feat: instrument telemetry for dev command

[Hweinstock]

Description

Add telemetry instrumentation to the dev command for all execution paths (invoke, exec, server modes).

Schema changes:

• Expanded Action enum: 'exec'
• Added UiMode enum: 'browser' | 'terminal'
• Added 'agui' to Protocol enum
• Added ui_mode field to DevAttrs

Attributes emitted: action, ui_mode, has_stream, protocol, invoke_count

Note: because the dev server runs indefinitely, we emit telemetry eagerly before the server starts. The alternative is to refactor how runWebUI works to return a result that allows us to determine if the error was a cancellation (ctrl + c) which would count as a success, or a crash, which would count as a failure. However, this is a large refactor and is therefore left out of scope.

The effect is that dev success metrics with browser correspond to "was the user able to launch the browser" rather than, "did the browser launch successfully".

Related Issue

Closes #

Documentation PR

N/A

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Other (please describe):

Testing

How have you tested the change?

• I ran npm run test:unit and npm run test:integ
• I ran npm run typecheck
• I ran npm run lint
• If I modified src/assets/, I ran npm run test:update-snapshots and committed the updated snapshots

Additional testing:

• E2E verified with tarball: telemetry JSONL correctly emitted for both success and failure paths
• Verified user-facing error messages preserved ("Dev server not running on port X")

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[github-actions]

Package Tarball

aws-agentcore-0.13.1.tgz

How to install

npm install https://github.com/aws/agentcore-cli/releases/download/pr-1223-tarball/aws-agentcore-0.13.1.tgz

[agentcore-cli-automation]

Telemetry instrumentation looks good overall, but a few concerns about the SIGINT/exit handling for the long-running server modes that I think need to be resolved before merging. The invoke and exec paths look clean.

[agentcore-cli-automation]

All four serious issues raised in earlier review comments appear to be addressed in the latest commit (53c5d65):

1. --logs SIGINT cleanup race — resolve() is no longer called from the SIGINT handler at lines 407–411; the promise now only resolves via onExit, so SIGTERM (with the 2s SIGKILL fallback in DevServer.kill) has time to drive the child to exit before process.exit(0) runs.
2. Browser-mode SIGINT race — the duplicate process.once('SIGINT', ...) { resolve() } was removed (lines 472–493). The code now just awaits runBrowserMode and includes an inline comment explicitly calling out the limitation that normal shutdown telemetry requires a follow-up refactor of runWebUI. Reasonable interim state.
3. TUI process.exit(0) from onBack — explicit collector?.stop(); process.exit(0) was added after the wrapper at lines 463–464, giving the --no-browser path a deterministic exit point.
4. --logs failure path silently exits — replaced with if (!devResult.success) throw devResult.error; process.exit(0); at lines 416–417, matching the pattern used by the other paths and surfacing the error via the outer catch.

The integ test exercises both success and failure telemetry through a real audit dir (no mocking of telemetry internals), which is good. Schema additions (exec action, UiMode, agui protocol, ui_mode field) and unit tests look correct.

LGTM.

[notgitika]

LGTM