Upload sort_danbooru2026_by_category.ipynb

sort_danbooru2026_by_category.ipynb

@@ -0,0 +1,649 @@
+{
+  "nbformat": 4,
+  "nbformat_minor": 0,
+  "metadata": {
+    "colab": {
+      "provenance": [],
+      "gpuType": "T4",
+      "collapsed_sections": [
+        "PvAtznhla3jS"
+      ]
+    },
+    "kernelspec": {
+      "name": "python3",
+      "display_name": "Python 3"
+    },
+    "language_info": {
+      "name": "python"
+    },
+    "accelerator": "GPU"
+  },
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "source": [
+        "Using https://huggingface.co/datasets/RicemanT/booru-essence-2026\n",
+        "\n",
+        "* Dowload the parquet from the huggingface huggingface repo and save to your google drive as /content/drive/MyDrive/booru_essence.parquet  and  /content/drive/MyDrive/e621_essence.parquet respectively"
+      ],
+      "metadata": {
+        "id": "oSQKdE5VYMoS"
+      }
+    },
+    {
+      "cell_type": "markdown",
+      "source": [
+        "# Parquet Explorer"
+      ],
+      "metadata": {
+        "id": "PvAtznhla3jS"
+      }
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "from google.colab import drive\n",
+        "drive.mount('/content/drive')"
+      ],
+      "metadata": {
+        "id": "HFy5aDxM3G7O",
+        "outputId": "6bab9c7a-5b63-4e1b-8698-5468b0715068",
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        }
+      },
+      "execution_count": null,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "Mounted at /content/drive\n"
+          ]
+        }
+      ]
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "# Step 1: Mount Google Drive\n",
+        "#from google.colab import drive\n",
+        "#drive.mount('/content/drive')\n",
+        "\n",
+        "#@markdown paste .parquet file stored on your Google Drive folder to see its characteristics\n",
+        "\n",
+        "# Step 2: Import required libraries\n",
+        "import pandas as pd\n",
+        "\n",
+        "# Step 3: Define the path to the Parquet file\n",
+        "file_path = '/content/drive/MyDrive/Saved from Chrome/danbooru.parquet' #@param {type:'string'}\n",
+        "\n",
+        "# Step 4: Read the Parquet file\n",
+        "df = pd.read_parquet(file_path)\n",
+        "\n",
+        "# Step 5: Basic exploration of the Parquet file\n",
+        "print(\"First 5 rows of the dataset:\")\n",
+        "print(df.head())\n",
+        "\n",
+        "print(\"\\nDataset Info:\")\n",
+        "print(df.info())\n",
+        "\n",
+        "print(\"\\nBasic Statistics:\")\n",
+        "print(df.describe())\n",
+        "\n",
+        "print(\"\\nColumn Names:\")\n",
+        "print(df.columns.tolist())\n",
+        "\n",
+        "print(\"\\nMissing Values:\")\n",
+        "print(df.isnull().sum())\n",
+        "\n",
+        "# Optional: Display number of rows and columns\n",
+        "print(f\"\\nShape of the dataset: {df.shape}\")"
+      ],
+      "metadata": {
+        "id": "So-PKtbo5AVA",
+        "colab": {
+          "base_uri": "https://localhost:8080/"
+        },
+        "outputId": "4b183874-827d-4822-872b-fea7e5d3ae03"
+      },
+      "execution_count": null,
+      "outputs": [
+        {
+          "output_type": "stream",
+          "name": "stdout",
+          "text": [
+            "First 5 rows of the dataset:\n",
+            "                                md5                               image_path  \\\n",
+            "0  0004100ffd82a677a13706c7a7049b3c  00/0004100ffd82a677a13706c7a7049b3c.png   \n",
+            "1  00041bad25f261b96eaa15517dd06cb6  00/00041bad25f261b96eaa15517dd06cb6.jpg   \n",
+            "2  000665c12cc1638ce15286a7972f5fc2  00/000665c12cc1638ce15286a7972f5fc2.jpg   \n",
+            "3  001193d9df978b62b0100d12fa6cb270  00/001193d9df978b62b0100d12fa6cb270.jpg   \n",
+            "4  0011caa6d25cab0016c0dedb523085b1  00/0011caa6d25cab0016c0dedb523085b1.jpg   \n",
+            "\n",
+            "     source     rating caption_bucket  \\\n",
+            "0  danbooru  sensitive           safe   \n",
+            "1  danbooru  sensitive           safe   \n",
