Files
Chronos/docs/CHANGELOG.md
T
Alby96andClaude Opus 5.5 efa87f68ce
CI / Build and test (push) Failing after 3m1s
CI / PostgreSQL 16 integration (push) Skipped
CI / PostgreSQL 17 integration (push) Skipped
CI / Docker image (push) Skipped
CI / Windows tests and installer (push) Skipped
CI / YAML check (push) Successful in 10s
Release / Build, test and publish (push) Successful in 15m31s
Release / Server image (push) Successful in 1m13s
feat: owner review 2 phase 4b: mass conversion after import (#13)
Convert photos already in the archive to JPEG, TIFF-16 or Linear DNG, from the panel's Library and the desktop's Archive and Library. Each run asks Replace (same frames, originals to the trash, recoverable) or New sequence. A dry run lists what converts, what is skipped and why, size before and after, and free space; the job converts exactly the confirmed photos, resumes, and commits all or nothing per sequence. A conversion never changes what Chronos reads from a photo. Replaced originals are erased only on the archive's own word that their conversion is gone. Desktop TIFFs that WIC showed black or as noise now go to the managed reader; the Edit picker no longer spins a core.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Alberto Balbo <alberto.balbo.96@gmail.com>
2026-11-03 10:41:50 +01:00

88 KiB
Raw Blame History

Changelog

All notable changes. Format: Keep a Changelog; versions follow SemVer; commits follow Conventional Commits (release notes are generated from them by Chronos.Tools changelog).

[Unreleased]

[0.1.0-rc.7] — 2026-11-02

Added

  • Convert on the desktop and the admin panel (owner review 2, R7): convert the photos of sequences already in the archive to JPEG (with a quality), 16-bit TIFF or linear DNG, the archival choice. At every run you choose what becomes of the originals: Replace (the converted files become the sequence's photos, the same frames, and the originals go to the archive's trash, from which Undo or Restore brings them back until the trash is emptied) or New sequence (converted copies in a new sequence beside each one, titled "… (DNG)", with its tags, rating, notes and project; the originals stay as they are). No mode is preselected and none is remembered.
  • A dry run first, following your choices: per sequence the photos converted, their size before and the estimate after, the photos that stay as they are and why (already in that format, a format this computer cannot read at full resolution, which the server answers with "convert these photos on a PC", read only from a preview, a RAW this computer develops at 8 bits only, a colour profile the format cannot keep, too large, unreadable, missing, never verified, held open by another program), the sequences left as they are (busy, offline, unreadable) and the free space the conversion needs. A PC asks its own decoders again about a photo the server found unreadable (a CMYK TIFF), so it converts what it reads. With New sequence it says which photos each new sequence lacks. Replace asks you to type the number of photos (desktop) or the word "replace" (admin panel); New sequence asks a confirmation that counts the photos not copied.
  • A Convert job then converts exactly the photos the dry run listed, one sequence at a time and all or nothing: photos uploaded meanwhile stay as they are, and a sequence whose photos changed since is left as it is and named in the job's log. Pause, resume and retry it in Jobs: it keeps what it converted, resumes after a crash, checks the free space as it goes (counting every photo still being written) and stops before the volume fills. A second conversion of the same sequences is refused while the first waits or runs. A retried New sequence conversion finishes the copy it already made and never makes a second one; a source another conversion copied in the same format since this one was confirmed gets no copy (the job's log says so), and while that other conversion has not finished, the source waits for its retry.
  • Every photo is decoded at full resolution only, never from an embedded preview, by Chronos' own decoders wherever they read it (RAWs developed at full precision, the same on every computer), and keeps its metadata (capture time to the tick, camera, lens, exposure, GPS); every converted file is read back and decoded by the server's decoders before it is kept. A linear DNG keeps a RAW's highlights above white (up to 4 stops, in BaselineExposure) and its colours outside sRGB (ProPhoto primaries).
  • A conversion never changes what Chronos shows of a photo: the converted file decodes to the values the original decoded to, within what the format holds. A JPEG or TIFF-16 keeps the photo's values and carries its colour profile; a float TIFF's values get a profile made for them, with the original's primaries, and a gray profile's tones are kept. A linear DNG takes only photos whose values are sRGB as Chronos reads them (RAWs, photos without a profile, sRGB): photos in Adobe RGB (camera JPEGs tagged so included), Display P3 (HEIC), ProPhoto or a gamma-1.0 profile stay as they are, with "a colour profile the format cannot keep"; TIFF-16 and JPEG convert them with their profile. A file whose pixels hold values that are not numbers fails as damaged and nothing is written.
  • Desktop: Convert… in the Archive manager's set bar and in the Library's mass bar (the sequences of one archive). When a Replace job ends, its toast offers Undo, which brings back the originals that job replaced.
  • Admin panel: Convert… in the Library's selection bar (up to 1,000 sequences, or every sequence after Select all without a filter), and a toast that opens the job. A dry run that takes longer than a proxy lets a request last is asked again until it is ready.
  • The archive's trash lists the originals a conversion replaced as their own kind: Restore puts them back in place of their conversions while those are unchanged (the conversions then go to the trash), otherwise as a new sequence. Empty… erases them only while each conversion still matches where the archive holds it now. A conversion counts as deleted only when the archive says so (a deletion recorded in its sequence's sidecar, the sequence in the trash, or the trash's record of erasing that sequence); such originals are named first in its confirmation ("conversion deleted"), since they may be the last copy of their photos. Whatever the archive cannot tell keeps the entry.
  • API v1: POST /api/v1/archive/convert (archive administrators): a dry run with the plan, its space line and its fingerprint, then the convert job (202) of exactly that plan; 409 plan_changed, nothing_to_convert, not_enough_space; 503 plan_dropped when the server let the plan go. The server keeps a dry run a few minutes, in the plan cache Reorganize uses. sequenceChanged with the changes converted and created; trash entries of kind replaced, and conversionDeleted in an emptying's dry run; the audit log and the Dashboard name conversions.
  • Known limitations: a photo that cannot be converted although the dry run listed it (a damaged RAW whose headers read well) leaves its whole sequence as it was, while the other sequences convert, and the job's retry fails the same way; a decode larger than the computer's memory stops the job itself at every retry, before the sequences after it (converted photos are capped at 268 MP for JPEG and about 179 MP for TIFF-16 and DNG, which makes it rare); a damaged file that still decodes (a truncated TIFF strip, a corrupt JPEG scan) converts as the partial picture its decoder gives; a read error while the dry run reads the headers of a TIFF, a DNG or a TIFF-based RAW lists that photo as needing a PC, preview-only or unreadable rather than held open (it stays as it is); JPEG and TIFF-16 clip a float TIFF's values above white and below black; a PNG's gamma and primaries without an embedded profile, JPEG XL, JPEG XR, BMP and GIF, and Exif colour spaces other than sRGB, Adobe RGB and uncalibrated are taken as sRGB; a linear DNG keeps the as-shot white balance baked in and clips noise below black; MakerNote and XMP are not carried; the server's kept dry run does not see a change made on disk alone (a photo another program held stays listed as held for up to 5 minutes, unless the index changes), and the job then converts only the photos the dry run listed, each only while it holds the content confirmed; a conversion of every sequence of an archive that keeps receiving uploads may never be confirmed, as each upload changes the plan; of two confirmations of the same sequences at once the second is refused with plan_changed; an emptying's dry run lists as erasable an entry the job then keeps (its conversion changed, never verified or not located, its sequence busy), and the job says why and fails at the end; erasing replaced originals holds their sequence while every conversion is re-hashed; the desktop's Undo lives in the toast of the job's end only: later, after a restart, or for a conversion the server ran, Restore in the trash brings the originals back.

Changed

  • Import still copies every photo byte for byte and converts nothing. Deleting imported photos from a card keeps, and formatting the card refuses, a photo converted since: its archived file no longer matches the card.
  • Sync and imports count a photo and its conversion as one photo: a photo converted on either side, or on both, compares as one; importing a card or a folder again, or downloading an original, files nothing for a photo converted since; an upload of such an original into its sequence is answered alreadyExists; Download originals puts the server's conversion in place of a photo missing here, instead of adding a frame.
  • Exif written by Chronos (proxies, exports, converted photos) includes the exposure program, metering mode, flash and white balance when known.
  • Server: Reorganize's and Convert's dry runs share one plan cache, two made at once and eight kept across both.
  • Emptying the trash says in its job log why it kept each entry and what settles it: changed after the plan, its sequence busy, a conversion that no longer matches, a photo never verified (verify the sequence), a sidecar missing or unreadable (scan the archive), an index row the archive does not bear out (rebuild the index), or a conversion whose fate the archive does not show (restore the entry to keep its photos).
  • Desktop: a confirmation with a long list scrolls it, so its buttons stay on screen; Tab reaches the list, which then scrolls with the arrow, page and Home/End keys.
  • Upgrade: there is no migration: the index stays at schema 8, and earlier versions keep the sidecar fields conversions write (frames[].source, tombstones of reason replaced). Undo conversions before going back to an older version: it cannot swap replaced originals back (its Restore keeps them in the trash) nor settle a replacement a crash interrupted.

