Repository navigation
Feat/new design - #10
Merged
Merged
Conversation
- styles/ds.css: the mockup stylesheet ported 1:1, every class prefixed with ds- so it lives next to the legacy primitives without collisions. Element resets are dropped (Tailwind preflight covers them). - tokens.css: semantic tokens take the mockup values (warm neutrals, lime accent for fills, deep green --accent-text for text, radii 7/10/16/20, light warm shadows, -1% tracking). Legacy primitives (.btn*, .card, .badge, .seg, .tab, .input, .kbd, .caption) follow the same look; ghost buttons get a hairline frame. - text-accent utilities and direct `color: var(--accent)` uses read --accent-text, so the light lime never ends up as text colour. - Shell as in the mockup: one white panel on the canvas, 236px sidebar with the collapse icon next to the name, project card, ВЕРСИЯ pill, step rail with round nodes and real counts, topbar with GPU · VRAM · CPU · RAM, task chip and runtime chip. All sidebar behaviour is kept (cursor advance/skip, version switch/new/export/delete, read-only outside the project, phone drawer, collapse). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… RAM) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Page head, search + sort + All/Active/Archive toolbar, four KPIs, project cards with the 5-step bar and continue action, and "What is happening right now". Everything is real data: - KPIs: active projects (+ created this week), queued/running training tasks, GPU hours from task durations (30 / 7 days), LoRA checkpoints across projects (+ added this week). - Cards: version status or queue position from the existing forecast, image and version counts, live step/total while training, start time while queued, checkpoint count when done. - The ⋯ menu archives / restores (existing API) or deletes. Sort, bundle import, create and delete behave as before. New ds/PageHead and ds/KebabMenu components; ru/en strings. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Page head (edit project / new version / continue), the 7-stage pipeline card with live training progress, four KPIs, a Tasks / LoRA files table and a right column with versions, set composition and top tags, all from real project data. The status-banner actions moved to each task's menu (pause, cancel, log, monitor); the full output folder stays reachable from the file menu. The kebab menu width is set inline so the mockup's card reset can't collapse it. New strings are ru/en only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Booru scraping (danbooru / gelbooru through the existing estimate and download endpoints, exclusion chips on the shared download setting, a limit stepper) next to the file import card, download stats with the format split, and the downloaded images as a grid or a list with a bulk bar and "in train" pins for the active version. Background jobs show in the new data-jobs strip at the bottom, with the log one click away. ImageGrid cells now use the mockup thumb (checkbox mark, pin, caption) on every page, and restore their src after StrictMode's dev re-mount, which used to leave thumbnails blank in dev. The download listing also returns each file's mtime so the grid can sort newest first. Also fills the Russian strings of the download page that were missing. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Step head with the Train set / Validation switch and the next-step action, four KPIs (dataset, train set with its groups, held-out validation, effective samples per epoch from the folder repeats), and the dataset pool next to the version's set, each with its own footer actions. Groups are chips with delete and a "+ New" chip; renaming stays inline. "To validation" moves selected train images to the held-out set after a confirmation that spells out that the train copies and their edits go. Sorting, Alt-hover preview and the keyboard preview are unchanged. Adds the Russian curate strings, which were missing entirely. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every version step now uses the mockup frame: StepShell renders the page head with the green step eyebrow, and the job log drawer is the mockup's bottom strip (dot, job, last line, expand) with the same behaviour. The preprocess tools share one frame: common head tools (restore all processed images, next: tags), four tool cards with each tool's state from the train manifest, and one card with the tool tabs. Upscale and Duplicates follow their artboards (settings column + images / group cards); Overview, Crop and Inpaint keep their editors inside the tabbed card with their actions in a toolbar row. Disabled ds controls now dim. Adds the Russian strings for all preprocess tools, which were missing. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Tag statistics, image grid and active image side by side, with the green unsaved bar at the bottom. Picking a tag selects its images; remove and replace work on the picked tag. Add and dedupe for the selection moved to a bar under the grid. Restore points open from the bottom bar. Russian strings for tagEdit, tagEditor and bulkAction. The autotagging card is left out: this fork has no train tagging job. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Status card with the four set numbers, a tabbed card (generate / images) and the build log beside it. Source picked with option cards; booru and AI prior settings as field rows with their parameter keys; excluded tags as chips with "+ tag" and the frequent train tags to click out. Top up and start-over are the two footer buttons (start-over asks first when a set exists). The images tab shows the grid with folder chips and the picked image with its caption. Russian strings for the reg namespace. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Config bar (version config, save state, change config, preset actions), four KPIs (samples per epoch, total steps and time, peak VRAM, live system load), the step budget from files to optimizer steps, the form in four tabs (model / data / training / system & output) with a field filter and section reset, the live preview (epoch mix, steps, memory, buckets / YAML), the config map and this session's changes with revert. The latest run of the version sits in the bottom strip. Schema fields render as the mockup's rows: label, parameter key, badges, description, control on the right; a changed field gets a dot and its old value. Russian labels for all schema fields and groups. The trigger word is a field row: gated under DOP when DOP is on, otherwise on the data tab. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Rows keep the mockup layout (name, key, control); the explanation and the recommended value move into a tooltip on the setting name, portalled so card overflow never clips it. Applies to schema fields, the trigger word row and the reg-set rows. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Summary card (count, plan bar to the end of the queue, current finish), training task cards with status, time, actions and notes, drag to reorder waiting tasks, and a data task table (download, preprocess, tagging, reg build, eval, upload). Kebab menus now portal so card overflow never clips them. Full Russian strings for the queue namespace. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Page head with status line and actions, tabs with a stat strip, overview grid (curves, log, samples, properties, outputs), metrics tab, and the outputs / config snapshot tabs restyled with the same logic. Time formatting shared with the queue page. Russian and English strings for the whole task page. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Page head with the single / XY switch and the generate button; three columns: settings (prompt, negative, LoRA slots, size / steps / guidance / seed / sampler, aspect and model), result card with a status footer, and the history thumbs; the daemon status strip at the bottom. Generation logic is unchanged. Russian strings for the whole page. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Presets table (type, rank, res, optimizer, updated, row menu) with a filter, and a summary card with key values and the pruned YAML / TOML. Import and export live in the head menu; the full editor keeps the schema form. Shared YAML highlighter moved to components/ds. Russian strings for the page. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Checkpoint metrics overview for a training task: evaluation progress note, KPI cards for the best checkpoint by DINO-I, checkpoints table (or per-metric chart) with selection for a manual evaluation, the DINO-I dynamics with an overtraining hint, and training samples of the best checkpoint. The live loss / lr dashboard stays on the task page (linked from the menu). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
"In the pot" card with checkpoint rows (share, network, weight slider), a drop zone for .safetensors, the compatibility note and the method / file name footer; result card and the list of merged files with a row menu. Merge logic and the source picker are unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Groups "projects and versions", "tasks and files", "actions and settings" with the match highlighted, the kind on the right and a match counter; training tasks are searchable too. Russian strings for the palette. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Header with saved/unsaved badge, section rail, flat sections with hover descriptions on field names, switches, runtime/tunnel option cards, model catalog rows and a footer with Save / Done. Adds the missing Russian settings strings and localizes the main-model names that the backend sends in Chinese. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.