v0.1.3

Changelog

All notable changes to the Databricks Bundle Inspector extension are documented in this file.

[0.1.3] - 2026-05-27

Fixed

• Parameter precedence corrected: job parameters now correctly take precedence over task base_parameters when computing effective parameter values in the graph, matching Databricks runtime behaviour. Previously, task base_parameters were incorrectly overriding job parameters with the same key. This fixes this issue

[0.1.2] - skipped

Tag v0.1.2 was tombstoned by GitHub after being created under a tag immutability rule and subsequently deleted. GitHub permanently prevents recreation of tag names that were deleted under such a rule, making v0.1.2 unpublishable. Skipped in favour of v0.1.3.

0.1.1 - 2026-05-04

Changed

• Simplified README to focus on structural DAG visualization and removed development/testing documentation.
• Disabled target selection dropdown in the panel. The extension now displays "structural preview" only, removing the ability to switch between targets. This aligns with the focus on structural inspection.

0.1.0 - 2026-05-04

Initial public release. The extension is a read-only inspector for Declarative Automation Bundles (previously known as Databricks Asset Bundles). It runs databricks bundle validate --output json, renders the resolved bundle as an interactive graph, and surfaces bundle issues in VS Code's native Problems panel. Everything operates on the CLI-resolved bundle. The extension does not modify YAML and does not call Databricks workspace APIs.

Added

Bundle inspection

• New command Inspect Databricks Bundle (databricksBundleInspector.inspectBundle). Opens an interactive graph of the resolved bundle in a VS Code webview.
• React Flow graph view with pan, zoom, search, header chip panel, node legend, and node selection. Built on @xyflow/react.
• First-class extraction for jobs, tasks, pipelines, and depends_on edges.
• Support for 17+ Databricks task types, including notebook_task, sql_task, spark_python_task, python_wheel_task, dbt_task, dbt_platform_task, for_each_task, condition_task, run_job_task, dashboard_task, pipeline_task, power_bi_task, clean_rooms_notebook_task, sql_alert_task, spark_jar_task, spark_submit_task, plus job_cluster_key and existing_cluster_id compute attachments.
• Click-to-open from graph nodes opens the source file in the editor at the correct line and column. Notebooks (.ipynb) open at the top via the Jupyter editor.

Issue detection and Problems panel integration

• New command Open Bundle Issues (databricksBundleInspector.openBundleIssues). Reveals the inspector with the issues panel focused.
• Six typed inspector issue kinds:
  • missing_file (error): local file references that do not exist on disk.
  • missing_library (error): local library artifacts that do not exist.
  • unresolved_variable (error): variable references not defined in the bundle.
  • unknown_or_deprecated_field (warning): fields the CLI flagged as unknown or deprecated.
  • unknown_task_type (warning): tasks the inspector does not recognize.
  • validation_diagnostic (severity from CLI): pass-through of CLI bundle diagnostics with file/line/column.
• Each issue carries severity, kind, title, detail, taskId, taskName, yamlPath, fixHint, and resolved file/line/column.
• Inspector issues are emitted to VS Code's Problems panel under the source label Databricks Bundle Inspector (<bundle name>) so they appear alongside other diagnostics.
• CLI bundle diagnostics are emitted under a separate source Databricks Bundle (<bundle name>).
• On-save diagnostics: after a bundle has been inspected, saving the bundle file, an included YAML file, or a tracked referenced source file re-runs validation for the owning bundle and clears stale diagnostics for files that are now clean.