Fixed

  • The managed TIFF decoder (the server's, and the desktop's after WIC) read CMYK, CIELab and palette TIFFs as the gray of their first channel; it now leaves them unreadable (a PC reads them through WIC). A TIFF with an impossible number of strips or tiles is refused as unreadable too, so a PC asks WIC, instead of failing with an arithmetic error.
  • Sync's job log lines (a photo missing, changed, deleted on this PC or on the server, frames without a proxy or thumbnail on the server) are in your language; they were English in every language.
  • On a PC, TIFFs that WIC decodes wrongly without an error (a compression it does not know, such as ZSTD or JPEG 2000, which came out black; Photoshop's 32-bit files with a floating-point predictor and 32-bit integer samples, which came out as noise) are left to Chronos's own TIFF reader, or are unreadable, in renders, thumbnails and analyses as in conversions.
  • Edit: the picker no longer spins a processor core and stops responding when the list is loading a page as it looks for the sequence edited last.

[0.1.0-rc.6] — 2026-10-27

Added

  • Reorganize on the desktop and the admin panel (owner review 2, R7): apply new folder, file and title templates to sequences already in the archive. Each template has its switch. A dry run lists, in the order the changes apply, each sequence's folder and title before and after, how many photos get a new name with the first three, the sequences that only follow a folder they are in, and the photos that keep their names with why (kept in place outside originals, never verified, missing), or why a whole sequence stays as it is (its sequence file cannot be read). Nothing changes until you apply: a Reorganize job then makes exactly the changes shown, and you can pause, resume and retry it in Jobs.
  • Desktop: Reorganize… in the Archive manager's set bar and in the Library's mass bar (the sequences of one archive) replaces Rename… and Move…. Folders and file names start from the archive's templates, titles off with {title}; the dry run follows your typing.
  • Desktop: after a folder or file template is changed in Settings › Archive, a toast offers Reorganize… for every sequence of the archive with the changed templates switched on. A dialog already open on the Archive screen is never replaced: Reorganize opens when it closes.
  • Admin panel: Reorganize… in the Library's selection bar (up to 1,000 selected sequences, or every sequence after Select all without a filter), and in the Saved toast after new folder or file templates on the Archive page. A live dry run as you type, a confirmation that counts the sequences, folders and photos, then a toast with Open job.
  • Reorganize is crash-safe: files move in two phases through hidden temporary names under a journal in the sequence (.chronos/rename-files.json), so chains and swaps of names work and no file is ever written over; a failure puts every file back, and a crash is completed by the next program that writes the sequence. XMP files follow their photo. The plan's journal lets an interrupted Reorganize finish at the next archive open; one whose job was paused, cancelled or failed waits for that job to be resumed or retried (a job that failed because the archive's drive went away completes the plan once the drive is back; retried while the drive is still away, it fails and keeps the plan).
  • A folder that moves takes along the sequences nested in it and the exports inside it; an editor open on a reorganized sequence reloads where it is now (desktop).
  • A sequence still receiving photos does not stop a Reorganize: it keeps its file names, and its folder and title change while its first photo still gives them, else it stays as it is; the others go on, and the job then says how many sequences were not reorganized as planned. A new Reorganize completes them. A selected sequence left as it is inside a selected folder that moves goes with that folder, keeping its title and file names.
  • API v1: POST /api/v1/archive/reorganize (archive administrators): a dry run with the plan and its fingerprint, then the change as a move job (202) for exactly that plan, 409 plan_changed when the archive changed since, 503 plan_dropped when the server stopped making a plan; the server keeps a dry run a few minutes for its confirmation. sequenceChanged with the change reorganized; the audit log and the Dashboard name reorganizations.
  • Known limitations: while one sequence's files are renamed, a program that reads them without the sequence's lease (an editor open on it, a render node or a download fetching originals) can miss a file for a moment; a change made on disk alone after a server dry run (a folder made by hand, a desktop writing to a shared archive before the server's scan) is found by the job, which then fails and changes nothing; a photo restored from the trash after a Reorganize comes back under its old name; a Reorganize job cancelled and never retried leaves its held plan in .chronos/journal (only the file renames a step left half-way are completed at an archive open); deleting sequences is refused while a Reorganize job runs or is paused; a render of another program whose folder moves while it runs is not stopped, and its output can land in a folder created again at the old place; a plan journal that cannot be read is logged and left in .chronos/journal: the archive opens, its plan is not resumed and its job fails with the reason; a render node retries a result the server cannot file for now for about 30 seconds, then its job fails and a retry renders it again; an export one sequence saved into another's folder (the Export folder picker) keeps naming its old path when another program then moves that other sequence away and the first one into its folder (the file is intact: it moved with the other sequence).

Changed

  • Import, upload and Reorganize name photos by one rule: every photo of a sequence is named with the sequence's camera and lens (its first photo's, or the camera its creator named), so a whole-folder import of two bodies, or a lens changed during a session, no longer shows in the file names; a sequence's folder comes from its first photo. A Reorganize with the archive's own templates changes nothing filing named; where filing could not know the sequence as it is now ({session} of a card's second session, numbers after a photo skipped as a duplicate), its dry run lists the renames.
  • Templates are checked alike everywhere (Settings › Archive, Import, Reorganize, the admin panel's Archive page): a folder template may not write a folder scans skip (one starting with ., or named originals, proxies, exports); a file template must keep the photo's extension ({n}{ext} is refused, {n}.{ext} is fine), tell a session's photos apart and hold no slash. A folder a title would hide that way gets a leading _, and a file name over 200 characters is shortened before its extension, never through it.
  • Upgrade: a template saved before these rules stays as it is through edits of other settings, on the desktop and the server; changing it holds it to the rules. Creating an archive (Welcome › Create, adopting a new folder) with this PC's templates that break a rule is refused, and the message names the rule: fix it in Settings › Archive. There is no migration.
  • Server: invalid_template from PUT /api/v1/archive/settings and POST /api/v1/archive/reorganize names the broken rule after a colon (fileTemplate:extension, fileTemplate:not_unique, fileTemplate:file_separator, folderTemplate:reserved_folder) with its sentence in the request's language; the admin panel shows it on the field.
  • Renaming a sequence's files raises its frames version: an editor or client holding the old one reloads once.
  • Proxy passes hold the thumbnail lease one batch (about 30 s) at a time, so another program's Reorganize, sequence deletion or pack compaction waits for one batch rather than the whole pass.
  • Server: a render upload or a render node's result that cannot be filed for now answers 503 and keeps the upload, to be completed again: sequence_busy (the sequence busy past the lease wait), archive_offline, or the new folder_moved (another program sharing the archive moved its folder, or moved another sequence into it, until the server's next scan); the node keeps its job. A proxy upload answers 503 sequence_busy while the sequence is busy.
  • The job kind Move is now Reorganize.

Fixed

  • After another program moved a sequence away and another sequence into its folder: a render resumed after a crash no longer deletes a file of the same name in the folder the sequence left, and Download originals restores missing photos only where the sequence is now (under its lease); until this PC's next scan it fails with "folder moved" and keeps the downloaded copies for the retry. Before, the photo could land in the other sequence's folder.
  • Renaming titles with a template (former Rename…) turned characters a file name cannot hold into _ (Paris: night became Paris_ night); titles now keep them.
  • Moving folders with a template (former Move…) could flip a sequence between "… (2)" and "… (3)" at every run, could give two sequences folders that differ only in case, named a folder after today's date when a sequence had no capture time in the index, and failed on sequences created without photos; folders now come from the sequence's first photo, and such sequences are left out.
  • Exports stay reachable when another program sharing the archive moves their sequence's folder (its Reorganize, or a folder moved by hand): the next scan moves their paths with the folder, image-sequence exports included, and when another sequence was moved into the old folder each sequence keeps naming its own exports. Before, they answered 404 there.
  • A writer that read a sequence before its folder moved (an upload, a proxy, an export, an editor save, a sync, a deep analysis) writes where the sequence is now, or, when another program moved it, refuses until the next scan instead of creating the old folder again, and also when that program moved another sequence into the folder it left, so nothing lands in the other sequence's folder. Desktop: Save both and Take server on such a sequence say that its folder is not where the index has it, and a render job of it does not start.

[0.1.0-rc.5] — 2026-10-27

Added

  • Admin panel: Library, right after the Dashboard (owner review 2, R6). It manages the library and never edits a video: browse the sequences as cards or a list, search, filter by year, camera, status and tag (each with its count), sort, and load more; select with check boxes, Shift ranges, the keyboard or Select all of a filter; then Analyse, Proxies, Thumbnails, Verify, add or remove tags, export a CSV, or delete into the archive's trash after a dry run that lists what goes and what cannot, with why.
  • Admin panel: a sequence's page: title edited in place, tags, rating and notes; its facts; its photos with a viewer (←/→, Esc, the photo's details); set the cover, download originals and renders, delete photos (their number typed above 20) or the whole sequence. Nothing here edits the video.
  • Admin panel: the archive's trash, from the Library and from Archive: restore an entry, or empty the trash after a dry run and typing "empty"; the emptying runs as a job that erases exactly what the dry run showed.
  • Admin panel: Upload photos from the browser into an existing or a new sequence, with the file picker or by drag and drop (owner review 2, R3). Photos are hashed in the browser by a self-tested in-house SHA-256, dated from their Exif (else the file's time), sent in resumable chunks and verified by the server. Each photo shows its state, with a total, Cancel, and a warning before leaving the page while photos are sent. Folders are skipped: importing a whole folder stays on the desktop.
  • Deleted photos stay deleted: a tombstone in the sequence's sidecar and the index keeps sync, the Uploader, watch folders and older clients from bringing them back, also after the trash is emptied. A photo comes back only by restoring it from the trash or with Upload again on the panel. Sync never deletes your local files. Known limitations: a photo never hashed whose file was already missing when it was deleted leaves no tombstone, so a copy elsewhere (another PC, a card) can come back; on a PC, a photo deleted there that the server has not verified yet still counts as missing in the sync table, and Download fetches and discards it until the server verifies it.
  • Desktop: Archive › Maintenance › Trash… lists deleted photos beside deleted sequences, restores both (photos after a hash check, into a new sequence when theirs is gone), says why a photo stayed in the trash, and Empty… erases the trash for good after a dry run and typing the number of entries, through a new Empty trash job that refuses when the trash changed since. No schedule ever empties the trash.
  • API v1: POST /api/v1/sequences/{id}/frames/delete, /api/v1/sequences/bulk/delete, /bulk/retag and /bulk/jobs (up to 1,000 sequences a request, answered per sequence), GET /api/v1/sequences/facets, GET /api/v1/sequences/{id}/thumb, rendered=true|false on GET /api/v1/sequences, /api/v1/archive/trash (list, restore, empty with a dry run and its fingerprint), removed in the sequence manifest, extensions in GET /api/v1/server/info, restore on an upload's start and completion, and 409 photo_removed.

