merge: tail-drain and chunked ASR hardening into main

Resolve FlowContinuousCapture by combining tail-drain gate/drain logic
with flow-abcd route-change and interruption recovery.
This commit is contained in:
Rocky
2026-07-06 18:21:34 +08:00
14 changed files with 696 additions and 53 deletions
@@ -0,0 +1,130 @@
// FlowCaptureTailDrain.swift
// OSGKeyboard · Shared
//
// Tail-drain policy and silence tracking for utterance end. After the user
// stops recording, capture keeps forwarding PCM until trailing speech drains
// or a safety timeout elapses (symmetric to pre-roll at utterance start).
import Foundation
/// Tunable tail-drain policy shared by Flow capture and preview dictation.
public struct FlowCaptureTailDrainPolicy: Sendable, Equatable {
/// RMS below this counts as silence while draining (16 kHz mono Float32).
public let silenceRMSThreshold: Float
/// Finish drain after this much continuous silence.
public let silenceDurationSeconds: TimeInterval
/// Hard cap so noisy environments cannot stall finalize forever.
public let maxDrainSeconds: TimeInterval
public init(
silenceRMSThreshold: Float,
silenceDurationSeconds: TimeInterval,
maxDrainSeconds: TimeInterval
) {
self.silenceRMSThreshold = silenceRMSThreshold
self.silenceDurationSeconds = silenceDurationSeconds
self.maxDrainSeconds = maxDrainSeconds
}
public static let flowDefault = FlowCaptureTailDrainPolicy(
silenceRMSThreshold: 0.015,
silenceDurationSeconds: 0.25,
maxDrainSeconds: 1.5
)
}
/// Metrics emitted when tail drain completes (for diagnostics and tests).
public struct FlowCaptureDrainReport: Sendable, Equatable {
public let drainDurationSeconds: Double
public let endedBySilence: Bool
public let tailSampleCount: Int
public init(
drainDurationSeconds: Double,
endedBySilence: Bool,
tailSampleCount: Int
) {
self.drainDurationSeconds = drainDurationSeconds
self.endedBySilence = endedBySilence
self.tailSampleCount = tailSampleCount
}
public static let skipped = FlowCaptureDrainReport(
drainDurationSeconds: 0,
endedBySilence: false,
tailSampleCount: 0
)
}
/// Thread-safe silence tracker used while draining trailing audio.
public final class FlowCaptureDrainTracker: @unchecked Sendable {
private let lock = OSAllocatedUnfairLock()
private var drainStartedAt: TimeInterval?
private var lastAudibleAt: TimeInterval?
public init() {}
public func reset() {
lock.withLock {
drainStartedAt = nil
lastAudibleAt = nil
}
}
public func beginDrain(now: TimeInterval = Date().timeIntervalSince1970) {
lock.withLock {
drainStartedAt = now
lastAudibleAt = now
}
}
public func noteAudio(
samples: [Float],
policy: FlowCaptureTailDrainPolicy,
now: TimeInterval = Date().timeIntervalSince1970
) {
guard !samples.isEmpty else { return }
let rms = Self.rms(of: samples)
lock.withLock {
guard drainStartedAt != nil else { return }
if rms >= policy.silenceRMSThreshold {
lastAudibleAt = now
}
}
}
public func shouldFinish(
policy: FlowCaptureTailDrainPolicy,
now: TimeInterval = Date().timeIntervalSince1970
) -> (finished: Bool, endedBySilence: Bool) {
lock.withLock {
guard let started = drainStartedAt else {
return (true, false)
}
let audible = lastAudibleAt ?? started
if now - started >= policy.maxDrainSeconds {
return (true, false)
}
if now - audible >= policy.silenceDurationSeconds {
return (true, true)
}
return (false, false)
}
}
public func elapsedSeconds(now: TimeInterval = Date().timeIntervalSince1970) -> Double {
lock.withLock {
guard let started = drainStartedAt else { return 0 }
return max(0, now - started)
}
}
public static func rms(of samples: [Float]) -> Float {
guard !samples.isEmpty else { return 0 }
var sum: Float = 0
for sample in samples {
sum += sample * sample
}
return sqrtf(sum / Float(samples.count))
}
}
@@ -0,0 +1,34 @@
// FlowPipelineDiagnostics.swift
// OSGKeyboard · Shared
//
// Structured Flow pipeline metrics for Console.app filtering.
import Foundation
import os
public enum FlowPipelineDiagnostics {
public static func logDrain(_ report: FlowCaptureDrainReport) {
OSGLog.flow.info(
"tailDrain duration=\(report.drainDurationSeconds, format: .fixed(precision: 2))s " +
"silenceEnd=\(report.endedBySilence) tailSamples=\(report.tailSampleCount)"
)
}
public static func logChunkFinalize(
chunkCount: Int,
lastChunkSamples: Int,
stitchedLength: Int,
chunkWarnings: Int
) {
OSGLog.flow.info(
"chunkPipeline chunks=\(chunkCount) lastChunkSamples=\(lastChunkSamples) " +
"stitchedLen=\(stitchedLength) warnings=\(chunkWarnings)"
)
}
public static func logStitcherSafeFallback(naiveLength: Int, mergedLength: Int) {
OSGLog.asr.warning(
"stitcher safe fallback naive=\(naiveLength) merged=\(mergedLength)"
)
}
}
@@ -21,6 +21,12 @@ public struct UtteranceTranscriptStitcher: Sendable {
}
}
/// Drop the highest-index segment (used when re-transcribing a merged tail chunk).
public mutating func removeLastSegment() {
guard !segments.isEmpty else { return }
segments.removeLast()
}
public func composed() -> String {
guard let first = segments.first else { return "" }
var result = first.text
@@ -30,6 +36,21 @@ public struct UtteranceTranscriptStitcher: Sendable {
return result
}
/// Prefer overlap-aware merge, but fall back to naive join when dedup would drop real content.
public func composedSafely() -> String {
let merged = composed()
guard segments.count >= 2 else { return merged }
let naive = segments.map(\.text).joined(separator: " ")
if merged.count + 16 < naive.count {
FlowPipelineDiagnostics.logStitcherSafeFallback(
naiveLength: naive.count,
mergedLength: merged.count
)
return naive
}
return merged
}
/// Merge `next` onto `previous`, dropping duplicated suffix/prefix overlap.
public static func mergeWithOverlap(previous: String, next: String) -> String {
let trimmedNext = next.trimmingCharacters(in: .whitespacesAndNewlines)