Requirements

A dependency panel for your project. It reads and writes your project's requirements.txt and installs, uninstalls and checks packages against a runtime you choose.

What you can do

• See every dependency your project declares, with its version specifier, read straight from requirements.txt
• Know where each one stands — installed, outdated, or not installed — against the runtime you selected
• Add a requirement, which writes the line and installs it in one step
• Install a single package, or install everything you have declared
• Remove a requirement, which deletes the line and uninstalls the package
• Choose whether all of that applies to your project environment or to a kernel, and always see which one you are acting on

Declaration and runtime

requirements.txt is the declaration — what your project needs. A runtime target is where that declaration is realized. You pick the target at the top of the panel:

• Your project environment — the Python environment selected in the status bar. This is what scripts, tests and the CLI use.
• A kernel — the environment a notebook actually runs in. It may be on this machine or on a remote server.

These are not rivals. A local .venv for scripts and a remote GPU kernel for training is an ordinary setup, and a package installed into one is genuinely not present in the other.

Good to know

Non-package lines (-r includes, -e editable, git+/URL requirements, pip options) are preserved verbatim and shown, but not interpreted.

Opening the panel never starts a kernel just to check status. If the target is a kernel that is not already running, status stays unknown until you press Refresh.

If an install fails, the line stays. A dependency you declared is still real, and the row simply shows not installed.

If the selected target cannot be used — no environment chosen, a kernel endpoint that no longer exists, a server that cannot be reached — the panel says so and stops. It will not quietly install somewhere else.

Packages installed into a remote or ephemeral kernel live in that server's environment and may not survive it being reset. requirements.txt keeps the declaration, so you can reinstall in one click.

The extension makes no network requests of its own.

Requirements

• An open workspace containing (or to contain) a requirements.txt.
• At least one runtime: a Python environment (see Python Environments) or a kernel endpoint (see Kernels).

Getting started

Open the Requirements panel, pick your runtime target at the top, and press Install all.