Repair ComfyUI Portable Dependencies One Missing Module At A Time

A tested Windows portable ComfyUI repair transcript for missing SQLAlchemy, torch, frontend package, kornia, PyOpenGL, and other startup dependencies.

If a Windows portable ComfyUI environment starts failing with missing core packages, do not begin with a broad reinstall.

The safer repair loop is:

start ComfyUI
read the first new blocker in the log
install only that package in python_embeded
start again
repeat until the server, /system_stats, /object_info, and pip check are healthy

This page is based on a real broken portable environment repaired on May 18, 2026. The lab started with ModuleNotFoundError: No module named 'sqlalchemy' and eventually recovered on an RTX 4070 Ti SUPER with CUDA PyTorch.

Practical rule

Run one install, then restart and read the new log. A copied error search should lead to the exact package that changed the next startup attempt, not to a giant command that hides what actually fixed the environment.

When To Use This Runbook

Use this page when:

ComfyUI portable starts with python_embeded\python.exe
startup stops before Starting server
you see copied errors such as ModuleNotFoundError: No module named 'sqlalchemy', No module named 'torch', or comfyui-frontend-package is not installed
your models, workflows, and custom nodes are worth preserving
you want to know which package changed the next failure, instead of reinstalling everything

Do not use this as a universal upgrade command. If only one custom node is missing one package, fix that custom node first.

The Tested Repair Chain

This was the actual one-package-at-a-time sequence from the repaired lab.

Attempt	First new signal in the startup log	One package action	What changed after restart
1	`ModuleNotFoundError: No module named 'sqlalchemy'`	`pip install "SQLAlchemy>=2.0"`	exposed `filelock`
2	`ModuleNotFoundError: No module named 'filelock'`	`pip install filelock`	exposed `alembic` and `yaml`
3	`Error importing dependencies: No module named 'alembic'`	`pip install alembic`	database setup moved forward
4	`ModuleNotFoundError: No module named 'yaml'`	`pip install PyYAML`	exposed `PIL`
5	`ModuleNotFoundError: No module named 'PIL'`	`pip install Pillow`	exposed `tqdm`
6	`ModuleNotFoundError: No module named 'tqdm'`	`pip install tqdm`	exposed `comfy_aimdo`
7	`ModuleNotFoundError: No module named 'comfy_aimdo'`	`pip install comfy-aimdo`	exposed `torch`
8	`ModuleNotFoundError: No module named 'torch'`	installed `torch`, then replaced CPU Torch with `torch==2.12.0+cu130` from the PyTorch CUDA index	CUDA became available
9	`ModuleNotFoundError: No module named 'safetensors'`	`pip install safetensors`	exposed `packaging`
10	`ModuleNotFoundError: No module named 'packaging'`	`pip install packaging`	exposed `numpy`
11	`ModuleNotFoundError: No module named 'numpy'`	`pip install "numpy>=1.25.0"`	exposed `einops`
12	`ModuleNotFoundError: No module named 'einops'`	`pip install einops`	exposed `psutil`
13	`ModuleNotFoundError: No module named 'psutil'`	`pip install psutil`	ComfyUI reached VRAM detection, then exposed `torchvision`
14	`ModuleNotFoundError: No module named 'torchvision'`	installed `torchvision` from the same CUDA wheel index	exposed `torchaudio`
15	`ModuleNotFoundError: No module named 'torchaudio'`	installed `torchaudio` from the same CUDA wheel index	exposed `scipy`
16	`ModuleNotFoundError: No module named 'scipy'`	`pip install scipy`	exposed `transformers`
17	`ModuleNotFoundError: No module named 'transformers'`	`pip install transformers`	exposed `torchsde`
18	`ModuleNotFoundError: No module named 'torchsde'`	`pip install torchsde`	exposed `av`
19	`ModuleNotFoundError: No module named 'av'`	`pip install av`	exposed `aiohttp`
20	`ModuleNotFoundError: No module named 'aiohttp'`	`pip install aiohttp`	exposed `requests`
21	`ModuleNotFoundError: No module named 'requests'`	`pip install requests`	exposed `pydantic`
22	`ModuleNotFoundError: No module named 'pydantic'`	`pip install pydantic`	exposed `comfyui-frontend-package`
23	`comfyui-frontend-package is not installed`	`pip install comfyui-frontend-package`	server started, but some built-in extra nodes failed to import

At that point ComfyUI was alive, but the log still showed missing optional or built-in-extra dependencies. The cleanup pass continued in the same style:

