Code Interpreter
Sandboxed Python execution via WASM Component Model. Two nodes: inline eval and project-based execution with dynamic PyPI package installation.
About
Python Interpreter Node
Run Python code directly inside your workflows — no setup, no servers, no hassle.
The Python Interpreter node brings the full power of Python right into Flow-Like. Whether you need to crunch some numbers, transform data on the fly, or let an AI agent write and execute code as part of a larger workflow — this node has you covered.
What does it do?
Think of it as a tiny Python environment living inside your workflow. You give it Python code (either written by you or generated dynamically by another node), and it runs it — safely, locally, and without needing Python installed on your machine. It even supports installing pure Python packages from pip, so you're not limited to just the basics.
When would I use this?
Agentic Code Evaluation — This is where things get really interesting. Imagine an AI agent that writes a small Python script to solve a problem, then passes it to this node to actually run it and check the result. The agent can iterate, fix mistakes, and try again — all within a single workflow. It's like giving your AI a scratch pad to think on.
Quick Data Transformations — Got some JSON that needs reshaping? A CSV that needs filtering? Instead of building a complex chain of nodes, sometimes a few lines of Python are just faster and more expressive.
Prototyping & Experimentation — Want to test a formula, try out a calculation, or validate some logic before committing to a full node pack? Drop in a Python Interpreter node, write your code, and see what happens.
Extending Your Workflows — Need something niche that doesn't have a dedicated node yet? If there's a pure Python library for it on pip, you can pull it in and use it right away — no need to wait for a custom node to be built.
What are "pure Python" pip packages?
Some Python packages are written entirely in Python (like requests, pydantic, or beautifulsoup4), while others rely on compiled C/C++ code under the hood (like numpy or pandas). This node supports the pure Python ones, which still covers a surprisingly large number of useful libraries — from web scraping to data validation to working with APIs.
Good to know
- Safe by design — Code runs inside Flow-Like's sandboxed environment, so it can't mess with your system or access things it shouldn't.
- Local-first — Everything executes on your machine. Your code and data never leave your device.
- Dynamic input — The code doesn't have to be hardcoded. It can come from another node's output, making it perfect for AI-driven workflows where the code is generated at runtime.
Example: Let an AI solve a math problem
- Prompt Node → Ask an LLM: "Write Python code that calculates compound interest for €10,000 at 5% over 10 years."
- Python Interpreter Node → Receives the generated code and executes it.
- Output Node → Returns the result: €16,288.95
No manual coding. No copy-pasting into a terminal. Just wire it up and let it flow.
Use Case
Python Code Execution
Release Notes
- Added two default modes
Provided Nodes
3 nodes included in this package.
Code / Python
LLM-powered agent that solves tasks by writing and executing Python code. Takes a model (Bit) and a task description, then uses the model to iteratively write and run Python code until it produces an answer. Compatible with any LLM that supports tool/function calling.
Execute inline Python code in a WASM sandbox. Available globals inside the script: - inputs — dict with values from the Inputs pin - outputs — write your results here - ctx — SDK Context (storage, HTTP, LLM, logging, …)
Execute a Python project from a FlowPath directory. Reads main.py (or a custom entry point) from the project root, along with any requirements.txt for automatic dependency installation. All .py files in the project are available for import.
