Pylance

Pylance provides IntelliSense completions by analyzing Python type hints (PEP 484/526) and performing static type inference across the entire workspace using Pyright's type inference engine. Completions are ranked by type compatibility and semantic relevance, with support for stub files (.pyi) and installed package introspection. The completion engine indexes workspace symbols and resolves imports to provide context-aware suggestions without executing code.

Pylance continuously analyzes Python code as you type, using Pyright's static type checker to identify type mismatches, undefined names, missing imports, and other errors. Diagnostics are reported in-line with red squiggles and appear in the Problems panel, with configurable severity levels (error/warning/information). The type checker respects Python's type system (PEP 484, PEP 586, PEP 589) and supports gradual typing, allowing mixed typed and untyped code in the same project.

Extends Pylance's analysis capabilities to Jupyter Notebooks in VS Code, providing type checking, code completion, and diagnostics for notebook cells. The engine treats each cell as a separate Python scope while maintaining context from previously executed cells, enabling accurate analysis of notebook code.

Supports VS Code multi-root workspaces where multiple folders are open simultaneously, with per-folder Python environment and configuration settings. The engine maintains separate symbol tables and analysis contexts for each folder, enabling accurate analysis of projects with different Python versions, dependencies, or configurations.

Pylance automatically generates and manages import statements by analyzing symbol usage and resolving them against the workspace and installed packages. When you use an undefined symbol, Pylance suggests adding the import; it can also remove unused imports and organize import statements. The auto-import engine resolves symbols using the Python import system (sys.path, PYTHONPATH, virtual environments) and respects __init__.py files and package structures.

Pylance provides code navigation capabilities including go-to-definition, find-all-references, and symbol outline/tree view. These features work by analyzing the workspace's symbol table (built from type inference and AST analysis) and resolving symbol references across files. Go-to-definition jumps to the source of a symbol (function, class, variable), find-references locates all usages, and the outline view displays the hierarchical structure of symbols in the current file.

Pylance provides semantic highlighting that colors code based on type information and semantic analysis, not just syntax rules. Variables, functions, classes, and other symbols are colored according to their semantic role (e.g., type parameters in a different color than variables). This highlighting is computed by analyzing the type graph and symbol table, enabling more nuanced and informative code visualization than traditional syntax highlighting.

Pylance offers three language server modes ('light', 'default', 'full') that trade off feature breadth against performance and resource usage. The 'light' mode disables some features to minimize overhead, 'default' provides a balanced set of features, and 'full' enables all features including advanced type checking. The mode is configured via the `python.analysis.languageServerMode` setting in workspace settings.json or VS Code Settings UI, allowing teams to tune Pylance's behavior for their hardware and project size.

Pylance extends its type checking and completion capabilities to Jupyter notebook cells, analyzing cell code in context and providing IntelliSense, diagnostics, and navigation within notebooks. The language server treats notebook cells as a continuous Python context, tracking variable definitions across cells and providing completion based on previously executed cells. This integration works with VS Code's built-in notebook editor.

Pylance supports VS Code's multi-root workspace feature, allowing analysis of multiple project folders in a single window. For each root, Pylance respects the configured Python interpreter and virtual environment, enabling accurate import resolution and type checking across different projects with different dependencies. The language server automatically detects and uses virtual environments (venv, conda, etc.) configured in the workspace.

Pylance uses Python type stub files (.pyi) to provide type information for packages without inline type hints, including third-party packages and C extensions. The language server searches for stubs in the workspace, installed packages (via typeshed), and the Python environment, enabling type checking and completion for untyped libraries. This allows Pylance to provide IDE support for popular packages like NumPy, Django, and others that may not have inline type hints.

Automatically discovers Python modules and packages in the workspace and installed environments, and resolves symbol references across multiple files and modules. The engine builds a workspace-wide symbol table that tracks all definitions and their locations, enabling accurate code completion, navigation, and refactoring across the entire project.