+            "2  danbooru  sensitive           safe   \n",
+            "3  danbooru  sensitive           safe   \n",
+            "4  danbooru  sensitive           safe   \n",
+            "\n",
+            "                                                tags  \\\n",
+            "0  zhixie jiaobu, flandre scarlet, touhou, 1girl,...   \n",
+            "1  kamishima kanon, gogeta, hong meiling, izayoi ...   \n",
+            "2  ichi (ichikai), kazami yuuka, kazami yuuka (pc...   \n",
+            "3  hoshikuzu pan, princess peach, princess peach ...   \n",
+            "4  chiwino, mimi houllier von schwarzlang, totoor...   \n",
+            "\n",
+            "                                             caption caption_model  \\\n",
+            "0  Drawn by zhixie jiaobu. Flandre Scarlet from T...            nl   \n",
+            "1  Drawn by kamishima kanon. Gogeta from Dragon B...            nl   \n",
+            "2  Drawn by ichi (ichikai). Kazami Yuuka and Koch...            nl   \n",
+            "3  Drawn by hoshikuzu pan. Princess Peach and Ros...            nl   \n",
+            "4  Drawn by chiwino. Totooria Helmold and Mimi Ho...            nl   \n",
+            "\n",
+            "                trigger_tag                               url  ...  \\\n",
+            "0    Drawn by zhixie jiaobu  0004100ffd82a677a13706c7a7049b3c  ...   \n",
+            "1  Drawn by kamishima kanon  00041bad25f261b96eaa15517dd06cb6  ...   \n",
+            "2   Drawn by ichi (ichikai)  000665c12cc1638ce15286a7972f5fc2  ...   \n",
+            "3    Drawn by hoshikuzu pan  001193d9df978b62b0100d12fa6cb270  ...   \n",
+            "4          Drawn by chiwino  0011caa6d25cab0016c0dedb523085b1  ...   \n",
+            "\n",
+            "   image_height  fav_count                       created_at solo_subject  \\\n",
+            "0          2480         35 2018-04-10 18:19:04.307000-07:00         True   \n",
+            "1          1662         83 2012-09-27 07:17:12.405000-07:00         True   \n",
+            "2          1692        135 2009-05-24 14:37:17.815000-07:00        False   \n",
+            "3          1275        153 2018-03-23 20:46:43.043000-07:00        False   \n",
+            "4           850         99 2011-06-22 21:33:11.525000-07:00        False   \n",
+            "\n",
+            "   multi_subject                              artist_weights  \\\n",
+            "0          False   {\"zhixie_jiaobu\": 2.6965785370217326e-05}   \n",
+            "1          False  {\"kamishima_kanon\": 1.632874605151334e-05}   \n",
+            "2           True  {\"ichi_(ichikai)\": 1.9545675811406287e-05}   \n",
+            "3           True    {\"hoshikuzu_pan\": 2.538654628444895e-05}   \n",
+            "4           True         {\"chiwino\": 1.8019746011716172e-05}   \n",
+            "\n",
+            "                                   character_weights species_weights  \\\n",
+            "0         {\"flandre_scarlet\": 5.549727045990053e-06}              {}   \n",
+            "1  {\"gogeta\": 2.1352518798382155e-05, \"hong_meili...              {}   \n",
+            "2  {\"kazami_yuuka\": 7.954217649439972e-06, \"kochi...              {}   \n",
+            "3  {\"princess_peach\": 9.696893506992817e-06, \"ros...              {}   \n",
+            "4  {\"mimi_houllier_von_schwarzlang\": 3.4177584640...              {}   \n",
+            "\n",
+            "  post_id_cur  post_id_old  \n",
+            "0   3082788.0         None  \n",
+            "1   1264640.0         None  \n",
+            "2    459758.0         None  \n",
+            "3   3060933.0         None  \n",
+            "4    940637.0         None  \n",
+            "\n",
+            "[5 rows x 21 columns]\n",
+            "\n",
+            "Dataset Info:\n",
+            "<class 'pandas.core.frame.DataFrame'>\n",
+            "RangeIndex: 19987 entries, 0 to 19986\n",
+            "Data columns (total 21 columns):\n",
+            " #   Column             Non-Null Count  Dtype                              \n",
