` nonfiling marks, place / publisher / year split from the imprint, CDD and Cutter, subjects from `` or `;` lists) with one item in 952 per tombo of the `registro` field: copies split on hyphens, commas and ISIS carets (`^f`), volumes (`v1`, `I`, `1ª série`) to $h, ranges (`669 a 672`) expanded, stray letters, `$` and call numbers typed in the field left out, `(Baixa)` as withdrawn, `ISIS` when there is no tombo and `ISIS-` when it repeats, the field as typed kept in the internal note; the occurrences table adds notes and marks lost copies. Text cleaned of `Informação não encontrada` fillers, control characters and `Ẃ` for `ª`. The export has no patrons or loans | Tables, columns and where they go, every cleaning count and the MFNs to check, a sample of the records; then Koha's staging report | `PRE-IMPORT` | - The Biblivre, SophiA and Pergamum item fields follow their usual export layout; the preview shows the result before anything is written. - The column scoring was built from the usual column names of Brazilian library programs and spreadsheets, and tested with synthetic files only. The preview lists every column, where it goes and the ones left out, so check it before confirming. - The Biblioteca Fácil database reader was built from a backup of the program's empty template database. The loan, hold and patron tables were checked with a synthetic backup only, so compare the preview's counts with the old program before confirming. - The ISIS reader was built from one library's export. Titles with a caret inside may keep an ISIS subfield letter glued to the next word, so the preview lists their MFNs for a check in Koha. ### Brazil: localization **Library tools > Brazil: localization**. Nothing here is applied during the installation or by a schedule: each option acts only when you choose it, follows the same steps as the other tools (lock, preview, confirmation, verified `PRE-*` backup) and can be undone. The migrations from other systems are in the [Magic Import Tool](#-magic-import-tool). | Tool | How it works | Preview | Backup | |------|--------------|---------|--------| | Check patron CPFs (read-only) | Card number, username, sort1/sort2 or the `CPF` attribute: valid, invalid, shared by two patrons | — | — | | Pimaco label templates | Sheets 6180, 6181, 6287 (Letter) and A4256, A4251 (A4) plus 3 layouts (spine, barcode, title and barcode) in Koha's label creator; updates or removes only its own templates | Summary before the change | `PRE-LABELS` | | Cataloguing card (ficha catalográfica) | Stylesheets that wrap Koha's default detail view (OPAC and staff interface, one per installed language) and add the card below the record, with a print button; set in `OPACXSLTDetailsDisplay` / `XSLTDetailsDisplay`. The previous values are kept and **Hide the card** puts them back; the files are rewritten each time the panel starts, so they follow Koha updates | Old and new values | `PRE-FICHA` | | Brazilian holidays in the calendar | National holidays of a year, Carnival (Monday and Tuesday), Good Friday and Corpus Christi (Easter by the Meeus/Jones/Butcher algorithm) and an optional municipal holiday, for one library or all; days already in the calendar are skipped and **Remove** takes out only the days added by the panel | Dates with the weekday | `PRE-CALENDAR` | | Government census reports (MEC/INEP/IBGE) | 10 read-only SQL reports in Koha's saved reports with the figures asked by the INEP School and Higher Education censuses, the IBGE surveys, the SNBP registry of public libraries and the MEC course evaluations: summary of the year, active collection by item type and library and by CDD class, volumes added, circulation by month, loans by item type and patron category, readers by category, sex and age group, losses and withdrawals (with the replacement value), age of the collection by CDD class and weeding candidates. Koha asks for the year when a report runs. Installed and removed separately from the essential pack (tag `[koha-easy-installer-censo:...]`) | The reports, new or updated; a report the Koha schema cannot run is skipped | `PRE-CENSO` | | ABNT references and bibliographies (read-only) | References by ABNT NBR 6023:2018 built from the MARC 21 records (books, chapters and articles through 773, theses through 502, online documents through 856; SURNAME, Forenames; up to three authors, then *et al.*; organizers; entry by title with the first word in capitals; `[S. l.]`, `[s. n.]