github_harvester #285

RobUmf · 2026-01-13T01:54:30Z

RobUmf
Jan 13, 2026

Ready for review.

I've been learning how to effectively use AI, specifically Gemini, to achieve my desired outcomes. I learn best through hands-on experience, so I've set myself three tasks that I can currently perform, albeit imperfectly.

I'm confident that the harvester is as complete as it can be given my current hardware limitations.

I wasn't satisfied with the default behavior of virtual environments, where downloads were deleted upon creation. This meant redownloading everything if I needed to rebuild.

I've strived to make it as universally compatible as possible and have included extensive comments throughout. I'm running this on a small HP Elitedesk 800 G3 Mini (no GPU) with Linux Mint.

The GitHub Harvester will:

Take your requirements.
Search for necessary dependencies.
First, it will check a local folder. If not found, it will download them.
Create a wheelhouse.
Provide an option for models (UltraSinger, for example, requires over 14 GB for the CPU version alone).
It will not build the venv. The necessary command is printed for easy copying and pasting.

You now have the option to copy the wheelhouse, then delete it. The models, don't delete, to a USB drive for later use or to install them on a similar system. If run again, it will locate needed components based on the requirement file. Don't forget the .json .yaml files

Commits are welcome. Thanks for participating.

The links didn't work. I'll come up with something

github_harvester.sh
github_harvester.mk.txt

RobUmf · 2026-01-13T04:17:33Z

RobUmf
Jan 13, 2026
Author

🚜 Universal GitHub Harvester v1.0

Generated by robumf with Gemini 3

Sealed Warehouse Manager for UltraSinger & AI Models

A robust automation tool designed to harvest, validate, and maintain high-performance AI models (Whisper, MDX-Net, Demucs) in a local "Warehouse." This tool is optimized for Python 3.10 environments and NTFS-shared storage, ensuring your AI workflow is completely offline-ready.

🌟 Key Features

Recursive Rescue: Automatically crawls source drives (e.g., Seagate Cabinet) to find and sort .pt, .onnx, .pth, .h5, and .bin models.
Folder Standardization: Auto-sorts assets into whisper/, mdx_networks/, and demucs/ to match UltraSinger's internal requirements.
Auto-Healing: Detects missing configuration files (like Kim_Vocal_2.json) and generates them automatically.
Sealed Environment: Keeps all heavy model assets outside the venv while maintaining a clean path for local installation.
Alignment Fix: Specifically harvests the Wav2Vec2 alignment kit to bypass Hugging Face connection errors during lyric transcription.

📂 Warehouse Architecture

After running the harvester, your /media/$USER/NTFSshare/Warehouse/UltraSinger-main/Models/ directory will be organized as follows:

Folder	Detected Files	Purpose
`whisper/`	`.pt`, `.h5`, `.pth`, `.bin`	Transcription weights & Wav2Vec2 alignment.
`mdx_networks/`	`.onnx`, `.json`	High-quality vocal/instrumental separation.
`demucs/`	`.th`, `.yaml`	Hybrid Transformer Demucs stem separation.

🚀 How to Use

1. Setup

Ensure your Seagate Cabinet (Source) and your NTFSshare (Warehouse) are both mounted.

2. Run the Harvester

Execute the script from your terminal:

bash github_harvester.sh

# Force Offline Mode
export HF_HUB_OFFLINE=1
export TRANSFORMERS_OFFLINE=1

# Execute with Local Alignment
python UltraSinger.py \
    --input "your_song.mp3" \
    --language en \
    --whisper_align_model "/media/robumf/NTFSshare/Warehouse/UltraSinger-main/Models/whisper"

## 🚀Admendments

	Due to Conflicting Info Between What I have and A.I Gemini 		
	How to find dirrect links of the biggest files for off location downloads	
			
