@hasna/recordings 0.1.15 → 0.1.17

package/dist/cli/index.js

@@ -7,7 +7,7 @@ var __require = import.meta.require;
 var require_package = __commonJS((exports, module) => {
   module.exports = {
     name: "@hasna/recordings",
-    version: "0.1.15",
+    version: "0.1.17",
     type: "module",
     description: "Speech-to-text recording tool with MCP and CLI \u2014 records, transcribes, and optionally enhances text using AI",
     repository: {
@@ -10466,7 +10466,7 @@ async function processText(rawText, config, systemPrompt) {
 }
 
 // src/version.ts
-var VERSION = "0.1.15";
+var VERSION = "0.1.17";
 
 // src/cli/index.ts
 var program = new Command;

package/dist/mcp/index.js

@@ -21,7 +21,7 @@ var __require = import.meta.require;
 var require_package = __commonJS((exports, module) => {
   module.exports = {
     name: "@hasna/recordings",
-    version: "0.1.15",
+    version: "0.1.17",
     type: "module",
     description: "Speech-to-text recording tool with MCP and CLI \u2014 records, transcribes, and optionally enhances text using AI",
     repository: {
@@ -14983,7 +14983,7 @@ async function processText(rawText, config, systemPrompt) {
 }
 
 // src/version.ts
-var VERSION = "0.1.15";
+var VERSION = "0.1.17";
 
 // src/mcp/index.ts
 var config = loadConfig();

package/dist/version.d.ts

@@ -1,2 +1,2 @@
-export declare const VERSION = "0.1.15";
+export declare const VERSION = "0.1.17";
 //# sourceMappingURL=version.d.ts.map

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hasna/recordings",
-  "version": "0.1.15",
+  "version": "0.1.17",
   "type": "module",
   "description": "Speech-to-text recording tool with MCP and CLI — records, transcribes, and optionally enhances text using AI",
   "repository": {

package/src/native/Recordings/RecordingsLib/RealtimeTranscriptionClient.swift

@@ -16,7 +16,7 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
 
     private var ws: URLSessionWebSocketTask?
     private var receiveTask: Task<Void, Never>?
-    private var audioSendTask: Task<Void, Never>?
+    private var outboundEventTask: Task<Void, Never>?
     private var isConfigured = false
     private var pendingAudioChunks: [Data] = []
     private var itemOrder: [String] = []
@@ -25,6 +25,7 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
     private var committedItemIDs = Set<String>()
     private var completedItemIDs = Set<String>()
     private var completedEventCount = 0
+    private var queuedCommitCount = 0
     private var uncommittedAudioBytes = 0
     private var lastRealtimeEventAt = Date.distantPast
 
@@ -62,6 +63,7 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         committedItemIDs.removeAll(keepingCapacity: true)
         completedItemIDs.removeAll(keepingCapacity: true)
         completedEventCount = 0
+        queuedCommitCount = 0
         uncommittedAudioBytes = 0
         lastRealtimeEventAt = Date()
         error = nil
@@ -114,27 +116,27 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
             "type": "input_audio_buffer.append",
             "audio": base64,
         ]
-        let previousSendTask = audioSendTask
-        audioSendTask = Task { [weak self] in
-            await previousSendTask?.value
-            try? await self?.sendEvent(msg)
-        }
+        enqueueOutboundEvent(msg)
     }
 
     /// Signal end of input — triggers final transcription completion.
     @discardableResult
-    public func commitInput() async -> Bool {
+    public func commitInput(reason: String = "final") async -> Bool {
         guard isStreaming, ws != nil else { return false }
         flushPendingAudio()
-        await audioSendTask?.value
-        guard Self.shouldManuallyCommit(uncommittedAudioBytes: uncommittedAudioBytes) else {
+        let bytesToCommit = uncommittedAudioBytes
+        guard Self.shouldManuallyCommit(uncommittedAudioBytes: bytesToCommit) else {
             return false
         }
         let msg: [String: Any] = [
             "type": "input_audio_buffer.commit",
         ]
+        uncommittedAudioBytes = 0
+        queuedCommitCount += 1
         lastRealtimeEventAt = Date()
-        try? await sendEvent(msg)
+        NativeAppLog.write("realtime commit queued reason=\(reason) bytes=\(bytesToCommit)", homePath: homePath)
+        let commitTask = enqueueOutboundEvent(msg)
+        await commitTask?.value
         return true
     }
 