+            "---  ------             --------------  -----                              \n",
+            " 0   md5                19987 non-null  object                             \n",
+            " 1   image_path         19987 non-null  object                             \n",
+            " 2   source             19987 non-null  object                             \n",
+            " 3   rating             19987 non-null  object                             \n",
+            " 4   caption_bucket     19987 non-null  object                             \n",
+            " 5   tags               19987 non-null  object                             \n",
+            " 6   caption            19987 non-null  object                             \n",
+            " 7   caption_model      19987 non-null  object                             \n",
+            " 8   trigger_tag        19987 non-null  object                             \n",
+            " 9   url                19987 non-null  object                             \n",
+            " 10  image_width        19987 non-null  int64                              \n",
+            " 11  image_height       19987 non-null  int64                              \n",
+            " 12  fav_count          19987 non-null  int64                              \n",
+            " 13  created_at         19987 non-null  datetime64[us, America/Los_Angeles]\n",
+            " 14  solo_subject       19987 non-null  bool                               \n",
+            " 15  multi_subject      19987 non-null  bool                               \n",
+            " 16  artist_weights     19987 non-null  object                             \n",
+            " 17  character_weights  19987 non-null  object                             \n",
+            " 18  species_weights    19987 non-null  object                             \n",
+            " 19  post_id_cur        19967 non-null  float64                            \n",
+            " 20  post_id_old        0 non-null      object                             \n",
+            "dtypes: bool(2), datetime64[us, America/Los_Angeles](1), float64(1), int64(3), object(14)\n",
+            "memory usage: 2.9+ MB\n",
+            "None\n",
+            "\n",
+            "Basic Statistics:\n",
+            "        image_width  image_height     fav_count   post_id_cur\n",
+            "count  19987.000000  19987.000000  19987.000000  1.996700e+04\n",
+            "mean    1631.031120   1929.382298    211.866563  4.612090e+06\n",
+            "std     1016.239287   1201.701401    208.898400  2.072274e+06\n",
+            "min      250.000000    240.000000      0.000000  1.465000e+03\n",
+            "25%      952.000000   1097.000000     79.000000  3.272472e+06\n",
+            "50%     1311.000000   1540.000000    152.000000  4.727361e+06\n",
+            "75%     2048.000000   2400.000000    278.000000  6.222154e+06\n",
+            "max    29700.000000  27772.000000   4962.000000  8.359647e+06\n",
+            "\n",
+            "Column Names:\n",
+            "['md5', 'image_path', 'source', 'rating', 'caption_bucket', 'tags', 'caption', 'caption_model', 'trigger_tag', 'url', 'image_width', 'image_height', 'fav_count', 'created_at', 'solo_subject', 'multi_subject', 'artist_weights', 'character_weights', 'species_weights', 'post_id_cur', 'post_id_old']\n",
+            "\n",
+            "Missing Values:\n",
+            "md5                      0\n",
+            "image_path               0\n",
+            "source                   0\n",
+            "rating                   0\n",
+            "caption_bucket           0\n",
+            "tags                     0\n",
+            "caption                  0\n",
+            "caption_model            0\n",
+            "trigger_tag              0\n",
+            "url                      0\n",
+            "image_width              0\n",
+            "image_height             0\n",
+            "fav_count                0\n",
+            "created_at               0\n",
+            "solo_subject             0\n",
+            "multi_subject            0\n",
