Fix glob pattern matching for filenames with special characters (#335 )

Fix model path breaking with commas in filenames
Fixes #332
2026-01-27 03:29:55 +00:00 · 2025-11-11 11:20:18 +01:00 · 2025-09-09 10:04:00 +02:00 · 2025-09-09 09:57:50 +02:00 · 2025-05-08 12:43:40 +02:00 · 2025-05-05 20:44:51 +02:00
30 changed files with 871150 additions and 198667 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -1,2 +1,3 @@
 tags/temp/
 __pycache__/
+tags/tag_frequency.db
--- a/README.md
+++ b/README.md
@@ -23,11 +23,12 @@ Booru style tag autocompletion for the AUTOMATIC1111 Stable Diffusion WebUI
 # 📄 Description

 Tag Autocomplete is an extension for the popular [AUTOMATIC1111 web UI](https://github.com/AUTOMATIC1111/stable-diffusion-webui) for Stable Diffusion.
+You can install it using the inbuilt available extensions list, clone the files manually as described [below](#-installation), or use a pre-packaged version from [Releases](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/releases).

 It displays autocompletion hints for recognized tags from "image booru" boards such as Danbooru, which are primarily used for browsing Anime-style illustrations.
-Since some Stable Diffusion models were trained using this information, for example [Waifu Diffusion](https://github.com/harubaru/waifu-diffusion) and many of the NAI-descendant models or merges, using exact tags in prompts can often improve composition and consistency.
+Since most custom Stable Diffusion models were trained using this information or merged with ones that did, using exact tags in prompts can often improve composition and consistency, even if the model itself has a photorealistic style.

-You can install it using the inbuilt available extensions list, clone the files manually as described [below](#-installation), or use a pre-packaged version from [Releases](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/releases).
+Disclaimer: The default tag lists contain NSFW terms, please use them responsibly.

 <br/>

@@ -74,6 +75,10 @@ Wildcard script support:

 https://user-images.githubusercontent.com/34448969/200128031-22dd7c33-71d1-464f-ae36-5f6c8fd49df0.mp4

+Extra Network preview support:
+
+https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/3c0cad84-fb5f-436d-b05a-28db35860d13
+
 Dark and Light mode supported, including tag colors:

 ![results_dark](https://user-images.githubusercontent.com/34448969/200128214-3b6f21b4-9dda-4acf-820e-5df0285c30d6.png)
@@ -123,24 +128,43 @@ Completion for these types is triggered by typing `<`. By default it will show t
   - Or `<lora:` and `<lyco:` respectively for the long form
 - `<h:` or `<hypernet:` will only show Hypernetworks

+### Live previews
+Tag Autocomplete will now also show the preview images used for the cards in the Extra Networks menu in a small window next to the regular popup.
+This enables quick comparisons and additional info for unclear filenames without having to stop typing to look it up in the webui menu.
+It works for all supported extra network types that use preview images (Loras/Lycos, Embeddings & Hypernetworks). The preview window will stay hidden for normal tags or if no preview was found.
+
+![extra_live_preview](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/6a5d81e6-b3a0-407b-8bac-c9790f86016c)
+
 ### Lora / Lyco trigger word completion
-This is an advanced feature that will try to add known trigger words on autocompleting a Lora/Lyco.
+This feature will try to add known trigger words on autocompleting a Lora/Lyco.

-It uses the list provided by the [model-keyword](https://github.com/mix1009/model-keyword/) extension, which thus needs to be installed to use this feature. The list is also regularly updated through it.
+It primarily uses the list provided by the [model-keyword](https://github.com/mix1009/model-keyword/) extension, which thus needs to be installed to use this feature. The list is also regularly updated through it.
 However, once installed, you can deactivate it if you want, since tag autocomplete only needs the local keyword lists it ships with, not the extension itself.
-
-The used files are `lora-keywords.txt` and `lora-keywords-user.txt` in the model-keyword installation folder.
+The used files are `lora-keyword.txt` and `lora-keyword-user.txt` in the model-keyword installation folder.
 If the main file isn't found, the feature will simply deactivate itself, everything else should work normally.

-To add custom mappings for unknown Loras, you can use the UI provided by model-keyword, it will automatically write it to the `lora-keywords-user.txt` for you (and create it if it doesn't exist).
-The only issue is that it has no official support for the Lycoris extension and doesn't scan its folder for files, so to add them through the UI you will have to temporarily move them into the Lora model folder to be able to select them in model-keywords dropdown.
-Some are already included in the default list though, so trying it out first is advisable.
-<details>
-<summary>Walkthrough to add custom keywords</summary>
+#### Note:
+As of [v1.5.0](https://github.com/AUTOMATIC1111/stable-diffusion-webui/commit/a3ddf464a2ed24c999f67ddfef7969f8291567be), the webui provides a native method to add activation keywords for Lora through the Extra networks config UI.
+These trigger words will always be preferred over the model-keyword ones and can be used without needing to install the model-keyword extension. This will however, obviously, be limited to those manually added keywords. For automatic discovery of keywords, you will still need the big list provided by model-keyword.

-![image](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/4302c44e-c632-473d-a14a-76f164f966cb)
-</details>
-After having added your custom keywords, you will need to either restart the UI or use the "Refresh TAC temp files" setting button.
+Custom trigger words can be added through two methods:
+1. Using the extra networks UI (recommended):
+   - Only works with webui version v1.5.0 upwards, but much easier to use and works without the model-keyword extension
+   - This method requires no manual refresh
+   - <details>
+     <summary>Image example</summary>
+	   
+     ![edit button](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/22e95040-1d85-4b7e-a005-1918fafec807)
+     ![lora_edit](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/3e6c5245-d3bc-498d-8cd2-26eadf8882e7)
+     </details>
+2. Through the model-keyword UI:
+   - One issue with this method is that it has no official support for the Lycoris extension and doesn't scan its folder for files, so to add them through the UI you will have to temporarily move them into the Lora model folder to be able to select them in model-keywords dropdown. Some are already included in the default list though, so trying it out first is advisable.
+   - After having added your custom keywords, you will need to either restart the UI or use the "Refresh TAC temp files" setting button.
+   - <details>
+     <summary>Image example</summary>
+
+     ![image](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/4302c44e-c632-473d-a14a-76f164f966cb)
+     </details>

 Sometimes the inserted keywords can be wrong due to a hash collision, however model-keyword and tag autocomplete take the name of the file into account too if the collision is known.

@@ -295,6 +319,14 @@ If this option is turned on, it will show a `?` link next to the tag. Clicking t

 ![wikiLink](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/733e1ba8-89e1-4c2b-8c4e-2d23352bd3d7)
 </details>
+<!-- Wiki links -->
+<details>
+<summary>Extra network live previews</summary>
+
+This option enables a small preview window alongside the normal completion popup that will show the card preview also usd in the extra networks tab for that file.
+
+![extraNetworkPreviews](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/72b5473f-563e-4238-a513-38b60ac87e96)
+</details>
 <!-- Insertion -->
 <details>
 <summary>Completion settings</summary>
@@ -443,7 +475,9 @@ You can also add this to your quicksettings bar to have the refresh button avail

 # Translations
 An additional file can be added in the translation section, which will be used to translate both tags and aliases and also enables searching by translation.
-This file needs to be a CSV in the format `<English tag/alias>,<Translation>`, but for backwards compatibility with older files that used a three column format, you can turn on `Translation file uses old 3-column translation format instead of the new 2-column one` to support them. In that case, the second column will be unused and skipped during parsing.
+This file needs to be a CSV in the format `<English tag/alias>,<Translation>`. Some older files use a three column format, which requires a compatibility setting to be activated.
+You can find it under `Settings > Tag autocomplete > Translation filename > Translation file uses old 3-column translation format instead of the new 2-column one`.
+With it on, the second column will be unused and skipped during parsing.

 Example with Chinese translation:

@@ -453,6 +487,7 @@ Example with Chinese translation:
 ## List of translations
 - [🇨🇳 Chinese tags](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/discussions/23) by @HalfMAI, using machine translation and manual correction for the most common tags (uses legacy format)
 - [🇨🇳 Chinese tags](https://github.com/sgmklp/tag-for-autocompletion-with-translation) by @sgmklp, smaller set of manual translations based on https://github.com/zcyzcy88/TagTable
+- [🇯🇵 Japanese tags](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/discussions/265) by @applemango, both machine and human translations available

 > ### 🫵 I need your help!
 > Translations are a community effort. If you have translated a tag file or want to create one, please open a Pull Request or Issue so your link can be added here.
--- a/README_JA.md
+++ b/README_JA.md
@@ -410,8 +410,9 @@ https://www.w3.org/TR/uievents-key/#named-key-attribute-value
 ![english-input](https://user-images.githubusercontent.com/34448969/200126513-bf6b3940-6e22-41b0-a369-f2b4640f87d6.png)

 ## 翻訳リスト
- [🇨🇳 Chinese tags](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/discussions/23) by @HalfMAI, 最も一般的なタグを機械翻訳と手作業で修正（レガシーフォーマットを使用）
- [🇨🇳 Chinese tags](https://github.com/sgmklp/tag-for-autocompletion-with-translation) by @sgmklp, [こちら](https://github.com/zcyzcy88/TagTable)をベースにして、より小さくした手動での翻訳セット。
+- [🇨🇳 中国語訳](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/discussions/23) by @HalfMAI, 最も一般的なタグを機械翻訳と手作業で修正（レガシーフォーマットを使用）
+- [🇨🇳 中国語訳](https://github.com/sgmklp/tag-for-autocompletion-with-translation) by @sgmklp, [こちら](https://github.com/zcyzcy88/TagTable)をベースにして、より小さくした手動での翻訳セット。
+- [🇯🇵 日本語訳](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/discussions/265) by @applemango, 機械翻訳と人力翻訳の両方が利用可能。

 > ### 🫵 あなたの助けが必要です！
 > 翻訳はコミュニティの努力により支えられています。もしあなたがタグファイルを翻訳したことがある場合、または作成したい場合は、あなたの成果をここに追加できるように、Pull RequestまたはIssueを開いてください。
--- a/README_ZH.md
+++ b/README_ZH.md
@@ -13,6 +13,12 @@
 你可以按照[以下方法](#installation)下载或拷贝文件，也可以使用[Releases](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/releases)中打包好的文件。

 ## 常见问题 & 已知缺陷:
+- 很多中国用户都报告过此扩展名和其他扩展名的 JavaScript 文件被阻止的问题。
+常见的罪魁祸首是 IDM / Internet Download Manager 浏览器插件，它似乎出于安全目的阻止了本地文件请求。
+如果您安装了 IDM，请确保在使用 webui 时禁用以下插件：
+
+    ![image](https://github.com/DominikDoom/a1111-sd-webui-tagcomplete/assets/34448969/d253a4a9-71ab-4b5f-80c4-5aa926fc1fc9)
+
 - 当`replaceUnderscores`选项开启时, 脚本只会替换Tag的一部分如果Tag包含多个单词,比如将`atago (azur lane)`修改`atago`为`taihou`并使用自动补全时.会得到 `taihou (azur lane), lane)`的结果, 因为脚本没有把后面的部分认为成同一个Tag。

 ## 演示与截图
--- a/javascript/__globals.js
+++ b/javascript/__globals.js
@@ -2,6 +2,7 @@
 var TAC_CFG = null;
 var tagBasePath = "";
 var modelKeywordPath = "";
+var tacSelfTrigger = false;

 // Tag completion data loaded from files
 var allTags = [];
@@ -11,12 +12,14 @@ var extras = [];
 var wildcardFiles = [];
 var wildcardExtFiles = [];
 var yamlWildcards = [];
+var umiWildcards = [];
 var embeddings = [];
 var hypernetworks = [];
 var loras = [];
 var lycos = [];
 var modelKeywordDict = new Map();
 var chants = [];
+var styleNames = [];

 // Selected model info for black/whitelisting
 var currentModelHash = "";
@@ -35,6 +38,7 @@ let hideBlocked = false;
 // Tag selection for keyboard navigation
 var selectedTag = null;
 var oldSelectedTag = null;
+var resultCountBeforeNormalTags = 0;

 // Lora keyword undo/redo history
 var textBeforeKeywordInsertion = "";
--- a/javascript/_result.js
+++ b/javascript/_result.js
@@ -8,10 +8,12 @@ const ResultType = Object.freeze({
    "wildcardTag": 4,
    "wildcardFile": 5,
    "yamlWildcard": 6,
-    "hypernetwork": 7,
-    "lora": 8,
-    "lyco": 9,
-    "chant": 10
+    "umiWildcard": 7,
+    "hypernetwork": 8,
+    "lora": 9,
+    "lyco": 10,
+    "chant": 11,
+    "styleName": 12
 });

 // Class to hold result data and annotations to make it clearer to use
@@ -22,10 +24,12 @@ class AutocompleteResult {

    // Additional info, only used in some cases
    category = null;
-    count = null;
+    count = Number.MAX_SAFE_INTEGER;
+    usageBias = null;
    aliases = null;
    meta = null;
    hash = null;
+    sortKey = null;

    // Constructor
    constructor(text, type) {
--- a/javascript/_textAreas.js
+++ b/javascript/_textAreas.js
@@ -9,7 +9,11 @@ const core = [
    "#img2img_prompt > label > textarea",
    "#txt2img_neg_prompt > label > textarea",
    "#img2img_neg_prompt > label > textarea",
-    ".prompt > label > textarea"
+    ".prompt > label > textarea",
+    "#txt2img_edit_style_prompt > label > textarea",
+    "#txt2img_edit_style_neg_prompt > label > textarea",
+    "#img2img_edit_style_prompt > label > textarea",
+    "#img2img_edit_style_neg_prompt > label > textarea"
 ];

 // Third party text area selectors
@@ -57,6 +61,38 @@ const thirdParty = {
            "[id^=MD-i2i][id$=prompt] textarea",
            "[id^=MD-i2i][id$=prompt] input[type='text']"
        ]
+    },
+    "adetailer-t2i": {
+        "base": "#txt2img_script_container",
+        "hasIds": true,
+        "onDemand": true,
+        "selectors": [
+            "[id^=script_txt2img_adetailer_ad_prompt] textarea",
+            "[id^=script_txt2img_adetailer_ad_negative_prompt] textarea"
+        ]
+    },
+    "adetailer-i2i": {
+        "base": "#img2img_script_container",
+        "hasIds": true,
+        "onDemand": true,
+        "selectors": [
+            "[id^=script_img2img_adetailer_ad_prompt] textarea",
+            "[id^=script_img2img_adetailer_ad_negative_prompt] textarea"
+        ]
+    },
+    "deepdanbooru-object-recognition": {
+        "base": "#tab_deepdanboru_object_recg_tab",
+        "hasIds": false,
+        "selectors": [
+            "Found tags",
+        ]
+    },
+    "TIPO": {
+        "base": "#tab_txt2img",
+        "hasIds": false,
+        "selectors": [
+            "Tag Prompt"
+        ]
    }
 }

