venv

Python officially provides the venv module for implementing basic virtual environments.

A common way to deal with environment configuration issues is to use virtual environments (such as venv or conda) or containers (such as Docker) to isolate projects, and to clearly tell the AI the current environment version in the prompt.

When determining environment ownership, pip relies on sys.prefix and the location of the EXTERNALLY-MANAGED file; if a venv is incompletely created, PYTHONPATH is inherited from outside, or site-packages symlinks are abnormal, pip may misjudge an activated venv as an externally managed environment and refuse to write.

The standard practice for self-hosting a Python project is to first create a virtual environment with python -m venv venv, activate it, and then run pip install -r requirements.txt, avoiding pollution of the system environment.

Creating and activating a virtual environment (venv) can effectively prevent dependency pollution between subsequent projects.

Python 虚拟环境（venv、conda）只能隔离 Python 包，无法隔离系统级依赖如 C 库、CUDA 运行时

venv is a virtual environment tool built into Python 3.3+, achieving dependency isolation by creating an independent interpreter copy and a site-packages directory under the project directory.