WHEELS			
			
			
cuDNN		663 M	https://pypi.org/project/nvidia-cudnn-cu12/9.1.0.70/#files
TF-CPU		260 M	https://pypi.org/project/tensorflow-cpu/2.18.0/#files
Torch		210 M	https://pypi.org/project/torch/2.5.1/#files
ONNX-OV		70 M	https://pypi.org/project/onnxruntime-openvino/1.16.0/#files
SciPy		36M	https://pypi.org/project/scipy/1.15.1/#files
			
tensorflow	tensorflow-2.19.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl	644 M	https://pypi.org/project/tensorflow/2.19.0/#files
torch	torch-2.9.1+cpu-cp310-cp310-manylinux_2_28_x86_64.whl	899.8 MB	https://pypi.org/project/torch/#files
torch	torch-2.7.1+cpu-cp310-cp310-manylinux_2_28_x86_64.whl	821.2 M	https://pypi.org/project/torch/2.7.1/#files
llvmlite	llvmlite-0.46.0-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl	56.3 M	https://pypi.org/project/llvmlite/#files
llvmlite	llvmlite-0.44.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl	30.3 M	https://pypi.org/project/llvmlite/0.44.0/#files
					
MODELS Whisper	python3 UltraSinger.py --input "song.mp3" --whisper_model 	
When you launch UltraSinger, use the --language flag. This prevents the auto-detection script from trying to download language-specific alignment models:
python UltraSinger.py --input "song.mp3" --language en --whisper_model large-v3
			
Version / Type	File Name	Approx. Size	Official Direct Download (Plain Text Only)
Large-v3	large-v3.pt	2.9 GB	https://openaipublic.azureedge.net/main/whisper/models/e5b1a55b89c1367dacf97e3e19bfd829a01529dbfdeefa8caeb59b3f1b81dadb/large-v3.pt
Large-v2	large-v2.pt	2.9 GB	https://openaipublic.azureedge.net/main/whisper/models/81f7c96c852ee8fc832187b0132e569d6c3065a3252ed18e56effd0b6a73e524/large-v2.pt
Turbo?	turbo.pt	1.6 GB	https://openaipublic.azureedge.net/main/whisper/models/aff26ae408abcba5fbf8813c21e62b0941638c5f6eebfb145be0c9839262a19a/large-v3-turbo.pt
Medium	medium.pt	1.4 GB	https://openaipublic.azureedge.net/main/whisper/models/345ae4da62f9b3d59415adc60127b97c714f32e89e936602e85993674d08dcb1/medium.pt
Medium.en	medium.en.pt	1.4 GB	https://openaipublic.azureedge.net/main/whisper/models/d7440d1dc186f76616474e0ff0b3b6b879abc9d1a4926b7adfa41db2d497ab4f/medium.en.pt
Small	small.pt	466 MB	https://openaipublic.azureedge.net/main/whisper/models/9ecf779972d90ba49c06d968637d720dd632c55bbf19d441fb42bf17a411e794/small.pt
Small.en	small.en.pt	466 MB	https://openaipublic.azureedge.net/main/whisper/models/f953ad0fd29cacd07d5a9eda5624af0f6bcf2258be67c92b79389873d91e0872/small.en.pt
Base	base.pt	142 MB	https://openaipublic.azureedge.net/main/whisper/models/ed3a0b6b1c0edf879ad9b11b1af5a0e6ab5db9205f891f668f8b0e6c6326e34e/base.pt
Base.en	base.en.pt	142 MB	https://openaipublic.azureedge.net/main/whisper/models/25a8566e1d0c1e2231d1c762132cd20e0f96a85d16145c3a00adf5d1ac670ead/base.en.pt
Tiny	tiny.pt	75 MB	https://openaipublic.azureedge.net/main/whisper/models/65147644a518d12f04e32d6f3b26facc3f8dd46e5390956a9424a650c0ce22b9/tiny.pt
Tiny.en	tiny.en.pt	75 MB	https://openaipublic.azureedge.net/main/whisper/models/d3dd57d32accea0b295c963554e173b78ee57fe7ca2ad289e47a2a4c2206430c/tiny.en.pt

MDX python UltraSinger.py --input "path/to/song.mp3" --demucs mdx_extra

.onnx and .json
models/mdx_networks
--demucs mdx_extra
Open your terminal.

