@markusylisiurunen/tau 0.3.77 → 0.3.79

package/dist/tui/session_chat_controller.js

@@ -1,6 +1,6 @@
 import { dirname } from "node:path/posix";
 import { createCommandRegistry, } from "../core/commands/index.js";
-import { getElevenLabsApiKey, getOpenAIApiKey, } from "../core/config/index.js";
+import { getElevenLabsApiKey, getOpenAIApiKey } from "../core/config/index.js";
 import { captureDiffReviewSnapshot, DiffReviewBridge, } from "../core/diff_review/index.js";
 import { buildDiffReviewInstructions } from "../core/diff_review/review_instructions.js";
 import { createDefaultCoreDeps } from "../core/runtime/deps.js";
@@ -14,6 +14,7 @@ import { getSpeechToTextRecordingMaxDurationMs, getSpeechToTextStreamingSampleRa
 import { collectSpeechToTextContext } from "../core/utils/speech_to_text_context.js";
 import { formatTauUserText, hasAutoCompactionContinuationMetadata, splitTauUserText, } from "../core/utils/user_metadata.js";
 import { APP_VERSION } from "../core/version.js";
+import { launchBuiltInDiffTool } from "../diff_tool/launcher.js";
 import { applySessionProtocolDelta, applySessionProtocolSubagentActivitiesMessage, } from "../protocol/session_protocol.js";
 import { TauSessionProtocolResponseError } from "../transport/errors.js";
 import { copyAssistantCodeToClipboard, copyAssistantTextToClipboard, } from "./chat_controller/assistant_clipboard.js";
@@ -37,7 +38,6 @@ export class SessionChatController {
     targetLabel;
     configuredClientToolNames;
     config;
-    defaultDiffTool;
     diffToolLauncher;
     deps;
     speechToTextDeps;
@@ -88,6 +88,9 @@ export class SessionChatController {
     submissionHiddenSystemMessages = [];
     pendingEditorMutations = 0;
     speechActivityLabel;
+    autoSpeak = false;
+    pendingAutoSpeak;
+    lastAutoSpeakMessageId;
     speakTask;
     constructor(options) {
         this.view = options.view;
@@ -98,7 +101,6 @@ export class SessionChatController {
         this.targetLabel = options.targetLabel;
         this.configuredClientToolNames = options.configuredClientToolNames ?? [];
         this.config = options.config ?? {};
-        this.defaultDiffTool = options.defaultDiffTool;
         this.diffToolLauncher = options.diffToolLauncher;
         this.deps = options.deps ?? createDefaultCoreDeps();
         this.speechToTextDeps = options.speechToTextDeps;
@@ -119,6 +121,7 @@ export class SessionChatController {
             reload: () => this.reloadContent(),
             listen: (action) => this.handleListenCommand(action),
             speak: () => this.speakLastAssistantMessage(),
+            autoSpeak: (enabled) => this.setAutoSpeak(enabled),
             persona: (id) => this.setPersona(id),
             prompt: (id) => this.insertPrompt(id),
             theme: (id) => this.switchTheme(id),
@@ -130,7 +133,6 @@ export class SessionChatController {
             refreshStatus: () => this.refreshStatus(),
             startTurnTimer: () => this.startTurnTimer(),
             stopTurnTimer: () => this.stopTurnTimer(),
-            getDiffToolConfig: () => this.resolveDiffToolConfig(),
             startSession: (args) => this.startDiffReviewBridge(args, this.session, this.snapshot),
             onReviewReturned: (review) => this.handleReturnedDiffReview(review, this.session),
         });
@@ -187,7 +189,13 @@ export class SessionChatController {
             onShiftTab: () => void this.cycleReasoningLevel(),
             onCtrlP: () => void this.cyclePersonality(),
             onCtrlS: () => void this.stashEditorToClipboard(),
-            onCtrlY: () => void this.toggleListenCapture(),
+            onToggleRecording: () => void this.toggleListenCapture(),
+            recordingShortcut: this.config.speech?.recordingShortcut,
+            onDisabledSubmit: (mode) => {
+                if (this.listenRecording && !this.listenTransition) {
+                    void this.runListenTransition(() => this.stopListenCapture(mode));
+                }
+            },
             onCtrlG: () => this.interruptSelectedSubagent(),
             onEscape: () => void this.interrupt(),
             beforeSubmit: (text) => {
@@ -197,7 +205,7 @@ export class SessionChatController {
             },
             onChange: (text) => this.handleEditorChange(text),
             onSubmit: (text) => void this.handleSubmit(text, this.takeHiddenSystemSubmission()),
-            onSteerSubmit: (text) => void this.submitSteeringMessage(text, this.takeHiddenSystemSubmission()),
+            onQueueSubmit: (text) => void this.handleSubmit(text, this.takeHiddenSystemSubmission(), "queue"),
             onAltUp: () => void this.cancelPendingMessagesIntoEditor(),
             onAltDown: () => this.cycleSubagentSelection(),
