Exploring the March 2026 Python Extension for VS Code: Q&A

The March 2026 release of the Python extension for Visual Studio Code brings two major enhancements: the ability to search for Python symbols across installed packages and an experimental Rust-based parallel indexer designed to accelerate IntelliSense. This Q&A covers what these features do, how to enable them, and what to expect in terms of performance and workflow improvements.

What are the key new features in the March 2026 Python extension release?

This release focuses on improving code navigation and exploration. The first feature lets you search for symbols (functions, classes, etc.) from packages installed in your active virtual environment using Workspace Symbol search (Cmd/Ctrl+T). This makes it easier to find definitions without leaving VS Code. The second is an experimental Rust-based parallel indexer that runs out-of-process, aiming to dramatically speed up completions, auto‑imports, and symbol search—especially on large projects. Both features are opt‑in to preserve default performance for all users.

How does the new symbol search for installed packages work?

When enabled via the Python › Analysis: Include Venv In Workspace Symbols setting, Pylance includes symbols from the site-packages directory of your current virtual environment in its workspace symbol index. This means you can quickly jump to function or class definitions inside third‑party libraries, even if they’re not part of your own workspace. For packages that lack a py.typed marker, only symbols exposed through __init__.py or __all__ are indexed, keeping results relevant. You can further control indexing depth per package with the Python › Analysis: Package Index Depths setting.

How do I enable the “Include Venv In Workspace Symbols” setting?

To activate this feature, open VS Code settings (Cmd+, on macOS or Ctrl+, on Windows/Linux). Search for Include Venv In Workspace Symbols. You’ll find the checkbox under the Python › Analysis section. Simply check it to turn on symbol indexing for your active virtual environment. After enabling, you may want to reload the window to ensure a fresh index. This setting is off by default to avoid any performance impact on projects where it isn’t needed.

What is the experimental Rust-based parallel indexer?

This new, optional indexer replaces Pylance’s existing indexing engine with a Rust‑based implementation that runs out‑of‑process and processes data in parallel. In internal tests, it shows an average 10× speed improvement on large Python projects, leading to faster completions after opening a workspace and a more responsive IntelliSense experience overall. Because it’s experimental, it is not enabled by default—the team wants to validate performance and reliability across diverse project configurations before making it the standard.

How can I try the Rust-based parallel indexer?

To test it, go to VS Code settings (Cmd+, / Ctrl+,) and search for Parallel Indexing. Check the box labeled Enable Parallel Indexing (Experimental) under the Python › Analysis section. Alternatively, you can add "python.analysis.enableParallelIndexing": true directly to your settings.json file. After enabling, reload VS Code (Cmd/Ctrl+Shift+P → Reload Window) to ensure the new indexer starts cleanly. This setting has the most noticeable effect on larger projects; small workspaces may see little difference.

What performance considerations should I be aware of?

The symbol search for installed packages is opt‑in because indexing many third‑party libraries can slow down the editor, especially on systems with limited resources. Use the Package Index Depths setting to fine‑tune how deeply Pylance inspects each package. Similarly, the parallel indexer is experimental and may not work perfectly in all environments. If you encounter slowdowns or unexpected behavior after enabling either feature, you can disable them at any time. The team welcomes feedback—if you try the parallel indexer, please share your experience via the VS Code repository or the Python extension issue tracker.