Run this command to create the file directly in your warehouse:

cat <<EOF > /media/robumf/NTFSshare/Warehouse/UltraSinger-main/Models/mdx/Kim_Vocal_2.json
{
    "model_path": "Kim_Vocal_2.onnx",
    "model_name": "Kim_Vocal_2",
    "is_kim_model": true,
    "mdx_dim_f_set": 3072,
    "mdx_dim_t_set": 8,
    "mdx_n_fft_scale_set": 6144,
    "mdx_hop_set": 1024,
    "mdx_sr_set": 44100
}
EOF

demucs

https://dl.fbaipublicfiles.com/demucs/hybrid_transformer/f7e0c4bc-ba3fe64a.th
https://dl.fbaipublicfiles.com/demucs/hybrid_transformer/d12395a8-e57c48e6.th
https://dl.fbaipublicfiles.com/demucs/hybrid_transformer/92cfc3b6-ef3bcb9c.th
https://dl.fbaipublicfiles.com/demucs/hybrid_transformer/04573f0d-f3cf25b2.th
https://raw.githubusercontent.com/facebookresearch/demucs/main/demucs/remote/htdemucs_ft.yaml

hugginface?
# 1. Create the standardized folder
mkdir -p /media/robumf/NTFSshare/Warehouse/UltraSinger-main/Models/mdx_networks

# 2. Move your existing ONNX and math files
mv /media/robumf/NTFSshare/Warehouse/UltraSinger-main/Models/mdx/* /media/robumf/NTFSshare/Warehouse/UltraSinger-main/Models/mdx_networks/

# 3. Remove the old directory
rmdir /media/robumf/NTFSshare/Warehouse/UltraSinger-main/Models/mdx

# Navigate to your warehouse
cd /media/robumf/NTFSshare/Warehouse/UltraSinger-main/Models/mdx_networks/

# Download the ONNX weights
wget https://huggingface.co/seanghay/uvr_models/resolve/main/MDX23C-InstVoc.onnx

# Download the required JSON config
wget https://raw.githubusercontent.com/Anjok07/ultimatevocalremovergui/master/gui_data/model_data/mdx_model_data/model_data/MDX23C-InstVoc.json
https://huggingface.co/facebook/wav2vec2-large-960h-lv60-self/resolve/main/pytorch_model.bin

HF_HUB_OFFLINE=1 TRANSFORMERS_OFFLINE=1 python UltraSinger.py --input "your_song.mp3" --language en --whisper_align_model "/media/robumf/NTFSshare/Warehouse/UltraSinger-main/Models/whisper"


https://github.com/ShadowLoveElysia/Whisper-vits-svc-LargeV3
https://github.com/Recordtini/MusicSepGUI/blob/main/models.json
https://github.com/sindresorhus/awesome

0 replies

RobUmf · 2026-01-13T05:52:30Z

RobUmf
Jan 13, 2026
Author

#!/bin/bash

--- UNIVERSAL GITHUB HARVESTER v1.0 (Public Beta) ---

Linux CPU plus A.I Models

Optimized for Python 3.10 "Sweet Spot" Environments

1. IDENTITY & DYNAMIC DETECTION

echo "--------------------------------------------------------"
echo "🚜 UNIVERSAL HARVESTER v1.0: (Public Beta)"
echo "--------------------------------------------------------"

CRITICAL FIX: Define CURRENT_DIR before the IF statement

CURRENT_DIR=$(pwd)

if [[ "$CURRENT_DIR" == "install/CPU" ]]; then
# Goes up 2 levels to find the repository root name
DETECTED_NAME=$(basename $(readlink -f "$CURRENT_DIR/../.."))
SEARCH_PATH=". ../.."
else
# Assumes current folder is the project root
DETECTED_NAME=$(basename "$CURRENT_DIR")
SEARCH_PATH="."
fi

The -i flag provides the detected name as a default,

but -e (readline) allows the user to backspace and change it.

read -p "🏷️ Project ID: " -i "$DETECTED_NAME" -e PROJECT_NAME

Scan for Requirements Files

mapfile -t FILES < <(find $SEARCH_PATH -maxdepth 2 -name "*.txt" 2>/dev/null | grep -i "requirement" | sort -u)
if [ ${#FILES[@]} -gt 0 ]; then
for i in "${!FILES[@]}"; do printf " [%d] %s\n" "$i" "${FILES[$i]}"; done
read -p ">> Select index: " USER_INPUT
REQ_FILE="${FILES[$USER_INPUT]}"
else
read -e -p ">> Input requirements file path: " REQ_FILE
fi

2. UNIVERSAL DRIVE MAPPING

Defaults to standard Linux media mount point

USER_HOME="/media/$USER"

A. FIND CABINET: Smart Detection for External Source

(Change 'Seagate Basic' to your specific drive label if needed, or leave dynamic)

DETECTED_SEAGATE=$(ls -d "$USER_HOME"/Seagate* 2>/dev/null | head -n 1)