@@ -90,7 +126,7 @@ function addOnDemandObservers(setupFunction) {

        let base = gradioApp().querySelector(entry.base);
        if (!base) continue;
-        
+
        let accordions = [...base?.querySelectorAll(".gradio-accordion")];
        if (!accordions) continue;

@@ -115,12 +151,12 @@ function addOnDemandObservers(setupFunction) {
                            [...gradioApp().querySelectorAll(entry.selectors.join(", "))].forEach(x => setupFunction(x));
                        } else { // Otherwise, we have to find the text areas by their adjacent labels
                            let base = gradioApp().querySelector(entry.base);
-                
+
                            // Safety check
                            if (!base) continue;
-                
+
                            let allTextAreas = [...base.querySelectorAll("textarea, input[type='text']")];
-                
+
                            // Filter the text areas where the adjacent label matches one of the selectors
                            let matchingTextAreas = allTextAreas.filter(ta => [...ta.parentElement.childNodes].some(x => entry.selectors.includes(x.innerText)));
                            matchingTextAreas.forEach(x => setupFunction(x));
@@ -165,4 +201,4 @@ function getTextAreaIdentifier(textArea) {
            break;
    }
    return modifier;
-}
+}
--- a/javascript/_utils.js
+++ b/javascript/_utils.js
@@ -1,13 +1,14 @@
 // Utility functions for tag autocomplete

 // Parse the CSV file into a 2D array. Doesn't use regex, so it is very lightweight.
+// We are ignoring newlines in quote fields since we expect one-line entries and parsing would break for unclosed quotes otherwise
 function parseCSV(str) {
-    var arr = [];
-    var quote = false;  // 'true' means we're inside a quoted field
+    const arr = [];
+    let quote = false;  // 'true' means we're inside a quoted field

    // Iterate over each character, keep track of current row and column (of the returned array)
-    for (var row = 0, col = 0, c = 0; c < str.length; c++) {
-        var cc = str[c], nc = str[c + 1];        // Current character, next character
+    for (let row = 0, col = 0, c = 0; c < str.length; c++) {
+        let cc = str[c], nc = str[c+1];        // Current character, next character
        arr[row] = arr[row] || [];             // Create a new row if necessary
        arr[row][col] = arr[row][col] || '';   // Create a new column (start with empty string) if necessary

@@ -22,14 +23,12 @@ function parseCSV(str) {
        // If it's a comma and we're not in a quoted field, move on to the next column
        if (cc == ',' && !quote) { ++col; continue; }

-        // If it's a newline (CRLF) and we're not in a quoted field, skip the next character
-        // and move on to the next row and move to column 0 of that new row
-        if (cc == '\r' && nc == '\n' && !quote) { ++row; col = 0; ++c; continue; }
+        // If it's a newline (CRLF), skip the next character and move on to the next row and move to column 0 of that new row
+        if (cc == '\r' && nc == '\n') { ++row; col = 0; ++c; quote = false; continue; }

-        // If it's a newline (LF or CR) and we're not in a quoted field,
-        // move on to the next row and move to column 0 of that new row
-        if (cc == '\n' && !quote) { ++row; col = 0; continue; }
-        if (cc == '\r' && !quote) { ++row; col = 0; continue; }
+        // If it's a newline (LF or CR) move on to the next row and move to column 0 of that new row
+        if (cc == '\n') { ++row; col = 0; quote = false; continue; }
+        if (cc == '\r') { ++row; col = 0; quote = false; continue; }

        // Otherwise, append the current character to the current column
        arr[row][col] += cc;
@@ -41,7 +40,7 @@ function parseCSV(str) {
 async function readFile(filePath, json = false, cache = false) {
    if (!cache)
        filePath += `?${new Date().getTime()}`;
-        
+
    let response = await fetch(`file=${filePath}`);

    if (response.status != 200) {
@@ -61,6 +60,92 @@ async function loadCSV(path) {
    return parseCSV(text);
 }

+// Fetch API
+async function fetchTacAPI(url, json = true, cache = false) {
+    if (!cache) {
+        const appendChar = url.includes("?") ? "&" : "?";
+        url += `${appendChar}${new Date().getTime()}`
+    }
+
+    let response = await fetch(url);
+
+    if (response.status != 200) {
+        console.error(`Error fetching API endpoint "${url}": ` + response.status, response.statusText);
+        return null;
+    }
+
+    if (json)
+        return await response.json();
+    else
+        return await response.text();
+}
+
+async function postTacAPI(url, body = null) {
+    let response = await fetch(url, {
+        method: "POST",
+        headers: {'Content-Type': 'application/json'},
+        body: body
+    });
+
+    if (response.status != 200) {
+        console.error(`Error posting to API endpoint "${url}": ` + response.status, response.statusText);
+        return null;
+    }
+
+    return await response.json();
+}
+
+async function putTacAPI(url, body = null) {
+    let response = await fetch(url, { method: "PUT", body: body });
+    
+    if (response.status != 200) {
+        console.error(`Error putting to API endpoint "${url}": ` + response.status, response.statusText);
+        return null;
+    }
+
+    return await response.json();
+}
+
+// Extra network preview thumbnails
+async function getTacExtraNetworkPreviewURL(filename, type) {
+    const previewJSON = await fetchTacAPI(`tacapi/v1/thumb-preview/${filename}?type=${type}`, true, true);
+    if (previewJSON?.url) {
+        const properURL = `sd_extra_networks/thumb?filename=${previewJSON.url}`;
+        if ((await fetch(properURL)).status == 200) {
+            return properURL;
+        } else {
+            // create blob url
+            const blob = await (await fetch(`tacapi/v1/thumb-preview-blob/${filename}?type=${type}`)).blob();
+            return URL.createObjectURL(blob);
+        }
+    } else {
+        return null;
+    }
+}
+
+lastStyleRefresh = 0;
+// Refresh style file if needed
+async function refreshStyleNamesIfChanged() {
+    // Only refresh once per second
+    currentTimestamp = new Date().getTime();
+    if (currentTimestamp - lastStyleRefresh < 1000) return;
+    lastStyleRefresh = currentTimestamp;
+
+    const response = await fetch(`tacapi/v1/refresh-styles-if-changed?${new Date().getTime()}`)
+    if (response.status === 304) {
+        // Not modified
+    } else if (response.status === 200) {
+        // Reload
+        QUEUE_FILE_LOAD.forEach(async fn => {
+            if (fn.toString().includes("styleNames"))
+                await fn.call(null, true);
+        })
+    } else {
+        // Error
+        console.error(`Error refreshing styles.txt: ` + response.status, response.statusText);
+    }
+}
+
 // Debounce function to prevent spamming the autocomplete function
 var dbTimeOut;
 const debounce = (func, wait = 300) => {
@@ -89,6 +174,136 @@ function difference(a, b) {
    )].reduce((acc, [v, count]) => acc.concat(Array(Math.abs(count)).fill(v)), []);
 }

+// Object flatten function adapted from https://stackoverflow.com/a/61602592
+// $roots keeps previous parent properties as they will be added as a prefix for each prop.
+// $sep is just a preference if you want to seperate nested paths other than dot.
+function flatten(obj, roots = [], sep = ".") {
+  return Object.keys(obj).reduce(
+    (memo, prop) =>
+      Object.assign(
+        // create a new object
+        {},
+        // include previously returned object
+        memo,
+        Object.prototype.toString.call(obj[prop]) === "[object Object]"
+          ? // keep working if value is an object
+            flatten(obj[prop], roots.concat([prop]), sep)
+          : // include current prop and value and prefix prop with the roots
+            { [roots.concat([prop]).join(sep)]: obj[prop] }
+      ),
+    {}
+  );
+}
+
+// Calculate biased tag score based on post count and frequent usage
+function calculateUsageBias(result, count, uses) {
+    // Check setting conditions
+    if (uses < TAC_CFG.frequencyMinCount) {
+        uses = 0;
+    } else if (uses != 0) {
+        result.usageBias = true;
+    }
+
+    switch (TAC_CFG.frequencyFunction) {
+        case "Logarithmic (weak)":
+            return Math.log(1 + count) + Math.log(1 + uses);
+        case "Logarithmic (strong)":
+            return Math.log(1 + count) + 2 * Math.log(1 + uses);
+        case "Usage first":
+            return uses;
+        default:
+            return count;
+    }
+}
+// Beautify return type for easier parsing
+function mapUseCountArray(useCounts, posAndNeg = false) {
+    return useCounts.map(useCount => {
+        if (posAndNeg) {
+            return {
+                "name": useCount[0],
+                "type": useCount[1],
+                "count": useCount[2],
+                "negCount": useCount[3],
+                "lastUseDate": useCount[4]
+            }
+        }
+        return {
+            "name": useCount[0],
+            "type": useCount[1],
+            "count": useCount[2],
+            "lastUseDate": useCount[3]
+        }
+    });
+}
+// Call API endpoint to increase bias of tag in the database
+function increaseUseCount(tagName, type, negative = false) {
+    postTacAPI(`tacapi/v1/increase-use-count?tagname=${tagName}&ttype=${type}&neg=${negative}`);
+}
+// Get use count of tag from the database
+async function getUseCount(tagName, type, negative = false) {
+    const response = await fetchTacAPI(`tacapi/v1/get-use-count?tagname=${tagName}&ttype=${type}&neg=${negative}`, true, false);
+    // Guard for no db
+    if (response == null) return null;
+    // Result
+    return response["result"];
+}
+async function getUseCounts(tagNames, types, negative = false) {
+    // While semantically weird, we have to use POST here for the body, as urls are limited in length
+    const body = JSON.stringify({"tagNames": tagNames, "tagTypes": types, "neg": negative});
+    const response = await postTacAPI(`tacapi/v1/get-use-count-list`, body)
+    // Guard for no db
+    if (response == null) return null;
+    // Results
+    return mapUseCountArray(response["result"]);
+}
+async function getAllUseCounts() {
+    const response = await fetchTacAPI(`tacapi/v1/get-all-use-counts`);
+    // Guard for no db
+    if (response == null) return null;
+    // Results
+    return mapUseCountArray(response["result"], true);
+}
+async function resetUseCount(tagName, type, resetPosCount, resetNegCount) {
+    await putTacAPI(`tacapi/v1/reset-use-count?tagname=${tagName}&ttype=${type}&pos=${resetPosCount}&neg=${resetNegCount}`);
+}
+
+function createTagUsageTable(tagCounts) {
+    // Create table
+    let tagTable = document.createElement("table");
+    tagTable.innerHTML =
+    `<thead>
+        <tr>
+            <td>Name</td>
+            <td>Type</td>
+            <td>Count(+)</td>
+            <td>Count(-)</td>
+            <td>Last used</td>
+        </tr>
+    </thead>`;
+    tagTable.id = "tac_tagUsageTable"
+
+    tagCounts.forEach(t => {
+        let tr = document.createElement("tr");
+        
+        // Fill values
+        let values = [t.name, t.type-1, t.count, t.negCount, t.lastUseDate]
+        values.forEach(v => {
+            let td = document.createElement("td");
+            td.innerText = v;
+            tr.append(td);
+        });
+        // Add delete/reset button
+        let delButton = document.createElement("button");
+        delButton.innerText = "🗑️";
+        delButton.title = "Reset count";
+        tr.append(delButton);
+        
+        tagTable.append(tr)
+    });
+
+    return tagTable;
+}
+
 // Sliding window function to get possible combination groups of an array
 function toNgrams(inputArray, size) {
    return Array.from(
@@ -97,7 +312,11 @@ function toNgrams(inputArray, size) {
    );
 }

-function escapeRegExp(string) {
+function escapeRegExp(string, wildcardMatching = false) {
+    if (wildcardMatching) {
+        // Escape all characters except asterisks and ?, which should be treated separately as placeholders.
+        return string.replace(/[-[\]{}()+.,\\^$|#\s]/g, '\\$&').replace(/\*/g, '.*').replace(/\?/g, '.');
+    }
    return string.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'); // $& means the whole matched string
 }
 function escapeHTML(unsafeText) {
@@ -141,6 +360,49 @@ function observeElement(element, property, callback, delay = 0) {
    }
 }

+// Sort functions
+function getSortFunction() {
+    let criterion = TAC_CFG.modelSortOrder || "Name";
+
+    const textSort = (a, b, reverse = false) => {
+        // Assign keys so next sort is faster
+        if (!a.sortKey) {
+            a.sortKey = a.type === ResultType.chant
+                ? a.aliases
+                : a.text;
+        }
+        if (!b.sortKey) {
+            b.sortKey = b.type === ResultType.chant
+                ? b.aliases
+                : b.text;
+        }
+
+        return reverse ? b.sortKey.localeCompare(a.sortKey) : a.sortKey.localeCompare(b.sortKey);
+    }
+    const numericSort = (a, b, reverse = false) => {
+        const noKey = reverse ? "-1" : Number.MAX_SAFE_INTEGER;
+        let aParsed = parseFloat(a.sortKey || noKey);
+        let bParsed = parseFloat(b.sortKey || noKey);
+
+        if (aParsed === bParsed) {
+            return textSort(a, b, false);
+        }
+        
+        return reverse ? bParsed - aParsed : aParsed - bParsed;
+    }
+
+    return (a, b) => {
+        switch (criterion) {
+            case "Date Modified (newest first)":
+                return numericSort(a, b, true);
+            case "Date Modified (oldest first)":
+                return numericSort(a, b, false);
+            default:
+                return textSort(a, b);
+        }
+    }
+}
+
 // Queue calling function to process global queues
 async function processQueue(queue, context, ...args) {
    for (let i = 0; i < queue.length; i++) {
--- a/javascript/ext_chants.js
+++ b/javascript/ext_chants.js
@@ -7,7 +7,10 @@ class ChantParser extends BaseTagParser {
        let tempResults = [];
        if (tagword !== "<" && tagword !== "<c:") {
            let searchTerm = tagword.replace("<chant:", "").replace("<c:", "").replace("<", "");
-            let filterCondition = x => x.terms.toLowerCase().includes(searchTerm) || x.name.toLowerCase().includes(searchTerm);
+            let filterCondition = x => {
+                let regex = new RegExp(escapeRegExp(searchTerm, true), 'i');
+                return regex.test(x.terms.toLowerCase()) || regex.test(x.name.toLowerCase());
+            };
            tempResults = chants.filter(x => filterCondition(x)); // Filter by tagword
        } else {
            tempResults = chants;
@@ -41,7 +44,7 @@ async function load() {

 function sanitize(tagType, text) {
    if (tagType === ResultType.chant) {
-        return text.replace(/^.*?: /g, "");
+        return text;
    }
    return null;
 }
@@ -51,4 +54,4 @@ PARSERS.push(new ChantParser(CHANT_TRIGGER));
 // Add our utility functions to their respective queues
 QUEUE_FILE_LOAD.push(load);
 QUEUE_SANITIZE.push(sanitize);
-QUEUE_AFTER_CONFIG_CHANGE.push(load);
+QUEUE_AFTER_CONFIG_CHANGE.push(load);
--- a/javascript/ext_embeddings.js
+++ b/javascript/ext_embeddings.js
@@ -1,5 +1,5 @@
 const EMB_REGEX = /<(?!l:|h:|c:)[^,> ]*>?/g;
-const EMB_TRIGGER = () => TAC_CFG.useEmbeddings && tagword.match(EMB_REGEX);
+const EMB_TRIGGER = () => TAC_CFG.useEmbeddings && (tagword.match(EMB_REGEX) || TAC_CFG.includeEmbeddingsInNormalResults);

 class EmbeddingParser extends BaseTagParser {
    parse() {
@@ -11,12 +11,18 @@ class EmbeddingParser extends BaseTagParser {
            if (searchTerm.startsWith("v1") || searchTerm.startsWith("v2")) {
                versionString = searchTerm.slice(0, 2);
                searchTerm = searchTerm.slice(2);
+            } else if (searchTerm.startsWith("vxl")) {
+                versionString = searchTerm.slice(0, 3);
+                searchTerm = searchTerm.slice(3);
            }