`; title of the book or journal in bold), in alphabetical order, from a Koha list, the items added in a period, a call number class, a library's whole collection or record numbers. Written as HTML with the NBR 14724 presentation (A4, Times 12, single spacing, a blank line between references; opens in Word or LibreOffice) and as plain text, optionally with the call numbers and copies under each reference (collection listing) | The first references | — | - The label sizes are the Avery equivalents of each Pimaco sheet: do a test print on plain paper first and adjust the template in Koha if your printer shifts the page. - The card starts closed on the record page, behind **Show the AACR2 catalogue card**, and opens with a click. - The card follows the AACR2 layout used by Brazilian libraries (call number column, hanging indent, numbered subjects, roman-numbered added entries). Below it, the reference of the record by ABNT NBR 6023 (essential elements) for readers to copy. - The census reports count what Koha records: in-library use needs Koha's *local use* check-ins, the sex and age groups need those patron fields, and the CDD class comes from the first three digits of the call number. Check each figure against the form of the census before sending it. - Carnival and Corpus Christi are *ponto facultativo*, not national holidays: remove them in Tools > Calendar if the library opens. Consciência Negra (20/11) is added from 2024 on. ### Messaging: WhatsApp and Telegram **Library tools > Messaging: WhatsApp and Telegram**. Koha already writes the notices (checkout, check-in, overdue, due, hold) and hands those of the SMS type to the SMS::Send driver named in the `SMSSendDriver` preference. The panel installs such a driver (`SMS::Send::KohaEasy::Gateway`): each notice goes to Telegram when the patron linked the library's bot, otherwise to WhatsApp. Nothing changes in Koha until **Send Koha notices** is turned on, and turning it off puts the previous `SMSSendDriver` back (verified `PRE-MESSAGING` backup both ways). | Option | What it does | |--------|--------------| | WhatsApp gateway | A gateway of the library's own: Evolution API v2 (`POST /message/sendText/`, `apikey` header) or another one that accepts `{"number", "to", "text"}` with a Bearer token | | Telegram bot | The token from @BotFather is checked with Telegram (`getMe`). Patrons open the bot, tap Start and share their own contact; a job every two minutes links the chat to the number (`/stop` undoes it) | | Country and area code | Numbers are completed and corrected before sending: country code (Brazil by default, or any other), area code (DDD) for numbers without one, trunk and carrier prefixes removed, and the 9th digit of Brazilian mobiles added | | Send a test message | Through the same SMS::Send path Koha uses, as the instance user | | What Koha needs | Checks `SMSSendDriver`, the SMS versions of the notices, overdue rules with SMS and patrons with SMS numbers and preferences; offers to create short SMS versions of the notices that have none (`PRE-NOTICES`) | | Check the patrons' phone numbers | Read-only report with the same rules; the corrected numbers are written only on confirmation (`PRE-PHONES`) | The settings (with the tokens) are in `/etc/koha/sites/library/kei-messaging.conf`, readable by root and Koha only. When the notices are on, the SMS queue is also sent every two minutes (`/etc/cron.d/koha_messaging`). ### Cataloguing aids: PHA, Cutter-Sanborn, CDD **Library tools > Cataloguing aids**, read-only for the catalogue. - **Author notation**: the entry right before the name in the table, the initial of the surname, the number and the initial of the title (the initial article never counts, and a title starting with "l" gets a capital L); prefixes read as one word (La Fonte, O'Donnel) and M' / Mc as Mac; institutions by the first word and anonymous works by the first word of the title, as in the PHA explanation. From a catalogue record it reads 100 / 110 / 111, 245 (with its nonfiling indicator) and 082, and lists the call numbers that already use the same number in that class, with the free numbers right before and after it. - **The tables are not distributed with the panel**: the PHA table (Heloísa de Almeida Prado, T. A. Queiroz) and the Cutter-Sanborn tables are copyrighted. Each library loads its own copy as a text file (one entry and its number per line; the PHA rows can be typed as printed, entry - number - entry; the PDF of the PHA book is also read, see [Cutter Calculator](#cutter-calculator)). The file is checked before it is kept: entries per letter, numbers out of order (typing mistakes) and repeated entries. - **CDD**: the ten main classes are built in; the library can load its own schedule (number and caption per line) to search by number or by word, and every lookup also shows how the catalogue already uses the number. ### Replace a MARC record (staff interface) **Library tools > Replace a MARC record** installs `marc_replace.pl` in Koha's staff interface (`/cgi-bin/koha/tools/marc_replace.pl`). Optionally it also adds shortcuts in the staff interface: **MARC Injection & AI** next to **New record** (and in the New menu of the record pages), which opens the AI cataloguing below to draft a new record; **Replace via MARC / AI** next to **Edit** on each record page, which opens the page with that record (MARC file or text, or photos of the book), always through the preview before anything is replaced; **Replace the record (MARC file)** in the Edit menu of each record; and **Replace a MARC record** and **AI cataloguing** in the tools of the cataloguing home page (`/cgi-bin/koha/cataloguing/cataloging-home.pl`). It replaces a record, found by its biblionumber, with an `.mrc` or `.xml` file or pasted text (Biblioteca Nacional `245 10 |a`, MarcEdit or yaz-marcdump lines): - login with the `edit_catalogue` permission and a CSRF token (Koha 24.05+ `cud-` operations); - a preview first; then, in one transaction, the current record is locked and compared with the preview (nothing is replaced if it changed), saved as MARCXML in `/var/lib/koha/library/kei-marc-replace/` (downloadable from the page, and accepted back to undo) and replaced with Koha's `ModBiblio`; - item fields in the file are always left out: the items are never touched; - MARC-8 / Latin-1 files are converted by Koha's own `MarcToUTF8Record`. The page is checked with Koha's Perl modules before it is installed; removing it keeps the saved versions. #### AI cataloguing (photos of the book) Two more tabs of the same page catalogue a book from photos of its cover, title page and verso of the title page (with the CIP block): - **AI settings** (staff allowed to change the system preferences only): the vision model, either a cloud API with a token or key (OpenAI, Anthropic, Google Gemini with a key from Google AI Studio, or any OpenAI-compatible service) or a model on the library's own network (Ollama `http://localhost:11434` or LM Studio `http://localhost:1234/v1` with a vision model such as `qwen2.5vl`, `llama3.2-vision` or `gemma3`). The settings are kept in `kei-marc-replace/vision.conf`, readable by the Koha instance user only; the token is never shown again, and a token is refused over plain `http` to a server outside the local network. **Test the connection** lists the models of the server. No token ships with the panel. - **Cataloguing instructions for the model** (same tab): the instructions sent with every book, before the fixed answer format. The default ones ask for cataloguing by MARC 21, AACR2 and ISBD in the language of cataloguing, from the chief sources and the context of the book, with a complete example record holding the fields in common use (020, 040, 041, 043, 082, 090, 100, 240, 245, 246, 250, 260, 300, 336-338, 490, 500, 504, 505, 520, 521, 546, 650, 651, 655, 700, 830, 942): the model fills those the photos support, and the library deletes the ones it does not use instead of adding fields by hand. The library can rewrite them (`{language}` becomes the language of cataloguing); they are kept in `kei-marc-replace/vision-prompt.txt`, and **Restore the default instructions** brings the default back. - **AI cataloguing**: the photos go to the model, with an optional **instructions for this book only** field (for example, "the author is also the illustrator"), which takes precedence over the general instructions, is sent with those photos only and is never saved: the next book starts with an empty field. The model which answers with the bibliographic data as JSON (checked against a fixed schema: nothing invented, placeholders dropped). The national rules are then applied by `KohaEasy::Cataloguing::Rules`: AACR2 and ISBD punctuation (245 `:` `/`, 260 with `[S.l.]` / `[s.n.]