Repository navigation
Legacy TagComplete
Note
Archived pre-merge history. This document describes sd-webui-tagcomplete-neo as a standalone extension, before it was merged into Prompt Studio Neo at v0.1. Kept for historical reference only β for the current, unified extension, see Home.
Tag autocompletion for Forge Neo β suggests Danbooru/e621 tags, LoRA/embedding names, wildcards, and chants as you type, with automatic trigger word injection and CivitAI lookup support.
Fork of a1111-sd-webui-tagcomplete by DominikDoom, maintained here exclusively for Forge Neo. If you are not running Forge Neo, use the original extension instead.
- Backspace no longer lags on mobile β holding the delete key no longer triggers heavy translation handlers, making deletion smooth even on older phones
- Mid-prompt LoRA keywords fixed β when using "After LORA/LyCO" insertion, trigger words now insert correctly even if the LoRA token is in the middle of the prompt (no more overwritten text)
-
Multi-word tag search β type any word from a tag with underscores and it will be found (e.g.
towardsβwalking_towards_viewer)
- Faster tag suggestions β the autocomplete list now appears more quickly, even with large tag databases like the merged Danbooru/e621 list
- Much smoother typing on mobile β typing in the prompt box no longer lags or stutters on phones and tablets
- Lower input delay β reduced waiting time between keystrokes and results appearing on screen
- Indexed search toggle β new setting to switch between fast prefix-indexed mode and legacy full-scan mode β
- Status indicator β a small colored dot in the toolbar shows when the extension is loading (orange), ready (green), or encountered an error (red) β
- Backspace no longer triggers heavy translation handlers, eliminating lag when holding delete on mobile
- Fixed "After LORA/LyCO" trigger word insertion corrupting text when the LoRA token is in the middle of the prompt
- Multi-word tag search β type any word from a tag with underscores (e.g.
towardsβwalking_towards_viewer)
- Tag search significantly faster thanks to an internal prefix index built on load
- Input debounce tightened to keep up with fast typists without causing frame drops
- Reduced DOM and memory churn during dropdown rendering
- Fixed a mouseover listener leak that could accumulate over long sessions
- Startup no longer blocks typing β the tag index is built in small chunks so the prompt box stays responsive while the extension initializes
- Extension parsers now load in parallel, cutting startup time from ~43s to ~7s on slower machines
- Status indicator dot shows loading / ready / error state directly in the Forge Neo toolbar
- Indexed search can be toggled on/off in Settings β Tag Autocomplete
- Fixed crash when inserting tags containing apostrophes
- Fixed broken bold highlight on tags with special characters
- Embedding manual refresh (
tac_forceRefreshEmbeddings) no longer throwsTypeErroron Forge Neo - LoRA list always populated from filesystem scan;
lora.available_lorasused only for alias enrichment -
get_embeddings()resolves symlinks beforerelative_to()β no more crash with linked embedding folders -
transaction()in frequency DB initialisesconn = Nonebeforetryblock β preventsNameErroron failed DB creation
- LoRA/LyCORIS completion now inserts the alias Forge Neo expects, eliminating token blink
- Respects the "Alias from file" / "Filename" setting in Extra Networks
- Full Forge Neo / Gradio 4 compatibility
- Booru tags, initialization, and embedding reload fixed
- CivitAI trigger word lookup with SHA256 cache
- "After LoRA/LyCO" insertion option
πΊοΈ Roadmap (as of the pre-merge fork β superseded by Roadmap)
- Embedding manual refresh fixed (Forge Neo removed
force_reloadkwarg) β - LoRA list always populated at startup β
- Symlinked embeddings no longer crash on model load β
- Frequency database
NameErroron first run fixed β
- LoRA/LyCORIS token matches Forge Neo's expected alias β
- Forge Neo / Gradio 4 compatibility β
- Booru tag display fixed (Gradio 4 selectors) β
- Reliable re-initialization after WebUI reconnect β
- Embedding reload hardened β
- Extension resilience after Forge updates β
- CivitAI trigger word lookup with SHA256 cache β
- "After LoRA/LyCO" insertion option β
- Backspace lag eliminated (updateRuby skipped on delete) β
- Mid-prompt "After LORA/LyCO" trigger word insertion fixed β
- Multi-word tag search (any word of a multi-word tag) β
- Prefix-indexed tag search (toggleable) β
- Tighter debounce and delete-event skipping β
- DOM batching + deferred rendering β
- Parallel extension loading (43s β ~7s startup) β
- Status indicator dot in toolbar β
- Apostrophe crash and highlight fixes β
- Mobile typing no longer lags or stutters β
(Everything planned beyond this point was carried into the unified Prompt Studio Neo roadmap β see Roadmap.)
β = added or fixed in this Neo fork Β· everything else is original work by DominikDoom
- Instant suggestions as you type, sourced from Danbooru, e621, or merged lists
- Indexed search β built-in prefix index for near-instant filtering on large tag lists; can be disabled in settings if needed β
- Keyboard navigation β arrow keys, Tab, Enter, Escape, all configurable
- Tag color coding by category, with post count for relevance
- Alias and translation search β find tags by their alternate names or translated terms
- Frequency sorting β remembers your most-used tags and promotes them to the top β
- Status indicator β colored dot in the toolbar shows when the extension is loading, ready, or in error state β
-
LoRA and LyCORIS autocomplete triggered by
< -
Embedding autocomplete triggered by
<e: - Thumbnail preview in the completion popup
-
Correct alias insertion β uses the same identifier Forge Neo expects (
ss_output_nameor filename, respecting your Extra Networks setting) so tokens never blink β -
Trigger word injection on LoRA selection β inserts activation keywords automatically
- Fetches from CivitAI if not set locally β
- Cached by SHA256 β only re-fetched when the model file changes β
- Configurable position: Start of prompt, End of prompt, Before or After the LoRA token β
-
Wildcard file autocomplete triggered by
__ - Nested folder support
- YAML wildcard format (UMI-compatible)
- Prompt preset completion for longer phrase templates stored in JSON files
- Triggered by
<c:or<chant:
- a1111-sd-webui-tagcomplete by DominikDoom β original project, all core functionality
- sd-webui-tagcomplete-neo by Eduardo Abreu β since merged into Prompt Studio Neo
- Forge Neo by Haoming02
Release
Features
- Prompt Editing
- Translation
- History & Favorites
- Tag Autocompletion
- Live Tag Colorizer
- Extra Networks
- ChatGPT Integration
- Quality Presets
- Format & Settings
Reference
Legacy (pre-merge)