@@ -143,13 +145,15 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         guard isStreaming else { return accumulatedText }
         let completedCountBeforeCommit = completedEventCount
         let didManualCommit = await commitInput()
+        let expectedCommitCount = queuedCommitCount
 
         let deadline = Date().addingTimeInterval(TimeInterval(timeoutMilliseconds) / 1_000)
         while Date() < deadline, isStreaming {
             let hasIncompleteCommittedItems = !committedItemIDs.subtracting(completedItemIDs).isEmpty
             let hasManualCommitCompletion = !didManualCommit || completedEventCount > completedCountBeforeCommit
+            let hasQueuedCommitCompletion = completedEventCount >= expectedCommitCount
             let quietLongEnough = Date().timeIntervalSince(lastRealtimeEventAt) >= 0.35
-            if !hasIncompleteCommittedItems, hasManualCommitCompletion, quietLongEnough {
+            if !hasIncompleteCommittedItems, hasManualCommitCompletion, hasQueuedCommitCompletion, quietLongEnough {
                 break
             }
             try? await Task.sleep(for: .milliseconds(50))
@@ -166,10 +170,10 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         isConfigured = false
         receiveTask?.cancel()
         ws?.cancel(with: .normalClosure, reason: nil)
-        audioSendTask?.cancel()
+        outboundEventTask?.cancel()
         ws = nil
         receiveTask = nil
-        audioSendTask = nil
+        outboundEventTask = nil
         pendingAudioChunks.removeAll(keepingCapacity: true)
         let text = accumulatedText
         return text
@@ -214,7 +218,6 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
                 registerItem(itemID, previousItemID: json["previous_item_id"] as? String)
                 committedItemIDs.insert(itemID)
             }
-            uncommittedAudioBytes = 0
 
         case "conversation.item.input_audio_transcription.delta":
             lastRealtimeEventAt = Date()
@@ -230,15 +233,19 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
             lastRealtimeEventAt = Date()
             let itemID = json["item_id"] as? String ?? "__default__"
             registerItem(itemID, previousItemID: nil)
+            completedItemIDs.insert(itemID)
+            completedEventCount += 1
             if let text = json["transcript"] as? String, !text.isEmpty {
                 completedTextByItem[itemID] = text
-                completedItemIDs.insert(itemID)
-                completedEventCount += 1
-                rebuildAccumulatedText()
             }
+            rebuildAccumulatedText()
 
         case "conversation.item.input_audio_transcription.failed":
             lastRealtimeEventAt = Date()
+            if let itemID = json["item_id"] as? String {
+                completedItemIDs.insert(itemID)
+                completedEventCount += 1
+            }
             if let msg = json["error"] as? [String: Any],
                let message = msg["message"] as? String {
                 self.error = message
@@ -257,6 +264,18 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         }
     }
 
+    @discardableResult
+    private func enqueueOutboundEvent(_ obj: [String: Any]) -> Task<Void, Never>? {
+        guard ws != nil else { return nil }
+        let previousTask = outboundEventTask
+        let task = Task { [weak self] in
+            await previousTask?.value
+            try? await self?.sendEvent(obj)
+        }
+        outboundEventTask = task
+        return task
+    }
+
     private func registerItem(_ itemID: String, previousItemID: String?) {
         guard !itemOrder.contains(itemID) else { return }
         if let previousItemID, let previousIndex = itemOrder.firstIndex(of: previousItemID) {

package/src/native/Recordings/RecordingsLib/RecordingEngine.swift

@@ -382,13 +382,12 @@ public final class RecordingEngine: ObservableObject {
     // MARK: - Real-time Streaming
 
     private func startRealtimeStreaming(apiKey: String) {
-        let systemPrompt = projectStore?.effectiveSystemPrompt ?? ""
         let client = RealtimeTranscriptionClient(apiKey: apiKey, homePath: home)
         realtimeClient = client
         log("realtime streaming task starting")
 
         streamingTask = Task {
-            await client.startStreaming(systemPrompt: systemPrompt)
+            await client.startStreaming()
             self.log("realtime start completed streaming=\(client.isStreaming) error=\(client.error ?? "")")
 
             // Receive deltas
@@ -453,25 +452,15 @@ public final class RecordingEngine: ObservableObject {
             self.streamingTask?.cancel()
             self.streamingTask = nil
 
-            let text = streamingResult.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty ? nil : streamingResult
+            let realtimeText = streamingResult.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty ? nil : streamingResult
 
             self.liveTranscriptionText = ""
 
-            if let text, !Self.shouldFallbackFromPartialRealtime(text: text, pcmByteCount: self.recordedPCM.count) {
-                self.log("finish using realtime text chars=\(text.count)")
-                self.isTranscribing = false
-                self.finishWithText(
-                    text,
-                    curMode: curMode,
-                    targetAppBundleIdentifier: targetAppBundleIdentifier,
-                    activeProjectId: activeProjectId,
-                    activeProjectName: activeProjectName
-                )
-            } else if let audioPath, self.writeCapturedWAV(to: audioPath) {
-                if let text {
-                    self.log("realtime text looked partial chars=\(text.count); falling back to CLI")
+            if let audioPath, self.writeCapturedWAV(to: audioPath) {
+                if let realtimeText {
+                    self.log("realtime preview ignored for final chars=\(realtimeText.count)")
                 }
-                self.log("falling back to CLI transcription audioPath=\(audioPath)")
+                self.log("transcribing captured full audio audioPath=\(audioPath)")
                 self.fallbackTranscribe(
                     audioPath: audioPath,
                     curMode: curMode,