-            let filterCondition = x => x[0].toLowerCase().includes(searchTerm) || x[0].toLowerCase().replaceAll(" ", "_").includes(searchTerm);
+            let filterCondition = x => {
+                let regex = new RegExp(escapeRegExp(searchTerm, true), 'i');
+                return regex.test(x[0].toLowerCase()) || regex.test(x[0].toLowerCase().replaceAll(" ", "_"));
+            };

            if (versionString)
-                tempResults = embeddings.filter(x => filterCondition(x) && x[1] && x[1] === versionString); // Filter by tagword
+                tempResults = embeddings.filter(x => filterCondition(x) && x[2] && x[2].toLowerCase() === versionString.toLowerCase()); // Filter by tagword
            else
                tempResults = embeddings.filter(x => filterCondition(x)); // Filter by tagword
        } else {
@@ -26,8 +32,13 @@ class EmbeddingParser extends BaseTagParser {
        // Add final results
        let finalResults = [];
        tempResults.forEach(t => {
-            let result = new AutocompleteResult(t[0].trim(), ResultType.embedding)
-            result.meta = t[1] + " Embedding";
+            let lastDot = t[0].lastIndexOf(".") > -1 ? t[0].lastIndexOf(".") : t[0].length;
+            let lastSlash = t[0].lastIndexOf("/") > -1 ? t[0].lastIndexOf("/") : -1;
+            let name = t[0].trim().substring(lastSlash + 1, lastDot);
+
+            let result = new AutocompleteResult(name, ResultType.embedding)
+            result.sortKey = t[1];
+            result.meta = t[2] + " Embedding";
            finalResults.push(result);
        });

@@ -38,9 +49,9 @@ class EmbeddingParser extends BaseTagParser {
 async function load() {
    if (embeddings.length === 0) {
        try {
-            embeddings = (await readFile(`${tagBasePath}/temp/emb.txt`)).split("\n")
-                .filter(x => x.trim().length > 0) // Remove empty lines
-                .map(x => x.trim().split(",")); // Split into name, version type pairs
+            embeddings = (await loadCSV(`${tagBasePath}/temp/emb.txt`))
+                .filter(x => x[0]?.trim().length > 0) // Remove empty lines
+                .map(x => [x[0].trim(), x[1], x[2]]); // Return name, sortKey, hash tuples
        } catch (e) {
            console.error("Error loading embeddings.txt: " + e);
        }
@@ -49,7 +60,7 @@ async function load() {

 function sanitize(tagType, text) {
    if (tagType === ResultType.embedding) {
-        return text.replace(/^.*?: /g, "");
+        return text;
    }
    return null;
 }
@@ -58,4 +69,4 @@ PARSERS.push(new EmbeddingParser(EMB_TRIGGER));

 // Add our utility functions to their respective queues
 QUEUE_FILE_LOAD.push(load);
-QUEUE_SANITIZE.push(sanitize);
+QUEUE_SANITIZE.push(sanitize);
--- a/javascript/ext_hypernets.js
+++ b/javascript/ext_hypernets.js
@@ -7,8 +7,11 @@ class HypernetParser extends BaseTagParser {
        let tempResults = [];
        if (tagword !== "<" && tagword !== "<h:" && tagword !== "<hypernet:") {
            let searchTerm = tagword.replace("<hypernet:", "").replace("<h:", "").replace("<", "");
-            let filterCondition = x => x.toLowerCase().includes(searchTerm) || x.toLowerCase().replaceAll(" ", "_").includes(searchTerm);
-            tempResults = hypernetworks.filter(x => filterCondition(x)); // Filter by tagword
+            let filterCondition = x => {
+                let regex = new RegExp(escapeRegExp(searchTerm, true), 'i');
+                return regex.test(x.toLowerCase()) || regex.test(x.toLowerCase().replaceAll(" ", "_"));
+            };
+            tempResults = hypernetworks.filter(x => filterCondition(x[0])); // Filter by tagword
        } else {
            tempResults = hypernetworks;
        }
@@ -16,8 +19,9 @@ class HypernetParser extends BaseTagParser {
        // Add final results
        let finalResults = [];
        tempResults.forEach(t => {
-            let result = new AutocompleteResult(t.trim(), ResultType.hypernetwork)
+            let result = new AutocompleteResult(t[0].trim(), ResultType.hypernetwork)
            result.meta = "Hypernetwork";
+            result.sortKey = t[1];
            finalResults.push(result);
        });

@@ -28,9 +32,9 @@ class HypernetParser extends BaseTagParser {
 async function load() {
    if (hypernetworks.length === 0) {
        try {
-            hypernetworks = (await readFile(`${tagBasePath}/temp/hyp.txt`)).split("\n")
-                .filter(x => x.trim().length > 0) //Remove empty lines
-                .map(x => x.trim()); // Remove carriage returns and padding if it exists
+            hypernetworks = (await loadCSV(`${tagBasePath}/temp/hyp.txt`))
+                .filter(x => x[0]?.trim().length > 0) //Remove empty lines
+                .map(x => [x[0]?.trim(), x[1]]); // Remove carriage returns and padding if it exists
        } catch (e) {
            console.error("Error loading hypernetworks.txt: " + e);
        }
@@ -48,4 +52,4 @@ PARSERS.push(new HypernetParser(HYP_TRIGGER));

 // Add our utility functions to their respective queues
 QUEUE_FILE_LOAD.push(load);
-QUEUE_SANITIZE.push(sanitize);
+QUEUE_SANITIZE.push(sanitize);
--- a/javascript/ext_loras.js
+++ b/javascript/ext_loras.js
@@ -7,7 +7,10 @@ class LoraParser extends BaseTagParser {
        let tempResults = [];
        if (tagword !== "<" && tagword !== "<l:" && tagword !== "<lora:") {
            let searchTerm = tagword.replace("<lora:", "").replace("<l:", "").replace("<", "");
-            let filterCondition = x => x.toLowerCase().includes(searchTerm) || x.toLowerCase().replaceAll(" ", "_").includes(searchTerm);
+            let filterCondition = x => {
+                let regex = new RegExp(escapeRegExp(searchTerm, true), 'i');
+                return regex.test(x.toLowerCase()) || regex.test(x.toLowerCase().replaceAll(" ", "_"));
+            };
            tempResults = loras.filter(x => filterCondition(x[0])); // Filter by tagword
        } else {
            tempResults = loras;
@@ -16,9 +19,15 @@ class LoraParser extends BaseTagParser {
        // Add final results
        let finalResults = [];
        tempResults.forEach(t => {
-            let result = new AutocompleteResult(t[0].trim(), ResultType.lora)
+            const text = t[0].trim();
+            let lastDot = text.lastIndexOf(".") > -1 ? text.lastIndexOf(".") : text.length;
+            let lastSlash = text.lastIndexOf("/") > -1 ? text.lastIndexOf("/") : -1;
+            let name = text.substring(lastSlash + 1, lastDot);
+
+            let result = new AutocompleteResult(name, ResultType.lora)
            result.meta = "Lora";
-            result.hash = t[1];
+            result.sortKey = t[1];
+            result.hash = t[2];
            finalResults.push(result);
        });

@@ -31,16 +40,22 @@ async function load() {
        try {
            loras = (await loadCSV(`${tagBasePath}/temp/lora.txt`))
                .filter(x => x[0]?.trim().length > 0) // Remove empty lines
-                .map(x => [x[0]?.trim(), x[1]]); // Trim filenames and return the name, hash pairs
+                .map(x => [x[0]?.trim(), x[1], x[2]]); // Trim filenames and return the name, sortKey, hash pairs
        } catch (e) {
            console.error("Error loading lora.txt: " + e);
        }
    }
 }

-function sanitize(tagType, text) {
+async function sanitize(tagType, text) {
    if (tagType === ResultType.lora) {
-        return `<lora:${text}:${TAC_CFG.extraNetworksDefaultMultiplier}>`;
+        let multiplier = TAC_CFG.extraNetworksDefaultMultiplier;
+        let info = await fetchTacAPI(`tacapi/v1/lora-info/${text}`)
+        if (info && info["preferred weight"]) {
+            multiplier = info["preferred weight"];
+        }
+
+        return `<lora:${text}:${multiplier}>`;
    }
    return null;
 }
@@ -49,4 +64,4 @@ PARSERS.push(new LoraParser(LORA_TRIGGER));

 // Add our utility functions to their respective queues
 QUEUE_FILE_LOAD.push(load);
-QUEUE_SANITIZE.push(sanitize);
+QUEUE_SANITIZE.push(sanitize);
--- a/javascript/ext_lycos.js
+++ b/javascript/ext_lycos.js
@@ -5,9 +5,12 @@ class LycoParser extends BaseTagParser {
    parse() {
        // Show lyco
        let tempResults = [];
-        if (tagword !== "<" && tagword !== "<l:" && tagword !== "<lyco:") {
-            let searchTerm = tagword.replace("<lyco:", "").replace("<l:", "").replace("<", "");
-            let filterCondition = x => x.toLowerCase().includes(searchTerm) || x.toLowerCase().replaceAll(" ", "_").includes(searchTerm);
+        if (tagword !== "<" && tagword !== "<l:" && tagword !== "<lyco:" && tagword !== "<lora:") {
+            let searchTerm = tagword.replace("<lyco:", "").replace("<lora:", "").replace("<l:", "").replace("<", "");
+            let filterCondition = x => {
+                let regex = new RegExp(escapeRegExp(searchTerm, true), 'i');
+                return regex.test(x.toLowerCase()) || regex.test(x.toLowerCase().replaceAll(" ", "_"));
+            };
            tempResults = lycos.filter(x => filterCondition(x[0])); // Filter by tagword
        } else {
            tempResults = lycos;
@@ -16,9 +19,15 @@ class LycoParser extends BaseTagParser {
        // Add final results
        let finalResults = [];
        tempResults.forEach(t => {
-            let result = new AutocompleteResult(t[0].trim(), ResultType.lyco)
+            const text = t[0].trim();
+            let lastDot = text.lastIndexOf(".") > -1 ? text.lastIndexOf(".") : text.length;
+            let lastSlash = text.lastIndexOf("/") > -1 ? text.lastIndexOf("/") : -1;
+            let name = text.substring(lastSlash + 1, lastDot);
+
+            let result = new AutocompleteResult(name, ResultType.lyco)
            result.meta = "Lyco";
-            result.hash = t[1];
+            result.sortKey = t[1];
+            result.hash = t[2];
            finalResults.push(result);
        });

@@ -31,16 +40,23 @@ async function load() {
        try {
            lycos = (await loadCSV(`${tagBasePath}/temp/lyco.txt`))
                .filter(x => x[0]?.trim().length > 0) // Remove empty lines
-                .map(x => [x[0]?.trim(), x[1]]); // Trim filenames and return the name, hash pairs
+                .map(x => [x[0]?.trim(), x[1], x[2]]); // Trim filenames and return the name, sortKey, hash pairs
        } catch (e) {
            console.error("Error loading lyco.txt: " + e);
        }
    }
 }

-function sanitize(tagType, text) {
+async function sanitize(tagType, text) {
    if (tagType === ResultType.lyco) {
-        return `<lyco:${text}:${TAC_CFG.extraNetworksDefaultMultiplier}>`;
+        let multiplier = TAC_CFG.extraNetworksDefaultMultiplier;
+        let info = await fetchTacAPI(`tacapi/v1/lyco-info/${text}`)
+        if (info && info["preferred weight"]) {
+            multiplier = info["preferred weight"];
+        }
+
+        let prefix = TAC_CFG.useLoraPrefixForLycos ? "lora" : "lyco";
+        return `<${prefix}:${text}:${multiplier}>`;
    }
    return null;
 }
@@ -49,4 +65,4 @@ PARSERS.push(new LycoParser(LYCO_TRIGGER));

 // Add our utility functions to their respective queues
 QUEUE_FILE_LOAD.push(load);
-QUEUE_SANITIZE.push(sanitize);
+QUEUE_SANITIZE.push(sanitize);
--- a/javascript/ext_modelKeyword.js
+++ b/javascript/ext_modelKeyword.js
@@ -20,7 +20,7 @@ async function load() {
            // Add to the dict
            csv_lines.forEach(parts => {
                const hash = parts[0];
-                const keywords = parts[1].replaceAll("| ", ", ").replaceAll("|", ", ").trim();
+                const keywords = parts[1]?.replaceAll("| ", ", ")?.replaceAll("|", ", ")?.trim();
                const lastSepIndex = parts[2]?.lastIndexOf("/") + 1 || parts[2]?.lastIndexOf("\\") + 1 || 0;
                const name = parts[2]?.substring(lastSepIndex).trim() || "none"

