Code Block

The Code Block node lets you write and execute custom Python code within your workflow, with dynamic inputs/outputs, 94+ packages, and a built-in test environment.

What does the Code Block node do?

The Code Block node lets you write and execute custom Python code directly in your workflow. When the built-in nodes don’t cover your specific use case, Code Block gives you full programming flexibility with access to 94+ Python packages.

Common use cases:

• Transforming data in ways that standard nodes can’t handle (custom parsing, reformatting, calculations)
• Running data analysis or statistical computations using NumPy, Pandas, or SciPy
• Scraping or processing web content with BeautifulSoup and Requests
• Applying custom business logic, scoring algorithms, or validation rules to workflow data

Quick setup

Configuration parameters

Code Editor

Available Python packages

The Code Block environment includes 94+ pre-installed packages. Here are the main packages you can use directly:

What does the node output?

The output format depends on how many outputs you define:

Single output (e.g., result):

"The processed value"

The value is returned directly to the connected downstream node.

Multiple outputs (e.g., summary and score):

{
  "summary": "The article covers SEO best practices...",
  "score": 85
}

Each output becomes a separate connection point on the node.

Usage examples

Example 1: Calculate readability score

You want to compute a custom readability score for article content that no built-in node provides.

Inputs: text (required) Outputs: score, level

Code:

words = text.split()
sentences = text.count('.') + text.count('!') + text.count('?')
if sentences == 0:
    sentences = 1
avg_words_per_sentence = len(words) / sentences
score = max(0, min(100, 100 - (avg_words_per_sentence - 15) * 5))
level = "Easy" if score > 70 else "Medium" if score > 40 else "Hard"

Example 2: Parse and restructure JSON data

You receive complex JSON from an API and need to extract and reshape specific fields.

Inputs: raw_data (required) Outputs: cleaned

Code:

import json
data = json.loads(raw_data)
cleaned = json.dumps([
    {"name": item["full_name"], "email": item["contact"]["email"]}
    for item in data["results"]
    if item.get("active", False)
])

Example 3: Analyze an image with OpenCV

Extract dominant colors from a product image.

Inputs: image_url (required) Outputs: colors

Code:

import requests
import numpy as np
import cv2
from collections import Counter

response = requests.get(image_url)
img_array = np.frombuffer(response.content, np.uint8)
img = cv2.imdecode(img_array, cv2.IMREAD_COLOR)
img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
pixels = img_rgb.reshape(-1, 3)
rounded = (pixels // 32 * 32).tolist()
most_common = Counter(map(tuple, rounded)).most_common(5)
colors = [{"rgb": list(c), "frequency": f} for c, f in most_common]

Best practices

Common issues

How does it fit into a workflow?

Code Block is a versatile node that can be placed anywhere in a workflow for custom data processing:

graph LR
    Input[Data Source] --> Code[Code Block: Custom Logic]
    Code --> Next[LLM / Output / Next Step]

Related nodes