Cleanup signal	One package action	Result
`No module named 'kornia'` in post-processing, canny, and morphology nodes	`pip install kornia`	node count rose from 657 to 685
`No module named 'spandrel'` in `nodes_upscale_model.py`	`pip install spandrel`	node count rose to 687
repeated `No module named 'pydantic_settings'` while parsing `pyproject.toml`	`pip install pydantic-settings`	the startup log shrank from noisy repeats to a short warning set
`OpenGL dependencies not available`	`pip install PyOpenGL`, restart, then `pip install glfw`	GLSL/OpenGL node imported
`No module named 'simpleeval'` in `nodes_math.py`	`pip install simpleeval`	no `IMPORT FAILED` lines remained
`comfyui-workflow-templates is not installed`	`pip install comfyui-workflow-templates`	workflow template warning disappeared
`comfyui-embedded-docs package not found`	`pip install "comfyui-embedded-docs==0.4.4"`	embedded docs warning disappeared
`WARNING: blake3 package not installed`	`pip install blake3`	`blake3` warning disappeared
`Failed to import comfy_kitchen`	`pip install "comfy-kitchen>=0.2.8"`	CUDA and eager `comfy_kitchen` backends loaded

The final log still said the comfy_kitchen Triton backend was unavailable because triton was not installed. That was not a startup blocker: the CUDA and eager backends were available and the server opened normally.

Step 0: Start From The Portable Root

Open PowerShell in the folder that contains:

run_nvidia_gpu.bat
run_cpu.bat
python_embeded
ComfyUI

Confirm the Python executable:

.\python_embeded\python.exe -s -c "import sys; print(sys.executable); print(sys.version)"

If this path is not the Python that starts ComfyUI, stop and switch to the right environment.

Step 1: Capture Each Attempt

For a one-by-one repair, save logs for each start:

$ts = Get-Date -Format 'yyyyMMdd_HHmmss'
New-Item -ItemType Directory -Force .\repair_logs | Out-Null
$out = ".\repair_logs\attempt_${ts}_stdout.log"
$err = ".\repair_logs\attempt_${ts}_stderr.log"
Start-Process -FilePath ".\python_embeded\python.exe" `
  -ArgumentList @("-s", "ComfyUI\main.py", "--windows-standalone-build") `
  -WorkingDirectory (Get-Location) `
  -RedirectStandardOutput $out `
  -RedirectStandardError $err `
  -PassThru

Then read the bottom of the stderr log:

Get-Content $err -Tail 200

Install only the package named by the first real blocker. Then start again and save a new log.

Step 2: Do Not Trust Port 8188 Until You Check Ownership

In the lab, http://127.0.0.1:8188/ returned 200, but it was another ComfyUI process from C:\Users\huhaibin\Documents\ComfyUI, not the broken lab.

Check the owner before declaring success:

Get-NetTCPConnection -LocalPort 8188 -ErrorAction SilentlyContinue |
  Select-Object LocalAddress,LocalPort,State,OwningProcess

Get-CimInstance Win32_Process |
  Where-Object { $_.CommandLine -like "*ComfyUI*main.py*" } |
  Select-Object ProcessId,CommandLine

If 8188 is already occupied, start the broken environment on a clean port while repairing:

.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --port 8191

Only the process from the same portable root counts as proof.

Step 3: Verify Torch Before And After Installing It

If the log says No module named 'torch', installing torch from the default index may satisfy the import while giving you a CPU-only build.

That happened in the lab:

2.12.0+cpu None False

For the tested Nvidia repair, Torch had to be replaced from the CUDA wheel index:

.\python_embeded\python.exe -s -m pip install `
  --index-url https://download.pytorch.org/whl/cu130 `
  --extra-index-url https://pypi.tuna.tsinghua.edu.cn/simple `
  "torch==2.12.0+cu130"

Then verify in the same embedded Python:

.\python_embeded\python.exe -s -c "import torch; print(torch.__version__, torch.version.cuda, torch.cuda.is_available()); print(torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'no cuda')"

The lab recovered this result:

2.12.0+cu130 13.0 True
NVIDIA GeForce RTX 4070 Ti SUPER

Use the CUDA wheel that matches your environment. Do not assume the default pip install torch is GPU-enabled.

Step 4: Treat Frontend And Template Packages As ComfyUI Versioned Packages

When the log says:

comfyui-frontend-package is not installed.

look at your own ComfyUI\requirements.txt before choosing versions.

