IntelliCode Completions

Generates up-to-one-line code predictions that appear as non-intrusive grey-text inline suggestions to the right of the cursor as the user types. The completion engine analyzes the current file context (cursor position, surrounding code tokens, language syntax) and triggers automatically without explicit user action. Predictions are rendered inline rather than in a popup menu, minimizing visual disruption while maintaining discoverability through standard Tab/ESC acceptance keybindings.

Provides granular configuration to enable or disable inline completion predictions on a per-language basis (Python, JavaScript, TypeScript) while preserving other IntelliCode features like IntelliSense ranking. Configuration is stored in VS Code Settings and discoverable via extension-specific settings search. Allows developers to use AI completions selectively — e.g., enable for Python but disable for TypeScript — without uninstalling the extension or affecting IntelliSense functionality.

Processes source code entirely on the developer's machine without transmitting code content to external servers. The extension explicitly guarantees that 'Your code does not leave your machine and is not used to train our model,' implying a pre-trained model architecture that performs inference locally or via a privacy-preserving remote endpoint that does not log or retain code. This design choice prioritizes data security for enterprises and developers working with proprietary or sensitive codebases.

Coordinates inline completion predictions with VS Code's native IntelliSense popup menu to prevent suggestion conflicts and enable sequential acceptance. When IntelliSense is open, the first Tab keypress accepts the token selected in the IntelliSense list, and the second Tab keypress accepts the remaining inline completion. This coordination pattern ensures that inline completions augment rather than compete with IntelliSense, creating a unified suggestion workflow that respects the user's existing IntelliSense muscle memory.

Generates and displays code predictions automatically as the user types, without requiring explicit trigger actions (e.g., Ctrl+Space or menu navigation). The prediction engine monitors keystroke events and cursor position changes, analyzes the current code context in real-time, and renders suggestions inline when confidence thresholds are met. This automatic trigger pattern minimizes friction in the coding workflow by eliminating the need for users to consciously request completions.

Provides AI-driven code completion predictions optimized for three specific programming languages: Python, JavaScript, and TypeScript. The underlying model(s) are pre-trained on code in these languages and tuned to understand language-specific syntax, idioms, and common patterns. Inference is performed per-language with language detection based on file extension or explicit language mode in VS Code, enabling language-appropriate suggestions that respect each language's conventions and standard libraries.

Collects usage telemetry and analytics data about IntelliCode Completions usage patterns (e.g., suggestion acceptance rates, language distribution, feature usage) and transmits this metadata to Microsoft servers. Telemetry collection respects VS Code's global `telemetry.enableTelemetry` setting, allowing users to disable all telemetry collection across VS Code and its extensions via a single configuration option. Specific telemetry fields and data retention policies are not documented.