Changed

  • Upgrade: the index (SQLite, or the server's PostgreSQL database) moves to schema 8 (migration 0008, photo tombstones) at the first start, one way: an older build refuses the upgraded index, so going back means restoring a copy made before the upgrade.
  • After photos are deleted from a sequence, its project's exclusions, hero and keyframes on those photos wait for them (a restore puts them back) and the others follow their photos; a cover whose photo went is cleared. Before, positions stayed where they were and could slide onto other photos.
  • Server: frame uploads whose file name starts with . (the ._ files a Mac leaves on cards, hidden files) are refused from every client, as the archive scan leaves them out.
  • Admin panel: its files are checked with the server before each use (no-cache with an ETag), so after an upgrade the browser never mixes new and old pages.
  • Admin panel: confirmations can list what they concern and ask for a name, a word or a count; arrows mirror in right-to-left languages, as on the desktop.
  • The audit log and the dashboard name photo deletions, trash restores and requests to empty the trash.

Fixed

  • Server: a verify of chosen sequences (POST /api/v1/jobs with verify) failed at once, because it targeted the sequences instead of their archive.

[0.1.0-rc.4] — 2026-10-27

Added

  • Jobs › Scheduled, on the desktop and in the admin panel: every scheduled job of this PC or of the server, its own system schedules included, with when it runs, its next run, its last result and an on/off switch; Run now, Edit, Reset (system schedules) or Delete (your own), and a schedule's last runs, each opening its job in the Queue tab. Jobs a schedule queued carry a Scheduled pill (owner review 2, R10).
  • Schedules you create: scan, integrity check (a share of the archive per run, the least recently checked first), thumbnails, proxies (when the archive makes them), analysis, maintenance, index backup, clean-up, and duplicates and orphans reports, every n minutes, hours or days, daily, on chosen weekdays or on a day of the month, at a time of the PC's or the server's time zone, with priority and catch-up (a run missed while the program was off runs once when it is back). A run is skipped while the previous one, or the same work, is still pending.
  • System schedules for everything that used to run out of sight. Server: Archive scan (every 10 minutes and at start), Upload clean-up (every 6 hours), Index backup (03:00, SQLite only), Job history clean-up (04:00) and Audit log clean-up (04:30, off by default). Desktop: Archive scan (every 10 minutes, at start and when an unplugged archive is back), Index backup (every day) and Housekeeping (04:00: job history, and the disk cache trimmed to Settings › Storage › Cache limit, oldest files first, never files in use, synced proxies or import staging). They can be changed, turned off and reset, never deleted.
  • New job kinds Index backup and Clean-up; duplicates and orphans reports also run as jobs, on the server too. The clean-up's trash report lists old trash entries and never deletes them.
  • Admin panel: the Scheduled tab refreshes live (skipped runs, moved next runs, other administrators' edits), shows every time in the server's time zone, and the Logs page and the dashboard describe schedule changes and runs in words, a system schedule by its name in your language.
  • API v1: /api/v1/schedules (list, create, PUT/PATCH, delete, run now, reset, runs), audited; the WebSocket event scheduleChanged; jobs carry scheduleId when a schedule queued them; GET/PUT /server/settings carries logs and lockedBy.
  • Server: log settings in the admin panel's Settings › Logs: file level, console level, days kept, file size, total size, daily or hourly rotation, audit retention and the diagnostics share, applied without a restart; and the variables CHRONOS_LOG_CONSOLE_LEVEL, CHRONOS_LOG_RETENTION_DAYS, CHRONOS_LOG_MAX_FILE_MB, CHRONOS_LOG_MAX_TOTAL_MB, CHRONOS_LOG_ROTATION, CHRONOS_AUDIT_RETENTION_DAYS and CHRONOS_DIAGNOSTICS_LOG_MB (owner review 2, R11).
  • Desktop: Settings › Advanced gains Log level, Keep logs, File size, Logs total, Rotation and Diagnostics logs, applied from the first log line, and a Reset section for them.

Changed

  • Upgrade: the docker-compose.yml and Unraid template of earlier versions set CHRONOS_LOG_LEVEL=Information. That variable now sets and locks both the file and the console level, so Settings › Logs shows them read-only, its tooltip naming the variable. To choose the levels in the admin panel, remove it: in compose delete the line; on Unraid edit the container and clear Log level.
  • Upgrade: the index (SQLite, or the server's PostgreSQL database) moves to schema 7 (migration 0007, schedules) at the first start, one way: an older build refuses the upgraded index, so going back means restoring a copy made before the upgrade.
  • No timer queues work out of sight any more: the periodic archive scan, the upload sweep, the daily index backup and the job history purge are system schedules in Jobs › Scheduled. The job engine no longer deletes finished jobs at start; the job history clean-up does, at 04:00.
  • Index backups: the server copies at 03:00 (and once after a missed night), the desktop every day, instead of at start when the newest copy was a day old. The first start after the upgrade still copies at once when the newest copy is a day old. Back up now, Back up database and the copy kept before a restore keep as many copies as the backup schedules (3 by default), so no copy deletes what a schedule keeps.
  • Admin panel: Settings no longer has Job history (days); it is the age limit of the Job history clean-up schedule, and the saved setting (server.json or API) is what Reset restores, without a restart. Keep audit in Settings › Logs is the Audit log clean-up schedule's age limit: the audit log is kept forever by default, and CHRONOS_AUDIT_RETENTION_DAYS fixes the limit at every start.
  • Server: an abandoned upload is swept every 6 hours, so it may expire up to 6 hours after its 48 idle hours.
  • Desktop: Settings › Archive › Rescan every is gone; its value became the interval of the Archive scan schedule, whose scans follow Settings › Analysis › After a scan as it is at every run. An archive unplugged at start is looked for every minute.
  • Desktop: a scheduled background job raises a toast when it fails (a backup, an integrity check), unless the next run repeats it (scans, thumbnails, proxies, light analysis, clean-ups).
  • Log files of every program are capped: a file continues in a numbered one (.1.log, .2.log…) at 50 MB, and the oldest are deleted when all of them exceed 1,024 MB (the file being written is always kept); hourly rotation names files prefix-yyyyMMdd-HH.log.
  • CHRONOS_LOG_LEVEL also accepts Critical and None (and info, warn, off); a number is refused at start. In server.json, LogLevel now sets the log files only and ConsoleLogLevel the console.
  • Admin panel: a setting fixed by an environment variable names that exact variable in its tooltip.
  • API v1: POST /api/v1/jobs with kind duplicates or orphans queues a report job instead of answering 409 job_unavailable.
  • Containers: TZ also sets the time zone of the server's schedules.

Fixed

  • Server: CHRONOS_LOG_LEVEL=Debug or Trace never reached the log files, which stopped at Information.
  • A log that cannot be written (its folder removed, a full disk) loses only that line: a failing job is still marked Failed, and the job engine and the schedules go on, instead of the job staying Running and the worker stopping.

[0.1.0-rc.3] — 2026-10-26

Added

  • Import › Folder…: Group › Whole folder imports the picked folder, with its subfolders when Subfolders is on, as one sequence per file format, whatever the cameras and pauses (RAW+JPEG stays two sequences). The title defaults to the folder's name, and an interrupted or repeated import of the folder continues the same sequence without renaming it. Desktop only (owner review 2, R9).
  • Server: CHRONOS_SPACE_PATH (SpacePath in server.json) makes the admin panel show the free space of that folder's volume instead of the archive's. The Unraid template maps the share's array view (/mnt/user0/Timelapse) read-only to /space (Array view), so the dashboard shows the array's free space, not the cache's (owner review 2, R2).
  • Admin panel: the storage figures' tooltip names the volume and its file system, and on an Unraid share without a space path explains that they are the share's primary storage; a Write volume line shows the archive's own volume when it differs.
  • Server diagnostics: the environment report lists the archive, space and data volumes with their free space, file system and mount.
  • Chronos.Tools release-channels: which channels a release tag moves (latest=, image_latest=), by SemVer 2.0 precedence among the published releases (every page of Gitea's list, one --releases file per page).