In the lab, installing the frontend package unpinned gave:

ComfyUI frontend version: 1.43.18
required_frontend_version: 1.42.15

It worked, but the safer public instruction is: copy the versions required by your local ComfyUI code when they are pinned.

The same applies to:

comfyui-frontend-package
comfyui-workflow-templates
comfyui-embedded-docs

Step 5: Verify Recovery With More Than The Browser

The final lab start used:

.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --port 8191

Healthy signals:

ComfyUI version: 0.20.1
pytorch version: 2.12.0+cu130
Device: cuda:0 NVIDIA GeForce RTX 4070 Ti SUPER : cudaMallocAsync
ComfyUI frontend version: 1.43.18
comfy-kitchen version: 0.2.8
Starting server
To see the GUI go to: http://127.0.0.1:8191

Then verify endpoints:

Invoke-RestMethod http://127.0.0.1:8191/system_stats
Invoke-RestMethod http://127.0.0.1:8191/object_info
.\python_embeded\python.exe -s -m pip check

The repaired lab returned:

ComfyUI 0.20.1
PyTorch 2.12.0+cu130
GPU: NVIDIA GeForce RTX 4070 Ti SUPER
object_info node count: 689
CheckpointLoaderSimple: present
KSampler: present
ComfyMathExpression: present
pip check: No broken requirements found.

What Is A Blocker And What Is Only A Warning?

Log line	Blocker?	What to do
`ModuleNotFoundError: No module named 'sqlalchemy'` before server start	Yes	Install SQLAlchemy in ComfyUI's embedded Python, then restart
`ModuleNotFoundError: No module named 'torch'`	Yes	Install Torch, then verify CUDA, not just import success
`comfyui-frontend-package is not installed`	Yes	Install the frontend package required by local ComfyUI
`IMPORT FAILED: nodes_math.py`	Partial	Server can start, but that node is missing; install `simpleeval`
`Starting server` appears	No	Server has started; continue with endpoint verification
`Found comfy_kitchen backend triton: {'available': False, ...}`	Usually no	Ignore unless a workflow actually requires Triton
`/system_stats` returns the expected GPU	No	Good sign that runtime is alive

One Package At A Time Or Requirements All At Once?

Use this decision:

Situation	Better move
You are diagnosing an unknown broken portable environment	One package, restart, read the next log
The first error is a custom node missing one package	Install that node's package or requirements
Torch imports but CUDA is false	Fix the CUDA Torch wheel first
Torch/CUDA is healthy	Avoid broad commands that may replace it
You are creating a fresh clean environment	`pip install -r ComfyUI\requirements.txt` can be normal

A blind pip install -r ComfyUI\requirements.txt can be acceptable for a clean manual install, but it is not the best forensic repair loop when you need to understand what broke.

What Not To Do

do not use system pip when ComfyUI uses python_embeded
do not assume a browser page on 8188 belongs to the environment you are repairing
do not treat CPU Torch import success as CUDA success
do not paste a long package list when you are trying to learn the next log change
do not treat every warning after Starting server as fatal
do not delete the environment before preserving workflows, models, and custom nodes

How Wonderful Launcher Helps

The hard part of this repair is not memorizing package names. It is the loop:

start
capture the first real error
repair the right Python
start again
distinguish blockers from warnings
verify the app is really alive

Wonderful Launcher is built around that recovery loop. It makes environment boundaries and startup logs easier to reason about, especially when a portable ComfyUI install has been modified by plugins over time.

Source References

Windows portable repair lab, May 18, 2026: E:\ComfyUI_portable\ComfyUI_SEO_Lab
First fatal log: ModuleNotFoundError: No module named 'sqlalchemy'
Important pitfall: 8188 belonged to another ComfyUI process, so the lab was verified on port 8191
Final lab launch: ComfyUI 0.20.1, frontend 1.43.18, PyTorch 2.12.0+cu130, GPU NVIDIA GeForce RTX 4070 Ti SUPER
Final endpoint check: /object_info reported 689 nodes and pip check returned No broken requirements found

You can fix it manually, or download Wonderful Launcher for Windows to diagnose plugin errors, missing dependencies, and broken ComfyUI environments without reinstalling.

Download free for Windows

Did this fix your issue?

Your answer helps prioritize verified ComfyUI repairs.

If a Windows portable ComfyUI environment starts failing with missing core packages, do not begin with a broad reinstall.

