Releases · microsoft/vscode-azuretools

Standardizes what it means to "skipExeucte" for AzureWizard, via adding an option to AzureWizard for skipExecute. When set to true this will:
- Remove all executeSteps and replaces them with an execute step which simply throws a UserCancelledError
- Replaces any ExecuteActivityContext properties from the context with ones that avoid outputting Azure activities
Have runQuickPickWizard respect showLoadingPrompt.
Add a SimpleCommandConfig which makes it easy to simply expose a command without have to make any wizard changes. Good for when just getting started and trying out exposing a command.
Change names of some types (technically breaking, but only agent extension has used these types so far in any "merged" capacity, do not need to move to a new major version number for utils).

Added various types and classes to use when integrating functionality into the Azure agent extension. This includes:

Integration

ExtensionAgentMetadata: the metadata an extension adds to its package.json which provides integration information for the agent extension.
WizardBasedCommandConfig: the config object extensions use to expose existing wizard command based functionality to the agent extension.
AgentBenchmarkConfig: the config object extensions use to expose benchmarks (similar to tests) to the agent extension.

IAzureAgentInput: the azure input class the the agent extension gives to service extensions when running command wizards in the background.
AzureUserInputQueue: the queue of azure inputs that the agent extensions gives to service extensions when a user chooses to run a command with the inputs the agent has determined.
AzExtUserInputWithInputQueue: the azure input class that service extensions uses when needing to use inputs from a AzureUserInputQueue.

AgentBenchmarkConfig: the config object extensions use to expose benchmarks (similar to tests) to the agent extension. Benchmarks define how the agent should respond given a certain prompt. Benchmarks can be run from the chat UX only. In the future, we will hopefully be able to run benchmarks from CICD. Benchmarks are not called tests because benchmarks are not expected to pass 100% of the time.