Changed

  • Server image: :latest now follows the highest stable release, or the highest pre-release while no stable release exists, so the Unraid template and the compose file find an image from the first pre-release on. The public releases repository still follows stable releases only, and re-publishing an older tag never moves either back. When the release workflow cannot decide (the releases list unreadable, or a tag older than release-channels such as v0.1.0-rc.2 published again) it warns and moves neither, and still pushes the version's image. :nightly stays the nightly build of main, for testing only (owner review 2, R1).
  • Release workflow: the tag must be SemVer 2.0 without build metadata (no leading zeros such as v01.2.0, no empty identifiers); any other tag is refused before anything is built.
  • Desktop: About is the last page of Settings and the rail keeps Settings alone at the bottom. Check for updates and the Releases link are on Settings › Updates; the data folder and the diagnostics bundle are on Settings › Advanced only (About's own copies of those buttons are gone) (owner review 2, R5).
  • Admin panel: Settings sits at the foot of the sidebar, above your account (at the end of the top bar on narrow screens); on a short window the pages above them scroll instead. About is the last section of Settings, without the server name and certificate fingerprint already shown above it, and its database and uptime fill in without holding up the form; old #/about links open it there (owner review 2, R4, R5).
  • The Settings icon is a toothed gear in Chronos, the Uploader and the admin panel; it no longer looks like a sun (owner review 2, R4).

Fixed

  • Import: a session that ends on another day shows both dates in its plan row and in Show plan, not only the first.
  • Server: completing an upload, a render upload or a render node's result now checks the archive volume's free space first. A full volume answers 507 insufficient_storage and the upload stays resumable, instead of the copy into staging failing (owner review 2, R2).

[0.1.0-rc.2] — 2026-10-23

Fixed

  • Update check: the release page and the download links come from the repository address the build names (ChronosReleasesUrl), not from Gitea's answer, whose links use Gitea's ROOT_URL; with the owner's plain-http LAN ROOT_URL every release was ignored (the check accepts only https links).

[0.1.0-rc.1] — 2026-10-23

First pre-release, built and published by Gitea Actions.

Added

Foundation

  • Solution and build rules: warnings as errors, nullable, .NET 10 SDK pin, zero-package guard (build target and Chronos.Tools package-guard).
  • Chronos.Testing: in-house test framework and runner (theories, categories, timeouts, environment and Windows gates, serial classes, parallel runs, JUnit XML, workflow annotations).
  • Localization runtime (Loc, CLDR plural rules with few/many forms, runtime culture switch, RTL), en + it catalogs, machine translations with per-language glossaries, i18n check/export/missing/import.
  • Brand mark geometry, SDF rasteriser, PNG encoder, ICO writer and make-icon (icons generated at build time).
  • VS Code debug configurations for the Windows programs (.vscode/launch.json, tasks.json): Chronos, the Uploader, the server on SQLite with its data in out/debug/server, the screenshot captures, Chronos.Tools and the test programs, each built in Debug first.

Engine (Chronos.Core)

  • Metadata: signature sniffing, TIFF/Exif/GPS/XMP/MakerNotes, JPEG segments, ISO-BMFF (HEIF, CR3), RAF, PNG, WebP and RIFF parsers with fuzz-tested partial results; MetadataBuilder fallback chains; ExifWriter.
  • Codecs: JPEG (baseline, progressive, lossless; libjpeg-exact integer IDCT), PNG, TIFF, linear DNG writer, GIF and APNG encoders, ICO; native RAW development for DNG (opcodes, AHD demosaic, DNG colour), CR2, NEF, ARW, PEF, RAF.
  • Imaging: planar linear-light buffers, decoder providers and decode tiers, separable resampler, frame window with read-ahead, memory budget and float32 spill, thumbs.pack, thumbnail cache, proxy builder.
  • Analysis and advisor: frame statistics, colour temperature, deflicker (global and regional), holy-grail step redistribution, white-balance smoothing, light analysis with findings, streaming deep analysis with checkpoints; Advisor recommendations with localised reasons, apply/optimise, sequence presets.
  • Motion: FFT and phase correlation, pyramidal Lucas–Kanade flow, motion blur, nine-patch stabiliser, virtual camera, frame interpolation, stacking and star trails.
  • Render and export: render plan (constant, adaptive, interpolated, remapped timing), preparation pass, geometry, pipeline, self-test suite (Chronos.Tools selftest); preset resolution, encoder selection with hardware preference and fallbacks, output naming.
  • Video: streaming MP4 muxer (H.264/HEVC/AV1) with inspector, AVI-MJPEG with OpenDML indexes, NV12/P010 conversion, level tables, managed encoder provider (MJPEG-AVI, image sequences, GIF, APNG).
  • Archive format: layout and naming templates, sidecars, leases, index snapshot, scanner, verify, adopt, orphans, rename plans; card import planning, sessions, hash-verified copy and verified erase; TPK1 installer payload.
  • Diagnostics: SecretScrubber and DiagnosticsArchive for diagnostics bundles that never carry passwords, tokens, keys or photos.

Data (Chronos.Data)

  • SQLite binding (winsqlite3 / libsqlite3) and in-house PostgreSQL wire protocol v3 client (pool, prepared statements, COPY, LISTEN, cancellation).
  • Migrations, repositories and the ChronosDatabase facade; one persistent job engine with checkpoints, pause/resume, priorities and crash recovery for client and server.
  • Archive engine: create/adopt/open, incremental and full scans, rebuild from sidecars, ingest of uploads and card imports, verify, thumbnails and proxies, maintenance, mass rename, job handlers.
  • Analysis and render engine: light/deep analysis and render job handlers, render node client.

Programs

  • Chronos desktop: shell with rail, job bar and toasts; Welcome, Library (virtualised grid and list, filters, sort, hover scrubbing), Jobs, Server (trust-on-first-use sign-in, DPAPI token), Settings (general, archive, rendering with encoder inventory and self-test, storage, server, updates, advanced) and About; --capture screenshots.
  • Update check against the releases repository on Gitea and Download: the setup (the portable zip for a portable copy) is fetched into %TEMP%\Chronos\updates, verified against the release's SHA256SUMS and shown in Explorer; an interrupted download resumes, Cancel stops it, and a transfer that stalls for 60 s fails.
  • Diagnostics bundle (Settings › Advanced, About): environment, codec and encoder inventory, scrubbed settings, recent failed jobs and logs.
  • Chronos Server: configuration from server.json and CHRONOS_*, TLS (auto, PEM, proxy), API v1 (auth, users, sequences, frames, uploads, sync, jobs, render nodes, events over WebSocket, archive), admin panel (dashboard, users, archive, jobs, settings, logs, about), audit log, LAN discovery, Windows service.
  • Server diagnostics: GET /api/v1/server/diagnostics (ManageArchive, audited) and a Diagnostics button in admin › Logs.
  • Admin panel: light and dark themes following the browser with a toggle, plural-aware counts, audit filter by several actions, password change, typed confirmations, right-to-left layout; an administrator cannot reset their own password from Users (409). DELETE /api/v1/sequences/{id} moves the sequence to the archive's trash and answers 409 or 503 with a reason when it cannot.
  • Chronos Uploader: tray program with resumable persisted queue, session grouping, watch folders, bandwidth limit, verified-only source deletion.
  • Chronos Setup: one-screen installer, per-user or all-users, components, shortcuts, file types, upgrade in place, silent mode, uninstaller that keeps user data unless asked.

Desktop workflows

  • Library: timeline with year bands (keyboard), saved views, quick actions reachable by keyboard, mass actions bar, Delete… into the archive's trash (dry run, typed confirmation, progress, Undo), Rendered / Not rendered / Synced and Origin filters, and cards for server-only sequences whose Render and Analyse run on the server.
  • Archive manager: tree by capture year, health with duplicate groups and Mark copies, statistics charts with values for screen readers, maintenance, Trash… with restore, mass rename and move behind a dry run; failures as one-sentence toasts with Retry, dialog errors inline, Cancel as the default button.
  • Import: devices and card pool, camera profiles, plan with sessions (merge, split, titles), hash-verified copy into the archive, a folder or the server, verified delete, format behind a typed confirmation, eject, auto-import.
  • Sequence editor: Frames, Exposure, Camera, Motion and Export tabs with the advisor's Auto values and reasons, Optimise, preview tools, section charts, curve editor (keyboard included), undo/redo and saving shortly after each change; Set hero updates the Library thumbnail; batch export with size and time estimates.
  • Server sequences open in the editor with proxies: saved, analysed and rendered on the server, with cancel and retry of server jobs, a read-only mode for accounts without Edit, and an explicit prompt when the same setting changed on the server; Export Sync sends local renders to the server; Save both copies can be restored from the Export tab.
  • Client sync (Server screen): sync table with filters, conflicts settled with Keep mine, Take server or Save both, uploads and downloads as resumable jobs that check the free space first, proxies kept offline, uploads paused on metered connections; this PC as a render node while signed in.

Localization and accessibility

  • 38 cultures: en, it and 36 machine translations with per-language glossaries and an independent review.
  • Right-to-left: windows mirror from startup and on every language switch (ChronosWindow), directional icons mirror with them, photos, charts, numbers, paths and templates stay left-to-right islands, category charts lay out from the reading side.
  • Accessibility: automation peers for sliders, filmstrip, curve editor, preview, charts, split buttons and the job bar; live regions for status lines; toasts announced (failures as errors); Field.Error for inputs without validation; charts take the keyboard (arrows, Home/End, Enter) where an item can be activated; two-word names for icon-only toggles.