--- a/javascript/ext_styles.js
+++ b/javascript/ext_styles.js
@@ -0,0 +1,70 @@
+const STYLE_REGEX = /(\$(\d*)\(?)[^$|\[\],\s]*\)?/;
+const STYLE_TRIGGER = () => TAC_CFG.useStyleVars && tagword.match(STYLE_REGEX);
+
+var lastStyleVarIndex = "";
+
+class StyleParser extends BaseTagParser {
+   async parse() {
+        // Refresh if needed
+        await refreshStyleNamesIfChanged();
+
+        // Show styles
+        let tempResults = [];
+        let matchGroups = tagword.match(STYLE_REGEX);
+        
+        // Save index to insert again later or clear last one
+        lastStyleVarIndex = matchGroups[2] ? matchGroups[2] : "";
+
+        if (tagword !== matchGroups[1]) {
+            let searchTerm = tagword.replace(matchGroups[1], "");
+            
+            let filterCondition = x => {
+                let regex = new RegExp(escapeRegExp(searchTerm, true), 'i');
+                return regex.test(x[0].toLowerCase()) || regex.test(x[0].toLowerCase().replaceAll(" ", "_"));
+            };
+            tempResults = styleNames.filter(x => filterCondition(x)); // Filter by tagword
+        } else {
+            tempResults = styleNames;
+        }
+
+        // Add final results
+        let finalResults = [];
+        tempResults.forEach(t => {
+            let result = new AutocompleteResult(t[0].trim(), ResultType.styleName)
+            result.meta = "Style";
+            finalResults.push(result);
+        });
+
+        return finalResults;
+    }
+}
+
+async function load(force = false) {
+    if (styleNames.length === 0 || force) {
+        try {
+            styleNames = (await loadCSV(`${tagBasePath}/temp/styles.txt`))
+                .filter(x => x[0]?.trim().length > 0) // Remove empty lines
+                .filter(x => x[0] !== "None") // Remove "None" style
+                .map(x => [x[0].trim()]); // Trim name
+        } catch (e) {
+            console.error("Error loading styles.txt: " + e);
+        }
+    }
+}
+
+function sanitize(tagType, text) {
+    if (tagType === ResultType.styleName) {
+        if (text.includes(" ")) {
+            return `$${lastStyleVarIndex}(${text})`;
+        } else {
+            return`$${lastStyleVarIndex}${text}`
+        }
+    }
+    return null;
+}
+
+PARSERS.push(new StyleParser(STYLE_TRIGGER));
+
+// Add our utility functions to their respective queues
+QUEUE_FILE_LOAD.push(load);
+QUEUE_SANITIZE.push(sanitize);
--- a/javascript/ext_umi.js
+++ b/javascript/ext_umi.js
@@ -7,7 +7,7 @@ class UmiParser extends BaseTagParser {
    parse(textArea, prompt) {
        // We are in a UMI yaml tag definition, parse further
        let umiSubPrompts = [...prompt.matchAll(UMI_PROMPT_REGEX)];
-                
+
        let umiTags = [];
        let umiTagsWithOperators = []

@@ -15,7 +15,7 @@ class UmiParser extends BaseTagParser {

        umiSubPrompts.forEach(umiSubPrompt => {
            umiTags = umiTags.concat([...umiSubPrompt[0].matchAll(UMI_TAG_REGEX)].map(x => x[1].toLowerCase()));
-            
+
            const start = umiSubPrompt.index;
            const end = umiSubPrompt.index + umiSubPrompt[0].length;
            if (textArea.selectionStart >= start && textArea.selectionStart <= end) {
@@ -74,7 +74,7 @@ class UmiParser extends BaseTagParser {
        //console.log({ matches })

        const filteredWildcards = (tagword) => {
-            const wildcards = yamlWildcards.filter(x => {
+            const wildcards = umiWildcards.filter(x => {
                let tags = x[1];
                const matchesNeg =
                    matches.negative.length === 0
@@ -113,7 +113,7 @@ class UmiParser extends BaseTagParser {
                    || !matches.all.includes(x[0])
                );
        }
-        
+
        if (umiTags.length > 0) {
            // Get difference for subprompt
            let tagCountChange = umiTags.length - umiPreviousTags.length;
@@ -129,7 +129,7 @@ class UmiParser extends BaseTagParser {
                return;
            }

-            let umiTagword = diff[0] || '';
+            let umiTagword = tagCountChange < 0 ? '' : diff[0] || '';
            let tempResults = [];
            if (umiTagword && umiTagword.length > 0) {
                umiTagword = umiTagword.toLowerCase().replace(/[\n\r]/g, "");
@@ -144,48 +144,53 @@ class UmiParser extends BaseTagParser {
                // Add final results
                let finalResults = [];
                tempResults.forEach(t => {
-                    let result = new AutocompleteResult(t[0].trim(), ResultType.yamlWildcard)
+                    let result = new AutocompleteResult(t[0].trim(), ResultType.umiWildcard)
                    result.count = t[1];
                    finalResults.push(result);
                });

+                finalResults = finalResults.sort((a, b) => b.count - a.count);
                return finalResults;
            } else if (showAll) {
                let filteredWildcardsSorted = filteredWildcards("");
-                
+
                // Add final results
                let finalResults = [];
                filteredWildcardsSorted.forEach(t => {
-                    let result = new AutocompleteResult(t[0].trim(), ResultType.yamlWildcard)
+                    let result = new AutocompleteResult(t[0].trim(), ResultType.umiWildcard)
                    result.count = t[1];
                    finalResults.push(result);
                });
-        
+
                originalTagword = tagword;
                tagword = "";
+
+                finalResults = finalResults.sort((a, b) => b.count - a.count);
                return finalResults;
            }
        } else {
            let filteredWildcardsSorted = filteredWildcards("");
-                
+
            // Add final results
            let finalResults = [];
            filteredWildcardsSorted.forEach(t => {
-                let result = new AutocompleteResult(t[0].trim(), ResultType.yamlWildcard)
+                let result = new AutocompleteResult(t[0].trim(), ResultType.umiWildcard)
                result.count = t[1];
                finalResults.push(result);
            });

            originalTagword = tagword;
            tagword = "";
+
+            finalResults = finalResults.sort((a, b) => b.count - a.count);
            return finalResults;
        }
    }
 }

-function updateUmiTags( tagType, sanitizedText, newPrompt, textArea) {
-    // If it was a yaml wildcard, also update the umiPreviousTags
-    if (tagType === ResultType.yamlWildcard && originalTagword.length > 0) {
+function updateUmiTags(tagType, sanitizedText, newPrompt, textArea) {
+    // If it was a umi wildcard, also update the umiPreviousTags
+    if (tagType === ResultType.umiWildcard && originalTagword.length > 0) {
        let umiSubPrompts = [...newPrompt.matchAll(UMI_PROMPT_REGEX)];

        let umiTags = [];
@@ -203,11 +208,11 @@ function updateUmiTags( tagType, sanitizedText, newPrompt, textArea) {
 }

 async function load() {
-    if (yamlWildcards.length === 0) {
+    if (umiWildcards.length === 0) {
        try {
-            let yamlTags = (await readFile(`${tagBasePath}/temp/wcet.txt`)).split("\n");
+            let umiTags = (await readFile(`${tagBasePath}/temp/umi_tags.txt`)).split("\n");
            // Split into tag, count pairs
-            yamlWildcards = yamlTags.map(x => x
+            umiWildcards = umiTags.map(x => x
                .trim()
                .split(","))
                .map(([i, ...rest]) => [
@@ -218,14 +223,14 @@ async function load() {
                    }, {}),
                ]);
        } catch (e) {
-            console.error("Error loading yaml wildcards: " + e);
+            console.error("Error loading umi wildcards: " + e);
        }
    }
 }

 function sanitize(tagType, text) {
-    // Replace underscores only if the yaml tag is not using them
-    if (tagType === ResultType.yamlWildcard && !yamlWildcards.includes(text)) {
+    // Replace underscores only if the umi tag is not using them
+    if (tagType === ResultType.umiWildcard && !umiWildcards.includes(text)) {
        return text.replaceAll("_", " "); 
    }
    return null;
--- a/javascript/ext_wildcards.js
+++ b/javascript/ext_wildcards.js
@@ -1,24 +1,52 @@
 // Regex
-const WC_REGEX = /\b__([^,]+)__([^, ]*)\b/g;
+const WC_REGEX = new RegExp(/__([^,]+)__([^, ]*)/g);

 // Trigger conditions
-const WC_TRIGGER = () => TAC_CFG.useWildcards && [...tagword.matchAll(WC_REGEX)].length > 0;
-const WC_FILE_TRIGGER = () => TAC_CFG.useWildcards && (tagword.startsWith("__") && !tagword.endsWith("__") || tagword === "__");
+const WC_TRIGGER = () => TAC_CFG.useWildcards && [...tagword.matchAll(new RegExp(WC_REGEX.source.replaceAll("__", escapeRegExp(TAC_CFG.wcWrap)), "g"))].length > 0;
+const WC_FILE_TRIGGER = () => TAC_CFG.useWildcards && (tagword.startsWith(TAC_CFG.wcWrap) && !tagword.endsWith(TAC_CFG.wcWrap) || tagword === TAC_CFG.wcWrap);

 class WildcardParser extends BaseTagParser {
    async parse() {
        // Show wildcards from a file with that name
-        let wcMatch = [...tagword.matchAll(WC_REGEX)]
+        let wcMatch = [...tagword.matchAll(new RegExp(WC_REGEX.source.replaceAll("__", escapeRegExp(TAC_CFG.wcWrap)), "g"))];
        let wcFile = wcMatch[0][1];
        let wcWord = wcMatch[0][2];

        // Look in normal wildcard files
-        let wcFound = wildcardFiles.find(x => x[1].toLowerCase() === wcFile);
+        let wcFound = wildcardFiles.filter(x => x[1].toLowerCase() === wcFile);
+        if (wcFound.length === 0) wcFound = null;
        // Use found wildcard file or look in external wildcard files
-        let wcPair = wcFound || wildcardExtFiles.find(x => x[1].toLowerCase() === wcFile);
+        let wcPairs = wcFound || wildcardExtFiles.filter(x => x[1].toLowerCase() === wcFile);

-        let wildcards = (await readFile(`${wcPair[0]}/${wcPair[1]}.txt`)).split("\n")
-            .filter(x => x.trim().length > 0 && !x.startsWith('#'));  // Remove empty lines and comments
+        if (!wcPairs) return [];
+
+        let wildcards = [];
+        for (let i = 0; i < wcPairs.length; i++) {
+            const basePath = wcPairs[i][0];
+            const fileName = wcPairs[i][1];
+            if (!basePath || !fileName) return;
+
+            // YAML wildcards are already loaded as json, so we can get the values directly.
+            // basePath is the name of the file in this case, and fileName the key
+            if (basePath.endsWith(".yaml")) {
+                const getDescendantProp = (obj, desc) => {
+                    const arr = desc.split("/");
+                    while (arr.length) {
+                      obj = obj[arr.shift()];
+                    }
+                    return obj;
+                }
+                wildcards = wildcards.concat(getDescendantProp(yamlWildcards[basePath], fileName));
+            } else {
+                const fileContent = (await fetchTacAPI(`tacapi/v1/wildcard-contents?basepath=${basePath}&filename=${fileName}.txt`, false))
+                    .split("\n")
+                    .filter(x => x.trim().length > 0 && !x.startsWith('#'));  // Remove empty lines and comments
+                wildcards = wildcards.concat(fileContent);
+            }
+        }
+
+        if (TAC_CFG.sortWildcardResults)
+            wildcards.sort((a, b) => a.localeCompare(b));

        let finalResults = [];
        let tempResults = wildcards.filter(x => (wcWord !== null && wcWord.length > 0) ? x.toLowerCase().includes(wcWord) : x) // Filter by tagword
@@ -36,21 +64,36 @@ class WildcardFileParser extends BaseTagParser {
    parse() {
        // Show available wildcard files
        let tempResults = [];
-        if (tagword !== "__") {
-            let lmb = (x) => x[1].toLowerCase().includes(tagword.replace("__", ""))
+        if (tagword !== TAC_CFG.wcWrap) {
+            let lmb = (x) => x[1].toLowerCase().includes(tagword.replace(TAC_CFG.wcWrap, ""))
            tempResults = wildcardFiles.filter(lmb).concat(wildcardExtFiles.filter(lmb)) // Filter by tagword
        } else {
            tempResults = wildcardFiles.concat(wildcardExtFiles);
        }

        let finalResults = [];
+        const alreadyAdded = new Map();
        // Get final results
        tempResults.forEach(wcFile => {
-            let result = new AutocompleteResult(wcFile[1].trim(), ResultType.wildcardFile);
-            result.meta = "Wildcard file";
+            // Skip duplicate entries incase multiple files have the same name or yaml category
+            if (alreadyAdded.has(wcFile[1])) return;
+
+            let result = null;
+            if (wcFile[0].endsWith(".yaml")) {
+                result = new AutocompleteResult(wcFile[1].trim(), ResultType.yamlWildcard);
+                result.meta = "YAML wildcard collection";
+            } else {
+                result = new AutocompleteResult(wcFile[1].trim(), ResultType.wildcardFile);
+                result.meta = "Wildcard file";
+                result.sortKey = wcFile[2].trim();
+            }
+
            finalResults.push(result);
+            alreadyAdded.set(wcFile[1], true);
        });

+        finalResults.sort(getSortFunction());
+
        return finalResults;
    }
 }
@@ -58,17 +101,19 @@ class WildcardFileParser extends BaseTagParser {
 async function load() {
    if (wildcardFiles.length === 0 && wildcardExtFiles.length === 0) {
        try {
-            let wcFileArr = (await readFile(`${tagBasePath}/temp/wc.txt`)).split("\n");
-            let wcBasePath = wcFileArr[0].trim(); // First line should be the base path
-            wildcardFiles = wcFileArr.slice(1)
-                .filter(x => x.trim().length > 0) // Remove empty lines
-                .map(x => [wcBasePath, x.trim().replace(".txt", "")]); // Remove file extension & newlines
+            let wcFileArr = await loadCSV(`${tagBasePath}/temp/wc.txt`);
+            if (wcFileArr && wcFileArr.length > 0) {
+                let wcBasePath = wcFileArr[0][0].trim(); // First line should be the base path
+                wildcardFiles = wcFileArr.slice(1)
+                    .filter(x => x[0]?.trim().length > 0) //Remove empty lines
+                    .map(x => [wcBasePath, x[0]?.trim().replace(".txt", ""), x[1]]); // Remove file extension & newlines
+            }

