VS Code task and Python virtual environment

It might be a little late to answer your question but the trick is to set the command field to point to the virtual environment's python executable instead of the default python. If you set up your .vscode's settings.json correctly you should have something like this in your file:

{
    "python.pythonPath": "env\\Scripts\\python.exe",
    // other settings ...
}

Having this config in your workspace, you can then create a custom task of type process using ${config:python.pythonPath} as it's command field.

You can read all about it in this section of VSCode's docs.

---

This example creates a Django python manage.py migrate task:

{
    "version": "2.0.0",
    "tasks": [
        {
            "label": "Migrate",
            "type": "process",
            "command": "${config:python.pythonPath}",
            "args": [
                "${workspaceFolder}/src/manage.py",
                "migrate"
            ],
            "presentation": {
                "echo": true,
                "reveal": "always",
                "focus": false,
                "panel": "shared",
                "showReuseMessage": true,
                "clear": true
            },
            "problemMatcher": []
        }
    ]
}

I tested with the following tasks.json and it works:

{
  "version": "2.0.0",
  "tasks": [
    {
      "label": "test",
      "type": "shell",
      "command": "source /home/me/.pyenvs/myenv/bin/activate; make"
    }
  ]
}

First activate virtual environment and then executes make.

I had quite a lot of trouble with this recently myself. I wanted to update package requirements before running code, but I didn't want to implement a complex or maintenance-heavy solution. The following works well for me across Windows, OSX, and Linux environments.

{
  "version": "2.0.0",
  "tasks": [
    {
      "label": "pipInstall",
      "type": "shell",
      "command": "${command:python.interpreterPath}",
      "args": ["-m", "pip", "install", "-r", "requirements.txt"]
    }
  ]
}

I believe ${command:python.interpreterPath} points to whatever interpreter path one selects while setting up the environment.

I'm late to the party, but this alternative might be useful. If you use pipenv in stead of standard venv, you can use pipenv run. It will activate the virtualenv before running the process. For example, this works for building sphinx:

{
    "version": "2.0.0",
    "tasks": [
        {
            "label": "build html",
            "type": "process",
            "command": "pipenv",
            "args": [
                "run", 
                "sphinx-build", 
                "-b", 
                "html", 
                "${workspaceFolder}", 
                "${workspaceFolder}/_build/html"
            ],
            "group": {
                "kind": "build",
                "isDefault": true
            },
            "presentation": {
                "reveal": "always",
                "panel": "new"
            }
        }
    ]
}