Distribution and CI

  • Docker image (PUID/PGID, health check), compose file with PostgreSQL 17, Unraid template.
  • CI on Gitea Actions (.gitea/workflows/ci.yml): Linux build and tests, the Windows programs cross-compiled and checked (pe-check), PostgreSQL 16/17, Docker smoke test, i18n, XAML lint, format, package and YAML guards; the Windows tests, screenshots and installer test on a Windows runner once WINDOWS_RUNNER is set.
  • Release workflow: tests, self-contained publish cross-compiled on Linux, optional Authenticode signing (osslsigncode), installers, portable and server zips, SHA256SUMS, release notes from Conventional Commits, the Gitea release (Chronos.Tools gitea-release), the server image in the Gitea container registry, and optionally a separate releases repository (README, EULA draft, SECURITY policy, issue forms, Unraid template, compose file, i18n/).
  • Chronos.Tools commands for the Linux runner: zip, gitea-release, postgres-ci (the wire tests' roles and TLS through SQL) and pe-check.
  • Performance gates (perf.yml, weekly): 500,000-photo library and 100,000-frame rebuild from sidecars (every column compared), each on SQLite and PostgreSQL 17.
  • Chronos.Tools.Tests: the fixtures command's warning annotations and exit codes.

Verification against the brief (2026-10-12)

  • Settings › Import, Analysis and Formats pages (defaults of new cards, deep-analysis width, thresholds and auto-run, decode tier per format with Store links for missing codec extensions); Archive thumbnail size and rescan interval; Advanced › Restore (daily backups, dry-run comparison, staged for the next start) and Reset all; Archive › Maintenance › Restore database… opens it.
  • Daily VACUUM INTO backups of the SQLite index (newest three kept) on the desktop and on a SQLite server; maintenance vacuums the index when a tenth of it is free and its dry run reports the reclaimable size.
  • An archive's templates, proxy size and thumbnail size live in its archive.json. Settings › Archive and PUT /api/v1/archive/settings rewrite it atomically under a lease, and new desktop archives start from Settings › Archive.
  • CI: pixel diff of the --capture screenshots against tests/visual/baselines (Chronos.Tools screenshots diff and accept; 8/255 per channel, at most 0.05 % of a screen; a screen without a baseline fails once any is committed); captures are reproducible (fixed sample clock, English formats, no animation).
  • CI: installer test (silent install, run, upgrade in place, Uploader-only install, uninstall, exit codes), now the last step of the windows job on a Windows runner.
  • CI: the PostgreSQL 16/17 jobs run the MD5, cleartext, trust and TLS (require, verify-full) tests; a server without TLS fails them.
  • Windows tests: Media Foundation output opened by MediaPlayer (H.264, and HEVC Main10 where hardware allows) and its first frame decoded by a source reader.
  • Proxies of originals the server cannot decode (HEIF, AVIF, WebP): manifests and render-node claims carry the archive's proxy policy and, per frame, its flags and whether the server wants its proxy; PUT /api/v1/frames/{id}/proxy takes them from clients that edit and from render nodes; desktop uploads and render nodes make and send them; renders, analyses and thumbnails of those frames use them.
  • Server: the archive is scanned incrementally at start and every ten minutes, so photos already in an adopted folder and sequences added by another writer are indexed without a manual rescan.
  • Shell: the taskbar button mirrors the job bar and turns red while a failure waits on the Jobs screen. Toasts also appear as Windows notifications while the window is minimised or in the background; a click brings the window back and opens what the notification names. Ctrl+1…6 and Ctrl+, open the sections the rail tooltips list, and the editor's 1:1 view moves to Ctrl+Shift+1.
  • Jobs: a toast when a long operation ends or fails, whichever screen started it, and one toast for a whole batch ("12 jobs finished"). Rail badges count running jobs and unseen failures, sync conflicts, Library sequences with warnings, and Archive health lines that need attention.
  • Settings › General: Start with Windows (minimised, for the copy that set it; the uninstaller removes it) and Windows notifications.
  • Server events: cardEvent (BRIEF §14.3) tells every connected client about a card import another client streams into the server (started, progress, finished). Clients report it through POST /api/v1/uploads/cards.
  • Decode tiers (BRIEF §8.2): frames that decode only through an embedded preview are marked preview-only, with a Library chip showing the count, an Info finding and the output capped at the preview size. Frames nothing decodes are flagged undecodable, listed as a sequence warning and left out of renders. A frame that stops decoding during a render is shown as its neighbour instead of failing the job.
  • MakerNotes: exposure mode (Canon, Olympus, Fujifilm, Panasonic, Sony), RAW size (Canon sRAW/mRAW, Nikon RAW M/S) and the Panasonic time-lapse hint. Sequences shot by the camera's own intervalometer get adaptive timing across their pauses.
  • The SQLite compile options appear in the log and in the diagnostics bundle.
  • .chronos project files. Save project file on the Export tab writes the sequence's reference and settings. Opening one (a double-click, or Chronos.exe <file>) opens the sequence with those settings as one undoable step, with one-sentence toasts when the file cannot be read or its sequence is not in the library. One Chronos runs per data folder, and a later launch hands its file to the running one or, while that one is closing, waits for it and starts with the file.
  • Server: PATCH /api/v1/sequences/{id} takes expectedFramesVersion and answers 409 frames_conflict when the sequence's frames changed since the settings' frame positions were written. GET /api/v1/sequences/{id}/manifest returns frames, frames version and settings of one frames version.

Changed

  • Chronos on Gitea (2026-10-23, docs/requests/2026-10-23-chronos-gitea.md): the product is renamed from Titano to Chronos everywhere (programs, namespaces, CHRONOS_* variables, %LOCALAPPDATA%\Chronos, .chronos files, the image chronos-server, every translation with its grammar); the repository, update check, licence, issue and support links, container registry and workflows move from GitHub to the self-hosted Gitea (https://gitea.encke-hake.ts.net/BalboSoft/Chronos); releases are published in that repository, with EULA.md and SECURITY.md at its root and the issue forms in .gitea/ISSUE_TEMPLATE.
  • Owner review 1 (2026-10-16, R2, docs/requests/2026-10-15-owner-review-1.md):
    • jobs run one at a time, on the desktop and on the server. Settings › Rendering › Parallelism (server: CHRONOS_PARALLELISM, admin panel › Settings › Jobs) is now how many photos the running job processes at once: thumbnail and proxy passes draw that many photos together, decode tiers are probed that many at a time, the deep analysis measures that many frames and a render decodes that many ahead (fewer wherever large originals are developed whole, within the memory budget); the server's renders get a frame window sized to the host (a quarter of the memory available) instead of 1 GiB; the running job has the whole memory budget;
    • a new job waits its turn, and pausing the running job lets the next one start. An optional priority queue (Settings › Rendering, admin panel › Settings › Jobs, CHRONOS_PRIORITY_QUEUE) gives each job kind a level; by default renders, imports and downloads are high and verifies and proxies low. A queued job of a higher level pauses the running one without asking, and that one resumes by itself afterwards from its checkpoint;
    • the Jobs screen shows "8 photos at once", and a Priority queue chip while the queue is on; the admin dashboard shows the parallelism;
    • Chronos.Tools i18n prune removes from the translations the keys English no longer has.
  • Owner review 1 (2026-10-16, R1): the rail is 200 px wide, so section names fit in every language, and collapses to icons with ≡ or Ctrl+B; collapsed, an icon's tooltip names its section. The choice is saved at once and survives Settings › Reset all… and Import…. The admin panel's sidebar is 256 px and collapses to icons the same way (≡ or Ctrl+B in any keyboard layout, kept in the browser). Beside the wider rail at the 1200 px minimum no action is cut off in any language: the Jobs table, the Archive tree, the Import plan and result, the trusted certificates, the cache folder, the editor's header (a server job's Cancel always stays), preview tools (a second line when needed) and frame table share their width differently; headers, titles and cells that may be shortened end in an ellipsis and show their text in their tooltip.
  • Owner review 1 (2026-10-16, R7): the Sequence section is now Edit (Ctrl+3) and edits one sequence at a time. Without an open sequence, or after Change sequence (the title's ▾, Ctrl+O, or Next sequence on the last step), it shows a picker of this PC's and the server's sequences, newest first, with a search; the open sequence (else the one edited last) is selected and in view. Enter, also right after typing, or a double-click opens one, and Esc goes back to the open one, which stays open behind the picker until another one has loaded: a server sequence whose proxies are still coming, or one that cannot open, leaves it open. Its last edit is saved first, or discarded on the user's word and then never written. The editor's tabs are now the steps of a guided path under the header, freely navigable: a new first step, Analysis, lists every finding of the analyses with a button to the step that deals with it, five measurements (light, flicker, exposure steps received, shake, flagged frames) and where each other step stands. Each step shows its status with its own glyph (to do, done, attention, running, waiting); its tooltip starts with its name, says why and names its shortcut. Previous · n / m · Next under the settings column, Alt+1…8, Ctrl+PgUp/PgDn and the arrow keys in the step bar move between steps, the keyboard following. The editor opens on the step last shown for that sequence in this run, else on Analysis while the sequence is not analysed or its analysis waits or runs, else on Frames, whose list is now the frame warnings.

