OpenCodeRunner

A fully free and open-source API that runs project-level (multi-file) code on both local machines and remote servers with sandboxes. It supports languages including Python JS/TS C/C++ Java Dafny Bash SQL etc.

OpenCodeRunner can be used in many ways:

• LLM, agent, reasoning, RL
• Project-level code evaluation
• Safe sandbox running
• Multi-language running
• Cloud or remote code running
• ...

Installation

1. Install OpenCodeRunner

# Install
git clone https://github.com/OpenCodeRunner/OpenCodeRunner
cd OpenCodeRunner
pip install -r requirements.txt
pip install -e .

# Uninstall
pip uninstall opencoderunner 
bash ./uninstall.sh

2. (Optional) Install Firejail sandbox for safety control (https://github.com/netblue30/firejail)

# Install via `apt`
sudo add-apt-repository ppa:deki/firejail
sudo apt-get update
sudo apt-get install firejail firejail-profiles

# Or install via building manually
git clone https://github.com/netblue30/firejail.git
cd firejail
./configure && make && sudo make install-strip

3. The installation guides for each code language's environment is in install_envs.md

Usage

1. (Optional) Start OpenCodeRunner Service

You can use firejail uvicorn to start the OpenCodeRunner service for remotely code execution. You can configure API key settings in .env for both client and server sides.

# Under the repo's root `OpenCodeRunner/` directory
firejail && bash start_server.sh

# Or if you don't have `firejail`
bash start_server.sh

2. Running Code

Then you can use either run_on_local or run_on_server for code running. If you use run_on_server, you need to specify host port api_key. You can configure API key settings in .env for both client and server sides.

• Code string running example.

from opencoderunner.run_info import RunInfo
from opencoderunner import run 

if __name__ == "__main__":
    run_info = RunInfo(
        code_str="import sys; print(sys.stdin.read())",
        language="python",
        project_root_name="project_root_name",  
        input_content="INPUT1\nINPUT2\n",
        timeout=1, # Test timeout
    )                    

    # # -- Run locally
    for i in range(3):
        result_info = run(run_info=run_info)
        print(run_info.command)
        print(result_info)

    # -- Or Run on server
    for i in range(3):
        result_info = run(run_info=run_info,
                        host="0.0.0.0",
                        port=8000,
                        )
        print(result_info)

• Quick run bash command example.

from opencoderunner import RunInfo
from opencoderunner import run as opencr_run 

if __name__ == "__main__":
    run_info = RunInfo(
        code_str="sleep 3",
        language="bash",
        timeout=2, # Test timeout
        is_quick_run=True, # Test quick run
    )                    
    # -- Run locally
    for i in range(3):
        result_info = opencr_run(run_info=run_info)
        print(run_info.command)
        print(result_info)

• Project-level example. It consists of multiple files.

from opencoderunner.run_info import RunInfo
from opencoderunner.file_info import FileInfo
from opencoderunner import run

if __name__ == "__main__":
    run_info = RunInfo(
        file_infos=[
            FileInfo(
                file_relpath="folder1/file1.py",
                file_content="""
def main1():
    import numpy as np
    arr = np.array([1, 2, 3])
    print("Hello World")
    output = arr.sum()
    return output
"""
            ),
            FileInfo(
                file_relpath="file2.py",
                file_content="""
from folder1.file1 import main1
import sys
import json

print(main1())
for line in sys.stdin:
    line = line.strip()
    print(line)
"""
            )
        ],
        language="python",
        project_root_name="project_root_name",                   
        entry_file_relpath="file2.py",
        input_content="INPUT1\nINPUT2\n",
    )                               
    run_info.print_tree()

    # -- Run locally
    for i in range(3):
        result_info = run(run_info=run_info)
        print(result_info)


    # -- Or Run on server
    for i in range(3):
        result_info = run(run_info=run_info,
                        host="0.0.0.0",
                        port=8000,
                        )
        print(result_info)

3. Configure Paths

You can configure the default paths for different languages in opencoderunner/infos/run_info.py. For example, python path, cmake path, java JDK path, node.js path, etc.

4. More Examples

More examples for various code languages are in examples/. You can run them easily.

TODO

• Sandbox for permission control
• Project structure visualization
• Input/argparse/stdin
• PyPi/Conda
• API key
• Code str running (simple running)
• Merge local+server
• os.getcwd issue
• remove too much print

Supported Languages

• Python
• JavaScript
• TypeScript
• C (CMake)
• C++ (CMake)
• Java
• Dafny
• Bash
• SQL
• ...

Update

• [2025-04-17] Launch OpenCodeRunner

Contributors

😊 We welcome passionate contributors! If you are interested, feel free to contact!