workflow-reader — ComfyUI Workflow Analysis

When you need to decode a ComfyUI workflow (inspect its nodes, expand subgraphs, or check pipeline feasibility), use the tool at tools/workflow.py — either run it from the CLI or import its functions directly in a Python script.

The tool is pure Python and requires no ComfyUI runtime.

---

CLI Usage

Run against one or more workflow JSON files to print a full node-type report:

python comfy_diffusion/skills/workflow-reader/tools/workflow.py \
    comfyui_official_workflows/video/ltx/ltx2/video_ltx2_t2v.json

# Multiple files at once
python comfy_diffusion/skills/workflow-reader/tools/workflow.py \
    comfyui_official_workflows/video/ltx/ltx2/*.json

Output per file:

• Subgraph names (if any)
• All node types after subgraph expansion (sorted)
• Subgraph I/O ports (inputs/outputs that map to pipeline parameters)

---

Python API

Import the four functions directly from the tool file:

import sys
sys.path.insert(0, "comfy_diffusion/skills/workflow-reader/tools")
from workflow import load_workflow, get_node_types, get_subgraph_names, get_subgraph_io

load_workflow(path) → dict

Load a workflow JSON file. Raises FileNotFoundError or ValueError on bad input.

wf = load_workflow("comfyui_official_workflows/video/ltx/ltx2/video_ltx2_t2v.json")

get_node_types(workflow) → set[str]

Return all ComfyUI node class names used, subgraphs fully expanded recursively. Display-only nodes are excluded (MarkdownNote, Note, Reroute, Primitive*).

types = get_node_types(wf)
# {'CheckpointLoaderSimple', 'CLIPTextEncode', 'LTXVConditioning', ...}

get_subgraph_names(workflow) → list[str]

Human-readable names of all subgraphs defined in the workflow.

get_subgraph_names(wf)
# ['Text to Video (LTX 2.0)']

get_subgraph_io(workflow) → list[dict]

Input/output port definitions for each subgraph — these become the pipeline parameters.

io = get_subgraph_io(wf)
# [{'name': 'Text to Video (LTX 2.0)', 'id': '...', 'inputs': [...], 'outputs': [...]}]

---

Pipeline Feasibility Check

Use this pattern to decide whether a workflow is directly convertible to a Python pipeline:

import sys
sys.path.insert(0, "comfy_diffusion/skills/workflow-reader/tools")
from workflow import load_workflow, get_node_types

SUPPORTED = {
    "CheckpointLoaderSimple", "CLIPLoader", "CLIPTextEncode",
    "LTXVConditioning", "LTXVImgToVideo", "EmptyLTXVLatentVideo",
    "LTXVScheduler", "KSamplerSelect", "SamplerCustom",
    "SamplerCustomAdvanced", "CFGGuider", "RandomNoise",
    "ManualSigmas", "VAEDecode", "VAEDecodeTiled",
    "LTXVCropGuides", "LTXVEmptyLatentAudio", "LTXVConcatAVLatent",
    "LTXVSeparateAVLatent", "LTXVAudioVAEDecode", "LTXVLatentUpsampler",
    "LoraLoaderModelOnly", "LoadImage", "LoadVideo", "LoadAudio",
    "SaveVideo", "CreateVideo", "GetImageSize", "ImageScaleBy",
    "ImageFromBatch", "GetVideoComponents",
}

def analyze(path):
    wf = load_workflow(path)
    types = get_node_types(wf)
    missing = types - SUPPORTED
    return {
        "path": path,
        "convertible": len(missing) == 0,
        "missing": sorted(missing),
    }

---

Subgraph Structure Reference

workflow["definitions"]["subgraphs"]  →  list of subgraph objects

Each subgraph:

• id — UUID matching the type field of the subgraph node in workflow["nodes"]
• name — human-readable label
• inputs / outputs — port definitions (name, type, linkIds)
• nodes — inner nodes (same format as top-level; may reference other UUIDs → resolved recursively with cycle detection)