+            "artist_weights           0\n",
+            "character_weights        0\n",
+            "species_weights          0\n",
+            "post_id_cur             20\n",
+            "post_id_old          19987\n",
+            "dtype: int64\n",
+            "\n",
+            "Shape of the dataset: (19987, 21)\n"
+          ]
+        }
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "source": [
+        "# Create text files from parquet\n",
+        "\n",
+        "Goals:\n",
+        "1) First , sort each parquet item inte separate groups. First by the parquet columns 'solo_subject'  vs 'multi_subject rating'. These columns are boolean values with True / False possible values. You will now have four boolean groups.\n",
+        "\n",
+        "2) Then for each of these four boolean groups , sort the contents  by its 'rating'  category. The ratings are a string , and can be 'explicit' , 'sensitive' , 'questionable' or 'general'.   You should now have 16 groups. Assign a name to each group.\n",
+        "\n",
+        "3) For each of the 16 groups ,  append the texts from the parquet columns 'brief_summary'  ,  'midjourney_style_summary' , and 'tags' into a single text for each item.  \n",
+        "\n",
+        "4) Once each item has the full descriptive text , replace all double quations \" with single quotations ' .  \n",
+        "\n",
+        "5) Make sure all commas , have one space before and one space after them.\n",
+        "\n",
+        "5) For each text , remove all symbols of the following types:  { , } ,] , [ , | , ^ , / , \\\n",
+        "\n",
+        "6) Replace all captilized letters for lowercase. Except for at the start of a sentence , or at the start of a quatation , for example 'Tree'\n",
+        "\n",
+        "7) Once the T5 model is loaded , truncate the text for each item to 500 tokens in length\n",
+        "\n",
+        "8) Encode the full descriptive texts using the T5 for each of the 12 group.\n",
+        "\n",
+        "9) For each of the 12 groups , put similiar texts into folders using K-means. Each group may contain a maximum of 100 folders of similiar texts.\n",
+        "\n",
+        "10) For each folder , create a txt file with the contents , using the format shown in the code. The dataset string should be enclosed with the {} symbols and each text should be separated by the | symbol , {text1|text2|....|text1000} . This txt file is javascript for a website. Make sure the indentation is correct.  The name of the group and folder id should be included in the javascript as strings.  Each txt file may contain a maximum of 1000 texts. Should the amount be exceeded , create more txt files until all texts within that similarity category are listed.\n",
+        "\n",
+        "11) Package the txt files into a single zip file. The folder structure within the zip file should be sorted by its 12 groups , then by the similiar texts found within each folder.\n",
+        "\n"
+      ],
+      "metadata": {
+        "id": "fB-SLVdVQvA4"
+      }
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "#@title Adapted Pipeline – Danbooru parquet → rating groups → clean → T5 → KMeans → JS TXT (flat, with prev/next) → ZIP\n",
+        "\n",
+        "!pip install -q transformers scikit-learn tqdm\n",
+        "\n",
+        "from google.colab import drive\n",
+        "drive.mount('/content/drive')\n",
+        "\n",
+        "import os\n",
+        "import re\n",
+        "import zipfile\n",
+        "import numpy as np\n",
+        "import pandas as pd\n",
+        "import torch\n",
+        "from pathlib import Path\n",
+        "from transformers import T5Tokenizer, T5Model\n",
+        "from sklearn.cluster import KMeans\n",
+        "from tqdm.auto import tqdm\n",
+        "import warnings\n",
+        "warnings.filterwarnings(\"ignore\")\n",
+        "\n",
+        "# ============================================================\n",
+        "# Configuration\n",
+        "# ============================================================\n",