Databricks CLI integration

• New configuration setting databricksBundleInspector.cliPath. When empty, the extension uses databricks from the system PATH.
• Probe target fallback: validation runs against a synthetic target (__bundle_inspector_probe__) by default so the CLI produces resolved bundle JSON without requiring workspace authentication.
• Target fallback path: when a user-requested target fails, the extension automatically falls back to the probe target and surfaces a warning so structural inspection still works.
• Auth-error recovery: if the CLI emits valid JSON on stdout but fails with cannot configure default credentials, the bundle is still parsed and surfaced with an AUTH_NOT_CONFIGURED issue.
• Structured error taxonomy:
  • CLI_NOT_FOUND: Databricks CLI could not be located.
  • CLI_NOT_EXECUTABLE: CLI was found but failed to execute.
  • VALIDATION_TIMEOUT: validation exceeded the 30 second timeout.
  • INVALID_BUNDLE_SHAPE: CLI returned JSON that did not match the expected schema.
  • VALIDATION_FAILED: validation completed with errors.
  • AUTH_NOT_CONFIGURED: parsed bundle returned despite missing credentials.
  • BUNDLE_DIAGNOSTICS: CLI reported diagnostics on stderr.
  • CLI_WARNING: validation completed with warnings.

Source-file enrichment

• Local task source files are read after graph extraction and used to enrich nodes with detected:
  • Secret scope references via dbutils.secrets.get(...) in Python and SQL, case-insensitive.
  • Widgets.
  • F-string expressions inside Python source.
• Enrichment is read-only and best-effort; unreadable files do not fail the inspector.

Editor surface

• Title bar and editor context menu entries for Inspect Databricks Bundle, gated to databricks.yml and databricks.yaml.
• Inspect Databricks Bundle and Open Bundle Issues are available from the Command Palette.

Engineering

• Zod-based runtime schema validation of CLI output (ParsedBundleConfigSchema). The cast from JSON.parse to ParsedBundleConfig is verified at runtime, not just at compile time.
• Content Security Policy on the webview with a cryptographically generated nonce per render. connect-src is set to 'none'. Webview resources are loaded through webview.asWebviewUri.
• 30 second timeout on every CLI invocation.
• Diagnostic collection lifecycle is managed: stale entries are cleared per file when validation no longer reports them.

Tooling

• Node 22 dev container (.devcontainer/) with the Databricks CLI installed inside the container and databricksBundleInspector.cliPath preset.
• Husky pre-commit and pre-push hooks. Lint-staged runs ESLint with autofix on staged TypeScript files.
• GitHub Actions CI workflow.
• Trufflehog secret scanning in CI.
• vsce-ready manifest with display name, categories, keywords, repository, homepage, bugs, and Q&A links.

Tests

• Unit tests for bundleGraph, issues, jobDocumentation, documentationPolicy, documentationSignals, parseBundleDiagnostics, semanticGraph, sourceLocations, taskFileDetections, taskNodeData, validateBundle, bundleContext, parsing, processRunner, and jobSelection.
• Integration tests for validateBundle against a real CLI binary (semanticCli.integration.test.ts, verifyCliPath.integration.test.ts).
• Golden semantic graph baselines for three fixture bundles (broken-job, multi-job-dag, secret-scope-example). Each fixture commits a validated-bundle.json and a validated-bundle.meta.json carrying CLI version, generation timestamp, target, and SHA-256 provenance of the JSON payload.
• Live CLI compatibility matrix runner (scripts/run-semantic-cli-matrix.mjs) and a script to install pinned Databricks CLI releases from GitHub for local matrix testing (scripts/install-databricks-cli-matrix.mjs).
• Compatibility floor for v0.1.0: Databricks CLI v0.270.1+, scoped to the fixtures covered by the committed matrix.

Known limitations

• Paths under /Workspace/..., /Repos/..., /Volumes/..., dbfs:/..., and cloud URIs (s3://, abfss://, gs://) cannot be validated locally. They are accepted as resolved references; only local paths are checked for existence.
• The extension does not call Databricks workspace APIs and does not detect drift between the local bundle and a deployed workspace. This is intentional.
• The extension does not edit databricks.yml or any included YAML. For graphical authoring, see complementary tools that operate on the YAML directly.
• The webview operates on one bundle at a time. Multi-bundle workspaces can be inspected by opening each bundle file and running the inspector.

Compatibility

• VS Code ^1.85.0.
• Node 22 (development).
• Databricks CLI v0.270.1 or newer.