            // To support multiple sources, we need to separate them using the provided "-----" strings
-            let wcExtFileArr = (await readFile(`${tagBasePath}/temp/wce.txt`)).split("\n");
+            let wcExtFileArr = await loadCSV(`${tagBasePath}/temp/wce.txt`);
            let splitIndices = [];
            for (let index = 0; index < wcExtFileArr.length; index++) {
-                if (wcExtFileArr[index].trim() === "-----") {
+                if (wcExtFileArr[index][0].trim() === "-----") {
                    splitIndices.push(index);
                }
            }
@@ -79,14 +124,25 @@ async function load() {
                let end = splitIndices[i];

                let wcExtFile = wcExtFileArr.slice(start, end);
-                let base = wcExtFile[0].trim() + "/";
-                wcExtFile = wcExtFile.slice(1)
-                    .filter(x => x.trim().length > 0) // Remove empty lines
-                    .map(x => x.trim().replace(base, "").replace(".txt", "")); // Remove file extension & newlines;
-
-                wcExtFile = wcExtFile.map(x => [base, x]);
-                wildcardExtFiles.push(...wcExtFile);
+                if (wcExtFile && wcExtFile.length > 0) {
+                    let base = wcExtFile[0][0].trim() + "/";
+                    wcExtFile = wcExtFile.slice(1)
+                        .filter(x => x[0]?.trim().length > 0) //Remove empty lines
+                        .map(x => [base, x[0]?.trim().replace(base, "").replace(".txt", ""), x[1]]);
+                    wildcardExtFiles.push(...wcExtFile);
+                }
            }
+
+            // Load the yaml wildcard json file and append it as a wildcard file, appending each key as a path component until we reach the end
+            yamlWildcards = await readFile(`${tagBasePath}/temp/wc_yaml.json`, true);
+
+            // Append each key as a path component until we reach a leaf
+            Object.keys(yamlWildcards).forEach(file => {
+                const flattened = flatten(yamlWildcards[file], [], "/");
+                Object.keys(flattened).forEach(key => {
+                    wildcardExtFiles.push([file, key]);
+                });
+            });
        } catch (e) {
            console.error("Error loading wildcards: " + e);
        }
@@ -94,19 +150,18 @@ async function load() {
 }

 function sanitize(tagType, text) {
-    if (tagType === ResultType.wildcardFile) {
-        return `__${text}__`;
+    if (tagType === ResultType.wildcardFile || tagType === ResultType.yamlWildcard) {
+        return `${TAC_CFG.wcWrap}${text}${TAC_CFG.wcWrap}`;
    } else if (tagType === ResultType.wildcardTag) {
-        return text.replace(/^.*?: /g, "");
+        return text;
    }
    return null;
 }

 function keepOpenIfWildcard(tagType, sanitizedText, newPrompt, textArea) {
    // If it's a wildcard, we want to keep the results open so the user can select another wildcard
-    if (tagType === ResultType.wildcardFile) {
+    if (tagType === ResultType.wildcardFile || tagType === ResultType.yamlWildcard) {
        hideBlocked = true;
-        autocomplete(textArea, newPrompt, sanitizedText);
        setTimeout(() => { hideBlocked = false; }, 450);
        return true;
    }
--- a/javascript/tagAutocomplete.js
+++ b/javascript/tagAutocomplete.js
--- a/scripts/model_keyword_support.py
+++ b/scripts/model_keyword_support.py
@@ -16,6 +16,8 @@ hash_dict = {}


 def load_hash_cache():
+    if not known_hashes_file.exists():
+        known_hashes_file.touch()
    with open(known_hashes_file, "r", encoding="utf-8") as file:
        reader = csv.reader(
            file.readlines(), delimiter=",", quotechar='"', skipinitialspace=True
@@ -28,9 +30,12 @@ def load_hash_cache():
 def update_hash_cache():
    global file_needs_update
    if file_needs_update:
-        with open(known_hashes_file, "w", encoding="utf-8") as file:
+        if not known_hashes_file.exists():
+            known_hashes_file.touch()
+        with open(known_hashes_file, "w", encoding="utf-8", newline='') as file:
+            writer = csv.writer(file)
            for name, (hash, mtime) in hash_dict.items():
-                file.write(f'"{name}",{hash},{mtime}\n')
+                writer.writerow([name, hash, mtime])


 # Copy of the fast inaccurate hash function from the extension
@@ -75,6 +80,6 @@ def write_model_keyword_path():
        return True
    else:
        print(
-            "Tag Autocomplete: Could not locate model-keyword extension, LORA/LYCO trigger word completion will be unavailable."
+            "Tag Autocomplete: Could not locate model-keyword extension, Lora trigger word completion will be limited to those added through the extra networks menu."
        )
        return False
--- a/scripts/shared_paths.py
+++ b/scripts/shared_paths.py
@@ -1,35 +1,57 @@
 from pathlib import Path
+
 from modules import scripts, shared

 try:
    from modules.paths import extensions_dir, script_path

    # Webui root path
-    FILE_DIR = Path(script_path)
+    FILE_DIR = Path(script_path).absolute()

    # The extension base path
-    EXT_PATH = Path(extensions_dir)
+    EXT_PATH = Path(extensions_dir).absolute()
 except ImportError:
    # Webui root path
    FILE_DIR = Path().absolute()
    # The extension base path
-    EXT_PATH = FILE_DIR.joinpath("extensions")
+    EXT_PATH = FILE_DIR.joinpath("extensions").absolute()

 # Tags base path
-TAGS_PATH = Path(scripts.basedir()).joinpath("tags")
+TAGS_PATH = Path(scripts.basedir()).joinpath("tags").absolute()

 # The path to the folder containing the wildcards and embeddings
-WILDCARD_PATH = FILE_DIR.joinpath("scripts/wildcards")
-EMB_PATH = Path(shared.cmd_opts.embeddings_dir)
-HYP_PATH = Path(shared.cmd_opts.hypernetwork_dir)
+try: # SD.Next
+    WILDCARD_PATH = Path(shared.opts.wildcards_dir).absolute()
+except Exception: # A1111
+    WILDCARD_PATH = FILE_DIR.joinpath("scripts/wildcards").absolute()
+EMB_PATH = Path(shared.cmd_opts.embeddings_dir).absolute()
+
+# Forge Classic detection
+try:
+    from modules_forge.forge_version import version as forge_version
+    IS_FORGE_CLASSIC = forge_version == "classic"
+except ImportError:
+    IS_FORGE_CLASSIC = False
+
+# Forge Classic skips it
+if not IS_FORGE_CLASSIC:
+    try:
+        HYP_PATH = Path(shared.cmd_opts.hypernetwork_dir).absolute()
+    except AttributeError:
+        HYP_PATH = None
+else:
+    HYP_PATH = None

 try:
-    LORA_PATH = Path(shared.cmd_opts.lora_dir)
+    LORA_PATH = Path(shared.cmd_opts.lora_dir).absolute()
 except AttributeError:
    LORA_PATH = None

 try:
-    LYCO_PATH = Path(shared.cmd_opts.lyco_dir)
+    try:
+        LYCO_PATH = Path(shared.cmd_opts.lyco_dir_backcompat).absolute()
+    except:
+        LYCO_PATH = Path(shared.cmd_opts.lyco_dir).absolute() # attempt original non-backcompat path
 except AttributeError:
    LYCO_PATH = None

@@ -37,6 +59,21 @@ except AttributeError:
 def find_ext_wildcard_paths():
    """Returns the path to the extension wildcards folder"""
    found = list(EXT_PATH.glob("*/wildcards/"))
+    # Try to find the wildcard path from the shared opts
+    try:
+        from modules.shared import opts
+    except ImportError:  # likely not in an a1111 context
+        opts = None
+
+    # Append custom wildcard paths
+    custom_paths = [
+        getattr(shared.cmd_opts, "wildcards_dir", None),    # Cmd arg from the wildcard extension
+        getattr(opts, "wildcard_dir", None),                # Custom path from sd-dynamic-prompts
+    ]
+    for path in [Path(p).absolute() for p in custom_paths if p is not None]:
+        if path.exists():
+            found.append(path)
+
    return found


@@ -45,8 +82,8 @@ WILDCARD_EXT_PATHS = find_ext_wildcard_paths()

 # The path to the temporary files
 # In the webui root, on windows it exists by default, on linux it doesn't
-STATIC_TEMP_PATH = FILE_DIR.joinpath("tmp")
-TEMP_PATH = TAGS_PATH.joinpath("temp")  # Extension specific temp files
+STATIC_TEMP_PATH = FILE_DIR.joinpath("tmp").absolute()
+TEMP_PATH = TAGS_PATH.joinpath("temp").absolute()  # Extension specific temp files

 # Make sure these folders exist
 if not TEMP_PATH.exists():
--- a/scripts/tag_autocomplete_helper.py
+++ b/scripts/tag_autocomplete_helper.py
@@ -2,68 +2,213 @@
 # to a temporary file to expose it to the javascript side

 import glob
+import importlib
+import json
+import sqlite3
+import sys
+import urllib.parse
+from asyncio import sleep
 from pathlib import Path

 import gradio as gr
 import yaml
-from modules import script_callbacks, sd_hijack, shared
+from fastapi import FastAPI
+from fastapi.responses import FileResponse, JSONResponse, Response
+from modules import hashes, script_callbacks, sd_hijack, sd_models, shared
+from pydantic import BaseModel

 from scripts.model_keyword_support import (get_lora_simple_hash,
                                           load_hash_cache, update_hash_cache,
                                           write_model_keyword_path)
 from scripts.shared_paths import *

+try:
+    try:
+        from scripts import tag_frequency_db as tdb
+    except ModuleNotFoundError:
+        from inspect import currentframe, getframeinfo
+        filename = getframeinfo(currentframe()).filename
+        parent = Path(filename).resolve().parent
+        sys.path.append(str(parent))
+        import tag_frequency_db as tdb
+
+    # Ensure the db dependency is reloaded on script reload
+    importlib.reload(tdb)
+
+    db = tdb.TagFrequencyDb()
+    if int(db.version) != int(tdb.db_ver):
+        raise ValueError("Database version mismatch")
+except (ImportError, ValueError, sqlite3.Error) as e:
+    print(f"Tag Autocomplete: Tag frequency database error - \"{e}\"")
+    db = None
+
+def get_embed_db(sd_model=None):
+    """Returns the embedding database, if available."""
+    try:
+        return sd_hijack.model_hijack.embedding_db
+    except Exception:
+        try: # sd next with diffusers backend
+            sdnext_model = sd_model if sd_model is not None else shared.sd_model
+            return sdnext_model.embedding_db
+        except Exception:
+            try: # forge webui
+                forge_model = sd_model if sd_model is not None else sd_models.model_data.get_sd_model()
+                if type(forge_model).__name__ == "FakeInitialModel":
+                    return None
+                else:
+                    processer = getattr(forge_model, "text_processing_engine", getattr(forge_model, "text_processing_engine_l"))
+                    return processer.embeddings
+            except Exception:
+                return None
+
+# Attempt to get embedding load function, using the same call as api.
+try:
+    embed_db = get_embed_db()
+    if embed_db is not None:
+        load_textual_inversion_embeddings = embed_db.load_textual_inversion_embeddings
+    else:
+        load_textual_inversion_embeddings = lambda *args, **kwargs: None
+except Exception as e: # Not supported.
+    load_textual_inversion_embeddings = lambda *args, **kwargs: None
+    print("Tag Autocomplete: Cannot reload embeddings instantly:", e)
+
+# Sorting functions for extra networks / embeddings stuff
+sort_criteria = {
+    "Name": lambda path, name, subpath: name.lower() if subpath else path.stem.lower(),
+    "Date Modified (newest first)": lambda path, name, subpath: path.stat().st_mtime if path.exists() else name.lower(),
+    "Date Modified (oldest first)": lambda path, name, subpath: path.stat().st_mtime if path.exists() else name.lower()
+}
+
+def sort_models(model_list, sort_method = None, name_has_subpath = False):
+    """Sorts models according to the setting.
+    
+    Input: list of (full_path, display_name, {hash}) models. 
+    Returns models in the format of name, sort key, meta.
+    Meta is optional and can be a hash, version string or other required info.
+    """
+    if len(model_list) == 0:
+        return model_list
+
+    if sort_method is None:
+        sort_method = getattr(shared.opts, "tac_modelSortOrder", "Name")
+
+    # Get sorting method from dictionary
+    sorter = sort_criteria.get(sort_method, sort_criteria["Name"])
+
+    # During merging on the JS side we need to re-sort anyway, so here only the sort criteria are calculated.
+    # The list itself doesn't need to get sorted at this point.
+    if len(model_list[0]) > 2:
+        results = [f'"{name}","{sorter(path, name, name_has_subpath)}",{meta}' for path, name, meta in model_list]
+    else:
+        results = [f'"{name}","{sorter(path, name, name_has_subpath)}"' for path, name in model_list]
+    return results
+

 def get_wildcards():
    """Returns a list of all wildcards. Works on nested folders."""
    wildcard_files = list(WILDCARD_PATH.rglob("*.txt"))
-    resolved = [w.relative_to(WILDCARD_PATH).as_posix(
-    ) for w in wildcard_files if w.name != "put wildcards here.txt"]
-    return resolved
+    resolved = [(w, w.relative_to(WILDCARD_PATH).as_posix())
+                for w in wildcard_files
+                if w.name != "put wildcards here.txt"
+                and w.is_file()]
+    return sort_models(resolved, name_has_subpath=True)


 def get_ext_wildcards():
    """Returns a list of all extension wildcards. Works on nested folders."""
    wildcard_files = []
-
+    excluded_folder_names = [s.strip() for s in getattr(shared.opts, "tac_wildcardExclusionList", "").split(",")]
    for path in WILDCARD_EXT_PATHS:
        wildcard_files.append(path.as_posix())
-        wildcard_files.extend(p.relative_to(path).as_posix() for p in path.rglob("*.txt") if p.name != "put wildcards here.txt")
+        resolved = [(w, w.relative_to(path).as_posix())
+                    for w in path.rglob("*.txt")
+                    if w.name != "put wildcards here.txt"
+                    and not any(excluded in w.parts for excluded in excluded_folder_names)
+                    and w.is_file()]
+        wildcard_files.extend(sort_models(resolved, name_has_subpath=True))
        wildcard_files.append("-----")

    return wildcard_files

+def is_umi_format(data):
+    """Returns True if the YAML file is in UMI format."""
+    issue_found = False
+    for item in data:
+        try:
+            if not (data[item] and 'Tags' in data[item] and isinstance(data[item]['Tags'], list)):
+                issue_found = True
+                break
+        except:
+            issue_found = True
+            break
+    return not issue_found