Deprecated

  • CHRONOS_WORKERS and the Workers key of server.json set the parallelism (photos the running job processes at once, 1 to 64); the server logs a warning at start when the variable is set. Use CHRONOS_PARALLELISM.

Fixed

  • Owner review 1 (2026-10-16, docs/requests/2026-10-15-owner-review-1.md):
    • the App test program could crash on exit after every test had passed (WPF's binding cleanup racing the process's end);
    • choosing Keep editing when the open sequence's last edit could not be saved left the other sequence as the one asked for, so every later visit to the section asked again (R7);
    • choosing Discard change for an edit that could not be saved still wrote it when the archive answered again while the question was open (R7 review);
    • a double-click on a row or card of the Library opened nothing (only one on the empty space under the list opened the selected sequence): the row took the mouse-down (R7 review);
    • thumbnail and proxy jobs over several sequences failed after 10–20 minutes with "TimeSpan overflowed" and then stayed Running; their progress now counts frames, a stalled job shows no time left, and a job's end is always recorded (R10);
    • the Jobs screen showed the speed of hash verification and card import, which count bytes, in fps;
    • choosing presets quickly in the editor made the VS Code debugger stop in WicDecoderProvider.cs: a cancelled preview decode left its background task (R4);
    • GoPro DNGs converted by Lightroom or DNG Converter were treated as GPR and read only through their 8-bit embedded preview, and DNGs Windows' preview-only DNG decoder read were reported as decoded natively. Videos now develop them natively at 16 bits, analyses and the editor preview use the embedded preview when it is large enough, and the archive's existing frames are reclassified by the next light analysis (R4);
    • an upload or a card import could fail with "Access to the path is denied" (server: 503 commit_failed) when a thumbnail or proxy pass, an analysis or a scan was reading the sequence's sidecar at that moment: the write now waits a moment for the reader (found while testing phase 2);
    • taking a sequence's lease could fail with "Access to the path is denied" when the previous holder had just released it while another process was reading it (Windows keeps the file pending deletion meanwhile): the lease is now taken once the reader lets go (found while testing phase 2).
  • Opening the repository in VS Code with C# Dev Kit before the first build listed MSB3073 for Chronos, the Uploader and Setup: design-time builds no longer generate the program icons, and log WPF's MC1000 about projects not built yet as a message.
  • Server: downloading the current log while it was being written aborted the response (Content-Length mismatch); it now serves the bytes listed at request time.
  • Mass rename and move: a dry run asked for right after typing was replaced by the pending debounced one, so Apply stayed disabled; the explicit dry run now cancels it.
  • Radio groups leaked between windows not shown yet (the Setup's capture overflowed the stack); screenshots of right-to-left windows were mirrored twice.
  • Switching the language left texts kept by view models in the old language until their next change: the job bar, the Jobs rows, the editor's chart legends and summaries, Auto chips and info line, the Archive figures and the Import session table now follow at once.
  • Right-to-left: in meta lines a number after a Latin field ("Canon EOS R5 · 2 h") joined the Latin run and split its own field; fields are now separated with right-to-left marks (left-to-right ones in the preview's frame data). Charts no longer switch to Arabic-Indic digits after Arabic words. Desktop file sizes use the language's units; table cells holding a size, speed or time left read in the language's direction and fit their column.
  • CI: a full artifact storage failed jobs whose checks had passed; uploads no longer fail a job, artifacts expire in days, and screenshots reach their branch through the cache.
  • Release workflow: signing and the public-repository publish never ran (step conditions cannot read the step's own secrets); the publish job lacked the .NET SDK for make-icon; re-publishing an existing tag failed; the portable zip lacked portable.txt.
  • CI fixtures step reported missing RAW samples as a silent pass; it now emits a warning per missing sample.
  • Release workflow: the server image and its :latest tag were pushed before the Windows tests passed, and publishing an older tag again moved :latest back; the image now follows the tests and :latest the newest stable release. Stable release notes started at the last release candidate instead of the previous stable tag.
  • Settings › Rendering quality, parallelism and background priority had no effect on jobs. Parallelism Auto is now half the logical cores (owner review 1 makes it the photos the running job processes at once; see Changed).
  • A restore staged in Settings › Advanced now runs before the index opens, so it recovers an index that no longer opens. A copy failing quick_check is set aside before any backup is made, and a restored copy is never applied twice.
  • Settings › Archive could not save archives with an empty name (made at a drive root), showed sizes outside the presets as 1920 or 320, and kept editing the previous archive after the archive folder changed. Reset all and Import settings rewrote the shared archive.json.
  • Server archive settings were lost on restart and ignored by proxy jobs. A damaged archive.json returned 500, and a busy one reported a busy sequence.
  • The naming preview's sample title follows the interface language.
  • Server (Windows): the data folder, the temporary admin password and the TLS key were readable by every local user, and any user could plant server.json. The data folder is now private to SYSTEM, Administrators and the server's account. The server refuses to start when anything in it belongs to another account or is a link, reads server.json only after checking its owner, and creates secrets anew with a private ACL.
  • Server (Windows): a new archive folder under C: let every local user change it, including the folders where the service stages uploads. It now lets only administrators and the server change it (other users may read it), and the log reports an existing archive that others can change.
  • Server (Windows): --install-service registered a LocalSystem service from any folder. It now refuses outside a Program Files folder that only administrators can change.
  • Server: in TLS mode proxy, any client could set X-Forwarded-For to escape the sign-in rate limit and forge its audit address. Now only the proxies in CHRONOS_TRUSTED_PROXIES (default: loopback) may. The admin panel shows Not via proxy, and a refused sign-in names the setting.
  • Docker: a fresh, root-owned archive folder made the container crash-loop. An empty or root-owned archive root is now given to PUID:PGID. Any other archive, or .chronos folder, that this user cannot write stops the container with a one-line error. CHRONOS_DATA_PATH is honoured, and Kestrel no longer warns about overridden addresses at every start.
  • Docker and Unraid: remapped host ports broke the admin panel's HTTPS redirect; CHRONOS_PUBLIC_HTTPS_PORT and CHRONOS_PUBLIC_HTTP_PORT fix it, and the LAN beacon uses them too. The Unraid template's default database host could not resolve on the bridge network. --help now lists every variable, the TZ description is accurate, and the README explains host networking for LAN discovery (with PostgreSQL on loopback).
  • Tests: four tests switched the UI language outside a serial class and could break English assertions running beside them; such a switch now fails the test that made it, after English is restored. The frame window's unobserved-exception check runs alone and counts only its own exceptions.
  • Test runner: a skip or failed assertion in a test class's constructor was reported as "Exception has been thrown by the target of an invocation"; it now keeps its own outcome and message.
  • CI: a crashed screenshot capture passed once it had written one picture; the MediaPlayer playback check could end the whole Windows test program on an exception.
  • Render nodes: frames missing, corrupt or excluded on the server, and originals it had not hashed yet, failed the whole job; they are now skipped or fetched with a size check, and damaged frames the server's own render skips are skipped too.
  • Sync: a download whose partial file already held the whole original failed on every retry; an edit saved on this PC while an automatic comparison ran could be replaced by the server's version unseen.
  • Sync: proxies the server asked for were neither counted nor sent for adopted (unverified) folders, and the count did not follow proxies sent from elsewhere.
  • Mass rename and move: sequences inside a moved folder kept their old paths in the index; they now move with it (listed in the dry run, locked while it moves), and a target inside a sequence folder is refused.
  • Archive maintenance: compacting a thumbnail pack could drop thumbnails a thumbnail pass was writing.
  • Server: proxies of HEIF frames made from embedded previews by earlier versions are asked for again on the next proxy pass; a proxy sent for a frame the server decodes itself, or a preview-sized one, is refused.
  • Exiting while signed in to a server no longer waits for the UI thread that the exit blocks. Closing the window stops the editor's player and runners before the index and the engine close. Cancelling or closing the sign-in dialog no longer leaves a half-signed-in session or crashes on a late answer.
  • The usage counts consent, which promised a daily count that nothing sends, is offered only in builds with ChronosAdsEnabled=true (BRIEF §21).
  • Error toasts say why an action failed in the app's language instead of showing .NET or OS text: a job posted to the server from the Library (including a server whose certificate changed since sign-in, which used to fail silently), the card check and delete after an import, and a failed editor save. The editor's own reasons (frames changed on the server, settings changed elsewhere at the same time, sequence gone, project unreadable or from a newer Chronos, account without edit rights) are no longer reported as 'The server does not answer'.
  • Texts that count now agree with the number: advisor findings and reasons, the rendering cores and the upload-all tooltip have singular and plural forms ('1 frame is black'), a single exposure change has its own finding, and tallies and job log lines read as labels ('Copied: 40 · failed: 1', 'Missing: 2 · damaged: 1'). i18n check fails on a text that puts a count before a counted noun without plural forms, or that tallies counts with participles.
  • Archive: picking another archive while the screen was reading no longer leaves the previous archive's tree, figures or search results actionable. Until the picked archive is read, Scan, Verify, Rebuild and the set actions wait and say so, and the rebuild dialog shows the picked archive's own state and count.
  • Import: a job that ended before the screen recorded it (a destination gone offline) no longer leaves the screen copying forever, and a replaced import result is released.
  • Library: the timeline's month readouts follow a language switch.
  • Server: a sequence created with POST /api/v1/sequences (the Uploader's "New sequence") could not take uploads. It now gets its folder with its first photo, named by the folder template like an upload-created sequence.
  • Server: "already on the server" was also answered for copies flagged missing or corrupt or whose file was gone, and the Uploader then deleted its source. Only an intact copy answers now. An upload of the same content repairs the damaged frame in place, and the damaged file is kept beside it as <name>.damaged-<time>. The Uploader deletes a source only after a verified upload.
