbox access, search index, analysis data). The default is to keep them, and your document archive is never touched either way.
- Changed: ORDNA no longer creates an inbox folder in your Documents. It only ever held the working copies ORDNA makes while converting and recognising a document; these now live beside the settings and are cleaned up automatically. Anything still lying in the old folder is picked up on the next run.
- Fixed: if the recognition model fails while a mail attachment is being processed, the mail is no longer marked as done — the next run fetches it again. Previously such an attachment could be missed.
- Fixed: in the English, French and Spanish manuals the signal word of a notice box (NOTICE / AVIS / AVISO) ran on as the first words of the text instead of standing on its own line as in the German manual.
0.13.2 2026-08-05
- Changed: ORDNA now asks for Ghostscript 10.07.1 instead of 10.05.1. Please install the new version — the setup offers to fetch it for you on the Additional Programs page. The versions 10.6.x remain unsuitable and are the only ones ORDNA warns about; an existing 10.05.1 keeps working, but 10.07.1 closes four security holes in the helper program.
- Fixed: scanned pages keep their original image quality again when a current Ghostscript is installed. Until now the text recognition re-compressed the embedded images as a precaution whenever it saw a Ghostscript newer than 10.5 — measured on ten scanned invoices, four of them visibly lost quality. With the new combination the image data passes through untouched.
0.13.1 2026-08-04
- Changed: the two source cards now describe what ORDNA actually does with your files. Search a folder and Search the mailbox still promised that source files are deleted and mails moved to the trash — but since 0.13.0 neither happens unless you tick the box yourself. The cards now name the real steps: analyse, index, rename and file.
- Fixed: a document recognised as an invoice is no longer filed when neither its date nor its sender can be read. Such a file used to end up in the invoice folder as
0000-00-00_Unknown.pdf— a name that says nothing at all. It now stays where it is, so you can see it. If only one of the two is missing, filing continues as before. - Fixed: switching the Invoice / receipt category off no longer swallows tax documents. If the recognition took a document for an invoice while that folder was disabled, it fell through to the catch-all — even a clear income tax assessment.
0.13.0 2026-08-04
- Changed: ORDNA no longer deletes anything unless you ask it to. The ticks Delete source file after archiving and Move mail to trash once its attachments are fully archived are now switched off by default — your source folders and your mailbox stay untouched until you set them yourself. The first tick now also applies to continuous folder monitoring, which used to delete regardless.
- Added: the Process folder and Fetch mailbox buttons now react to the click straight away. The progress bar appears at once and the button is locked until the run has started, so a slow start no longer looks like nothing is happening.
- Fixed: credit card statements from Deutsche Bank are now filed under Bank documents instead of Invoices. They call themselves "Umsatzabrechnung", a term the recognition did not know, and carry the service provider rather than the bank in the institution field.
0.12.1 2026-08-04
- Fixed: the automatic Ghostscript installation on the Additional Programs page could stall forever: the Ghostscript installer ignores its silent switch and was waiting for clicks in a hidden window. Both helper installers now run visibly — Ghostscript's own installation window opens and is clicked through to "Finish" — and any message from the setup is brought to the front instead of opening invisibly behind the wizard. Every step is also written to the setup log.
0.12.0 2026-08-04
- Added: the setup now has an Additional Programs page. It checks whether Ghostscript and Ollama are already installed, shows the result — including a warning if an unsuitable Ghostscript version is found — and offers to download and install what is missing. Ghostscript needs administrator rights for this; Windows asks for confirmation at that moment. If the download is skipped or fails, the installation simply continues as before.
- Added: if you pick an installation folder that requires administrator rights (such as
C:\Program Files), the setup now offers to continue with those rights instead of failing. Declining lets you choose a different folder. - Changed: the document classes are presented to the recognition model in the order of the built-in priority chain. Measured on the full archive this yields more complete details for bank documents and slightly better filing overall.
- Fixed: two kinds of bank documents no longer lose their filing: statements whose bank name comes back without spaces (
DeutscheBankAG), and statements where the recognition returned a full date instead of the statement month.
0.11.1 2026-08-03
- Fixed: dates written in the German style (
03.04.2018) are now read instead of discarded. Where the recognition returned such a date, the document previously counted as missing a mandatory detail and ended up in a catch-all folder. Where the order of day and month could be read either way, the date now has to appear in the document itself — otherwise it is left out rather than guessed. - Fixed: bank names written as one word — Kreditbank, Raiffeisenbank, Commerzbank, Postbank and the like — were not recognised as banks at all. A credit card statement could therefore end up among the invoices even though the document names itself clearly.
- Fixed: a till receipt no longer disappears into a catch-all folder when the recognition names a class but cannot fill in a single detail for it. Such an answer no longer blocks the invoice route.
0.11.0 2026-08-03
- Added: four new document classes, each with its own folder — Health, Education & Qualification, Certificate & Record and Correspondence. They are checked after the established classes, so an invoice from a doctor stays an invoice and a care level decision stays an official notice. Existing installations receive the four folders automatically; nothing already archived is renamed.
- Changed: the class formerly shown as Bank statement is now called Bank documents. The old name described only one of its kinds and led to wrong expectations — account openings, card applications and letters from your bank have always belonged here. Only the label changed; your folder keeps its current name until you change it in the settings.
- Fixed: a document is no longer treated as an invoice merely because its file name looks like one. Re-importing documents that ORDNA had already archived previously filled the invoice folder with documents that were not invoices — the content now decides.