The safer repair loop is:

start ComfyUI
read the first new blocker in the log
install only that package in python_embeded
start again
repeat until the server, /system_stats, /object_info, and pip check are healthy

Practical rule

When To Use This Runbook

Use this page when:

ComfyUI portable starts with python_embeded\python.exe
startup stops before Starting server
you see copied errors such as ModuleNotFoundError: No module named 'sqlalchemy', No module named 'torch', or comfyui-frontend-package is not installed
your models, workflows, and custom nodes are worth preserving
you want to know which package changed the next failure, instead of reinstalling everything

Do not use this as a universal upgrade command. If only one custom node is missing one package, fix that custom node first.

The Tested Repair Chain

This was the actual one-package-at-a-time sequence from the repaired lab.

Attempt	First new signal in the startup log	One package action	What changed after restart
1	`ModuleNotFoundError: No module named 'sqlalchemy'`	`pip install "SQLAlchemy>=2.0"`	exposed `filelock`
2	`ModuleNotFoundError: No module named 'filelock'`	`pip install filelock`	exposed `alembic` and `yaml`
3	`Error importing dependencies: No module named 'alembic'`	`pip install alembic`	database setup moved forward
4	`ModuleNotFoundError: No module named 'yaml'`	`pip install PyYAML`	exposed `PIL`
5	`ModuleNotFoundError: No module named 'PIL'`	`pip install Pillow`	exposed `tqdm`
6	`ModuleNotFoundError: No module named 'tqdm'`	`pip install tqdm`	exposed `comfy_aimdo`
7	`ModuleNotFoundError: No module named 'comfy_aimdo'`	`pip install comfy-aimdo`	exposed `torch`
8	`ModuleNotFoundError: No module named 'torch'`	installed `torch`, then replaced CPU Torch with `torch==2.12.0+cu130` from the PyTorch CUDA index	CUDA became available
9	`ModuleNotFoundError: No module named 'safetensors'`	`pip install safetensors`	exposed `packaging`
10	`ModuleNotFoundError: No module named 'packaging'`	`pip install packaging`	exposed `numpy`
11	`ModuleNotFoundError: No module named 'numpy'`	`pip install "numpy>=1.25.0"`	exposed `einops`
12	`ModuleNotFoundError: No module named 'einops'`	`pip install einops`	exposed `psutil`
13	`ModuleNotFoundError: No module named 'psutil'`	`pip install psutil`	ComfyUI reached VRAM detection, then exposed `torchvision`
14	`ModuleNotFoundError: No module named 'torchvision'`	installed `torchvision` from the same CUDA wheel index	exposed `torchaudio`
15	`ModuleNotFoundError: No module named 'torchaudio'`	installed `torchaudio` from the same CUDA wheel index	exposed `scipy`
16	`ModuleNotFoundError: No module named 'scipy'`	`pip install scipy`	exposed `transformers`
17	`ModuleNotFoundError: No module named 'transformers'`	`pip install transformers`	exposed `torchsde`
18	`ModuleNotFoundError: No module named 'torchsde'`	`pip install torchsde`	exposed `av`
19	`ModuleNotFoundError: No module named 'av'`	`pip install av`	exposed `aiohttp`
20	`ModuleNotFoundError: No module named 'aiohttp'`	`pip install aiohttp`	exposed `requests`
21	`ModuleNotFoundError: No module named 'requests'`	`pip install requests`	exposed `pydantic`
22	`ModuleNotFoundError: No module named 'pydantic'`	`pip install pydantic`	exposed `comfyui-frontend-package`
23	`comfyui-frontend-package is not installed`	`pip install comfyui-frontend-package`	server started, but some built-in extra nodes failed to import

At that point ComfyUI was alive, but the log still showed missing optional or built-in-extra dependencies. The cleanup pass continued in the same style:

Cleanup signal	One package action	Result
`No module named 'kornia'` in post-processing, canny, and morphology nodes	`pip install kornia`	node count rose from 657 to 685
`No module named 'spandrel'` in `nodes_upscale_model.py`	`pip install spandrel`	node count rose to 687
repeated `No module named 'pydantic_settings'` while parsing `pyproject.toml`	`pip install pydantic-settings`	the startup log shrank from noisy repeats to a short warning set
`OpenGL dependencies not available`	`pip install PyOpenGL`, restart, then `pip install glfw`	GLSL/OpenGL node imported
`No module named 'simpleeval'` in `nodes_math.py`	`pip install simpleeval`	no `IMPORT FAILED` lines remained
`comfyui-workflow-templates is not installed`	`pip install comfyui-workflow-templates`	workflow template warning disappeared
`comfyui-embedded-docs package not found`	`pip install "comfyui-embedded-docs==0.4.4"`	embedded docs warning disappeared
`WARNING: blake3 package not installed`	`pip install blake3`	`blake3` warning disappeared
`Failed to import comfy_kitchen`	`pip install "comfy-kitchen>=0.2.8"`	CUDA and eager `comfy_kitchen` backends loaded