-def get_ext_wildcard_tags():
+count = 0
+def parse_umi_format(umi_tags, data):
+    global count
+    for item in data:
+        umi_tags[count] = ','.join(data[item]['Tags'])
+        count += 1
+
+
+def parse_dynamic_prompt_format(yaml_wildcards, data, path):
+    # Recurse subkeys, delete those without string lists as values
+    def recurse_dict(d: dict):
+        for key, value in d.copy().items():
+            if isinstance(value, dict):
+                recurse_dict(value)
+            elif not (isinstance(value, list) and all(isinstance(v, str) for v in value)):
+                del d[key]
+
+    try:
+        recurse_dict(data)
+        # Add to yaml_wildcards
+        yaml_wildcards[path.name] = data
+    except:
+        return
+
+
+def get_yaml_wildcards():
    """Returns a list of all tags found in extension YAML files found under a Tags: key."""
-    wildcard_tags = {} # { tag: count }
    yaml_files = []
    for path in WILDCARD_EXT_PATHS:
-        yaml_files.extend(p for p in path.rglob("*.yml"))
-        yaml_files.extend(p for p in path.rglob("*.yaml"))
-    count = 0
+        yaml_files.extend(p for p in path.rglob("*.yml") if p.is_file())
+        yaml_files.extend(p for p in path.rglob("*.yaml") if p.is_file())
+
+    yaml_wildcards = {}
+
+    umi_tags = {} # { tag: count }
+
    for path in yaml_files:
        try:
            with open(path, encoding="utf8") as file:
                data = yaml.safe_load(file)
-                if data:
-                    for item in data:
-                        if data[item] and 'Tags' in data[item] and isinstance(data[item]['Tags'], list):
-                            wildcard_tags[count] = ','.join(data[item]['Tags'])
-                            count += 1
-                        else:
-                            print('Issue with tags found in ' + path.name + ' at item ' + item)
+                if (data):
+                    if (is_umi_format(data)):
+                        parse_umi_format(umi_tags, data)
+                    else:
+                        parse_dynamic_prompt_format(yaml_wildcards, data, path)
                else:
                    print('No data found in ' + path.name)
-        except yaml.YAMLError:
-            print('Issue in parsing YAML file ' + path.name                       )
+        except (yaml.YAMLError, UnicodeDecodeError, AttributeError, TypeError) as e:
+            # YAML file not in wildcard format or couldn't be read
+            print(f'Issue in parsing YAML file {path.name}: {e}')
            continue
+        except Exception as e:
+            # Something else went wrong, just skip
+            continue
+
    # Sort by count
-    sorted_tags = sorted(wildcard_tags.items(), key=lambda item: item[1], reverse=True)
-    output = []
-    for tag, count in sorted_tags:
-        output.append(f"{tag},{count}")
-    return output
+    umi_sorted = sorted(umi_tags.items(), key=lambda item: item[1], reverse=True)
+    umi_output = []
+    for tag, count in umi_sorted:
+        umi_output.append(f"{tag},{count}")
+
+    if (len(umi_output) > 0):
+        write_to_temp_file('umi_tags.txt', umi_output)
+
+    with open(TEMP_PATH.joinpath("wc_yaml.json"), "w", encoding="utf-8") as file:
+        json.dump(yaml_wildcards, file, ensure_ascii=False)


 def get_embeddings(sd_model):
@@ -72,48 +217,59 @@ def get_embeddings(sd_model):
    # Version constants
    V1_SHAPE = 768
    V2_SHAPE = 1024
+    VXL_SHAPE = 2048
    emb_v1 = []
    emb_v2 = []
+    emb_vXL = []
+    emb_unknown = []
    results = []

    try:
-        # Get embedding dict from sd_hijack to separate v1/v2 embeddings
-        emb_type_a = sd_hijack.model_hijack.embedding_db.word_embeddings
-        emb_type_b = sd_hijack.model_hijack.embedding_db.skipped_embeddings
-        # Get the shape of the first item in the dict
-        emb_a_shape = -1
-        emb_b_shape = -1
-        if (len(emb_type_a) > 0):
-            emb_a_shape = next(iter(emb_type_a.items()))[1].shape
-        if (len(emb_type_b) > 0):
-            emb_b_shape = next(iter(emb_type_b.items()))[1].shape
+        embed_db = get_embed_db(sd_model)
+        # Re-register callback if needed
+        global load_textual_inversion_embeddings
+        if embed_db is not None and load_textual_inversion_embeddings != embed_db.load_textual_inversion_embeddings:
+            load_textual_inversion_embeddings = embed_db.load_textual_inversion_embeddings
+        
+        loaded = embed_db.word_embeddings
+        skipped = embed_db.skipped_embeddings

        # Add embeddings to the correct list
-        if (emb_a_shape == V1_SHAPE):
-            emb_v1 = list(emb_type_a.keys())
-        elif (emb_a_shape == V2_SHAPE):
-            emb_v2 = list(emb_type_a.keys())
+        for key, emb in (skipped | loaded).items():
+            filename = getattr(emb, "filename", None)
+            
+            if filename is None:
+                if emb.shape is None:
+                    emb_unknown.append((Path(key), key, ""))
+                elif emb.shape == V1_SHAPE:
+                    emb_v1.append((Path(key), key, "v1"))
+                elif emb.shape == V2_SHAPE:
+                    emb_v2.append((Path(key), key, "v2"))
+                elif emb.shape == VXL_SHAPE:
+                    emb_vXL.append((Path(key), key, "vXL"))
+                else:
+                    emb_unknown.append((Path(key), key, ""))
+            
+            else:
+                if emb.filename is None:
+                    continue

-        if (emb_b_shape == V1_SHAPE):
-            emb_v1 = list(emb_type_b.keys())
-        elif (emb_b_shape == V2_SHAPE):
-            emb_v2 = list(emb_type_b.keys())
+                if emb.shape is None:
+                    emb_unknown.append((Path(emb.filename), Path(emb.filename).relative_to(EMB_PATH).as_posix(), ""))
+                elif emb.shape == V1_SHAPE:
+                    emb_v1.append((Path(emb.filename), Path(emb.filename).relative_to(EMB_PATH).as_posix(), "v1"))
+                elif emb.shape == V2_SHAPE:
+                    emb_v2.append((Path(emb.filename), Path(emb.filename).relative_to(EMB_PATH).as_posix(), "v2"))
+                elif emb.shape == VXL_SHAPE:
+                    emb_vXL.append((Path(emb.filename), Path(emb.filename).relative_to(EMB_PATH).as_posix(), "vXL"))
+                else:
+                    emb_unknown.append((Path(emb.filename), Path(emb.filename).relative_to(EMB_PATH).as_posix(), ""))

-        # Get shape of current model
-        #vec = sd_model.cond_stage_model.encode_embedding_init_text(",", 1)
-        #model_shape = vec.shape[1]
-        # Show relevant entries at the top
-        #if (model_shape == V1_SHAPE):
-        #    results = [e + ",v1" for e in emb_v1] + [e + ",v2" for e in emb_v2]
-        #elif (model_shape == V2_SHAPE):
-        #    results = [e + ",v2" for e in emb_v2] + [e + ",v1" for e in emb_v1]
-        #else:
-        #    raise AttributeError # Fallback to old method
-        results = sorted([e + ",v1" for e in emb_v1] + [e + ",v2" for e in emb_v2], key=lambda x: x.lower())
+        results = sort_models(emb_v1) + sort_models(emb_v2) + sort_models(emb_vXL) + sort_models(emb_unknown)
    except AttributeError:
        print("tag_autocomplete_helper: Old webui version or unrecognized model shape, using fallback for embedding completion.")
        # Get a list of all embeddings in the folder
-        all_embeds = [str(e.relative_to(EMB_PATH)) for e in EMB_PATH.rglob("*") if e.suffix in {".bin", ".pt", ".png",'.webp', '.jxl', '.avif'}]
+        all_embeds = [str(e.relative_to(EMB_PATH)) for e in EMB_PATH.rglob("*") if e.suffix in {".bin", ".pt", ".png",'.webp', '.jxl', '.avif'} and e.is_file()]
        # Remove files with a size of 0
        all_embeds = [e for e in all_embeds if EMB_PATH.joinpath(e).stat().st_size > 0]
        # Remove file extensions
@@ -127,51 +283,129 @@ def get_hypernetworks():

    # Get a list of all hypernetworks in the folder
    hyp_paths = [Path(h) for h in glob.glob(HYP_PATH.joinpath("**/*").as_posix(), recursive=True)]
-    all_hypernetworks = [str(h.name) for h in hyp_paths if h.suffix in {".pt"}]
-    # Remove file extensions
-    return sorted([h[:h.rfind('.')] for h in all_hypernetworks], key=lambda x: x.lower())
+    all_hypernetworks = [(h, h.stem) for h in hyp_paths if h.suffix in {".pt"} and h.is_file()]
+    return sort_models(all_hypernetworks)

 model_keyword_installed = write_model_keyword_path()
+
+
+def _get_lora():
+    """
+    Write a list of all lora.
+    Fallback method for when the built-in Lora.networks module is not available.
+    """
+    # Get a list of all lora in the folder
+    lora_paths = [
+        Path(l)
+        for l in glob.glob(LORA_PATH.joinpath("**/*").as_posix(), recursive=True)
+    ]
+    # Get hashes
+    valid_loras = [
+        lf
+        for lf in lora_paths
+        if lf.suffix in {".safetensors", ".ckpt", ".pt"} and lf.is_file()
+    ]
+
+    return valid_loras
+
+
+def _get_lyco():
+    """
+    Write a list of all LyCORIS/LOHA from https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris
+    Fallback method for when the built-in Lora.networks module is not available.
+    """
+    # Get a list of all LyCORIS in the folder
+    lyco_paths = [
+        Path(ly)
+        for ly in glob.glob(LYCO_PATH.joinpath("**/*").as_posix(), recursive=True)
+    ]
+
+    # Get hashes
+    valid_lycos = [
+        lyf
+        for lyf in lyco_paths
+        if lyf.suffix in {".safetensors", ".ckpt", ".pt"} and lyf.is_file()
+    ]
+    return valid_lycos
+
+
+# Attempt to use the build-in Lora.networks Lora/LyCORIS models lists.
+try:
+    import sys
+    from modules import extensions
+    sys.path.append(Path(extensions.extensions_builtin_dir).joinpath("Lora").as_posix())
+    import lora  # pyright: ignore [reportMissingImports]
+
+    def _get_lora():
+        return [
+            Path(model.filename).absolute()
+            for model in lora.available_loras.values()
+            if Path(model.filename).absolute().is_relative_to(LORA_PATH)
+        ]
+
+    def _get_lyco():
+        return [
+            Path(model.filename).absolute()
+            for model in lora.available_loras.values()
+            if Path(model.filename).absolute().is_relative_to(LYCO_PATH)
+        ]
+
+except Exception as e:
+    pass
+    # no need to report
+    # print(f'Exception setting-up performant fetchers: {e}')
+
+
+def is_visible(p: Path) -> bool:
+    if getattr(shared.opts, "extra_networks_hidden_models", "When searched") != "Never":
+        return True
+    for part in p.parts:
+        if part.startswith('.'):
+            return False
+    return True
+
 def get_lora():
    """Write a list of all lora"""
-    global model_keyword_installed
-
-    # Get a list of all lora in the folder
-    lora_paths = [Path(l) for l in glob.glob(LORA_PATH.joinpath("**/*").as_posix(), recursive=True)]
    # Get hashes
-    valid_loras = [lf for lf in lora_paths if lf.suffix in {".safetensors", ".ckpt", ".pt"}]
-    hashes = {}
+    valid_loras = _get_lora()
+    loras_with_hash = []
    for l in valid_loras:
-        name = l.name[:l.name.rfind('.')]
+        if not l.exists() or not l.is_file() or not is_visible(l):
+            continue
+        name = l.relative_to(LORA_PATH).as_posix()
        if model_keyword_installed:
-            hashes[name] = get_lora_simple_hash(l)
+            hash = get_lora_simple_hash(l)
        else:
-            hashes[name] = ""
-    
+            hash = ""
+        loras_with_hash.append((l, name, hash))
    # Sort
-    sorted_loras = dict(sorted(hashes.items()))
-    # Add hashes and return
-    return [f"\"{name}\",{hash}" for name, hash in sorted_loras.items()]
+    return sort_models(loras_with_hash)


 def get_lyco():
    """Write a list of all LyCORIS/LOHA from https://github.com/KohakuBlueleaf/a1111-sd-webui-lycoris"""
-
-    # Get a list of all LyCORIS in the folder
-    lyco_paths = [Path(ly) for ly in glob.glob(LYCO_PATH.joinpath("**/*").as_posix(), recursive=True)]
-    
    # Get hashes
-    valid_lycos = [lyf for lyf in lyco_paths if lyf.suffix in {".safetensors", ".ckpt", ".pt"}]
-    hashes = {}
+    valid_lycos = _get_lyco()
+    lycos_with_hash = []
    for ly in valid_lycos:
-        name = ly.name[:ly.name.rfind('.')]
-        hashes[name] = get_lora_simple_hash(ly)
-    
+        if not ly.exists() or not ly.is_file() or not is_visible(ly):
+            continue
+        name = ly.relative_to(LYCO_PATH).as_posix()
+        if model_keyword_installed:
+            hash = get_lora_simple_hash(ly)
+        else:
+            hash = ""
+        lycos_with_hash.append((ly, name, hash))
    # Sort
-    sorted_lycos = dict(sorted(hashes.items()))
-    # Add hashes and return
-    return [f"\"{name}\",{hash}" for name, hash in sorted_lycos.items()]
+    return sort_models(lycos_with_hash)

+def get_style_names():
+    try:
+        style_names: list[str] = shared.prompt_styles.styles.keys()
+        style_names = sorted(style_names, key=len, reverse=True)
+        return style_names
+    except Exception:
+        return None

 def write_tag_base_path():
    """Writes the tag base path to a fixed location temporary file"""
@@ -187,19 +421,19 @@ def write_to_temp_file(name, data):

 csv_files = []
 csv_files_withnone = []
-def update_tag_files():
+def update_tag_files(*args, **kwargs):
    """Returns a list of all potential tag files"""
    global csv_files, csv_files_withnone
-    files = [str(t.relative_to(TAGS_PATH)) for t in TAGS_PATH.glob("*.csv")]
+    files = [str(t.relative_to(TAGS_PATH)) for t in TAGS_PATH.glob("*.csv") if t.is_file()]
    csv_files = files
    csv_files_withnone = ["None"] + files

 json_files = []
 json_files_withnone = []
-def update_json_files():
+def update_json_files(*args, **kwargs):
    """Returns a list of all potential json files"""
    global json_files, json_files_withnone
-    files = [str(j.relative_to(TAGS_PATH)) for j in TAGS_PATH.glob("*.json")]
+    files = [str(j.relative_to(TAGS_PATH)) for j in TAGS_PATH.glob("*.json") if j.is_file()]
    json_files = files
    json_files_withnone = ["None"] + files