+        "parquet_path = \"/content/drive/MyDrive/Saved from Chrome/danbooru.parquet\"   #@param {type:\"string\"}\n",
+        "t5_model_name = \"t5-base\"                                                   #@param [\"t5-small\", \"t5-base\"]\n",
+        "max_tokens = 500                                                            #@param {type:\"slider\", min:128, max:512, step:32}\n",
+        "batch_size = 1                                                              # keep 1 for safety on Colab\n",
+        "max_clusters = 20                                                          # max folders per rating group\n",
+        "max_texts_per_file = 500\n",
+        "min_texts_per_file = 300                                                    # merge small clusters\n",
+        "\n",
+        "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
+        "print(f\"Device: {device}\")\n",
+        "\n",
+        "output_base = Path(\"/content/output\")\n",
+        "js_root = output_base / \"js_files\"\n",
+        "os.makedirs(js_root, exist_ok=True)\n",
+        "\n",
+        "# ============================================================\n",
+        "# 0. Load your parquet\n",
+        "# ============================================================\n",
+        "print(f\"Loading: {parquet_path}\")\n",
+        "df = pd.read_parquet(parquet_path)\n",
+        "print(f\"Shape: {df.shape}\")\n",
+        "print(f\"Columns: {df.columns.tolist()}\")\n",
+        "print(df.head(3))\n",
+        "\n",
+        "print(\"\\nUnique ratings and counts:\")\n",
+        "print(df[\"rating\"].value_counts(dropna=False))\n",
+        "\n",
+        "# ============================================================\n",
+        "# Helper: Text cleaning\n",
+        "# ============================================================\n",
+        "def clean_text(text: str) -> str:\n",
+        "    if not isinstance(text, str) or not text.strip():\n",
+        "        return \"\"\n",
+        "\n",
+        "    text = text.strip()\n",
+        "\n",
+        "    # Replace double quotes with single quotes\n",
+        "    text = text.replace('\"', \"'\")\n",
+        "\n",
+        "    # Ensure commas have one space before and after\n",
+        "    text = re.sub(r'\\s*,\\s*', ' , ', text)\n",
+        "\n",
+        "    # Remove forbidden symbols: { } [ ] | ^ / \\\n",
+        "    text = re.sub(r'[{}\\[\\]\\|\\^/\\\\]', '', text)\n",
+        "\n",
+        "    # Normalize whitespace\n",
+        "    text = re.sub(r'\\s+', ' ', text).strip()\n",
+        "\n",
+        "    # Lowercase everything except start of sentence / start of quote\n",
+        "    def smart_case(s):\n",
+        "        parts = re.split(r'([.!?]\\s+)', s)\n",
+        "        result = []\n",
+        "        capitalize_next = True\n",
+        "\n",
+        "        for part in parts:\n",
+        "            if not part:\n",
+        "                continue\n",
+        "\n",
+        "            if capitalize_next and part and part[0].isalpha():\n",
+        "                part = part[0].upper() + part[1:].lower()\n",
+        "            else:\n",
+        "                part = part.lower()\n",
+        "\n",
+        "            # Capitalize letter after a single quote (e.g. 'Tree)\n",
+        "            part = re.sub(\n",
+        "                r\"(['‘’])([a-z])\",\n",
+        "                lambda m: m.group(1) + m.group(2).upper(),\n",
+        "                part\n",
+        "            )\n",
+        "\n",
+        "            result.append(part)\n",
+        "            capitalize_next = part.endswith(('.', '!', '?')) or bool(re.search(r'[.!?]\\s*$', part))\n",
+        "\n",
+        "        return ''.join(result)\n",
+        "\n",
+        "    return smart_case(text)\n",
+        "\n",
+        "\n",
+        "def prepare_text(row) -> str:\n",
+        "    \"\"\"Combine caption + trigger_tag + tags in the requested format\"\"\"\n",
+        "    caption = str(row.get(\"caption\", \"\") or \"\")\n",
+        "    trigger = str(row.get(\"trigger_tag\", \"\") or \"\")\n",