Step 0: Start From The Portable Root

Open PowerShell in the folder that contains:

run_nvidia_gpu.bat
run_cpu.bat
python_embeded
ComfyUI

Confirm the Python executable:

.\python_embeded\python.exe -s -c "import sys; print(sys.executable); print(sys.version)"

If this path is not the Python that starts ComfyUI, stop and switch to the right environment.

Step 1: Capture Each Attempt

For a one-by-one repair, save logs for each start:

$ts = Get-Date -Format 'yyyyMMdd_HHmmss'
New-Item -ItemType Directory -Force .\repair_logs | Out-Null
$out = ".\repair_logs\attempt_${ts}_stdout.log"
$err = ".\repair_logs\attempt_${ts}_stderr.log"
Start-Process -FilePath ".\python_embeded\python.exe" `
  -ArgumentList @("-s", "ComfyUI\main.py", "--windows-standalone-build") `
  -WorkingDirectory (Get-Location) `
  -RedirectStandardOutput $out `
  -RedirectStandardError $err `
  -PassThru

Then read the bottom of the stderr log:

Get-Content $err -Tail 200

Install only the package named by the first real blocker. Then start again and save a new log.

Step 2: Do Not Trust Port 8188 Until You Check Ownership

In the lab, http://127.0.0.1:8188/ returned 200, but it was another ComfyUI process from C:\Users\huhaibin\Documents\ComfyUI, not the broken lab.

Check the owner before declaring success:

Get-NetTCPConnection -LocalPort 8188 -ErrorAction SilentlyContinue |
  Select-Object LocalAddress,LocalPort,State,OwningProcess

Get-CimInstance Win32_Process |
  Where-Object { $_.CommandLine -like "*ComfyUI*main.py*" } |
  Select-Object ProcessId,CommandLine

If 8188 is already occupied, start the broken environment on a clean port while repairing:

.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --port 8191

Only the process from the same portable root counts as proof.

Step 3: Verify Torch Before And After Installing It

If the log says No module named 'torch', installing torch from the default index may satisfy the import while giving you a CPU-only build.

That happened in the lab:

2.12.0+cpu None False

For the tested Nvidia repair, Torch had to be replaced from the CUDA wheel index:

.\python_embeded\python.exe -s -m pip install `
  --index-url https://download.pytorch.org/whl/cu130 `
  --extra-index-url https://pypi.tuna.tsinghua.edu.cn/simple `
  "torch==2.12.0+cu130"

Then verify in the same embedded Python:

.\python_embeded\python.exe -s -c "import torch; print(torch.__version__, torch.version.cuda, torch.cuda.is_available()); print(torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'no cuda')"

The lab recovered this result:

2.12.0+cu130 13.0 True
NVIDIA GeForce RTX 4070 Ti SUPER

Use the CUDA wheel that matches your environment. Do not assume the default pip install torch is GPU-enabled.

Step 4: Treat Frontend And Template Packages As ComfyUI Versioned Packages

When the log says:

comfyui-frontend-package is not installed.

look at your own ComfyUI\requirements.txt before choosing versions.

In the lab, installing the frontend package unpinned gave:

ComfyUI frontend version: 1.43.18
required_frontend_version: 1.42.15

It worked, but the safer public instruction is: copy the versions required by your local ComfyUI code when they are pinned.

The same applies to:

comfyui-frontend-package
comfyui-workflow-templates
comfyui-embedded-docs

Step 5: Verify Recovery With More Than The Browser

The final lab start used:

.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build --port 8191

Healthy signals:

ComfyUI version: 0.20.1
pytorch version: 2.12.0+cu130
Device: cuda:0 NVIDIA GeForce RTX 4070 Ti SUPER : cudaMallocAsync
ComfyUI frontend version: 1.43.18
comfy-kitchen version: 0.2.8
Starting server
To see the GUI go to: http://127.0.0.1:8191

Then verify endpoints:

Invoke-RestMethod http://127.0.0.1:8191/system_stats
Invoke-RestMethod http://127.0.0.1:8191/object_info
.\python_embeded\python.exe -s -m pip check

The repaired lab returned:

ComfyUI 0.20.1
PyTorch 2.12.0+cu130
GPU: NVIDIA GeForce RTX 4070 Ti SUPER
object_info node count: 689
CheckpointLoaderSimple: present
KSampler: present
ComfyMathExpression: present
pip check: No broken requirements found.

