Miscellaneous minor cleanup and add event log support

[fpfp100]

1. Logger set up md updated based on feedback and cleaned up content following the LangChain PR.
2. Add event log support in ILogger and also log events for exporter.

[Copilot]

Pull request overview

Minor cleanup across observability core + LangChain extension, plus documentation updates for logger setup.

Changes:

• Fix minor formatting/indentation issue in SpanProcessor attribute setting.
• Update LOGGER_SETUP.md examples to import ILogger, setLogger, and resetLogger from the public package entrypoint.
• Harden LangChain extension utilities with safer optional chaining and an empty-input guard; align an error message with the actual instrument() API name.

Reviewed changes

Copilot reviewed 3 out of 4 changed files in this pull request and generated 1 comment.

File	Description
packages/agents-a365-observability/src/tracing/processors/SpanProcessor.ts	Minor cleanup to correctly align/format attribute setting for operation source.
packages/agents-a365-observability/LOGGER_SETUP.md	Documentation updates to use public exports instead of internal source paths.
packages/agents-a365-observability-extensions-langchain/src/Utils.ts	Safer attribute extraction (optional chaining) and guard against empty messages in inference preprocessing.
packages/agents-a365-observability-extensions-langchain/src/LangChainTraceInstrumentor.ts	Error message corrected to reference instrument() (actual API).

[fpfp100]

@copilot fix test failure FAIL tests/observability/core/custom-logger.test.ts

[Copilot]

@fpfp100 I've opened a new pull request, #186, to work on those changes. Once the pull request is ready, I'll request review from you.

[Copilot]

Pull request overview

Copilot reviewed 8 out of 9 changed files in this pull request and generated 7 comments.

Comments suppressed due to low confidence (1)

packages/agents-a365-observability/LOGGER_SETUP.md:244

• The custom logger examples are missing the event method, which is now a required part of the ILogger interface. These examples will fail at runtime because they don't implement the complete interface. Add the event method to both logger examples.

if (process.env.NODE_ENV === 'production') {
  setLogger({
    info: (msg, ...args) => console.log(`[Prod-INFO] ${msg}`, ...args),
    warn: (msg, ...args) => console.warn(`[Prod-WARN] ${msg}`, ...args),
    error: (msg, ...args) => console.error(`[Prod-ERROR] ${msg}`, ...args)
  });
}

// Development: selective logging (warn and error only)
if (process.env.NODE_ENV === 'development') {
  setLogger({
    info: () => {},
    warn: (msg, ...args) => console.warn(`[Dev-WARN] ${msg}`, ...args),
    error: (msg, ...args) => console.error(`[Dev-ERROR] ${msg}`, ...args)
  });
}

[Copilot]

Pull request overview

Copilot reviewed 9 out of 10 changed files in this pull request and generated 7 comments.

[Copilot]

Pull request overview

Copilot reviewed 9 out of 10 changed files in this pull request and generated 5 comments.

Comments suppressed due to low confidence (1)

packages/agents-a365-observability-extensions-langchain/src/Utils.ts:70

• preprocess can become messages[0] when scope is "inference". Since messages is already validated as an array, messages[0] is typically a single message object (not an array), which makes preprocess?.map(...) undefined and then .filter(...) will throw at runtime. Consider keeping preprocess as an array (e.g., wrap the first element in an array) or branching so .map/.filter only run on arrays.

  const preprocess = getScopeType(run) === "inference" && messages.length > 0 ? messages[0] : messages;
  const processed = preprocess?.map((msg: Record<string, unknown>) => {
      const content = extractMessageContent(msg);
      if (!content) return null;

[fpfp100]

@copilot resolve merge conflict

[Copilot]

@fpfp100 I've opened a new pull request, #188, to work on those changes. Once the pull request is ready, I'll request review from you.

[Copilot]

Pull request overview

Copilot reviewed 91 out of 94 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (5)

packages/agents-a365-runtime/src/power-platform-api-discovery.ts:1

• Changing ClusterCategory from a string-literal union to a string enum is a breaking TypeScript change for consumers who previously passed string literals (e.g. 'prod') into APIs typed as ClusterCategory (notably PowerPlatformApiDiscovery). To preserve backward compatibility, consider either (a) accepting ClusterCategory | string in public APIs and validating/normalizing at runtime, or (b) exporting a separate type ClusterCategoryString = ${ClusterCategory}`` and using that in public API surfaces where strings were previously accepted.
  packages/agents-a365-observability/src/utils/logging.ts:1
• Adding a new required ILogger.event(...) method is a compile-time breaking change for any existing consumers providing custom loggers (even if they don’t care about events). To keep backward compatibility, consider making event optional on ILogger (e.g., event?: (...) => void) and treating missing event as a no-op, or keep ILogger unchanged and introduce a separate IEventLogger (or adapter wrapper) while still allowing older ILogger implementations in setLogger.
  packages/agents-a365-runtime/src/utility.ts:1
• jwt.JwtPayload claim values can be non-strings (commonly string | string[] | undefined). The current implementation can return a non-string truthy value (e.g., a string[]), which violates the function’s return type and could lead to invalid header serialization. A concrete fix is to check typeof value === 'string' (and optionally value.trim().length > 0) for each claim before returning, otherwise fall through to ''.
  packages/agents-a365-observability/src/utils/logging.ts:1
• getEnabledLogLevels() re-parses the log-level string on every info/warn/error/event call, allocating a new Set each time. If logging is frequent, this can add noticeable overhead. Consider caching the parsed Set keyed by the last seen observabilityLogLevel string (e.g., store { lastLevel: string; parsed: Set<number> } on the logger instance), recomputing only when the level string changes. Also, avoid magic numbers in event() by using LOG_LEVELS.info / LOG_LEVELS.error.
  packages/agents-a365-tooling/src/McpToolServerConfigurationService.ts:1
• The thrown error is generic and loses key debugging context. Consider including actionable details like the authHandlerName and whether scopes were configured (without logging the token). For example, mention the handler name and scopes count/value (scope strings are typically non-secret) and/or suggest checking configuration (mcpPlatformAuthenticationScope) to reduce support time.