+        "    tags    = str(row.get(\"tags\", \"\") or \"\")\n",
+        "    combined = f\"{caption} , {trigger} and {tags}\"\n",
+        "    return clean_text(combined)\n",
+        "\n",
+        "# ============================================================\n",
+        "# Prepare cleaned texts\n",
+        "# ============================================================\n",
+        "print(\"\\nCleaning texts...\")\n",
+        "df = df.copy()\n",
+        "df[\"cleaned_text\"] = df.apply(prepare_text, axis=1)\n",
+        "df = df[df[\"cleaned_text\"].str.len() > 10].reset_index(drop=True)\n",
+        "print(f\"Usable rows after cleaning: {len(df)}\")\n",
+        "\n",
+        "if len(df) < 2:\n",
+        "    raise ValueError(\"Not enough usable texts to continue.\")\n",
+        "\n",
+        "# ============================================================\n",
+        "# Load T5 encoder (once)\n",
+        "# ============================================================\n",
+        "print(f\"\\nLoading {t5_model_name}...\")\n",
+        "tokenizer = T5Tokenizer.from_pretrained(t5_model_name)\n",
+        "model = T5Model.from_pretrained(t5_model_name).to(device)\n",
+        "model.eval()\n",
+        "\n",
+        "def get_t5_embeddings(texts):\n",
+        "    all_embs = []\n",
+        "    for i in tqdm(range(0, len(texts), batch_size), desc=\"Encoding with T5\"):\n",
+        "        batch = texts[i : i + batch_size]\n",
+        "        inputs = tokenizer(\n",
+        "            batch,\n",
+        "            return_tensors=\"pt\",\n",
+        "            padding=True,\n",
+        "            truncation=True,\n",
+        "            max_length=max_tokens,\n",
+        "        ).to(device)\n",
+        "\n",
+        "        with torch.no_grad():\n",
+        "            outputs = model.encoder(**inputs)\n",
+        "            last_hidden = outputs.last_hidden_state\n",
+        "            mask = inputs.attention_mask.unsqueeze(-1).float()\n",
+        "            summed = (last_hidden * mask).sum(dim=1)\n",
+        "            counts = mask.sum(dim=1).clamp(min=1e-9)\n",
+        "            mean_pooled = summed / counts\n",
+        "            all_embs.append(mean_pooled.cpu().numpy())\n",
+        "    return np.vstack(all_embs)\n",
+        "\n",
+        "# ============================================================\n",
+        "# Process each rating group separately\n",
+        "# ============================================================\n",
+        "# Preferred order + any extra ratings that appear\n",
+        "preferred_order = [\"safe\", \"questionable\", \"sensitive\", \"explicit\"]\n",
+        "all_ratings = list(df[\"rating\"].dropna().unique())\n",
+        "# Keep preferred first, then any unexpected ratings\n",
+        "ordered_ratings = [r for r in preferred_order if r in all_ratings] + \\\n",
+        "                  [r for r in all_ratings if r not in preferred_order]\n",
+        "\n",
+        "print(f\"\\nWill process these rating groups: {ordered_ratings}\")\n",
+        "\n",
+        "# We will collect every generated file in writing order so we can set prev/next\n",
+        "# Each entry: dict with everything needed to write the file later\n",
+        "all_file_specs = []\n",
+        "\n",
+        "NAME = \"danbooru\"\n",
+        "EMOJI = \"🎨\"\n",
+        "\n",
+        "for rating in ordered_ratings:\n",
+        "    subset = df[df[\"rating\"] == rating].copy().reset_index(drop=True)\n",
+        "    print(f\"\\n{'='*60}\")\n",
+        "    print(f\"Processing rating = '{rating}'  →  {len(subset)} rows\")\n",
+        "\n",
+        "    if len(subset) < 2:\n",
+        "        print(\"  Skipping – too few rows\")\n",
+        "        continue\n",
+        "\n",
+        "    texts = subset[\"cleaned_text\"].tolist()\n",
+        "\n",
+        "    # Encode\n",
+        "    embeddings = get_t5_embeddings(texts)\n",
+        "    print(f\"  Embeddings shape: {embeddings.shape}\")\n",
+        "\n",
+        "    # K-Means\n",
+        "    k = min(max_clusters, max(2, len(texts) // 8))\n",
