@hasna/recordings 0.1.16 → 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: "0.1.16",
+    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 = "0.1.16";
+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: "0.1.16",
+    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 = "0.1.16";
+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 = "0.1.16";
+export declare const VERSION = "0.1.17";
 //# sourceMappingURL=version.d.ts.map

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hasna/recordings",
-  "version": "0.1.16",
+  "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

@@ -17,7 +17,6 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
     private var ws: URLSessionWebSocketTask?
     private var receiveTask: Task<Void, Never>?
     private var outboundEventTask: Task<Void, Never>?
-    private var liveCommitTask: Task<Void, Never>?
     private var isConfigured = false
     private var pendingAudioChunks: [Data] = []
     private var itemOrder: [String] = []
@@ -29,11 +28,8 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
     private var queuedCommitCount = 0
     private var uncommittedAudioBytes = 0
     private var lastRealtimeEventAt = Date.distantPast
-    private var lastCommitQueuedAt = Date.distantPast
 
     private nonisolated static let minimumManualCommitBytes = 5_760
-    private nonisolated static let minimumLiveCommitBytes = 38_400
-    private nonisolated static let liveCommitIntervalSeconds: TimeInterval = 0.8
 
     private let apiKey: String
     private let homePath: String
@@ -70,7 +66,6 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         queuedCommitCount = 0
         uncommittedAudioBytes = 0
         lastRealtimeEventAt = Date()
-        lastCommitQueuedAt = Date()
         error = nil
 
         var request = URLRequest(url: Self.transcriptionURL)
@@ -122,7 +117,6 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
             "audio": base64,
         ]
         enqueueOutboundEvent(msg)
-        scheduleLiveCommitIfNeeded()
     }
 
     /// Signal end of input — triggers final transcription completion.
@@ -140,7 +134,6 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         uncommittedAudioBytes = 0
         queuedCommitCount += 1
         lastRealtimeEventAt = Date()
-        lastCommitQueuedAt = Date()
         NativeAppLog.write("realtime commit queued reason=\(reason) bytes=\(bytesToCommit)", homePath: homePath)
         let commitTask = enqueueOutboundEvent(msg)
         await commitTask?.value
@@ -178,11 +171,9 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         receiveTask?.cancel()
         ws?.cancel(with: .normalClosure, reason: nil)
         outboundEventTask?.cancel()
-        liveCommitTask?.cancel()
         ws = nil
         receiveTask = nil
         outboundEventTask = nil
-        liveCommitTask = nil
         pendingAudioChunks.removeAll(keepingCapacity: true)
         let text = accumulatedText
         return text
@@ -285,25 +276,6 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         return task
     }
 
-    private func scheduleLiveCommitIfNeeded(now: Date = Date()) {
-        guard liveCommitTask == nil else { return }
-        let secondsSinceLastCommit = now.timeIntervalSince(lastCommitQueuedAt)
-        guard Self.shouldAutoCommitLiveInput(
-            uncommittedAudioBytes: uncommittedAudioBytes,
-            secondsSinceLastCommit: secondsSinceLastCommit
-        ) else { return }
-
-        liveCommitTask = Task { [weak self] in
-            _ = await self?.commitLiveInput()
-        }
-    }
-
-    @discardableResult
-    private func commitLiveInput() async -> Bool {
-        defer { liveCommitTask = nil }
-        return await commitInput(reason: "live")
-    }
-
     private func registerItem(_ itemID: String, previousItemID: String?) {
         guard !itemOrder.contains(itemID) else { return }
         if let previousItemID, let previousIndex = itemOrder.firstIndex(of: previousItemID) {
@@ -383,14 +355,6 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
         uncommittedAudioBytes >= minimumManualCommitBytes
     }
 
-    private nonisolated static func shouldAutoCommitLiveInput(
-        uncommittedAudioBytes: Int,
-        secondsSinceLastCommit: TimeInterval
-    ) -> Bool {
-        uncommittedAudioBytes >= minimumLiveCommitBytes &&
-            secondsSinceLastCommit >= liveCommitIntervalSeconds
-    }
-
     private nonisolated static func transcriptionSessionUpdateEvent(transcription: [String: Any]) -> [String: Any] {
         [
             "type": "session.update",
@@ -403,7 +367,12 @@ public final class RealtimeTranscriptionClient: ObservableObject, @unchecked Sen
                             "rate": 24_000,
                         ],
                         "transcription": transcription,
-                        "turn_detection": NSNull(),
+                        "turn_detection": [
+                            "type": "server_vad",
+                            "threshold": 0.5,
+                            "prefix_padding_ms": 300,
+                            "silence_duration_ms": 350,
+                        ],
                         "noise_reduction": [
                             "type": "near_field",
                         ],
@@ -458,16 +427,6 @@ extension RealtimeTranscriptionClient {
         shouldManuallyCommit(uncommittedAudioBytes: uncommittedAudioBytes)
     }
 
-    public nonisolated static func shouldAutoCommitLiveInputTestHelper(
-        uncommittedAudioBytes: Int,
-        secondsSinceLastCommit: TimeInterval
-    ) -> Bool {
-        shouldAutoCommitLiveInput(
-            uncommittedAudioBytes: uncommittedAudioBytes,
-            secondsSinceLastCommit: secondsSinceLastCommit
-        )
-    }
-
     public nonisolated static func sessionUpdateTestHelper(prompt: String, language: String = "") -> [String: Any] {
         var transcription: [String: Any] = [
             "model": modelID,

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,

package/src/native/Recordings/RecordingsTests/RealtimeTranscriptionTests.swift

@@ -1,4 +1,3 @@
-import Foundation
 import Testing
 @testable import RecordingsLib
 
@@ -99,7 +98,8 @@ struct RealtimeTranscriptionTests {
         #expect(transcription?["prompt"] as? String == "Use Alumia as vocabulary")
         #expect(transcription?["language"] as? String == "en")
 
-        #expect(input?["turn_detection"] is NSNull)
+        let turnDetection = input?["turn_detection"] as? [String: Any]
+        #expect(turnDetection?["type"] as? String == "server_vad")
 
         let include = session?["include"] as? [String]
         #expect(include?.contains("item.input_audio_transcription.logprobs") == true)
@@ -117,22 +117,6 @@ struct RealtimeTranscriptionTests {
         #expect(RealtimeTranscriptionClient.shouldManuallyCommitTestHelper(uncommittedAudioBytes: 5_760) == true)
     }
 
-    @Test("Live commits wait for enough audio and elapsed time")
-    func liveCommitThreshold() {
-        #expect(RealtimeTranscriptionClient.shouldAutoCommitLiveInputTestHelper(
-            uncommittedAudioBytes: 38_399,
-            secondsSinceLastCommit: 1.0
-        ) == false)
-        #expect(RealtimeTranscriptionClient.shouldAutoCommitLiveInputTestHelper(
-            uncommittedAudioBytes: 38_400,
-            secondsSinceLastCommit: 0.79
-        ) == false)
-        #expect(RealtimeTranscriptionClient.shouldAutoCommitLiveInputTestHelper(
-            uncommittedAudioBytes: 38_400,
-            secondsSinceLastCommit: 0.8
-        ) == true)
-    }
-
     @Test("Partial realtime text falls back for longer recordings")
     func partialRealtimeFallback() {
         #expect(RecordingEngine.shouldFallbackFromPartialRealtime(text: "Hi", pcmByteCount: 96_000) == true)