B. FIND WAREHOUSE: Defaults to a "Warehouse" folder in Home if NTFSshare isn't found

if [ -d "$USER_HOME/NTFSshare" ]; then
DEFAULT_WAREHOUSE="$USER_HOME/NTFSshare/Warehouse/$PROJECT_NAME"
else
DEFAULT_WAREHOUSE="$HOME/Warehouse/$PROJECT_NAME"
fi

C. FIND BUILD DIR: NVMe/Root Detection

NVME_PATH=$(df "$PWD" | tail -1 | awk '{print $6}')

echo "--------------------------------------------------------"
read -e -p "🔍 Source (External Cabinet): " -i "${DETECTED_SEAGATE:-$PWD}" SOURCE_ROOT
read -e -p "📂 Warehouse (Destination): " -i "$DEFAULT_WAREHOUSE" WAREHOUSE
read -e -p "🔨 Build Dir (Fast Drive): " -i "$NVME_PATH/$PROJECT_NAME" BUILD_DIR

3. SAFETY CHECKS (The "Stop the User" Logic)

echo "--------------------------------------------------------"
echo "🛡️ SAFETY CHECK: Verifying Drive Health..."

Verify Build Dir

if [ ! -w "$NVME_PATH" ]; then
echo "❌ ERROR: Build Path ($NVME_PATH) is NOT writable."
exit 1
fi

Verify Warehouse Root

WAREHOUSE_ROOT=$(dirname "$WAREHOUSE")

If parent doesn't exist, check if we can create it

if [ ! -d "$WAREHOUSE_ROOT" ]; then
if [ ! -w "$(dirname "$WAREHOUSE_ROOT")" ]; then
echo "❌ ERROR: Cannot create Warehouse at $WAREHOUSE_ROOT."
echo " Please check your drive mounts."
exit 1
fi
fi

WHEEL_DEST="$WAREHOUSE/Wheelhouse"
MODEL_DEST="$WAREHOUSE/Models"
mkdir -p "$WHEEL_DEST" "$MODEL_DEST/checkpoints" "$MODEL_DEST/mdx_networks" "$MODEL_DEST/whisper" "$MODEL_DEST/demucs" "$BUILD_DIR"

4. PYTHON ENGINE & SWEET SPOT LOCK

echo "--------------------------------------------------------"
echo "🐍 DETECTING PYTHON ENGINE"
PY_BIN=$(which python3.10)

if [ -n "$PY_BIN" ]; then
echo "🎯 SWEET SPOT ACTIVE: $($PY_BIN --version)"
else
PY_BIN=$(which python3)
echo "⚠️ WARNING: Python 3.10 (The Sweet Spot) was not found."
echo " Detected: $($PY_BIN --version)"
read -p ">> Continue with $($PY_BIN --version) anyway? (y/n): " VER_CONT
if [[ "$VER_CONT" != "y" ]]; then exit 1; fi
fi

5. SANITIZATION CHAMBER (v6.9 - Build-Proof)

OFFLINE_REQ="$WAREHOUSE/requirements_offline.txt"
echo "🧹 Sanitizing requirements for Universal compatibility..."

Fixes: AV build error, +cpu 404 error, and Windows bloat

grep -vE "nvidia-|file://|winshell|pywin32" "$REQ_FILE" |
sed -e 's/av==./av==12.3.0/g'
-e 's/+cu[0-9]//g'
-e 's/+cpu//g'
-e 's/onnxruntime./onnxruntime==1.16.2/g'
-e 's/numpy./numpy==1.26.4/g'
> "$OFFLINE_REQ"

6. THE HARVEST

echo "--------------------------------------------------------"
HARVEST_START=$(date +%s)
echo "⏱️ Harvest Start: $(date +%H:%M:%S)"
echo "🚜 STARTING HARVEST"
echo "--------------------------------------------------------"

A. CABINET SYNC (Local Import)

if [ -d "$SOURCE_ROOT/Wheelhouse" ]; then
echo "📦 Importing local wheels from Source..."
$PY_BIN -m pip download -d "$WHEEL_DEST" --no-index --find-links "$SOURCE_ROOT/Wheelhouse" -r "$OFFLINE_REQ" --quiet 2>/dev/null
fi

B. WEB DOWNLOAD LOOP (Dual-Index + Mount Guard)

while read -r line; do
p=$(echo "$line" | tr -d '\r' | xargs)
if [[ -z "$p" || "$p" == #* || "$p" == -* ]]; then continue; fi

