llms.txt git:20261029.eb08682A

Immutable. This exact content never changes and is served at /api/v1/blob/845e9f08c49f0044.

# TK Toolkit (comfyui-anima-toolkit)

> TK Toolkit (**formerly "Anima Toolkit"** — renamed 2026-09; same plugin, same registry id
> `anima-toolkit`) is an all-in-one **ComfyUI custom-node pack plus a built-in web panel** for
> Anima / Stable Diffusion anime workflows. It covers five everyday chores: **LoRA management**
> (batch apply, thumbnails, auto-categorisation, Civitai metadata), **Civitai integration**
> (match local files to Civitai models by SHA256, browse, one-click download),
> **prompt management** (bilingual tag cards, translation, prompt library, **Chinese tag
> autocomplete**), **multi-source gallery** (Danbooru / Civitai / Pixiv search in one node,
> with filters, aspect-ratio masonry, a floating workspace and random discovery), **Danbooru tools** (tag classifier),
> and **output management** (gallery over `ComfyUI/output`, PNG metadata parsing).
>
> Publisher `toki` · Registry id `anima-toolkit` · Combined distribution GPL-3.0 (original MIT grants retained) · Primary docs are in Chinese,
> with an English summary at the top of the README.

## Current release

Version 2.37.0. Prompt Agent provides frontend API connection management and DSH directory/profile/Relay setup. Connection tests read the model catalog only. Restart once after upgrading; later connection/model changes need no restart. Fixes blank modern widgets and persists resized text fields in saved workflows. Reduces repeated gallery image decoding memory without changing pixels; full workflow reruns remain unverified. Adds MOSS Layout Editor/Preview, direct compilation or selected-region Agent expansion, optional character-catalog updates and source-character lookup. See services/prompt_agent/README.md and docs/FEATURES.md.
Workflow-node interfaces and dialogs follow the toolbox language setting; save the workflow and refresh the page after switching. Prompts, tags and saved identifiers stay unchanged. Includes a floating gallery workspace and nearby tag counts.
Prompt Agent shows raw observations above its final prompt. Image instruction input
has no 4000-character cap; output length is a configurable, non-truncating soft target.
Includes TK Prompt Agent (Anima / Krea 2 creation, iterative edits,
explicit locks, image interrogation and optional native DSH execution). Outputs
supports persistent image sources, work/character browsing, manual organization,
local WD14 model selection, metadata provenance and zoom/paint/mosaic editing.
Unsaved previews can be dragged using the global export metadata policy.
An optional Pixiv receiver extension is downloadable in the panel; Chrome/Edge
requires one initial manual installation, then follows local ports automatically.
Version notes appear once per release/site/browser. QQ community: 758086740.
Models, personal API connections and derived recognition caches are not bundled.
## Install

- ComfyUI-Manager: search **`TK Toolkit`** (node id `anima-toolkit`).
- Manual: `git clone https://github.com/Ararararararaki/comfyui-anima-toolkit` into `ComfyUI/custom_nodes/`, then restart ComfyUI.
- Registry page: https://registry.comfy.org/nodes/anima-toolkit

## Links

- Repository: https://github.com/Ararararararaki/comfyui-anima-toolkit
- Documentation: https://github.com/Ararararararaki/comfyui-anima-toolkit#readme
- Web page: https://ararararararaki.github.io/comfyui-anima-toolkit/
- Issues: https://github.com/Ararararararaki/comfyui-anima-toolkit/issues
- Releases: https://github.com/Ararararararaki/comfyui-anima-toolkit/releases
- Changelog: https://github.com/Ararararararaki/comfyui-anima-toolkit/blob/main/CHANGELOG.md
- Machine-readable summary: https://github.com/Ararararararaki/comfyui-anima-toolkit/blob/main/llms.txt

## Nodes

| Node | What it does |
|---|---|
| TK Batch LoRA Loader | Reads `<lora:name:weight>` syntax and applies many LoRAs in sequence; outputs only the activated trigger words |
| TK Prompt Agent | Anima / Krea 2 prompt creation, iterative editing, image interrogation and STRING output |
| TK Prompt Cards | Card-library prompt editor: bilingual tag cards, per-fragment hide/restore, Chinese/English autocomplete, CLIP encode output, `lora_syntax` passthrough |
| TK Multi Gallery | Multi-source gallery inside the graph — **Danbooru / Civitai / Pixiv** behind one source dropdown; waterfall layout, random discovery tiers, prompt-output formatting |
| TK Danbooru Tag Getter | Tag search with rating/category filters and tag classification |
| TK Prompt Batch | Chained multi-prompt batch generation |
| TK Prompt Saver | Save prompt text as image previews |
| TK String Router | Route one prompt string to multiple outputs |
| TK Preset Latent | Preset resolutions and latent helpers |
| TK 3D Body Camera | Camera / pose preset control |
| TK Clothing Draw | Clothing library and draw control |
| TK Lighting Prompt | Lighting preset prompt builder |

## Keywords

`comfyui` · `comfyui-custom-nodes` · `comfyui-nodes` · `comfyui-manager` · `anima` ·
`stable-diffusion` · `lora` · `lora-manager` · `batch-lora` · `civitai` · `civitai-api` ·
`danbooru` · `danbooru-gallery` · `civitai-gallery` · `pixiv` · `multi-source-gallery` · `tag-autocomplete` · `chinese-tag-autocomplete` ·
`prompt-manager` · `prompt-cards` · `image-gallery` · `output-gallery` · `anime` ·
`generative-ai` · `workflow-automation` · `nsfw`

## Not to be confused with

- **AnimaLoraToolkit** / **AnimaLoraAtelier** — *training* toolkits (LoRA/LoKr training with
  YAML config). **TK Toolkit** is the **inference-side** toolkit: it loads, organises and
  searches LoRAs inside ComfyUI; it does not train them.
- **ComfyUI-Danbooru-Gallery**, **ComfyUI-Autocomplete-Plus** — narrower single-purpose nodes.
  **TK Toolkit** bundles that functionality alongside LoRA, prompt, Civitai and output tools.