Non-idiomatic force-unwraps inside useElevenLabs guard

input.apiKey! on the same line as the != nil check, and both input.apiKey! / input.voiceId! on line 457, are safe because useElevenLabs can only be true when both are non-nil — but the pattern is non-idiomatic Swift and slightly obscures the intent.

A guard let at the call site would be more conventional and removes the forced unwraps entirely:

And in the if useElevenLabs branch on line 456–457, you could either shadow-bind there or just rely on the already-safe ! given the guard. Either way, a small refactor would make the intent clearer for future readers.

This is a comment left during a code review.
Path: apps/macos/Sources/OpenClaw/TalkModeRuntime.swift
Line: 454-457

Comment:
**Non-idiomatic force-unwraps inside `useElevenLabs` guard**

`input.apiKey!` on the same line as the `!= nil` check, and both `input.apiKey!` / `input.voiceId!` on line 457, are safe because `useElevenLabs` can only be `true` when both are non-nil — but the pattern is non-idiomatic Swift and slightly obscures the intent.

A `guard let` at the call site would be more conventional and removes the forced unwraps entirely:

```suggestion
        let useElevenLabs = !(input.apiKey?.isEmpty ?? true) && input.voiceId != nil
```

And in the `if useElevenLabs` branch on line 456–457, you could either shadow-bind there or just rely on the already-safe `!` given the guard. Either way, a small refactor would make the intent clearer for future readers.

How can I resolve this? If you propose a fix, please make it concise.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

Addressed — the current revision uses if let apiKey = input.apiKey, !apiKey.isEmpty, let voiceId = input.voiceId to bind safely without force-unwraps. No useElevenLabs boolean or ! operators remain.

Remove unsupported timeout arg from system TTS call

TalkSystemSpeechSynthesizer.shared.speak is called with timeout: here, but the OpenClawKit API only defines speak(text:language:onStart:) (see apps/shared/OpenClawKit/Sources/OpenClawKit/TalkSystemSpeechSynthesizer.swift). In macOS builds this produces an extra argument 'timeout' compile error, so the talk-mode fix cannot ship until this call matches the current method signature (or the API is updated in the same commit).

Useful? React with 👍 / 👎.

Addressed — the timeout: parameter has been removed from the playSystemVoice call. The watchdog timeout is now handled entirely inside TalkSystemSpeechSynthesizer using language-aware per-character estimates. No compile error.

Preserve default system voice when no directive language is set

Passing voiceWakeLocaleID as language here changes system-TTS behavior, not just watchdog timing: replies without a lang directive no longer use the macOS default voice and instead force the recognition locale. In setups where wake-word recognition locale differs from preferred speech voice, this makes Talk mode speak in the wrong language/voice for every system-provider reply. This regression comes from using locale fallback in the synthesis language: argument rather than only for timeout estimation.

Useful? React with 👍 / 👎.

Make watchdog timeout tests actor-safe

TalkSystemSpeechSynthesizer is @MainActor, so watchdogTimeoutSeconds is also main-actor isolated, but this synchronous XCTest method calls it from a nonisolated context. With Swift 6 strict concurrency enabled for OpenClawKitTests, this produces an actor-isolation compile error (call to main actor-isolated static method ... in a synchronous nonisolated context), so the new test target can fail to build until the call site is @MainActor (or the helper is marked nonisolated).

Useful? React with 👍 / 👎.

Keep system voice fallback from forcing a locale voice

playSystemVoice now always supplies input.language ?? AppStateStore.shared.voiceWakeLocaleID to TalkSystemSpeechSynthesizer.speak, and speak sets utterance.voice whenever language is non-nil. In the common case where no lang directive is present, this changes behavior from “use macOS default voice” to “force wake-locale voice”, so users with different recognition and preferred TTS locales will hear the wrong voice/language on every system-voice reply; the locale fallback should be used for timeout estimation only, not passed as synthesis language.

Useful? React with 👍 / 👎.

Mark watchdog timeout tests as main-actor isolated

TalkSystemSpeechSynthesizer is @MainActor, so watchdogTimeoutSeconds is main-actor isolated; these synchronous XCTest methods call it directly from a nonisolated context. With StrictConcurrency enabled in apps/shared/OpenClawKit/Package.swift, this becomes a Swift 6 actor-isolation compile error and can fail the OpenClawKitTests target build until the tests (or helper API) are explicitly main-actor safe.

Useful? React with 👍 / 👎.