Skip to main content

comfy-test

Testing infrastructure for ComfyUI custom nodes.

Test your nodes install and work correctly across Linux, macOS, Windows, and Windows Portable. No pytest code needed.

Quick Start

Add these files to your custom node repository:

1. comfy-test.toml

[test]
# Name is auto-detected from directory

[test.workflows]
cpu = "all"  # Run all workflows in workflows/ folder

2. .github/workflows/test-install.yml

name: Test Installation
on: [push, pull_request]

jobs:
  test:
    uses: PozzettiAndrea/comfy-test/.github/workflows/test-matrix.yml@main

3. workflows/test.json

A minimal ComfyUI workflow that uses your nodes. Export from ComfyUI.

Done! Push to GitHub and your tests will run automatically on all platforms.

Test Levels

comfy-test runs up to 10 test levels in sequence:

Level Name What It Does
1 SYNTAX Check project structure (pyproject.toml/requirements.txt), CP1252 compatibility, forbidden patterns
2 COVERAGE Every registered node is used by at least one bundled workflow (opt-in: fails on unused nodes)
3 INSTALL Clone ComfyUI, create environment, install node + dependencies
4 REGISTRATION Start server, verify nodes appear in /object_info
5 INSTANTIATION Test each node's constructor
6 STATIC_CAPTURE Screenshot workflows (no execution)
7 VALIDATION 4-level workflow validation (schema, graph, introspection, partial execution)
8 EXECUTION_LIGHT Run workflows end-to-end, one screenshot each (no video; for weak runners — use instead of EXECUTION)
9 EXECUTION Run workflows end-to-end, capture outputs + per-frame video
10 CUSTOM Your own hook ([test] custom = "tests/my_check.py" exposing run(ctx)); runs last against the live server

The default set is levels 1, 3-7, 9 (coverage and execution_light are opt-in; levels = "all" runs everything). Each level depends on previous levels. You can run up to a specific level with --level:

comfy-test run --level registration  # Runs: SYNTAX -> INSTALL -> REGISTRATION

Workflow Validation (4 Levels)

The VALIDATION level runs comprehensive checks before execution:

Level Name What It Checks
1 Schema Widget values match allowed enums, types, and ranges
2 Graph Connections are valid, all referenced nodes exist
3 Introspection Node definitions are well-formed (INPUT_TYPES, RETURN_TYPES, FUNCTION)
4 Partial Execution Runs non-CUDA nodes to verify they work

Detecting CUDA Nodes

To mark nodes as requiring CUDA (excluded from partial execution), use comfy-env.toml:

[cuda]
packages = ["nvdiffrast", "flash-attn"]

Configuration Reference

Minimal Config

[test]
# Levels default to: syntax, install, registration, instantiation,
# static_capture, validation, execution

[test.workflows]
cpu = "all"

[test.platforms]
# Explicit opt-in allowlist — only listed platforms run.
platforms = ["linux-cpu", "macos-cpu", "windows-cpu", "windows-portable-cpu"]

Full Config Example

[test]
# Name is auto-detected from directory name (e.g., "ComfyUI-MyNode")

# ComfyUI version to test against
comfyui_version = "latest"  # or a tag like "v0.2.0" or commit hash

# Python version (default: random from 3.11, 3.12, 3.13)
python_version = "3.11"

# Test levels to run. Default: syntax, install, registration, instantiation,
# static_capture, validation, execution. Options additionally: coverage,
# execution_light, custom. Or run everything:
levels = "all"

# Optional custom hook: a Python file exposing run(ctx) — raise to fail.
# Runs last, with the live server available via ctx.server / ctx.api.
custom = "tests/my_check.py"

# Platforms are an explicit opt-in ALLOWLIST — only listed targets run.
# Valid tokens: linux-cpu, macos-cpu, windows-cpu, windows-portable-cpu,
# macos-desktop, windows-desktop, linux-cuda, windows-cuda,
# windows-portable-cuda, windows-desktop-cuda
# (bare "linux"/"macos"/"windows"/"windows_portable" mean the cpu variant)
[test.platforms]
platforms = ["linux-cpu", "macos-cpu", "windows-cpu", "windows-portable-cpu"]

# Workflow configuration — accelerator is named by backend: cpu / cuda / rocm.
[test.workflows]
# Workflows to run on CPU runners (GitHub-hosted)
cpu = "all"  # or a list: ["test_basic.json"], or all-except: ["!heavy.json"]

# Workflows to run on CUDA runners (self-hosted); rocm reserved for later
cuda = ["test_cuda.json"]

# Timeout for workflow execution in seconds (default: 3600)
timeout = 120

# Platform-specific settings (enablement comes from the allowlist above)
[test.linux]
skip_workflow = false  # Skip workflow execution, only verify registration

[test.windows_portable]
comfyui_portable_version = "latest"  # Portable-specific version

Workflow Discovery

Workflows are auto-discovered from the workflows/ folder:

  • All .json files in workflows/ are found automatically
  • Use cpu = "all" to run all discovered workflows on CPU
  • Use cuda = "all" to run all discovered workflows on CUDA runners
  • Or specify individual files: cpu = ["basic.json", "advanced.json"]
  • Or exclude: cpu = ["!heavy.json"] runs everything except heavy.json

CLI

# Install
pip install comfy-test

# Initialize config and GitHub workflow
comfy-test init

# Run tests locally
comfy-test run --platform linux

# Run specific level only
comfy-test run --level registration

# Publish results to GitHub Pages
comfy-test publish ./results --repo owner/repo

CUDA Packages on CPU-only CI

comfy-test runs on CPU-only GitHub Actions runners. For nodes that use CUDA packages:

  1. Installation works - comfy-test sets COMFY_ENV_CUDA_VERSION=12.8 so comfy-env can resolve wheel URLs
  2. Import may fail - CUDA packages typically fail to import without a GPU

For full CUDA testing, use a self-hosted runner with a GPU.

License

MIT

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

comfy_test-0.4.7.tar.gz (271.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

comfy_test-0.4.7-py3-none-any.whl (332.4 kB view details)

Uploaded Python 3

File details

Details for the file comfy_test-0.4.7.tar.gz.

File metadata

  • Download URL: comfy_test-0.4.7.tar.gz
  • Upload date:
  • Size: 271.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for comfy_test-0.4.7.tar.gz
Algorithm Hash digest
SHA256 485f7228380e65cd33e6086c1ad8d59f5ad6fceb03a0ec1c6302a9d4e278d04a
MD5 c4d34d55418f0676f698006961ad0f3f
BLAKE2b-256 2b892e59f0f13568e36d7fb503f2fb5ca81efa275af903d4a6c4ccc877c24798

See more details on using hashes here.

Provenance

The following attestation bundles were made for comfy_test-0.4.7.tar.gz:

Publisher: publish.yml on PozzettiAndrea/comfy-test

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file comfy_test-0.4.7-py3-none-any.whl.

File metadata

  • Download URL: comfy_test-0.4.7-py3-none-any.whl
  • Upload date:
  • Size: 332.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for comfy_test-0.4.7-py3-none-any.whl
Algorithm Hash digest
SHA256 cfdd772a6efe2b82ad40cfd51dd015d7e528024b274501248657ee85449a3d8f
MD5 6df392f28bfbd8538d090aa5fbc0eadb
BLAKE2b-256 4517820203808532ee76ad051206910c930594a96f36eedc7b676103f71f991b

See more details on using hashes here.

Provenance

The following attestation bundles were made for comfy_test-0.4.7-py3-none-any.whl:

Publisher: publish.yml on PozzettiAndrea/comfy-test

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Release history Release notifications | RSS feed

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page