`, 300), names inverted as Brazilian cataloguers do (`Assis, Machado de`, `Andrade Filho, José de`), 245 non-filing characters, ISBN check digits (wrong ones go to 020 `$z`), the CDD printed in the CIP block (otherwise the model's suggestion, flagged for checking) in 082 and 090, and the **author notation of the library's own PHA or Cutter-Sanborn table**, computed by a Perl port of the panel's algorithm, so the page and **Cataloguing aids** give the same notation. - The librarian corrects the draft next to the photos, and a **mandatory preview** follows. Only then is the record added with Koha's `AddBiblio`, in one transaction under a database lock, after the ISBN is checked against the catalogue again (a record with the same ISBN needs an explicit tick), with a copy of the record and of the model's answer saved in `kei-marc-replace/vision/`. A form sent twice adds nothing. With a biblionumber, the draft goes to the replacement above instead (with its lock and saved version). The two modules are installed with the page in `/usr/local/lib/site_perl/KohaEasy/Cataloguing/` and removed with it (the settings stay with the saved versions). ### CDD lookup in the cataloguing (staff interface) **Library tools > CDD lookup** puts the whole CDD (Dewey) in Koha's cataloguing, at `/cgi-bin/koha/cataloguing/cdd_lookup.pl`. - **The CDD is not distributed with the panel**: the DDC is copyrighted by OCLC. **Build the index** asks for the library's own copy (the PDF, or a `.txt` with its text; a PDF is read with `pdftotext`) and builds a SQLite index on the server (`/var/lib/koha//kei-cdd/cdd.sqlite`, readable only by root and the instance). The schedules, Tables 1 to 6 and the alphabetical list of subjects are read, with bracketed (not used) and optional numbers marked. A file with fewer than 100 schedule entries is refused and the previous index is kept. A scanned PDF has no text: use the text of an OCR. - **The page** searches by number (`869.3`, `869,3`, `T1—09`) or by words of the subject (accents and plurals ignored). Each number shows its place in the hierarchy, the notes (with the numbers in them as links), the subdivisions, the subjects of the list, and, for a number not printed in the schedules, **how it was probably built** (base number plus Table 1, Table 2 through —09, Table 3 in 810–899, Table 4 in 420–499), marked as a hint. **Build a number** adds a table notation to the number as you type. - **In your library**: next to each result, how many titles and items of the catalogue use that class (with and without its subdivisions, from the call numbers of the items), the titles themselves, and how the library classified titles with the searched words. - **Install the page** optionally adds a **CDD** button next to 082, 083, 090, 092 `$a` and the item call number (952 `$o`) in the cataloguing forms, and **CDD lookup** in the tools of the cataloguing home page (a marked block in `IntranetUserJS`, verified `PRE-CDD` backup first). The button opens the lookup with the number already in the field or with the subjects and title of the record as suggestions; **Use in the record** puts the number in the field (082 and 083 get the number and `$2` the edition when empty; 090, 092 and 952 keep the author notation after it). **Copy** copies it. - **Remove the page** removes the page, the buttons and the module `KohaEasy/CDD.pm`; the index is kept. ### Cutter Calculator **Library tools > Cutter Calculator** (Calculadora Cutter) works out the call number of a record: the class and the author number from the library's own Cutter-Sanborn or PHA table, in the panel and in Koha's cataloguing, at `/cgi-bin/koha/cataloguing/cutter_calculator.pl`. - **The tables are not distributed with the panel**: the editions of the Cutter-Sanborn table and the PHA table (Heloísa de Almeida Prado, T. A. Queiroz) are copyrighted. **Load the Cutter-Sanborn table** asks for the library's own copy as a text file, one number and its entry per line as the three-figure table prints them (`848 Assis`, ` 127 Abbot, J.