  • Archive: excluded frames, the hero frame and virtual-camera keyframes stayed at their positions when earlier photos were inserted (late uploads, sync downloads, scans of added photos), so renders dropped the wrong frames. They now move with their frames, also when a scan finds the photos while a project is being saved.
  • Server: render outputs could be uploaded by upload-only accounts. Offering and completing a render upload now needs the Edit permission.
  • Server: an upload resumed after the chunk size setting was lowered had its chunks refused (413). Chunks keep the size their upload started with.
  • Server: the render_exists error had no message. i18n check now requires a message for every server error code.
  • Server: re-uploading a photo to repair its damaged copy counted against the uploader's quota again, and could be refused for quota. Repairs are free now. A copy in another sequence no longer stands in for the damaged one of the target sequence, neither at upload start nor after a crash.
  • Server: one account's card import reports could push job and sequence events out of every client's event queue. They are now limited to one progress report a second and 120 events a minute per account. A card import into the server no longer waits up to 20 s for an unreachable server when it ends.
  • GoPro GPR frames decode through the camera's JPEG even after an import or upload kept the JPEGs in a sequence of their own.
  • A file locked by another program or a network hiccup during analysis or a render no longer flags the frame undecodable, and neither does a file deleted after indexing. A resumed deep analysis checks such frames again.
  • A small render (720p, GIF) no longer lowers the output cap of preview-only sequences to the small preview it decoded.
  • Thumbnails of RAWs whose embedded preview is smaller than 320 px are developed from the RAW.
  • Frames both analyses could not decode are listed once, with the right count, including missing and corrupt ones.
  • Render logs no longer name frames a speed-up skipped as shown by a neighbour.
  • A failure while probing the SQLite compile options can no longer stop the index from opening.
  • The PostgreSQL server writes decode tiers in batches instead of one statement per frame. The 0003 migration headers name the right mirror file.
  • Upgrading or uninstalling while the Chronos Server service ran failed, and uninstalling left the LocalSystem service registered. Services of the installed programs are now stopped once the programs have closed, started again after the upgrade (or when it fails) and deleted on uninstall. Only services, sign-in entries and shortcuts that start a program setup installed are changed, so installing into a shared folder never touches other programs'.
  • The Server component installs only for all users in a folder inside Program Files, where standard users cannot replace the program its service starts.
  • Uninstalling left the Uploader's and the client's sign-in entries behind. 'Remove my data' deleted the whole data folder, also another installed Chronos product's index or upload queue, and now removes only the data of programs no other installed product uses. Uninstalling one product deleted the other's Uploader shortcuts; they now move to the copy that stays.
  • The .chronos file type name is written in setup's language.
  • Titled notes with no listed changes now say 'Maintenance release.'.
  • Server: HEIF frames whose proxies had arrived stayed preview-only, capped at the size of their Exif thumbnail (often 160 × 120); the advisor then chose the 720p preview preset and the Library showed the preview-only chip. A frame read from its proxy is no longer preview-only, an arriving proxy has the sequence analysed again, and a deep analysis or render that was running when the proxies came no longer flags the frames again.
  • Proxies: renders, analyses, the server's manifests and proxy downloads used a proxy another photo had left at a frame's place after frames were inserted before it, or a thumbnail-sized one made by an earlier version. Proxies of frames whose capture time has a seventh sub-second digit (file-system times) were rejected. One rule now decides: the frame's own capture time, to the microsecond, and the policy's size for frames the host reads through proxies.
  • Server: frames added to a sequence whose proxies had all arrived were not asked for their own until the next proxy pass. Manifests now read a proxy's header only when the file has changed.
  • Deep analysis: a run resumed after proxies arrived kept its checkpoint's 'unreadable' verdicts, and a saved report kept listing frames that now decode from their proxies (warning badge, exclusion advice). Such checkpoints start over, such reports no longer count, and the server no longer sends them to remote editors. Checkpoints written by earlier versions still resume.
  • Thumbnails drawn from proxies ignored the archive's thumbnail size. Below 320 px they were drawn again and appended to the pack on every pass; above 320 px they were never redrawn at the larger size.
  • A proxy or thumbnail pass resumed after a pause, crash or retry could mark the sequence complete while frames before the resume point lacked them, so the server stopped asking for those proxies. The proxy pass log now counts frames waiting for a proxy apart from unreadable ones.
  • Sync: exclusions, the hero frame and camera keyframes taken from the server, or sent to it, could land on other photos. This happened when the photos they were written against arrived later (a download or upload of earlier photos), when photos were imported or uploaded while the settings were compared or sent, or when a hero was chosen while another waited for its photo. Settings now move between copies translated to the receiving copy's frames, matched by photo and read as one state with those frames. Positions on photos that copy does not hold yet wait in the project until the photos arrive.
  • Server (PostgreSQL): settings saved through the API while photos were being filed into the sequence could be silently undone.
  • Sync: downloading a sequence the server adopted but never verified failed with 'did not arrive intact'. Its originals now download, are size-checked and hashed on this PC, and are matched to this PC's copies by size and capture time. They are not uploaded again, and their proxies are sent like the others. A photo the server's folder holds twice no longer leaves Download on offer forever.
  • Sync: when the server's settings version moved without a change, the comparison failed while the local archive was offline or read-only. Identical settings are now agreed without touching the archive.
  • Sync: thumbnails cached in place of proxies the server did not have yet were kept after the server got its proxies. The next proxy download replaces them, and offline caches refresh when the proxies arrive.
  • Server: a sequence created with POST /api/v1/sequences that was still waiting for its first photo could lend its id to loose photos at the archive root during a scan, also when the root's sidecar was damaged. The Uploader's uploads then went to the root. A rebuild could delete such a sequence, and maintenance compacted the root's thumbnails under its id. Scans, rebuilds, the orphans check, verify and maintenance now leave it alone until its first photo gives it a folder. A root sidecar the server may not read now fails that folder only, not the whole scan.
  • Server: sequences found by a scan now get thumbnails and proxies like uploaded ones. This covers photos already in an adopted folder, folders another writer adds and photos added to a folder. Each sequence gets its own background pass, so a folder renamed or deleted before its pass no longer leaves the others without images. Remote cards and proxy downloads no longer stay empty until an administrator queues the jobs.
  • Mass rename and move: a folder that gained a sequence after the dry run (a scan of photos copied into a subfolder) no longer moves while that sequence's index row stays behind. Such a plan is refused as out of date.
  • Server: sending a large chunk again after a lost answer, to an unknown or finished upload, or with a session that had signed out or lost the upload permission closed the connection instead of answering. The answer now reaches the client, including 401 and 403, so the Uploader asks the user to sign in instead of retrying.
  • Server: PUT /api/v1/archive/settings answered 500 when archive.json held a thumbnail size or proxy quality out of range. It now answers 409 archive_policy and names the policy to correct.
  • Desktop archive polling started only at launch, for the archive folder known then. An archive created on the first run, chosen in Settings or relinked was never rescanned, and Rescan every had no effect, until a restart. Polling now follows the archive folder and survives a drive that is offline at start.
  • When the server or another PC changed an archive's templates, proxy size or name in archive.json, imports, uploads and the admin panel kept the old values until a restart, and saving the admin panel's archive form wrote the old templates back. The index now follows the file whenever Settings › Archive opens and at every rescan, on the desktop and on the server, and the admin form reads the file when it opens. A poll no longer puts back values an administrator had just saved, and an archive.json written by a newer Chronos is no longer registered.
  • Settings › Archive saved 320 as this PC's thumbnail size whenever the open archive was offline. With an empty archive folder it edited the first archive of the index. It silently dropped an edit made while a newly chosen archive was still opening, and an edit made just before a second folder change went to the archive of the later folder.
  • Start with Windows (client and Uploader) stayed ticked after Task Manager disabled the entry, and ticking it again could not turn the entry back on. The Uploader now also reads it again when its settings open.
  • A portable copy unpacked in a OneDrive folder, including a Desktop or Documents folder backed up by OneDrive, kept its SQLite index there. It now keeps its data in %LOCALAPPDATA%\Chronos, says so at start and names the data it kept beside it before, which it leaves untouched. A per-user data folder that OneDrive synchronises is now reported with a warning.
  • Import: a photo that could not be copied, a card file the plan could not read, and every photo a stopped import never reached now say why in the app's language: destination full, access denied, card gone, the server's own reason, server not answering, untrusted certificate. The system's text no longer appears, on the Import screen or in the import job's log.
  • Errors: a server behind a reverse proxy that answers with its own error page (for example 502 while the server restarts, or 413 for a large photo) now reads 'The server does not answer' or 'The request is larger than the server accepts' instead of the English 'Bad Gateway' or 'Request Entity Too Large'.
  • Archive: picking another archive no longer leaves the previous archive's tiles, charts, last scan, recent runs and rail badge on screen, or its maintenance actions enabled, while the new one is read or after its read fails.
  • Jobs: a render that ends while its sequence is open in the editor always gets exactly one completion toast, with Show in folder. It still gets one, without Show in folder, when the editor cannot read the sequence again or closes as the render ends, and a render the editor did not follow is announced too.