@@ -221,10 +455,12 @@ if not TEMP_PATH.exists():
 # even if no wildcards or embeddings are found
 write_to_temp_file('wc.txt', [])
 write_to_temp_file('wce.txt', [])
-write_to_temp_file('wcet.txt', [])
+write_to_temp_file('wc_yaml.json', [])
+write_to_temp_file('umi_tags.txt', [])
 write_to_temp_file('hyp.txt', [])
 write_to_temp_file('lora.txt', [])
 write_to_temp_file('lyco.txt', [])
+write_to_temp_file('styles.txt', [])
 # Only reload embeddings if the file doesn't exist, since they are already re-written on model load
 if not TEMP_PATH.joinpath("emb.txt").exists():
    write_to_temp_file('emb.txt', [])
@@ -234,28 +470,59 @@ if EMB_PATH.exists():
    # Get embeddings after the model loaded callback
    script_callbacks.on_model_loaded(get_embeddings)

-def refresh_temp_files():
-    write_temp_files()
-    get_embeddings(shared.sd_model)
+def refresh_embeddings(force: bool, *args, **kwargs):
+    try:
+        # Fix for SD.Next infinite refresh loop due to gradio not updating after model load on demand.
+        # This will just skip embedding loading if no model is loaded yet (or there really are no embeddings).
+        # Try catch is just for safety incase sd_hijack access fails for some reason.
+        embed_db = get_embed_db()
+        if embed_db is None:
+            return
+        loaded = embed_db.word_embeddings
+        skipped = embed_db.skipped_embeddings
+        if len((loaded | skipped)) > 0:
+            load_textual_inversion_embeddings(force_reload=force)
+            get_embeddings(None)
+    except Exception:
+        pass

-def write_temp_files():
+def refresh_temp_files(*args, **kwargs):
+    global WILDCARD_EXT_PATHS
+    skip_wildcard_refresh = getattr(shared.opts, "tac_skipWildcardRefresh", False)
+    if skip_wildcard_refresh:
+        WILDCARD_EXT_PATHS = find_ext_wildcard_paths()
+    write_temp_files(skip_wildcard_refresh)
+    force_embed_refresh = getattr(shared.opts, "tac_forceRefreshEmbeddings", False)
+    refresh_embeddings(force=force_embed_refresh)
+
+def write_style_names(*args, **kwargs):
+    styles = get_style_names()
+    if styles:
+        write_to_temp_file('styles.txt', styles)
+
+def write_temp_files(skip_wildcard_refresh = False):
    # Write wildcards to wc.txt if found
-    if WILDCARD_PATH.exists():
-        wildcards = [WILDCARD_PATH.relative_to(FILE_DIR).as_posix()] + get_wildcards()
+    if WILDCARD_PATH.exists() and not skip_wildcard_refresh:
+        try:
+            # Attempt to create a relative path, but fall back to an absolute path if not possible
+            relative_wildcard_path = WILDCARD_PATH.relative_to(FILE_DIR).as_posix()
+        except ValueError:
+            # If the paths are not relative, use the absolute path
+            relative_wildcard_path = WILDCARD_PATH.as_posix()
+
+        wildcards = [relative_wildcard_path] + get_wildcards()
        if wildcards:
            write_to_temp_file('wc.txt', wildcards)

    # Write extension wildcards to wce.txt if found
-    if WILDCARD_EXT_PATHS is not None:
+    if WILDCARD_EXT_PATHS is not None and not skip_wildcard_refresh:
        wildcards_ext = get_ext_wildcards()
        if wildcards_ext:
            write_to_temp_file('wce.txt', wildcards_ext)
-        # Write yaml extension wildcards to wcet.txt if found
-        wildcards_yaml_ext = get_ext_wildcard_tags()
-        if wildcards_yaml_ext:
-            write_to_temp_file('wcet.txt', wildcards_yaml_ext)
+        # Write yaml extension wildcards to umi_tags.txt and wc_yaml.json if found
+        get_yaml_wildcards()

-    if HYP_PATH.exists():
+    if HYP_PATH is not None and HYP_PATH.exists():
        hypernets = get_hypernetworks()
        if hypernets:
            write_to_temp_file('hyp.txt', hypernets)
@@ -268,7 +535,7 @@ def write_temp_files():
        lora = get_lora()
        if lora:
            write_to_temp_file('lora.txt', lora)
-            
+
    lyco_exists = LYCO_PATH is not None and LYCO_PATH.exists()
    if lyco_exists and not (lora_exists and LYCO_PATH.samefile(LORA_PATH)):
        lyco = get_lyco()
@@ -280,6 +547,8 @@ def write_temp_files():
    if model_keyword_installed:
        update_hash_cache()

+    if shared.prompt_styles is not None:
+        write_style_names()

 write_temp_files()

@@ -299,6 +568,13 @@ def on_ui_settings():
            return self
        shared.OptionInfo.needs_restart = needs_restart

+    # Dictionary of function options and their explanations
+    frequency_sort_functions = {
+        "Logarithmic (weak)": "Will respect the base order and slightly prefer often used tags",
+        "Logarithmic (strong)": "Same as Logarithmic (weak), but with a stronger bias",
+        "Usage first": "Will list used tags by frequency before all others",
+    }
+
    tac_options = {
        # Main tag file
        "tac_tagFile": shared.OptionInfo("danbooru.csv", "Tag filename", gr.Dropdown, lambda: {"choices": csv_files_withnone}, refresh=update_tag_files),
@@ -317,18 +593,36 @@ def on_ui_settings():
        "tac_resultStepLength": shared.OptionInfo(100, "How many results to load at once"),
        "tac_delayTime": shared.OptionInfo(100, "Time in ms to wait before triggering completion again").needs_restart(),
        "tac_useWildcards": shared.OptionInfo(True, "Search for wildcards"),
+        "tac_sortWildcardResults": shared.OptionInfo(True, "Sort wildcard file contents alphabetically").info("If your wildcard files have a specific custom order, disable this to keep it"),
+        "tac_wildcardExclusionList": shared.OptionInfo("", "Wildcard folder exclusion list").info("Add folder names that shouldn't be searched for wildcards, separated by comma.").needs_restart(),
+        "tac_skipWildcardRefresh": shared.OptionInfo(False, "Don't re-scan for wildcard files when pressing the extra networks refresh button").info("Useful to prevent hanging if you use a very large wildcard collection."),
        "tac_useEmbeddings": shared.OptionInfo(True, "Search for embeddings"),
+        "tac_forceRefreshEmbeddings": shared.OptionInfo(False, "Force refresh embeddings when pressing the extra networks refresh button").info("Turn this on if you have issues with new embeddings not registering correctly in TAC. Warning: Seems to cause reloading issues in gradio for some users."),
+        "tac_includeEmbeddingsInNormalResults": shared.OptionInfo(False, "Include embeddings in normal tag results").info("The 'JumpTo...' keybinds (End & Home key by default) will select the first non-embedding result of their direction on the first press for quick navigation in longer lists."),
        "tac_useHypernetworks": shared.OptionInfo(True, "Search for hypernetworks"),
        "tac_useLoras": shared.OptionInfo(True, "Search for Loras"),
        "tac_useLycos": shared.OptionInfo(True, "Search for LyCORIS/LoHa"),
+        "tac_useLoraPrefixForLycos": shared.OptionInfo(True, "Use the '<lora:' prefix instead of '<lyco:' for models in the LyCORIS folder").info("The lyco prefix is included for backwards compatibility and not used anymore by default. Disable this if you are on an old webui version without built-in lyco support."),
        "tac_showWikiLinks": shared.OptionInfo(False, "Show '?' next to tags, linking to its Danbooru or e621 wiki page").info("Warning: This is an external site and very likely contains NSFW examples!"),
+        "tac_showExtraNetworkPreviews": shared.OptionInfo(True, "Show preview thumbnails for extra networks if available"),
+        "tac_modelSortOrder": shared.OptionInfo("Name", "Model sort order", gr.Dropdown, lambda: {"choices": list(sort_criteria.keys())}).info("Order for extra network models and wildcards in dropdown"),
+        "tac_useStyleVars": shared.OptionInfo(False, "Search for webui style names").info("Suggests style names from the webui dropdown with '$'. Currently requires a secondary extension like <a href=\"https://github.com/SirVeggie/extension-style-vars\" target=\"_blank\">style-vars</a> to actually apply the styles before generating."),
+        # Frequency sorting settings
+        "tac_frequencySort": shared.OptionInfo(True, "Locally record tag usage and sort frequent tags higher").info("Will also work for extra networks, keeping the specified base order"),
+        "tac_frequencyFunction": shared.OptionInfo("Logarithmic (weak)", "Function to use for frequency sorting", gr.Dropdown, lambda: {"choices": list(frequency_sort_functions.keys())}).info("; ".join([f'<b>{key}</b>: {val}' for key, val in frequency_sort_functions.items()])),
+        "tac_frequencyMinCount": shared.OptionInfo(3, "Minimum number of uses for a tag to be considered frequent").info("Tags with less uses than this will not be sorted higher, even if the sorting function would normally result in a higher position."),
+        "tac_frequencyMaxAge": shared.OptionInfo(30, "Maximum days since last use for a tag to be considered frequent").info("Similar to the above, tags that haven't been used in this many days will not be sorted higher. Set to 0 to disable."),
+        "tac_frequencyRecommendCap": shared.OptionInfo(10, "Maximum number of recommended tags").info("Limits the maximum number of recommended tags to not drown out normal results. Set to 0 to disable."),
+        "tac_frequencyIncludeAlias": shared.OptionInfo(False, "Frequency sorting matches aliases for frequent tags").info("Tag frequency will be increased for the main tag even if an alias is used for completion. This option can be used to override the default behavior of alias results being ignored for frequency sorting."),
        # Insertion related settings
        "tac_replaceUnderscores": shared.OptionInfo(True, "Replace underscores with spaces on insertion"),
+        "tac_undersocreReplacementExclusionList": shared.OptionInfo("0_0,(o)_(o),+_+,+_-,._.,<o>_<o>,<|>_<|>,=_=,>_<,3_3,6_9,>_o,@_@,^_^,o_o,u_u,x_x,|_|,||_||", "Underscore replacement exclusion list").info("Add tags that shouldn't have underscores replaced with spaces, separated by comma."),
        "tac_escapeParentheses": shared.OptionInfo(True, "Escape parentheses on insertion"),
        "tac_appendComma": shared.OptionInfo(True, "Append comma on tag autocompletion"),
        "tac_appendSpace": shared.OptionInfo(True, "Append space on tag autocompletion").info("will append after comma if the above is enabled"),
        "tac_alwaysSpaceAtEnd": shared.OptionInfo(True, "Always append space if inserting at the end of the textbox").info("takes precedence over the regular space setting for that position"),
-        "tac_modelKeywordCompletion": shared.OptionInfo("Never", "Try to add known trigger words for LORA/LyCO models", gr.Dropdown, lambda: {"interactive": model_keyword_installed, "choices": ["Never","Only user list","Always"]}).info("Requires the <a href=\"https://github.com/mix1009/model-keyword\" target=\"_blank\">model-keyword</a> extension to be installed, but will work with it disabled.").needs_restart(),
+        "tac_modelKeywordCompletion": shared.OptionInfo("Never", "Try to add known trigger words for LORA/LyCO models", gr.Dropdown, lambda: {"choices": ["Never","Only user list","Always"]}).info("Will use & prefer the native activation keywords settable in the extra networks UI. Other functionality requires the <a href=\"https://github.com/mix1009/model-keyword\" target=\"_blank\">model-keyword</a> extension to be installed, but will work with it disabled.").needs_restart(),
+        "tac_modelKeywordLocation": shared.OptionInfo("Start of prompt", "Where to insert the trigger keyword", gr.Dropdown, lambda: {"choices": ["Start of prompt","End of prompt","Before LORA/LyCO"]}).info("Only relevant if the above option is enabled"),
        "tac_wildcardCompletionMode": shared.OptionInfo("To next folder level", "How to complete nested wildcard paths", gr.Dropdown, lambda: {"choices": ["To next folder level","To first difference","Always fully"]}).info("e.g. \"hair/colours/light/...\""),
        # Alias settings
        "tac_alias.searchByAlias": shared.OptionInfo(True, "Search by alias"),
@@ -385,6 +679,37 @@ def on_ui_settings():
        "6": ["red", "maroon"],
        "7": ["whitesmoke", "black"],
        "8": ["seagreen", "darkseagreen"]
+    },
+    "derpibooru": {
+        "-1": ["red", "maroon"],
+        "0": ["#60d160", "#3d9d3d"],
+        "1": ["#fff956", "#918e2e"],
+        "3": ["#fd9961", "#a14c2e"],
+        "4": ["#cf5bbe", "#6c1e6c"],
+        "5": ["#3c8ad9", "#1e5e93"],
+        "6": ["#a6a6a6", "#555555"],
+        "7": ["#47abc1", "#1f6c7c"],
+        "8": ["#7871d0", "#392f7d"],
+        "9": ["#df3647", "#8e1c2b"],
+        "10": ["#c98f2b", "#7b470e"],
+        "11": ["#e87ebe", "#a83583"]
+    },
+    "danbooru_e621_merged": {
+        "-1": ["red", "maroon"],
+        "0": ["lightblue", "dodgerblue"],
+        "1": ["indianred", "firebrick"],
+        "3": ["violet", "darkorchid"],
+        "4": ["lightgreen", "darkgreen"],
+        "5": ["orange", "darkorange"],
+        "6": ["red", "maroon"],
+        "7": ["lightblue", "dodgerblue"],
+        "8": ["gold", "goldenrod"],
+        "9": ["gold", "goldenrod"],
+        "10": ["violet", "darkorchid"],
+        "11": ["lightgreen", "darkgreen"],
+        "12": ["tomato", "darksalmon"],
+        "14": ["whitesmoke", "black"],
+        "15": ["seagreen", "darkseagreen"]
    }
 }\
 """
@@ -399,5 +724,188 @@ def on_ui_settings():
        shared.opts.add_option("tac_colormap", shared.OptionInfo(colorDefault, colorLabel, gr.Textbox, section=TAC_SECTION))

