> ## Documentation Index
> Fetch the complete documentation index at: https://docs.markifact.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Python Code

> Execute custom Python code with access to data analysis libraries. Perfect for custom calculations, data transformations, and advanced processing.

**Python Code** executes arbitrary Python code and returns the result. Use this node when you need custom logic, calculations, or data processing that isn't available in existing nodes. Supports all standard libraries plus popular data analysis tools like pandas, numpy, and more.

***

## When to Use It

**⚠️ Advanced Node:** Use this only when existing nodes cannot accomplish your task. Always check if a specialized node exists first.

* Perform complex calculations not available in existing math/utility nodes
* Transform data in highly specific ways not supported by existing data nodes
* Apply advanced statistical analysis, machine learning, or scientific computing
* Parse or manipulate complex data structures requiring custom logic
* Create sophisticated business logic that combines multiple operations
* Process data with specialized libraries (computer vision, audio analysis, NLP)

**Before using this node:** Review available nodes in AI, Utilities, and other categories to ensure no existing solution meets your needs.

***

## Inputs

| Field         | Type      | Required | Description                                                   |
| ------------- | --------- | -------- | ------------------------------------------------------------- |
| **Variables** | Variables | No       | Dynamic data from other nodes injected into your code context |
| **Code**      | Code      | Yes      | The Python code to execute (result must be on the last line)  |
| **Packages**  | List      | No       | Additional Python packages to install (max 5 packages)        |

***

## Outputs

| Output   | Description                                       |
| -------- | ------------------------------------------------- |
| **Data** | The results value returned by your code execution |

***

## Credit Cost

2 credits per execution.

***

## How It Works

Python Code executes your custom Python script in a Jupyter notebook-like environment with access to extensive data science libraries. Like in Jupyter notebooks, only the result of the last line is returned as output. Variables from other nodes are automatically injected into the execution context, so you can reference them directly in your code.

**Key Features:**

* Jupyter notebook-style execution (last line becomes the output)
* Access to 30+ pre-installed data science libraries
* Automatic variable injection from workflow data
* JSON-safe output formatting
* Secure execution environment

***

## Available Libraries

The Python Code node includes 30+ pre-installed libraries for data science, web scraping, visualization, and more:

| Library            | Version      | Use Case                                       |
| ------------------ | ------------ | ---------------------------------------------- |
| **aiohttp**        | v3.9.3       | Asynchronous HTTP client/server                |
| **beautifulsoup4** | v4.12.3      | Web scraping and HTML/XML parsing              |
| **bokeh**          | v3.3.4       | Interactive visualization                      |
| **gensim**         | v4.3.2       | Topic modeling and document analysis           |
| **imageio**        | v2.34.0      | Image I/O operations                           |
| **joblib**         | v1.3.2       | Parallel computing and model persistence       |
| **librosa**        | v0.10.1      | Audio analysis and music information retrieval |
| **matplotlib**     | v3.8.3       | Data visualization and plotting                |
| **nltk**           | v3.8.1       | Natural language processing                    |
| **numpy**          | v1.26.4      | Numerical computing and arrays                 |
| **opencv-python**  | v4.9.0.80    | Computer vision and image processing           |
| **openpyxl**       | v3.1.2       | Excel file reading and writing                 |
| **pandas**         | v1.5.3       | Data manipulation and analysis                 |
| **plotly**         | v5.19.0      | Interactive web-based visualizations           |
| **pytest**         | v8.1.0       | Testing framework                              |
| **python-docx**    | v1.1.0       | Microsoft Word document manipulation           |
| **pytz**           | v2024.1      | Timezone handling                              |
| **requests**       | v2.26.0      | HTTP requests and API calls                    |
| **scikit-image**   | v0.22.0      | Image processing algorithms                    |
| **scikit-learn**   | v1.4.1.post1 | Machine learning library                       |
| **scipy**          | v1.12.0      | Scientific computing                           |
| **seaborn**        | v0.13.2      | Statistical data visualization                 |
| **soundfile**      | v0.12.1      | Audio file I/O                                 |
| **spacy**          | v3.7.4       | Advanced natural language processing           |
| **sympy**          | v1.12        | Symbolic mathematics                           |
| **textblob**       | v0.18.0      | Simple text processing                         |
| **tornado**        | v6.4         | Web framework and networking                   |
| **urllib3**        | v1.26.7      | HTTP client library                            |
| **xarray**         | v2024.2.0    | Multi-dimensional arrays and datasets          |
| **xlrd**           | v2.0.1       | Excel file reading                             |

Plus all standard Python libraries (json, datetime, math, statistics, etc.)

***

## Installing Additional Packages

Need a package not included in the pre-installed libraries? You can install up to 5 additional Python packages using the **Packages** field.