+        "    k = min(k, len(texts))\n",
+        "    print(f\"  Using k = {k} clusters\")\n",
+        "\n",
+        "    kmeans = KMeans(n_clusters=k, random_state=42, n_init=10)\n",
+        "    labels = kmeans.fit_predict(embeddings)\n",
+        "    subset[\"cluster\"] = labels\n",
+        "\n",
+        "    # Collect texts per cluster\n",
+        "    cluster_texts_list = []\n",
+        "    for c in range(k):\n",
+        "        cluster_texts = subset.loc[subset[\"cluster\"] == c, \"cleaned_text\"].tolist()\n",
+        "        if cluster_texts:\n",
+        "            cluster_texts_list.append(cluster_texts)\n",
+        "\n",
+        "    print(f\"  Raw clusters: {len(cluster_texts_list)}\")\n",
+        "\n",
+        "    # Merge small clusters\n",
+        "    large_clusters = []\n",
+        "    small_pool = []\n",
+        "\n",
+        "    for texts_list in cluster_texts_list:\n",
+        "        if len(texts_list) >= min_texts_per_file:\n",
+        "            large_clusters.append(texts_list)\n",
+        "        else:\n",
+        "            small_pool.extend(texts_list)\n",
+        "\n",
+        "    print(f\"  Large clusters (≥{min_texts_per_file}): {len(large_clusters)}\")\n",
+        "    print(f\"  Texts in small pool: {len(small_pool)}\")\n",
+        "\n",
+        "    merged_small_clusters = []\n",
+        "    if small_pool:\n",
+        "        for start in range(0, len(small_pool), max_texts_per_file):\n",
+        "            chunk = small_pool[start : start + max_texts_per_file]\n",
+        "            if (len(chunk) < min_texts_per_file and merged_small_clusters\n",
+        "                    and len(merged_small_clusters[-1]) + len(chunk) <= max_texts_per_file):\n",
+        "                merged_small_clusters[-1].extend(chunk)\n",
+        "            else:\n",
+        "                merged_small_clusters.append(chunk)\n",
+        "\n",
+        "    final_folders = large_clusters + merged_small_clusters\n",
+        "    print(f\"  Final folders after merging: {len(final_folders)}\")\n",
+        "\n",
+        "    # Build file specs for this rating (we write later so we can set prev/next)\n",
+        "    gname = str(rating).lower().replace(\" \", \"_\")\n",
+        "\n",
+        "    for folder_idx, folder_texts in enumerate(final_folders):\n",
+        "        folder_id = f\"cluster_{folder_idx:03d}\"\n",
+        "\n",
+        "        for part_idx, start in enumerate(range(0, len(folder_texts), max_texts_per_file)):\n",
+        "            chunk = folder_texts[start : start + max_texts_per_file]\n",
+        "            file_index = part_idx + 1\n",
+        "\n",
+        "            # Compute the exact pageTitle string that will appear in the file\n",
+        "            page_title = f\"{EMOJI} fusion-t2i-{NAME}-{gname}-{folder_id}-{file_index}\"\n",
+        "\n",
+        "            all_file_specs.append({\n",
+        "                \"gname\": gname,\n",
+        "                \"folder_id\": folder_id,\n",
+        "                \"part_idx\": part_idx,\n",
+        "                \"file_index\": file_index,\n",
+        "                \"chunk\": chunk,\n",
+        "                \"page_title\": page_title,\n",
+        "            })\n",
+        "\n",
+        "print(f\"\\nTotal TXT files that will be written: {len(all_file_specs)}\")\n",
+        "\n",
+        "# ============================================================\n",
+        "# Write all files FLAT (no subfolders) with correct backward / forward links\n",
+        "# ============================================================\n",
+        "def write_js_file(spec, backward_title: str, forward_title: str):\n",
+        "    gname      = spec[\"gname\"]\n",
+        "    folder_id  = spec[\"folder_id\"]\n",
+        "    part_idx   = spec[\"part_idx\"]\n",
+        "    file_index = spec[\"file_index\"]\n",
+        "    chunk      = spec[\"chunk\"]\n",
+        "    page_title = spec[\"page_title\"]\n",
+        "\n",