- Fixed: where the issuer of a health, education, certificate or correspondence document was reported under a different heading — a registry office is, after all, an authority — it is now used as the issuer instead of leaving the document unnamed.
- Changed: sharper boundaries between classes. A tax document must have an actual tax relevance, not merely a tax context; an official notice has to decide something, so a plain confirmation is a certificate and your own application is correspondence; letters about a contract or an insurance policy are correspondence, not the contract or the policy itself.
0.10.0 2026-08-02
- Changed: document types are now decided in a fixed order of precedence — tax before payroll, bank statement before invoice, invoice before official notice, policy before contract. Borderline documents therefore land in the same folder every time: an insurance premium invoice is filed as an invoice, an annual tax statement from your employer as a tax document, a credit card statement under bank statements.
- Changed: each override now requires positive evidence in the document itself. Wording that merely looks related — such as the account holder line of a direct debit block on an invoice — no longer moves a document to another folder.
0.9.1 2026-08-02
- Changed: refreshed brand colours throughout — dark blue
#245273and light blue#A3C1D5replace the previous palette in the interface, the logo, the application icon and the manual PDFs. - Fixed: checkboxes and the refresh buttons now follow the brand colours in both light and dark mode instead of the browser's default blue.
0.9.0 2026-08-01
- New: the user manual now opens as a print-quality PDF in its own window sized to the document, in all four languages. The manual's title page carries the ORDNA logo.
0.8.1 2026-08-01
- Fixed: clicking a document class in the archive overview now brings the Explorer window to the front. Previously it could open behind the browser.
0.8.0 2026-08-01
- New: the user manual has been rebuilt from scratch and is now included in all four interface languages (English, German, French, Spanish). The menu opens the manual in the language of the interface. The new manual is task-oriented, starts with what ORDNA does for you, and avoids technical jargon.
- New: the German edition of the appendix "How ORDNA recognises and sorts documents" is now shipped alongside the English one.
- Changed: documentation pages now show the document title in the browser tab instead of the file name.
0.7.0 2026-08-01
- New: the installer now asks for the root folder of your document archive on its own wizard page (default
Documents\ORDNA\Archiv). On first start ORDNA creates one subfolder per document type below it; the folders can still be changed at any time in the settings. When installing over an existing installation the page is skipped — your existing settings remain in charge. Silent installations can pass/ARCHIV="D:\path". - New: the installation folder can now also be chosen when installing over an existing installation; previously the page only appeared on a fresh installation.
0.6.2 2026-08-01
- Changed: the former working title "Archivar" has been removed from all documentation and document metadata; the product is consistently called ORDNA. No change in behaviour.
0.6.1 2026-08-01
- Changed: internal cleanup after the 0.6.0 work; no change in behaviour.
0.6.0 2026-08-01
- New: help menu in the header with direct access to the user manual and this change log — rendered locally, no internet required.
- Fixed: the invoice analysis cache could be wiped when the invoice folder appeared empty or unreachable (fresh installation, disconnected drive, test environment). The cleanup of orphaned entries now only runs when the folder scan actually finds files, and the cache honours the data-directory isolation used by test environments.
0.5.0 2026-08-01
- Changed: there is now a single Excel workbook,
Document Overview.xlsx, in the root folder of the archive — one sheet per document class. The separateRechnungsliste.xlsxin the invoice folder is gone; its twelve columns live on unchanged as the invoices sheet of the workbook. Leftover files from earlier versions are removed automatically after the new workbook has been written.
0.4.3 2026-08-01
- Fixed: before every run ORDNA now asks the recognition model a short test question. A model that is listed in Ollama but cannot load (typically an outdated graphics driver) stops the run with a clear message instead of letting documents fall silently into "Not archived".
- Fixed: if the language model fails in the middle of a run (connection or server error), unrecognized documents are no longer moved to the catch-all folders — they stay at their source and run through again next time.
0.4.2 2026-07-31
- Fixed: the installer now closes the running background server itself before updating or uninstalling. Previously an update could fail with "file in use" and the server had to be ended by hand.
0.4.1 2026-07-31
- Fixed: an Ollama server started by ORDNA no longer inherits the program folder in its DLL search path — this could lock the folder and make the next update fail.
0.4.0 2026-07-31
- New: prerequisite check before every run (text recognition, Ghostscript, Ollama, recognition model). ORDNA starts Ollama by itself and downloads a missing model in the background; what it cannot fix blocks the run with a message naming the missing piece.
0.3.0 2026-07-31
- Changed: internal cleanup; the legacy
/rechnung/…links were removed in favour of the universal document links.
0.2.1 2026-07-31
- Changed: the distribution package contains only the setup file — licences, prerequisites and manuals are shown and installed by the setup itself.
0.2.0 2026-07-31
- New: the interface is available in four languages — English (default), German, Spanish and French — and the language chosen during installation becomes the initial interface language.
- New: the version number is shown in the header.
- Changed: the end-user documentation shipped with the setup is English.
Before 0.2.0 2026-07-19 to 2026-07-30
Initial development: two sources (mail account, watched folder), one shared pipeline (conversion, OCR + PDF/A-2b, local recognition via Ollama, renaming, duplicate handling, filing), nine document classes with their own target folders, full-text search across all classes, Excel overviews, provenance journal, and the Windows installer.
Don't have ORDNA yet? Download it here, free of charge.