**Guidelines:**

* **Maximum 5 packages** per execution
* Common data science packages (pandas, numpy, requests, etc.) are already available - don't reinstall them
* Use specific versions when needed: `"package==1.2.3"`
* Packages are installed at runtime before code execution, which may slightly increase execution time
* Packages are temporary and only available for the current execution

***

## Code Requirements

**Jupyter-Style Execution:**

* Works like a Jupyter notebook - only the last line's result is returned
* No need for explicit `return` statements
* The final expression becomes the output automatically

**Return Value:**

* Place your result as the final line of code
* Must be JSON-safe types: dict, list, str, int, float, bool, or None
* No `print()` statements for output - use the last line instead

**Valid Examples:**

```python theme={"dark"}
# Simple calculation (last line is the result)
total_sales = sum(row['amount'] for row in sales_data)
average_sales = total_sales / len(sales_data)
average_sales  # This becomes the output

# Multiple values as dict
data_summary = {
    "total": sum(amounts), 
    "count": len(amounts),
    "average": sum(amounts) / len(amounts)
}
data_summary  # This becomes the output

# Processed data list
processed = [{"name": item["name"].upper(), "score": item["score"] * 1.1} for item in raw_data]
processed  # This becomes the output
```

**Invalid Returns:**

* pandas DataFrames (convert to dict/list first: `df.to_dict('records')`)
* numpy arrays (use `.tolist()`)
* Custom objects or functions
* Complex nested objects

***

## Variables Usage

Variables from other nodes are automatically available in your code context:

**Setup Variables:**

```json theme={"dark"}
{
  "sales_data": "{{sheetsNode.data}}",
  "target_amount": "{{inputNode.data}}",
  "account_ids": "{{adsNode.data}}"
}
```

**Use in Code:**

```python theme={"dark"}
# Variables are already available - no need to define them
total_sales = sum(row['amount'] for row in sales_data)
performance = total_sales / target_amount

{"total": total_sales, "performance_ratio": performance}
```

***

## Examples

### Example 1: Calculate Campaign Performance

**Variables:**

```json theme={"dark"}
{
  "campaigns": "{{adsNode.data}}",
  "target_roas": "{{inputNode.data}}"
}
```

**Code:**

```python theme={"dark"}
results = []
for campaign in campaigns:
    roas = campaign['revenue'] / campaign['cost'] if campaign['cost'] > 0 else 0
    performance = "Above Target" if roas >= target_roas else "Below Target"
    
    results.append({
        "campaign_name": campaign['name'],
        "roas": round(roas, 2),
        "status": performance
    })

results
```

### Example 2: Data Cleaning and Transformation

**Variables:**

```json theme={"dark"}
{
  "raw_data": "{{dataNode.data}}"
}
```

**Code:**

```python theme={"dark"}
import pandas as pd

# Convert to DataFrame for easier processing
df = pd.DataFrame(raw_data)

# Clean and transform
df['email'] = df['email'].str.lower().str.strip()
df['phone'] = df['phone'].str.replace(r'[^\d]', '', regex=True)
df = df.dropna(subset=['email'])

# Convert back to list of dicts
df.to_dict('records')
```

***

## FAQ

<Accordion title="What happens if my code has an error?">
  The node will fail and display the Python error message. Check your syntax, variable names, and ensure all required libraries are available.
</Accordion>

<Accordion title="Can I install additional packages?">
  Yes! Use the **Packages** field to install up to 5 additional Python packages. Specify package names as a list (e.g., `["countryinfo==0.1.2"]`). Common packages like pandas, numpy, and requests are already pre-installed, so don't reinstall them.
</Accordion>

<Accordion title="How do I return multiple values?">
  Combine them into a single dictionary: `{"value1": result1, "value2": result2}`. You can also return a list of dictionaries for multiple records.
</Accordion>

<Accordion title="Why can't I return a pandas DataFrame directly?">
  DataFrames aren't JSON-serializable. Convert them first: `df.to_dict('records')` for row-based data or `df.to_dict()` for column-based data.
</Accordion>

<Accordion title="How do I debug my code?">
  Use simple return statements to check intermediate values. For example, return `{"debug": variable_name}` to see what data you're working with.
</Accordion>

<Accordion title="Can I make HTTP requests in my code?">
  Yes, use the `requests` library to make API calls: `import requests; response = requests.get('https://api.example.com')`.
</Accordion>

<Accordion title="What if I need to process very large datasets?">
  Consider using pandas for efficient data operations, or break processing into smaller chunks. Be mindful of memory usage and execution time limits.
</Accordion>

<Accordion title="Can I save files or write to disk?">
  No, the execution environment is read-only. All data must be returned through the result value for use in subsequent workflow steps.
</Accordion>