`; the leading spaces may vary). **Load the PHA table** takes a text file or the PDF of the book, even a scanned one: the panel reads its text with `pdftotext` or, when the pages are images, with the Tesseract OCR (Portuguese) on the server, page by page. The rows of the book are put in its order and a number the OCR misread is corrected when one similar digit puts it back in the decimal order of the table; the report lists each correction to check against the book. Each table is checked before it is kept (entries per letter, numbers out of order, repeated entries) and saved in `/etc/koha-easy-install/tables/`, the same tables the AI cataloguing and Cataloguing aids use. - **The notation** follows the same rules everywhere (`KohaEasy::Cataloguing::Rules`): the entry right before the name in alphabetical order, the initial of the surname, the number and the initial of the title without its article (`Assis, Machado de` + *Dom Casmurro* = `A848d` with Cutter-Sanborn). The main entry is automatic: the author, or the first word of the title (without its article) when the work has no author; it can also be set to a person, an institution (first word) or the title. On the page a name typed in direct order (`Machado de Assis`) is read as `Assis, Machado de`. - **Options on the page**: the table (PHA or Cutter-Sanborn, when both are loaded), the classification, **Without the letter of the title** (`A848` instead of `A848d`), **Edition** (from the 2nd on: `A848d 2. ed.`; read from 250) and **Copy** (`ex. 2`; the copy number of the item, 952 `$t`, when the button is used there). - **The gear** (Settings) keeps the defaults of each librarian in the browser: the table, the classification, whether the letter of the title is added and whether the edition and the copy number are added when the record has them. They apply every time the page opens; a field changed by hand on the page wins. - **The classification** of the call number is CDD, CDU or the library's own codes (`LIT`, `INF J`). It is the one chosen in the settings; otherwise it is read from the class already typed, then from the record (082 CDD, 080 CDU, 084 other), and then from how most of the catalogue's call numbers look. Without a class in the field, the class of the record is used (082, 080 or 084, as the classification says). - **Suggested classes**: when the record has no class, or only a CDD class without decimals (`869`), the page lists classes from the catalogue: the classes of other titles of the same author, of titles with the same subjects (650), and of titles with the same words in the title, plus, for CDD, the numbers the CDD index of **CDD lookup** gives for the subjects and the title when that index was built. They are hints counted from the catalogue, not a classification: the best one fills a **Suggested call number**, and **Use this class** puts any of them in the form. The panel has no CDU table, so for CDU the class comes only from 080 or from the catalogue's own CDU call numbers. The AI cataloguing (Replace a MARC record) is the place where a model suggests a CDD from the photos of the book; this page sends nothing out. - **The page** shows the number as the author is typed, the table entry used and, with a class number, the call number and the call numbers of the catalogue that already use the same number in that class, with the numbers right before and after it and whether they are free to use. **Copy** copies the number or the call number. `?format=json` gives the same result as JSON (`code` is the full number with the edition and the copy, `notation` the number alone). - **Install the page** optionally adds a **Call number** button next to 090 and 092 `$a` and the item call number (952 `$o`), and a **Cutter** button next to 090 and 092 `$b`, in the cataloguing forms, after the CDD button when there is one, and **Cutter Calculator** in the tools of the cataloguing home page (a marked block in `IntranetUserJS`, verified `PRE-CUTTER` backup first). The buttons open the page with the author (100, or 110/111), title (245 and its non-filing indicator), class (the field, or 090), 080 / 082 / 084, subjects (650) and edition (250) of the form; in the item editor they are read from the record. **Use in the record** puts the class in `$a` and the author number in `$b` of 090 / 092, or the whole call number in the item call number (keeping a location prefix such as `R`). - **Calculate a Cutter number** does the same in the panel with the table you pick, with the numbers used in the class. **Remove the page** removes the page and the buttons; the tables are kept. ### Koha plugins (on or off) **Library tools > Koha plugins** turns on or off Koha's own plugin system (**Administration > Manage plugins**), which a new Debian instance ships turned off (`0` in `koha-conf.xml`). - **Turn Koha plugins on** sets `enable_plugins` to 1, adds `` when it is missing (`/var/lib/koha//plugins`, created for the instance user), sets `UseKohaPlugins` on Koha versions that still have that preference, and restarts Plack and the background workers. - **Turn Koha plugins off** sets `enable_plugins` back to 0 and changes