rememberStreamingHighlightedCode

Runs the debounce + syntax-highlight pipeline for streaming / real-time code (such as LLM responses or live terminal logs) and returns an AnnotatedString ready for rendering.

This composable is marked experimental (ExperimentalHighlightApi). Call sites must opt in with @OptIn(ExperimentalHighlightApi::class) or propagate the annotation.

Unlike rememberHighlightedCode, which cancels and resets to null on every string change, this function preserves syntax highlighting during active streaming using span transfer (applySnapshotSpans):

1. Immediate render (0 ms latency): As new tokens arrive, the full string is returned immediately.

2. Continuous styling: Spans from the last completed highlight run are carried forward onto the unchanged prefix of the new text, keeping existing lines colored.

3. Newline-aware progressive backfilling: When triggerOnNewline is enabled, completed lines are progressively highlighted in the background as newlines (\n) arrive (throttled by minThrottleMs), snapping finished lines to full syntax colors while subsequent tokens continue streaming.

4. Debounced engine calls: Idle pauses are debounced by debounceMs, coalescing rapid token emissions into a single highlight engine run when generation pauses or finishes.

Usage

HighlightThemeProvider(
    lightHighlightTheme = rememberTomorrowLightTheme(),
    darkHighlightTheme  = rememberAtomOneDarkTheme(),
) {
    val highlightedCode = rememberStreamingHighlightedCode(
        code     = streamingLlmText,
        language = "kotlin",
    )
    Text(text = highlightedCode)
}

Parameters