What Is A Blocker And What Is Only A Warning?

Log line	Blocker?	What to do
`ModuleNotFoundError: No module named 'sqlalchemy'` before server start	Yes	Install SQLAlchemy in ComfyUI's embedded Python, then restart
`ModuleNotFoundError: No module named 'torch'`	Yes	Install Torch, then verify CUDA, not just import success
`comfyui-frontend-package is not installed`	Yes	Install the frontend package required by local ComfyUI
`IMPORT FAILED: nodes_math.py`	Partial	Server can start, but that node is missing; install `simpleeval`
`Starting server` appears	No	Server has started; continue with endpoint verification
`Found comfy_kitchen backend triton: {'available': False, ...}`	Usually no	Ignore unless a workflow actually requires Triton
`/system_stats` returns the expected GPU	No	Good sign that runtime is alive

One Package At A Time Or Requirements All At Once?

Use this decision:

Situation	Better move
You are diagnosing an unknown broken portable environment	One package, restart, read the next log
The first error is a custom node missing one package	Install that node's package or requirements
Torch imports but CUDA is false	Fix the CUDA Torch wheel first
Torch/CUDA is healthy	Avoid broad commands that may replace it
You are creating a fresh clean environment	`pip install -r ComfyUI\requirements.txt` can be normal

A blind pip install -r ComfyUI\requirements.txt can be acceptable for a clean manual install, but it is not the best forensic repair loop when you need to understand what broke.

What Not To Do

do not use system pip when ComfyUI uses python_embeded
do not assume a browser page on 8188 belongs to the environment you are repairing
do not treat CPU Torch import success as CUDA success
do not paste a long package list when you are trying to learn the next log change
do not treat every warning after Starting server as fatal
do not delete the environment before preserving workflows, models, and custom nodes

How Wonderful Launcher Helps

The hard part of this repair is not memorizing package names. It is the loop:

start
capture the first real error
repair the right Python
start again
distinguish blockers from warnings
verify the app is really alive

Source References

Windows portable repair lab, May 18, 2026: E:\ComfyUI_portable\ComfyUI_SEO_Lab
First fatal log: ModuleNotFoundError: No module named 'sqlalchemy'
Important pitfall: 8188 belonged to another ComfyUI process, so the lab was verified on port 8191
Final lab launch: ComfyUI 0.20.1, frontend 1.43.18, PyTorch 2.12.0+cu130, GPU NVIDIA GeForce RTX 4070 Ti SUPER
Final endpoint check: /object_info reported 689 nodes and pip check returned No broken requirements found

You can fix it manually, or download Wonderful Launcher for Windows to diagnose plugin errors, missing dependencies, and broken ComfyUI environments without reinstalling.

Download free for Windows

Did this fix your issue?

Your answer helps prioritize verified ComfyUI repairs.

Repair ComfyUI Portable Dependencies One Missing Module At A Time

When To Use This Runbook

The Tested Repair Chain

Step 0: Start From The Portable Root

Step 1: Capture Each Attempt

Step 2: Do Not Trust Port 8188 Until You Check Ownership

Step 3: Verify Torch Before And After Installing It

Step 4: Treat Frontend And Template Packages As ComfyUI Versioned Packages

Step 5: Verify Recovery With More Than The Browser

What Is A Blocker And What Is Only A Warning?

One Package At A Time Or Requirements All At Once?

What Not To Do

How Wonderful Launcher Helps

Source References

Table of Contents

Repair ComfyUI Portable Dependencies One Missing Module At A Time

When To Use This Runbook

The Tested Repair Chain

Step 0: Start From The Portable Root

Step 1: Capture Each Attempt

Step 2: Do Not Trust Port 8188 Until You Check Ownership

Step 3: Verify Torch Before And After Installing It

Step 4: Treat Frontend And Template Packages As ComfyUI Versioned Packages

Step 5: Verify Recovery With More Than The Browser

What Is A Blocker And What Is Only A Warning?

One Package At A Time Or Requirements All At Once?

What Not To Do

How Wonderful Launcher Helps

Source References

Table of Contents