@livx.cc/agentx 0.96.15 → 0.96.17

package/dist/native/mic-aec.swift

@@ -34,9 +34,16 @@ FileHandle.standardError.write("mic access granted: \(granted)\n".data(using: .u
 
 let engine = AVAudioEngine()
 let input = engine.inputNode
-do { try input.setVoiceProcessingEnabled(true) } catch {
-    FileHandle.standardError.write("AEC unavailable: \(error)\n".data(using: .utf8)!)
-    exit(1)
+// VPIO gives hardware echo cancellation, but on some Macs (observed: macOS 26.5.x + a built-in mic that
+// VPIO exposes as a multichannel format) the VP input tap NEVER delivers buffers — capture is dead. The
+// host detects that (no audio) and respawns us with MIC_NO_VPIO=1: plain capture, no AEC (heuristic-echo
+// tier, gated below). Skipping VPIO trades echo cancellation for a mic that actually works.
+let useVPIO = ProcessInfo.processInfo.environment["MIC_NO_VPIO"] != "1"
+if useVPIO {
+    do { try input.setVoiceProcessingEnabled(true) } catch {
+        FileHandle.standardError.write("AEC unavailable: \(error)\n".data(using: .utf8)!)
+        exit(1)
+    }
 }
 // the VP I/O unit is full-duplex: input stays muted unless the output side renders too. The mixer→
 // output format must match the VP output's hardware format (nil/default fails kAUInitialize -10875).
@@ -44,9 +51,10 @@ let outHw = engine.outputNode.outputFormat(forBus: 0)
 engine.connect(engine.mainMixerNode, to: engine.outputNode, format: outHw)
 engine.mainMixerNode.outputVolume = 1 // playback is audible now (was 0 when this path was only a keep-alive)
 // VP ducks all other system audio by default — disable (we ARE the audio).
-if #available(macOS 14.0, *) {
+if useVPIO, #available(macOS 14.0, *) {
     input.voiceProcessingOtherAudioDuckingConfiguration = .init(enableAdvancedDucking: false, duckingLevel: .min)
 }
+FileHandle.standardError.write("mode: \(useVPIO ? "vpio" : "no-vpio")\n".data(using: .utf8)!)
 
 // ---- playback: framed s16le 44.1k mono on stdin → player node (the VPIO reference signal) ----
 let playFmt = AVAudioFormat(commonFormat: .pcmFormatFloat32, sampleRate: 44100, channels: 1, interleaved: false)!
@@ -92,6 +100,14 @@ func playbackEnvNow() -> Float {
 let inFmt = input.outputFormat(forBus: 0)
 let decim = max(1, Int(inFmt.sampleRate / 16000))
 let out = FileHandle.standardOutput
+FileHandle.standardError.write("input format: \(inFmt.sampleRate)Hz \(inFmt.channelCount)ch decim=\(decim)\n".data(using: .utf8)!)
+
+// Diagnostic (stderr → surfaced only under DEBUG=VoiceIO): log from INSIDE the tap, throttled to ~once
+// every 2s of audio. A GCD/runloop timer can't be used — AVAudioEngine+VPIO blocks background-queue
+// timers in this process. If these lines appear → the tap fires (rms tells silence vs signal); if NONE
+// appear → the CoreAudio input tap never fires (device/format dead) = no audio can reach STT.
+var diagTaps = 0
+var diagAcc = 0 // frames since last log (×decim ≈ input samples) → throttle by input sample rate
 
 input.installTap(onBus: 0, bufferSize: 2048, format: inFmt) { buf, _ in
     guard let ch = buf.floatChannelData else { return }
@@ -141,6 +157,11 @@ input.installTap(onBus: 0, bufferSize: 2048, format: inFmt) { buf, _ in
     let gate = gateEnabled && audible && rms < threshold // user speech must clearly dominate the expected residue
     gateLock.unlock()
     if gate { for k in 0..<pcm.count { pcm[k] = 0 } } // keep the stream's timing — send silence, not nothing
+    diagTaps += 1; diagAcc += n
+    if diagAcc >= Int(inFmt.sampleRate) * 2 { // ~every 2s of input audio
+        FileHandle.standardError.write("tap: \(diagTaps) firings/2s, rms=\(Int(rms)), gate=\(gate)\n".data(using: .utf8)!)
+        diagTaps = 0; diagAcc = 0
+    }
     pcm.withUnsafeBufferPointer { p in
         out.write(Data(buffer: p))
     }
@@ -152,6 +173,17 @@ do { try engine.start() } catch {
 }
 player.play()
 
+// Non-VPIO mode only: the input tap won't fire unless the output graph is actively rendering. Keep the
+// MAIN player warm with a continuous silent loop WHEN IDLE — when real TTS arrives the stdin handler
+// stops this loop, plays the TTS, then resumes the loop on FLUSH. (A separate node does NOT wake the
+// tap — empirically only the main player does.) Pure zeros → inaudible; gate envelope untouched.
+let keepSilence = AVAudioPCMBuffer(pcmFormat: playFmt, frameCapacity: 4410)!; keepSilence.frameLength = 4410
+var loopingSilence = false
+func armIdleKeepAlive() { if !useVPIO { player.scheduleBuffer(keepSilence, at: nil, options: .loops, completionHandler: nil); loopingSilence = true } }
+// Clear the idle loop so a real TTS buffer can play (a .loops buffer never yields the queue otherwise).
+func stopIdleKeepAlive() { if loopingSilence { player.stop(); player.play(); loopingSilence = false } }
+armIdleKeepAlive()
+
 // stdin reader (background): framed playback + in-band FLUSH. EOF = parent gone → clean exit.
 DispatchQueue.global(qos: .userInteractive).async {
     let stdin = FileHandle.standardInput
@@ -171,6 +203,7 @@ DispatchQueue.global(qos: .userInteractive).async {
         if len == 0 { // FLUSH: stop drops every scheduled buffer; restart for the next turn
             player.stop()
             player.play()
+            armIdleKeepAlive() // non-VPIO: resume the idle silence loop so the input tap stays awake between turns
             pausedAt = 0
             gateLock.lock(); pausedSince = 0; audibleUntil = CACurrentMediaTime(); schedEnv.removeAll(); gateLock.unlock() // gate tail only; envelope timeline resets with playerTime
             continue
@@ -224,6 +257,7 @@ DispatchQueue.global(qos: .userInteractive).async {
             }
             gateLock.unlock()
         }
+        stopIdleKeepAlive() // non-VPIO: drop the idle silence loop before the first real TTS buffer
         player.scheduleBuffer(buf, completionHandler: nil)
         // extend the audible window by this buffer's duration (+0.8s tail covers device latency)
         gateLock.lock()

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@livx.cc/agentx",
-  "version": "0.96.15",
+  "version": "0.96.17",
   "description": "Edge-native AI agent runtime — drives a virtual filesystem via any LLM (ai.libx.js). Same bytes run in node, browser, or edge.",
   "type": "module",
   "main": "./dist/index.js",