Remove unused VS Code telemetry and azext dependencies

[Shenglong Li (shenglol)]

Summary

Remove unused telemetry and broad utility dependencies from the Bicep VS Code extension.

• Replace azext authentication, subscription, UI, command, and logging helpers with focused local implementations.
• Replace shared action contexts with explicit error handling, prompt, command, and logging services.
• Remove the VS Code telemetry reporter, activation and command events, and language-server telemetry forwarding.
• Replace fs-extra and rimraf with Node's built-in filesystem APIs.
• Preserve user-facing logging and error handling.

Telemetry removed

The extension collected the following Bicep-specific telemetry. None of these events is monitored or used to make product or operational decisions.

Area	Data collected	Why it is not useful
Extension activation	Activation duration and generic error type on failure	A duration and error class do not identify an actionable failure, and there is no dashboard or alert consuming the data.
Command failures	Command ID and generic error type	Only failures are recorded, so there is no success-rate denominator. The generic error class lacks the context needed to diagnose a defect.
Language-server connection	Seconds since startup, generic error type, or closed reason	This reports coarse transport health without an error message, stack, or actionable failure context.
Decompile	Success or failure, output file count, conflicting file count, and failure type	These are feature counters without an active product question or decision tied to them.
Paste as Bicep	Paste context, paste type, source language, input JSON size, and output Bicep size	The size and context dimensions are not reviewed and do not indicate whether a product change is needed.
Insert resource	Resource type, API version, success or failure, and failure type	These usage and outcome counters have no owner or decision process, while actionable failures are reported with context through GitHub issues.
Import Kubernetes manifest	Success or failure and failure type	This is an unmonitored feature counter and does not provide enough diagnostic context to fix a failure.
Deployment	Succeeded or failed	A deployment result alone cannot distinguish extension defects from template, authorization, policy, or Azure service failures.
External source requests	Source availability, file extension, request type, archive file count, and failure type	These are internal protocol and archive implementation details rather than actionable user outcomes.
Unhandled exceptions	Event count with exception details removed	Without the exception message or stack, the event cannot be diagnosed. Including those details would introduce unnecessary privacy risk.

The language server still emits its existing telemetry notifications. This PR only removes collection and forwarding by the VS Code extension; language-server telemetry can be evaluated separately.

Dependency reduction

The production dependency footprint drops by 84 packages (58%):

• Production package entries: 146 -> 62
• Total lockfile package entries: 727 -> 678, so 49 packages are removed entirely
• Additional packages remain only as development dependencies; for example, @azure/identity remains through @vscode/vsce

Direct dependencies removed:

• @azure/identity
• @microsoft/vscode-azext-azureauth
• @microsoft/vscode-azext-azureutils
• @microsoft/vscode-azext-utils
• fs-extra
• @types/fs-extra
• rimraf

@azure/core-auth is added as a focused replacement, resulting in a net reduction of 6 direct dependencies.

The azext packages share much of their dependency graph, and several Azure SDK packages remain required by the ARM clients. This is why removing the direct dependencies does not remove independent copies of every transitive package.

Bundle size

Production webpack builds were compared by summing all root-level JavaScript assets loaded by the extension runtime. Separately copied deploy-pane and visual-designer assets are excluded.

Revision	Runtime JS files	Raw size	Gzip size
Before the Jest-to-Vitest migration (02a54d985)	16	2,411,767 bytes	534,569 bytes
Current main	16	2,412,621 bytes	534,768 bytes
This PR	5	917,343 bytes	211,070 bytes

Compared with current main, this PR reduces runtime JavaScript by 1,495,278 bytes (62.0%) raw and 323,698 bytes (60.5%) gzip. The primary extension.js file decreases from 1,364,040 bytes to 704,355 bytes (48.4%).

The pre-Vitest and current-main bundles are effectively the same size, confirming that the reduction comes from removing runtime dependencies and telemetry rather than from changing the test framework.

Validation

• npm run clean
• npm run test:unit - 66 tests passed
• npm run build
• npm run build:e2e
• npm run lint

[github-actions]

Test this change out locally with the following install scripts (Action run 32211670799)

VSCode

• Mac/Linux

  bash <(curl -Ls https://aka.ms/bicep/nightly-vsix.sh) --run-id 32211670799

• Windows

  iex "& { $(irm https://aka.ms/bicep/nightly-vsix.ps1) } -RunId 32211670799"

Azure CLI

• Mac/Linux

  bash <(curl -Ls https://aka.ms/bicep/nightly-cli.sh) --run-id 32211670799

• Windows

  iex "& { $(irm https://aka.ms/bicep/nightly-cli.ps1) } -RunId 32211670799"