    shared.opts.add_option("tac_refreshTempFiles", shared.OptionInfo("Refresh TAC temp files", "Refresh internal temp files", gr.HTML, {}, refresh=refresh_temp_files, section=TAC_SECTION))
-    
+
 script_callbacks.on_ui_settings(on_ui_settings)
+
+def get_style_mtime():
+    try:
+        style_file = getattr(shared, "styles_filename", "styles.csv")
+        # Check in case a list is returned
+        if isinstance(style_file, list):
+            style_file = style_file[0]
+        
+        style_file = Path(FILE_DIR).joinpath(style_file)
+        if Path.exists(style_file):
+            return style_file.stat().st_mtime
+    except Exception:
+        return None
+
+last_style_mtime = get_style_mtime()
+
+def api_tac(_: gr.Blocks, app: FastAPI):
+    async def get_json_info(base_path: Path, filename: str = None):
+        if base_path is None or (not base_path.exists()):
+            return Response(status_code=404)
+
+        try:
+            json_candidates = glob.glob(base_path.as_posix() + f"/**/{glob.escape(filename)}.json", recursive=True)
+            if json_candidates is not None and len(json_candidates) > 0 and Path(json_candidates[0]).is_file():
+                return FileResponse(json_candidates[0])
+        except Exception as e:
+            return JSONResponse({"error": e}, status_code=500)
+
+    async def get_preview_thumbnail(base_path: Path, filename: str = None, blob: bool = False):
+        if base_path is None or (not base_path.exists()):
+            return Response(status_code=404)
+
+        try:
+            img_glob = glob.glob(base_path.as_posix() + f"/**/{glob.escape(filename)}.*", recursive=True)
+            img_candidates = [img for img in img_glob if Path(img).suffix in [".png", ".jpg", ".jpeg", ".webp", ".gif"] and Path(img).is_file()]
+            if img_candidates is not None and len(img_candidates) > 0:
+                if blob:
+                    return FileResponse(img_candidates[0])
+                else:
+                    return JSONResponse({"url": urllib.parse.quote(img_candidates[0])})
+        except Exception as e:
+            return JSONResponse({"error": e}, status_code=500)
+
+    @app.post("/tacapi/v1/refresh-temp-files")
+    async def api_refresh_temp_files():
+        await sleep(0) # might help with refresh blocking gradio
+        refresh_temp_files()
+
+    @app.post("/tacapi/v1/refresh-embeddings")
+    async def api_refresh_embeddings():
+        refresh_embeddings(force=False)
+
+    @app.get("/tacapi/v1/lora-info/{lora_name}")
+    async def get_lora_info(lora_name):
+        return await get_json_info(LORA_PATH, lora_name)
+
+    @app.get("/tacapi/v1/lyco-info/{lyco_name}")
+    async def get_lyco_info(lyco_name):
+        return await get_json_info(LYCO_PATH, lyco_name)
+
+    @app.get("/tacapi/v1/lora-cached-hash/{lora_name}")
+    async def get_lora_cached_hash(lora_name: str):
+        path_glob = glob.glob(LORA_PATH.as_posix() + f"/**/{glob.escape(lora_name)}.*", recursive=True)
+        paths = [lora for lora in path_glob if Path(lora).suffix in [".safetensors", ".ckpt", ".pt"] and Path(lora).is_file()]
+        if paths is not None and len(paths) > 0:
+            path = paths[0]
+            hash = hashes.sha256_from_cache(path, f"lora/{lora_name}", path.endswith(".safetensors"))
+            if hash is not None:
+                return hash
+        
+        return None
+
+    def get_path_for_type(type):
+        if type == "lora":
+            return LORA_PATH
+        elif type == "lyco":
+            return LYCO_PATH
+        elif type == "hypernetwork":
+            return HYP_PATH
+        elif type == "embedding":
+            return EMB_PATH
+        else:
+            return None
+
+    @app.get("/tacapi/v1/thumb-preview/{filename}")
+    async def get_thumb_preview(filename, type):
+        return await get_preview_thumbnail(get_path_for_type(type), filename, False)
+
+    @app.get("/tacapi/v1/thumb-preview-blob/{filename}")
+    async def get_thumb_preview_blob(filename, type):
+        return await get_preview_thumbnail(get_path_for_type(type), filename, True)
+
+    @app.get("/tacapi/v1/wildcard-contents")
+    async def get_wildcard_contents(basepath: str, filename: str):
+        if basepath is None or basepath == "":
+            return Response(status_code=404)
+
+        base = Path(basepath)
+        if base is None or (not base.exists()):
+            return Response(status_code=404)
+
+        try:
+            wildcard_path = base.joinpath(filename)
+            if wildcard_path.exists() and wildcard_path.is_file():
+                return FileResponse(wildcard_path)
+            else:
+                return Response(status_code=404)
+        except Exception as e:
+            return JSONResponse({"error": e}, status_code=500)
+
+    @app.get("/tacapi/v1/refresh-styles-if-changed")
+    async def refresh_styles_if_changed():
+        global last_style_mtime
+        
+        mtime = get_style_mtime()
+        if mtime is not None and mtime > last_style_mtime:
+            last_style_mtime = mtime
+            # Update temp file
+            if shared.prompt_styles is not None:
+                write_style_names()
+            
+            return Response(status_code=200) # Success
+        else:
+            return Response(status_code=304) # Not modified
+    def db_request(func, get = False):
+        if db is not None:
+            try:
+                if get:
+                    ret = func()
+                    if ret is list:
+                        ret = [{"name": t[0], "type": t[1], "count": t[2], "lastUseDate": t[3]} for t in ret]
+                    return JSONResponse({"result": ret})
+                else:
+                    func()
+            except sqlite3.Error as e:
+                return JSONResponse({"error": e.__cause__}, status_code=500)
+        else:
+            return JSONResponse({"error": "Database not initialized"}, status_code=500)
+
+    @app.post("/tacapi/v1/increase-use-count")
+    async def increase_use_count(tagname: str, ttype: int, neg: bool):
+        db_request(lambda: db.increase_tag_count(tagname, ttype, neg))
+
+    @app.get("/tacapi/v1/get-use-count")
+    async def get_use_count(tagname: str, ttype: int, neg: bool):
+        return db_request(lambda: db.get_tag_count(tagname, ttype, neg), get=True)
+    
+    # Small dataholder class
+    class UseCountListRequest(BaseModel):
+        tagNames: list[str]
+        tagTypes: list[int]
+        neg: bool = False
+
+    # Semantically weird to use post here, but it's required for the body on js side
+    @app.post("/tacapi/v1/get-use-count-list")
+    async def get_use_count_list(body: UseCountListRequest):
+        # If a date limit is set > 0, pass it to the db
+        date_limit = getattr(shared.opts, "tac_frequencyMaxAge", 30)
+        date_limit = date_limit if date_limit > 0 else None
+
+        if db:
+            count_list = list(db.get_tag_counts(body.tagNames, body.tagTypes, body.neg, date_limit))
+        else:
+            count_list = None
+    
+        # If a limit is set, return at max the top n results by count
+        if count_list and len(count_list):
+            limit = int(min(getattr(shared.opts, "tac_frequencyRecommendCap", 10), len(count_list)))
+            # Sort by count and return the top n
+            if limit > 0:
+                count_list = sorted(count_list, key=lambda x: x[2], reverse=True)[:limit]
+
+        return db_request(lambda: count_list, get=True)
+
+    @app.put("/tacapi/v1/reset-use-count")
+    async def reset_use_count(tagname: str, ttype: int, pos: bool, neg: bool):
+        db_request(lambda: db.reset_tag_count(tagname, ttype, pos, neg))
+
+    @app.get("/tacapi/v1/get-all-use-counts")
+    async def get_all_tag_counts():
+        return db_request(lambda: db.get_all_tags(), get=True)
+
+script_callbacks.on_app_started(api_tac)
--- a/scripts/tag_frequency_db.py
+++ b/scripts/tag_frequency_db.py
@@ -0,0 +1,190 @@
+import sqlite3
+from contextlib import contextmanager
+
+from scripts.shared_paths import TAGS_PATH
+
+db_file = TAGS_PATH.joinpath("tag_frequency.db")
+timeout = 30
+db_ver = 1
+
+
+@contextmanager
+def transaction(db=db_file):
+    """Context manager for database transactions.
+    Ensures that the connection is properly closed after the transaction.
+    """
+    try:
+        conn = sqlite3.connect(db, timeout=timeout)
+        
+        conn.isolation_level = None
+        cursor = conn.cursor()
+        cursor.execute("BEGIN")
+        yield cursor
+        cursor.execute("COMMIT")
+    except sqlite3.Error as e:
+        print("Tag Autocomplete: Frequency database error:", e)
+    finally:
+        if conn:
+            conn.close()
+
+
+class TagFrequencyDb:
+    """Class containing creation and interaction methods for the tag frequency database"""
+
+    def __init__(self) -> None:
+        self.version = self.__check()
+
+    def __check(self):
+        if not db_file.exists():
+            print("Tag Autocomplete: Creating frequency database")
+            with transaction() as cursor:
+                self.__create_db(cursor)
+                self.__update_db_data(cursor, "version", db_ver)
+            print("Tag Autocomplete: Database successfully created")
+
+        return self.__get_version()
+
+    def __create_db(self, cursor: sqlite3.Cursor):
+        cursor.execute(
+            """
+        CREATE TABLE IF NOT EXISTS db_data (
+            key TEXT PRIMARY KEY,
+            value TEXT
+        )
+        """
+        )
+
+        cursor.execute(
+            """
+        CREATE TABLE IF NOT EXISTS tag_frequency (
+            name TEXT NOT NULL,
+            type INT NOT NULL,
+            count_pos INT,
+            count_neg INT,
+            last_used TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+            PRIMARY KEY (name, type)
+        )
+        """
+        )
+
+    def __update_db_data(self, cursor: sqlite3.Cursor, key, value):
+        cursor.execute(
+            """
+        INSERT OR REPLACE
+        INTO db_data (key, value)
+        VALUES (?, ?)
+        """,
+            (key, value),
+        )
+
+    def __get_version(self):
+        db_version = None
+        with transaction() as cursor:
+            cursor.execute(
+                """
+            SELECT value
+            FROM db_data
+            WHERE key = 'version'
+            """
+            )
+            db_version = cursor.fetchone()
+
+        return db_version[0] if db_version else 0
+
+    def get_all_tags(self):
+        with transaction() as cursor:
+            cursor.execute(
+                f"""
+            SELECT name, type, count_pos, count_neg, last_used
+            FROM tag_frequency
+            WHERE count_pos > 0 OR count_neg > 0
+            ORDER BY count_pos + count_neg DESC
+            """
+            )
+            tags = cursor.fetchall()
+
+        return tags
+
+    def get_tag_count(self, tag, ttype, negative=False):
+        count_str = "count_neg" if negative else "count_pos"
+        with transaction() as cursor:
+            cursor.execute(
+                f"""
+            SELECT {count_str}, last_used
+            FROM tag_frequency
+            WHERE name = ? AND type = ?
+            """,
+                (tag, ttype),
+            )
+            tag_count = cursor.fetchone()
+
+        if tag_count:
+            return tag_count[0], tag_count[1]
+        else:
+            return 0, None
+
+    def get_tag_counts(self, tags: list[str], ttypes: list[str], negative=False, date_limit=None):
+        count_str = "count_neg" if negative else "count_pos"
+        with transaction() as cursor:
+            for tag, ttype in zip(tags, ttypes):
+                if date_limit is not None:
+                    cursor.execute(
+                        f"""
+                    SELECT {count_str}, last_used
+                    FROM tag_frequency
+                    WHERE name = ? AND type = ?
+                    AND last_used > datetime('now', '-' || ? || ' days')
+                    """,
+                        (tag, ttype, date_limit),
+                    )
+                else:
+                    cursor.execute(
+                        f"""
+                    SELECT {count_str}, last_used
+                    FROM tag_frequency
+                    WHERE name = ? AND type = ?
+                    """,
+                        (tag, ttype),
+                    )
+                tag_count = cursor.fetchone()
+                if tag_count:
+                    yield (tag, ttype, tag_count[0], tag_count[1]) 
+                else:
+                    yield (tag, ttype, 0, None)
+
+    def increase_tag_count(self, tag, ttype, negative=False):
+        pos_count = self.get_tag_count(tag, ttype, False)[0]
+        neg_count = self.get_tag_count(tag, ttype, True)[0]
+
+        if negative:
+            neg_count += 1
+        else:
+            pos_count += 1
+
+        with transaction() as cursor:
+            cursor.execute(
+                f"""
+            INSERT OR REPLACE
+            INTO tag_frequency (name, type, count_pos, count_neg)
+            VALUES (?, ?, ?, ?)
+            """,
+                (tag, ttype, pos_count, neg_count),
+            )
+
+    def reset_tag_count(self, tag, ttype, positive=True, negative=False):
+        if positive and negative:
+            set_str = "count_pos = 0, count_neg = 0"
+        elif positive:
+            set_str = "count_pos = 0"
+        elif negative:
+            set_str = "count_neg = 0"
+
+        with transaction() as cursor:
+            cursor.execute(
+                f"""
+            UPDATE tag_frequency
+            SET {set_str}
+            WHERE name = ? AND type = ?
+            """,
+                (tag, ttype),
+            )
--- a/tags/EnglishDictionary.csv
+++ b/tags/EnglishDictionary.csv
--- a/tags/danbooru.csv
+++ b/tags/danbooru.csv
--- a/tags/danbooru_e621_merged.csv
+++ b/tags/danbooru_e621_merged.csv
--- a/tags/demo-chants.json
+++ b/tags/demo-chants.json
@@ -28,5 +28,17 @@
        "terms": "Water, Magic, Fancy",
        "content": "(extremely detailed CG unity 8k wallpaper), (masterpiece), (best quality), (ultra-detailed), (best illustration),(best shadow), (an extremely delicate and beautiful), classic, dynamic angle, floating, fine detail, Depth of field, classic, (painting), (sketch), (bloom), (shine), glinting stars,\n\na girl, solo, bare shoulders, flat chest, diamond and glaring eyes, beautiful detailed cold face, very long blue and sliver hair, floating black feathers, wavy hair, extremely delicate and beautiful girls, beautiful detailed eyes, glowing eyes,\n\nriver, (forest),palace, (fairyland,feather,flowers, nature),(sunlight),Hazy fog, mist",
        "color": 5
+    },
+    {
+        "name": "Pony-Positive",
+        "terms": "Pony,Score,Positive,Quality",
+        "content": "score_9, score_8_up, score_7_up, score_6_up, source_anime, source_furry, source_pony, source_cartoon",
+        "color": 1
+    },
+    {
+        "name": "Pony-Negative",
+        "terms": "Pony,Score,Negative,Quality",
+        "content": "score_1, score_2, score_3, score_4, score_5, source_anime, source_furry, source_pony, source_cartoon",
+        "color": 3
    }
 ]
--- a/tags/derpibooru.csv
+++ b/tags/derpibooru.csv
--- a/tags/e621.csv
+++ b/tags/e621.csv
--- a/tags/e621_sfw.csv
+++ b/tags/e621_sfw.csv
--- a/tags/noob_characters-chants.json
+++ b/tags/noob_characters-chants.json