DVC (deprecated)

Captures and organizes ML experiment runs (parameters, metrics, outputs) as Git commits, enabling version control of experiments alongside code. The extension reads DVC metadata files (.dvc, dvc.yaml) and Git commit history to reconstruct experiment lineage, displaying experiments in a hierarchical tree view within VS Code's Activity Bar. Each experiment is tied to a specific Git commit, allowing reproducibility by checking out historical commits.

Versions large files and datasets (outside Git's practical limits) by storing them in DVC's local cache and syncing to remote storage backends (S3, Azure Blob, GCS, NFS). The extension displays tracked data files in the Explorer View with version status indicators, allowing developers to pull/push specific datasets without cloning entire repositories. DVC uses content-addressable storage (file hashes) to deduplicate data across experiments and versions.

Enables one-click checkout of historical experiments by switching to the corresponding Git commit and pulling the associated data versions. The extension reads the Git commit hash from the selected experiment and executes git checkout followed by dvc pull, restoring both code and data to the experiment's state. This allows developers to reproduce results or inspect experiment artifacts without manual command execution.

Renders interactive dashboards within VS Code displaying experiment metrics (loss, accuracy, F1 score) and custom plots (training curves, confusion matrices) side-by-side for comparison. The extension parses metrics from JSON/CSV files logged during training and overlays them on a configurable grid layout. Plots are updated in real-time as training runs progress, with support for filtering by experiment branch or commit.

Monitors metric files (JSON, CSV) in real-time as training scripts write to them, updating the metrics dashboard in VS Code without requiring manual refresh. The extension watches the file system for changes to configured metric files and re-renders plots within 1-5 seconds of new data being written. This enables developers to observe training progress live without switching to terminal or external monitoring tools.

Adds a 'DVC' panel to VS Code's Source Control View showing the current state of tracked files and datasets (cached, remote, missing, modified). The extension reads DVC metadata and compares file hashes against the local cache and remote storage, displaying status indicators and file paths. This integrates DVC status alongside Git status, allowing developers to see both code and data versioning in one place.

Registers DVC commands in VS Code's Command Palette (accessible via Ctrl+Shift+P), allowing developers to execute DVC operations (dvc pull, dvc push, dvc repro, dvc dag) without opening a terminal. Commands are context-aware, operating on the current workspace or selected files. The extension translates user selections in the UI into corresponding DVC CLI invocations, capturing output and displaying results in the DVC output channel.

Displays a hierarchical tree of DVC-tracked files and directories in VS Code's Explorer View, showing version status (cached, remote, missing) and file sizes. The extension reads .dvc and dvc.yaml files to populate the tree, allowing developers to navigate tracked data without using the terminal. Right-click context menus provide quick access to pull/push operations for individual files or directories.

Enables side-by-side comparison of experiments by displaying metrics and hyperparameters in a table format, with support for sorting and filtering by metric values or parameter ranges. The extension extracts parameters from dvc.yaml and metrics from dvc.lock or metric files, aligning them by experiment (Git commit). Developers can select multiple experiments and view their differences highlighted in the comparison table.

Renders the DVC pipeline dependency graph (dvc dag) as a visual diagram within VS Code, showing data sources, processing stages, and outputs. The extension parses dvc.yaml to extract stage definitions and their dependencies, rendering them as a directed acyclic graph (DAG) with clickable nodes. Developers can click nodes to navigate to the corresponding stage definition in dvc.yaml.

Provides UI for configuring DVC remote storage backends (S3, Azure Blob, GCS, NFS) through VS Code settings or a configuration wizard. The extension stores remote credentials securely using VS Code's secret storage API and validates connectivity to configured remotes. Developers can switch between remotes and view remote storage status without editing configuration files manually.