# Mount Guard
until [ -d "$WHEEL_DEST" ]; do
    echo "🚨 WAREHOUSE DISCONNECTED! Reconnect drive..."
    sleep 10
done

echo "⬇️  Processing: $p"

# Dual-Index Strategy: Checks PyPI AND Torch CPU shelf
$PY_BIN -m pip download \
    -d "$WHEEL_DEST" \
    --index-url https://pypi.org/simple \
    --extra-index-url https://download.pytorch.org/whl/cpu \
    --timeout 120 --retries 10 "$p" --quiet

done < "$OFFLINE_REQ"

7. VALIDATION & SPECULATION

echo "--------------------------------------------------------"
REQ_COUNT=$(grep -vE "^\s*(#|$)" "$OFFLINE_REQ" | wc -l)
WHEEL_COUNT=$(ls -1 "$WHEEL_DEST"/*.whl 2>/dev/null | wc -l)

echo "📊 SUMMARY: $WHEEL_COUNT wheels collected for $REQ_COUNT core requirements."
if [ "$WHEEL_COUNT" -ge "$REQ_COUNT" ]; then
echo "🟢 STATUS: Solid. (Dependencies likely included)"
else
echo "🟡 STATUS: Lean. Check logs for missing packages."
fi

8. RECURSIVE RESCUE (Consolidated Model Harvest)

echo "--------------------------------------------------------"
read -p "🧠 Harvest AI Models (Whisper/MDX/Demucs/Align)? (y/n): " DO_MODELS

if [[ "$DO_MODELS" == "y" || "$DO_MODELS" == "Y" ]]; then
# Dynamic path detection for beta testers
DEFAULT_MODEL_SRC="$SOURCE_ROOT/Models"
read -e -p "🔍 Where are the source models?: " -i "$DEFAULT_MODEL_SRC" MODEL_SRC

if [ ! -d "$MODEL_SRC" ]; then
    echo "❌ ERROR: Model folder not found at $MODEL_SRC"
else
    MODEL_DEST_OUT="$WAREHOUSE/Models"
    mkdir -p "$MODEL_DEST_OUT/whisper" "$MODEL_DEST_OUT/mdx_networks" "$MODEL_DEST_OUT/demucs"

    echo "🚜 STARTING MODEL HARVEST FROM $MODEL_SRC..."
    
    # Recursive search for all model types (pt, onnx, json, bin, etc.)
    find "$MODEL_SRC" -type f \( -name "*.pt" -o -name "*.th" -o -name "*.yaml" -o -name "*.onnx" -o -name "*.json" -o -name "*.h5" -o -name "*.pth" -o -name "*.bin" \) | while read -r file; do
        fname=$(basename "$file")
        case "$fname" in
            # Whisper + Alignment + Wav2Vec2
            *whisper*|*large*|*medium*|*base*|*tiny*|*small*|*.h5|*pyannote*|*wav2vec2*|*preprocessor_config*|*pytorch_model.bin*|*config.json*) 
                cp -nv "$file" "$MODEL_DEST_OUT/whisper/" ;;
            
            # MDX Networks (Separation)
            *.onnx*|*.json*) 
                cp -nv "$file" "$MODEL_DEST_OUT/mdx_networks/" ;;
            
            # Demucs Stems
            *.th*|*.yaml*) 
                cp -nv "$file" "$MODEL_DEST_OUT/demucs/" ;;
        esac
    done

    # Fix for Discussion #285: Ensuring Kim_Vocal_2 has its config
    if [ -f "$MODEL_DEST_OUT/mdx_networks/Kim_Vocal_2.onnx" ] && [ ! -f "$MODEL_DEST_OUT/mdx_networks/Kim_Vocal_2.json" ]; then
        echo "📝 Auto-creating missing Kim_Vocal_2.json..."
        cat <<EOF > "$MODEL_DEST_OUT/mdx_networks/Kim_Vocal_2.json"

{
"model_path": "Kim_Vocal_2.onnx", "model_name": "Kim_Vocal_2",
"is_kim_model": true, "mdx_dim_f_set": 3072, "mdx_dim_t_set": 8,
"mdx_n_fft_scale_set": 6144, "mdx_hop_set": 1024, "mdx_sr_set": 44100
}
EOF
fi
fi
else
echo "⏭️ Skipping Model Harvest."
fi