nothing else: the plugin files and their data stay, and turning plugins on again brings them back as they were. - **Where plugins can be installed from**: `.kpz` upload plus the plugin repositories, or only the plugin repositories (`plugins_restricted`). When no repository is set, the panel offers to add the ones `koha-conf.xml` lists as examples (ByWater Solutions, Theke Solutions, PTFS Europe). - **Register plugins copied into the folder** runs Koha's `koha-plugins --install` (or `install_plugins.pl` on older Koha) after a verified `PRE-PLUGINS` backup, for plugins unpacked into the plugins folder by hand. - Every change writes a new `koha-conf.xml` only when it is well-formed and reads back with the new values; otherwise the file is left as it was. The previous file stays next to it as `koha-conf.xml.bak-`. ## Automated tasks Installed in `/etc/cron.d/koha_tasks`: | When | Task | |------|------| | Daily 23:00 | Compressed SQL backup (verified, optional cloud upload) | | Sundays 03:00 | MARC21 export of bibliographic and authority records | | Daily 01:30 | Session and Zebra queue cleanup (`cleanup_database.pl --confirm`) | | Daily 05:00 | Plack restart (keeps memory low) | | Every 2 min | Zebra indexing watchdog: keeps Koha's indexer daemon (`koha-indexer`) running and restarts it if records wait more than 10 min | | Every 5 min (Elasticsearch only) | Elasticsearch indexer watchdog | E-mail notices are left to Koha's own schedule (`koha-common`): overdue and advance notices once a day and the message queue every 15 minutes, for the instance enabled with `koha-email-enable` (done at installation and in **Koha settings > Configure email**). Schedules written by older versions of the panel are upgraded automatically: the 01:30 cleanup gets `--confirm` (without it, it only reported what it would delete) and the old 08:00/08:05 e-mail jobs, which duplicated Koha's, are removed once e-mail is enabled. ## Important files | Path | Content | |------|---------| | `/root/koha_credentials.txt` | First-access credentials | | `/etc/koha/sites/library/koha-conf.xml` | Koha instance configuration | | `/var/backups/koha_sql`, `/var/backups/koha_marc` | Local backups | | `/etc/koha-easy-install/` | Panel settings (language, backup) | | `/var/log/koha-easy-install/` | Panel, APT and validation logs (`tools/`: library tools, root only) | | `~/importar` (`C:\Koha\Importar` on Windows) | Drop folder of the Magic Import Tool | | `/etc/koha-easy-install/import-profiles/` | Column answers the Magic Import Tool remembers | | `/var/lib/koha/library/kei-xslt/` | Cataloguing card stylesheets (only while the card is enabled) | | `/etc/koha/sites/library/kei-messaging.conf` | WhatsApp / Telegram settings and tokens (root and Koha only) | | `/usr/local/lib/site_perl/SMS/Send/KohaEasy/Gateway.pm` | The messaging driver of Koha (with `KohaEasy/Messaging.pm`) | | `/etc/koha-easy-install/tables/` | Author tables and CDD schedule loaded by the library | | `/var/lib/koha/library/kei-marc-replace/` | Records as they were before each replacement | | `/etc/koha-easy-install/windows.conf` | Windows only: what the Windows tools tell the panel (network mode, automatic start...) | | `C:\Koha\` | Windows only: `Uninstall-Koha.cmd`, `bin\` (scripts, `KohaEasy.exe` and `koha.ico`), `logs\`, `Backups\`, `state.json` | The installation shows seven numbered steps, one line per task with a small Pac-Man progress bar, and a final screen with the addresses of the catalog and the staff interface. The commands' own output never reaches the screen: it is kept in `/var/log/koha-easy-install/apt.log`. If a task fails, the panel shows the failed step, what to check and the last lines of that log, which is the file to send to IT support. The panel's other long tasks look the same: switching the search engine, repairing the search index, setting up the Cloudflare Tunnel, updating the system, validation, service repair, deep maintenance and language updates. Their output goes to `/var/log/koha-easy-install/` (for example `search-engine.log`, `zebra-rebuild.log`, `cloudflare.log`, `restore.log` and `apt.log`). Backup restores and the tools the panel installs on demand (Midnight Commander, links, htop and nethogs) look the same. The panel's dialogs are drawn with `dialog`, which the panel installs, so buttons, menu items and check boxes also take mouse clicks in terminals that report the mouse: Windows Terminal, the Linux desktop terminals (GNOME Terminal, Konsole, xterm), PuTTY