+        "    dataset_str = \"{\" + \"|\".join(chunk) + \"}\"\n",
+        "\n",
+        "    # Plain-text previous / next pageTitle\n",
+        "    backward_str = f'[\"{backward_title}\"]' if backward_title else '[\"\"]'\n",
+        "    forward_str  = f'[\"{forward_title}\"]'  if forward_title  else '[\"\"]'\n",
+        "\n",
+        "    content = f\"\"\"//start\n",
+        "group = \"{gname}\"\n",
+        "folder_id = \"{folder_id}\"\n",
+        "file_index = {file_index}\n",
+        "name = \"{NAME}\"\n",
+        "emoji = \"{EMOJI}\"\n",
+        "showComments = true\n",
+        "showDemo = true\n",
+        "\n",
+        "pageTitle = {page_title}\n",
+        "pageDescription = Captioned works from danbooru parquet – {gname} – {folder_id}\n",
+        "\n",
+        "tagBlackListRaw = [\"\"]\n",
+        "tagBlackList = [tagBlackListRaw.replaceText(' ' , ',')]\n",
+        "\n",
+        "// Generator browse destinations\n",
+        "backward = {backward_str}\n",
+        "forward = {forward_str}\n",
+        "backward_em = {{emoji}}\n",
+        "forward_em = {{emoji}}\n",
+        "numItemsStart = 1\n",
+        "//----//\n",
+        "\n",
+        "image_url = [\"https://www.iguides.ru/upload/iblock/ccf/ccf3e83b95a8b350d5089424687fc133.jpg\"]\n",
+        "image_description = [\"placeholder\"]\n",
+        "optional_image_caption = [\"placeholder\"]\n",
+        "\n",
+        "$meta\n",
+        "  title = {EMOJI} {NAME} {gname} {folder_id} {file_index}\n",
+        "  description = Subset of {gname} – {folder_id} – part {file_index}\n",
+        "  image = https://i.imgur.com/TPHIzaa.jpeg\n",
+        "$output\n",
+        "  [dataset]\n",
+        "\n",
+        "data = [(($output.selectMany(document.numItems).join(\" \")+' '))]\n",
+        "\n",
+        "dataset\n",
+        "  [\"{dataset_str}\"]\n",
+        "//----//\n",
+        "\"\"\"\n",
+        "\n",
+        "    # Flat structure: all files directly under js_root\n",
+        "    filename = f\"fusion-t2i-{NAME}-{gname}-{folder_id}-part{file_index:02d}.txt\"\n",
+        "    with open(js_root / filename, \"w\", encoding=\"utf-8\") as f:\n",
+        "        f.write(content)\n",
+        "\n",
+        "\n",
+        "print(\"\\nWriting TXT files (flat) with sequential backward/forward links...\")\n",
+        "for i, spec in enumerate(all_file_specs):\n",
+        "    prev_title = all_file_specs[i-1][\"page_title\"] if i > 0 else \"\"\n",
+        "    next_title = all_file_specs[i+1][\"page_title\"] if i < len(all_file_specs)-1 else \"\"\n",
+        "    write_js_file(spec, prev_title, next_title)\n",
+        "\n",
+        "print(f\"→ All files saved flat under {js_root}\")\n",
+        "\n",
+        "# ============================================================\n",
+        "# Create ZIP (all files at top level, no folders)\n",
+        "# ============================================================\n",
+        "zip_path = \"/content/output/danbooru_js_files.zip\"\n",
+        "\n",
+        "with zipfile.ZipFile(zip_path, \"w\", zipfile.ZIP_DEFLATED) as zipf:\n",
+        "    for file in os.listdir(js_root):\n",
+        "        if file.endswith(\".txt\"):\n",
+        "            full = js_root / file\n",
+        "            # arcname = just the filename → top level of the zip\n",
+        "            zipf.write(full, arcname=file)\n",
+        "\n",
+        "print(\"\\n\" + \"=\"*70)\n",
+        "print(\"✅ ALL DONE\")\n",
+        "print(f\"ZIP file: {zip_path}\")\n",
+        "print(\"Structure inside ZIP: completely flat\")\n",
+        "print(\"  fusion-t2i-danbooru-safe-cluster_000-part01.txt\")\n",
+        "print(\"  fusion-t2i-danbooru-safe-cluster_000-part02.txt\")\n",
+        "print(\"  fusion-t2i-danbooru-sensitive-cluster_000-part01.txt\")\n",
+        "print(\"  ...\")\n",
+        "print(\"=\"*70)\n",
+        "print(\"Download the ZIP from the Colab file browser (left panel) or copy it to Drive.\")"
+      ],
+      "metadata": {
+        "id": "wsdSxW74jzn2"
+      },
+      "execution_count": null,
+      "outputs": []
+    }
+  ]
+}