9. FINAL OUTPUT & INVENTORY

END_TIME=$(date +%s)
DURATION=$((END_TIME - HARVEST_START))
MINUTES=$((DURATION / 60))
SECONDS=$((DURATION % 60))

echo "--------------------------------------------------------"
echo "✅ HARVEST COMPLETE"
echo "⏱️ Total Duration: ${MINUTES}m ${SECONDS}s"
echo "📦 Warehouse Size: $(du -sh "$WAREHOUSE" | awk '{print $1}')"
echo "--------------------------------------------------------"

Report from the specific Warehouse paths

echo "📂 MODEL REPOSITORY SUMMARY:"
W_MODELS=$(ls -1 "$WAREHOUSE/Models/whisper" 2>/dev/null | wc -l)
M_MODELS=$(ls -1 "$WAREHOUSE/Models/mdx_networks" 2>/dev/null | wc -l)
D_MODELS=$(ls -1 "$WAREHOUSE/Models/demucs" 2>/dev/null | wc -l)

echo " 🎙️ Whisper/Align Assets: ${W_MODELS:-0}"
echo " 🎵 MDX Networks: ${M_MODELS:-0}"
echo " 🥁 Demucs Stems: ${D_MODELS:-0}"
echo "--------------------------------------------------------"
echo "🚀 OFFLINE INSTALLER READY"
echo " To finalize your Sealed Environment, follow these steps:"
echo ""
echo " 1. Enter the Project Folder:"
echo " cd $BUILD_DIR"
echo ""
echo " 2. Activate your venv (Sweet Spot 3.10):"
echo " source .venv/bin/activate"
echo ""
echo " 3. Run the Offline Install command:"
echo " pip install --no-index --find-links='$WHEEL_DEST' -r '$OFFLINE_REQ'"
echo "--------------------------------------------------------"

1 reply

RobUmf Jan 13, 2026
Author

ugg that idea don't work either it change the remarks.
didn't copy whispers. Guess not ready.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

github_harvester #285

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Replies: 2 comments 1 reply

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Select a reply

Uh oh!

Uh oh!

github_harvester #285

Uh oh!

Uh oh!

RobUmf Jan 13, 2026

Replies: 2 comments · 1 reply

Uh oh!

RobUmf Jan 13, 2026 Author

🚜 Universal GitHub Harvester v1.0

🌟 Key Features

📂 Warehouse Architecture

🚀 How to Use

1. Setup

2. Run the Harvester

Uh oh!

Uh oh!

RobUmf Jan 13, 2026 Author

--- UNIVERSAL GITHUB HARVESTER v1.0 (Public Beta) ---

Linux CPU plus A.I Models

Optimized for Python 3.10 "Sweet Spot" Environments

1. IDENTITY & DYNAMIC DETECTION

CRITICAL FIX: Define CURRENT_DIR before the IF statement

The -i flag provides the detected name as a default,

but -e (readline) allows the user to backspace and change it.

Scan for Requirements Files

2. UNIVERSAL DRIVE MAPPING

Defaults to standard Linux media mount point

A. FIND CABINET: Smart Detection for External Source

(Change 'Seagate Basic' to your specific drive label if needed, or leave dynamic)

B. FIND WAREHOUSE: Defaults to a "Warehouse" folder in Home if NTFSshare isn't found

C. FIND BUILD DIR: NVMe/Root Detection

3. SAFETY CHECKS (The "Stop the User" Logic)

Verify Build Dir

Verify Warehouse Root

If parent doesn't exist, check if we can create it

4. PYTHON ENGINE & SWEET SPOT LOCK

5. SANITIZATION CHAMBER (v6.9 - Build-Proof)

Fixes: AV build error, +cpu 404 error, and Windows bloat

6. THE HARVEST

A. CABINET SYNC (Local Import)

B. WEB DOWNLOAD LOOP (Dual-Index + Mount Guard)

7. VALIDATION & SPECULATION

8. RECURSIVE RESCUE (Consolidated Model Harvest)

9. FINAL OUTPUT & INVENTORY

Report from the specific Warehouse paths

Uh oh!

Uh oh!

RobUmf Jan 13, 2026 Author

RobUmf
Jan 13, 2026

Replies: 2 comments 1 reply

RobUmf
Jan 13, 2026
Author

RobUmf
Jan 13, 2026
Author

RobUmf Jan 13, 2026
Author