and most SSH clients. The Linux text console only does so with `gpm` installed. `KEI_UI=whiptail` keeps the older keyboard-only look. On a Linux PC with a graphical desktop, the panel adds a **Koha** launcher (`koha-descomplicado.desktop`) to the applications menu of the user who ran it with `sudo`, and, the first time, to that user's desktop. It opens the panel in a terminal and asks for the `sudo` password. ## Uninstall `uninstall.sh` **permanently deletes** Koha, its databases, local backups and settings, and asks for confirmation first: ```bash sudo bash uninstall.sh # asks you to type APAGAR to confirm sudo bash uninstall.sh --yes # no questions (automation) ``` Copy your backups somewhere else before running it. On Windows, double-click **Uninstall-Koha.cmd** in `C:\Koha`, or use **Settings > Apps > Koha (koha.nexus) > Uninstall**. It asks you to type `KOHA` to confirm, then removes the Debian `koha` (with the Koha databases), the shortcuts, the status icon, the scheduled tasks, the firewall rule (Windows asks for permission once) and the folder. The `Backups` folder is kept unless you choose to delete it too. WSL itself and your `.wslconfig` stay. ## Tests (for contributors) `tests/` has a [bats-core](https://github.com/bats-core/bats-core) battery that runs the panel against a real MariaDB: corrupt, truncated and empty backups, MariaDB down or refusing the login, full or unwritable disks, CTRL+C / lost SSH connection in the middle of a restore, locks shared with the nightly backups, indexing after engine switches and restores, Debian/Ubuntu releases on amd64/arm64, the library tools (dry run before any change, lock, verified backup, `koha-shell` quoting), the Brazil tools (Latin-1 MARC migration to 952, CPF check digits, movable holidays, label templates, cataloguing card, collection spreadsheets, census reports, ABNT references), the messaging driver (against a WhatsApp / Telegram test double), the author notation, the Cutter Calculator and CDD lookup, the Koha plugins switch in `koha-conf.xml`, `marc_replace.pl` (run as a CGI), the schedule upgrade, the WSL mode, the QR code / browser / link sign-in, and the Windows tools (their PowerShell tests run with [Pester 5](https://pester.dev) when `pwsh` is installed). ```bash sudo apt-get install bats mariadb-server memcached whiptail yaz xsltproc python3 \ libmarc-record-perl libmarc-xml-perl libsms-send-perl libcgi-pm-perl libmodern-perl-perl sudo KEI_TEST_SANDBOX=1 tests/run.sh ``` **Only on a disposable container or VM:** the tests replace the `koha_library` database and install test doubles for the `koha-*` tools and `systemctl`. `tests/run.sh` refuses to run next to a real Koha. ## Translations (for contributors) The panel texts are in English inside `installer`; each language has a dictionary in `lang/.cache` (`base64(English)|base64(translation)`). The panel itself is pure Bash; the Python scripts below are optional contributor tools. - Wrap every user-visible text in `$(t "...")`. Variables must be escaped so the English text reaches `t()` unchanged: `$(t "Backup saved in \${file}")`. - Check coverage and repair dictionaries: `python3 i18n_common.py` / `python3 i18n_common.py --fix` - Translate only what is missing: `python3 gen_all_langs.py` (offline, Argos Translate) or `python3 gen_lang.py` (online). - Translations that lose a variable or `%s` are rejected automatically and English is shown instead. After changing `PANEL_VERSION`, run `python3 i18n_common.py --fix` (each dictionary is stamped with the panel version, and outdated dictionaries are only used as a last resort). After changing `installer`, regenerate the checksum used by the self-update: ```bash sha256sum installer > installer.sha256 ``` ## License koha.nexus is free software under the [GNU General Public License v3.0 or later](LICENSE) (GPL-3.0-or-later), the same license as Koha itself. You can use, study, share and modify it; if you distribute modified versions, they must stay under the GPL with their source code available. It comes with **no warranty**. ## Support the project - Pix (Brazil): `076.650.449.21` - Bitcoin (BTC): `bc1qw0kvacdkzul0panuppxcv90y08ah443m2z89tx` - ⭐ Star the repository and share it with other libraries. --- Created with dedication by **Paulo F. Baldi FH** — Library Assistant, Castro Alves Public Library, Palotina, Paraná, Brazil. The Koha name and logo belong to the Koha community (koha-community.org); this installer is an independent project.