  • Jobs: Open on a job over several sequences (a deep analysis of a Library selection), from its toast, its Windows notification or its row, shows all of them selected in the Library instead of opening only the first in the editor. A job over a whole archive opens the Library without freezing it.
  • Server (Windows): a service that refused its data folder or server.json stopped with error 1064 and gave no reason anywhere. Start and stop failures now go to the Application event log (source ChronosServer) with the entry to check. The README gives the full steps to fix a data folder created by a console run.
  • Docker and Unraid: the CHRONOS_TRUSTED_PROXIES guidance suggested trusting a whole Docker network (Unraid: 172.17.0.0/16), whose gateway carries relayed IPv6 and host connections. It now names the proxy container's own fixed address and warns against subnets, the gateway and the default bridge.
  • Setup: an upgrade that renamed the client or the Uploader deleted its sign-in entry, which silently turned Start with Windows off. The entry now moves to the new program with its options, and is removed only when the component is dropped.
  • Setup: uninstalling removed the client's and the Uploader's sign-in entries but left the on/off state Task Manager keeps for them. It now goes with the entry; an entry an upgrade moves keeps it.
  • Server: when a desktop sharing the archive imported photos into an existing sequence, they got no thumbnails or proxies, and the frames they renumbered answered 404, until a pass was forced by hand. A scan now notices when the sidecar's frames changed behind the index and queues the analysis and both image passes. Analyses, image passes, verify, maintenance, rating and tag edits and renames no longer hide such a change, or a new title set on another PC, from the next scan.
  • Server and desktop sharing an archive: each poll of the desktop after the server's uploads made the server analyse that sequence and draw its thumbnails and proxies again.
  • Archive: a rating, tag or title edit, a mass rename, a decode tier recorded by an analysis, image pass or render, or a verify stamp wrote the whole sequence row back from what had been read earlier. Settings saved meanwhile (Auto values, a server PATCH) were lost, or the frame count went back. These writers now write only their own fields, under the sequence's lease.
  • Server: a thumbnail or proxy pass that read the frames before a scan found photos added to the folder marked the sequence complete again, so no pass was queued for the new photos. A pass now marks it only while the sidecar still lists the photos it read, or while the next pass is already queued.
  • Server: when a camera uploaded faster than a pass could finish, the sequence offered no thumbnails or proxies until uploads paused.
  • Server: after a verify gave photos their first hash, their thumbnails answered 404 and the next pass drew them all again. They are served and reused again.
  • Deep analysis: when the proxies of frames a run had found unreadable arrived during that run, it still recorded the deep status, its warnings and its advice, which every later reader discards. It now checks its report against the frames as they are when it records.
  • Rebuild: a sequence that got its first photo, or was created by an upload session, while a rebuild ran lost its index row. The Uploader got 404 and the session continued in a second sequence. A row is now checked again under its lease before it is removed.
  • Editor: an edit made while earlier photos were being added to an open sequence of this PC (a sync download, an import, a scan) was saved in the old frame order, so other photos were excluded and the hero and keyframes moved. The editor now reads the frames again before it saves and moves its positions with them, undo steps included.
  • Editor: a render queued after earlier photos were added to an open sequence, without another edit, rendered the old frame order. Render now reads the frames again first. On a server sequence whose frames changed it is not queued, and the editor goes read-only with Get proxies.
  • Editor: when earlier photos were added to an open sequence, the frame on screen became another photo, and Exclude, Set hero and keyframes applied to it. The frame on screen now stays on its photo.
  • Editor (server sequences): a save made after the server's frames changed, before the editor heard of it (scans send no event; events are lost while reconnecting), overwrote the positions the server had moved. Saves now name their frames version, so the server refuses them and the editor goes read-only with Get proxies. Frames are compared again after a reconnect.
  • Sync: when a server folder held the same photo twice, settings crossing to this PC put every exclusion, hero and keyframe of that photo on its first copy, and a later push changed the server's. Each copy now keeps its own positions, and a photo not hashed yet is never taken for another of the same size and capture time.
  • Sync: with Keep proxies offline, thumbnails cached in place of proxies were never replaced when a server sequence could not get all its proxies (a missing or damaged frame). They are now downloaded again once the server offers their proxies, and again after a failed download.
  • Server: a proxy upload above 1 MiB that the server answered without reading it closed the connection, and clients retried a network error. This happened when the session was signed out, the Edit role was lost, the frame was gone or the proxy was not wanted. The answer now reaches the client.
  • Settings › Archive: a template, proxy size or thumbnail size committed while the archive's settings were being read could be shown reverted, together with this PC's default, while the archive kept the edit. This happened on a share that was waking up. Such a read now waits for the edit and is made again, so choosing another archive folder also finishes only once the rows show that folder's archive.
  • Import: a server import that could not stage its copies on this PC said the destination was full or refused writing. It now names this PC's cache folder.
  • Import: a failed archive import showed the job's raw system or database text as its line. It now shows the archive's own sentence or 'The import stopped before the end.', with the text in the tooltip.
  • Portable copy: a SharePoint or Teams library synced with OneDrive's Sync button was not recognised as synchronised, so a portable copy unpacked there kept its index inside it. Its data now goes to %LOCALAPPDATA%\Chronos.
  • Docs: the Windows README's recovery of a data folder created by an unelevated console run keeps the folder private. It no longer resets the permissions, which let every user read the TLS key and the temporary admin password, and it removes the account the service names, also when a standard user ran the console.
  • Analysis: an analysis that read a sequence's frames before photos were added, removed or renumbered (an upload, an import, a scan) recorded its advice, warnings and analysed status for the old frames. Its Auto frame rate and warning count overwrote a newer light analysis, and the new photos were never analysed. It now records nothing in that case, unless a light analysis is already queued to replace what it records. It says so in the job log and queues a light analysis of the photos as they are. When another Chronos sharing the archive changed the photos and this PC has not indexed them yet, the analysis no longer writes its old results over that Chronos's in the shared project and sidecar. The next scan indexes the photos.
  • Sync: with Keep proxies offline, a proxy download made while the server was still drawing newly uploaded photos cached thumbnails or nothing for them, and marked the cache up to date. The editor then kept those thumbnails or blank frames until the frames changed again. The cache now stays stale until every photo, missing or damaged ones aside, has its proxy. It is downloaded again once the server's sequence changes.
  • Sync: a title, tags, rating or notes taken from the server could hide photos that another Chronos sharing this PC's archive had just added to the sequence. Scans then skipped the folder until it changed again. The next scan now indexes them.
  • Server (PostgreSQL): an analysis, image pass or render could record decode tiers or light statistics while a scan or rebuild indexed the same sequence. The two could deadlock, and one of the jobs failed. Every writer now takes the sequence before its frames.
  • Server and desktop sharing an archive: an upload could reach a sequence after the desktop had verified it or changed its tiers, duplicates or title. The upload indexed only the new photo and hid the rest from later scans. Until a full scan, the server kept those photos unhashed and counted missing or corrupt ones as healthy. Such an upload now indexes the whole sequence from its sidecar.
  • Uploads, and photos a desktop downloads from its server: a photo copied by hand into the sequence's folder since the last scan stayed out of the index. Only a full scan or the orphans check found it, because later incremental scans skipped the folder. The next scan now indexes it.
  • Scans and rebuilds: a scan could read a sidecar just before an upload or another writer committed, then write it over the newer index. That dropped the new photos or tier flags, and incremental scans then skipped the folder. The scan now reads the sidecar again under the sequence's lease. An older sidecar put back by hand or by a file-sync tool no longer takes the frames version back.
  • Verify: flags recorded while a verify ran (a preview-only or undecodable photo, a duplicate) were undone by it. A photo an upload repaired during the run was flagged missing again, so the next copy moved the good file aside. A verify now writes only the missing and corrupt flags, and only for files still as it checked them.
  • Tags: on PostgreSQL, two writers of sequences sharing tags could deadlock, and one of them failed. This happened when an edit through the API listed the tags in another order than an upload, a scan or another edit. Tags are now written in one order, and an unchanged set is not written again.
  • Archive › Orphans: photos the fix indexed were never analysed, so the Auto values, warnings and preview-only count stayed those of the old frames. The fix now clears the analysis and image flags as a scan does, and queues the light analysis. When Settings › Analysis › After a scan is on, it also queues the background deep analysis, as a scan does.
  • Deep analysis: a pass overtaken by new photos replaced the current analysis.json that a later pass had made. The report endpoint then answered 404, and the next light analysis dropped the deep status. The report is now saved under the sequence's lock, and never over a more current one. A pass that leaves the saved report in place records that report's summary and warnings, not its own.
  • Scans of an archive shared by two Chronos indexes (a desktop and its server): one index could write a sidecar at the frames version the other held for other frames. The other index then took the frames as unchanged. When the versions match, a scan now also compares the frames themselves and writes changed ones one version past its row.
  • Orphans fix and scans: a photo that could not be read yet (held open by another program, no read permission, a OneDrive placeholder offline) made the folder count as changed, so every fix rewrote the sidecar, cleared the thumbnails, proxies and analysis status and queued the analyses again. It now changes nothing until the photo can be read.
  • Deep analysis: a pass that started before a proxy arrived could overwrite the report a later pass made from the proxies and drop the sequence's deep status. It now leaves a report that describes the frames as they are in place.