         };
@@ -234,7 +242,7 @@ export class SessionChatController {
     formatSubmittedText(text, hiddenSystemMessages) {
         return formatTauUserText({ text, hiddenSystemMessages });
     }
-    async handleSubmit(text, hiddenSystemMessages = []) {
+    async handleSubmit(text, hiddenSystemMessages = [], mode = "steer") {
         const trimmed = text.trim();
         if (!trimmed) {
             if (this.isStreaming || this.submittedTurnInProgress) {
@@ -272,7 +280,11 @@ export class SessionChatController {
             if (trimmed.startsWith("!")) {
                 return;
             }
-            this.submitQueuedText(this.formatSubmittedText(trimmed, hiddenSystemMessages));
+            const submittedText = this.formatSubmittedText(trimmed, hiddenSystemMessages);
+            if (mode === "steer")
+                this.submitSteeringText(submittedText);
+            else
+                this.submitQueuedText(submittedText);
             return;
         }
         if (trimmed.startsWith("/") && this.isSingleLineInput(text)) {
@@ -314,6 +326,9 @@ export class SessionChatController {
                 contextFiles: this.getContextFilePaths(),
                 skills: this.snapshot.catalog.skills,
                 themes: this.themeIds,
+                recordingShortcut: this.config.speech?.recordingShortcut
+                    ? `${this.config.speech.recordingShortcut.gesture === "double-tap" ? "double-tap " : ""}${this.config.speech.recordingShortcut.key}`
+                    : "ctrl+y",
                 formatPath: (path) => formatPathForSessionDisplay(path, this.snapshot.executionEnvironment.home),
             }),
         });
@@ -353,7 +368,6 @@ export class SessionChatController {
             themes: () => this.themeIds.map((id) => ({ id })),
             autocompletePaths: async (query, limit) => (await this.session.autocompletePaths({ query, limit })).paths,
             skills: () => this.snapshot.catalog.skills.map((skill) => skill.name),
-            subagents: () => Object.keys(this.getCurrentPersonaSnapshot()?.subagents ?? {}),
         };
     }
     buildStartupIntroTitle() {
@@ -473,21 +487,6 @@ export class SessionChatController {
             this.view.requestRender();
         }
     }
-    async submitSteeringMessage(text, hiddenSystemMessages = []) {
-        const trimmed = text.trim();
-        if (!trimmed) {
-            return;
-        }
-        if (!this.isSessionOperationActive()) {
-            await this.runSessionTurn(() => this.session.submit(this.formatSubmittedText(trimmed, hiddenSystemMessages)));
-            return;
-        }
-        if (this.isBlockingSessionOperationActive()) {
-            this.view.showFooterNotice("wait for tau to become idle before submitting input", "default");
-            return;
-        }
-        this.submitSteeringText(this.formatSubmittedText(trimmed, hiddenSystemMessages));
-    }
     submitQueuedText(text) {
         void this.session.queue(text).catch((error) => {
             if (!this.isPendingMessageCancellation(error)) {
@@ -553,8 +552,10 @@ export class SessionChatController {
         this.view.startWorkingIcon();
         this.refreshStatus();
         try {
-            await task();
+            const result = await task();
             await this.syncFromSessionSnapshot();
+            if (result.turn.status === "completed")
+                this.queueAutoSpeak();
         }
         catch (error) {
             this.view.addTranscriptNotice("failed to run assistant turn", "error", [
@@ -570,6 +571,7 @@ export class SessionChatController {
         }
     }
     async interrupt() {
+        this.pendingAutoSpeak = undefined;
         if (this.interruptLifecycle.interruptActiveTask()) {
             this.view.showFooterNotice("interrupted", "default");
             return;
@@ -655,7 +657,15 @@ export class SessionChatController {
             this.view.showFooterNotice("speech transcription already in progress", "default");
             return;
         }
-        await this.runListenTransition(() => this.startListenCapture());
+        this.pendingAutoSpeak = undefined;
+        await this.runListenTransition(async () => {
+            if (this.speakTask) {
+                this.speakTask.abortController.abort();
+                await this.speakTask.completion.catch(() => { });
+            }
+            if (!this.disposed)
+                await this.startListenCapture();
+        });
     }
     async runListenTransition(task) {
         if (this.listenTransition) {
@@ -790,7 +800,7 @@ export class SessionChatController {
                 clearTimeout(timeout);
         }
     }
-    async stopListenCapture() {
+    async stopListenCapture(mode) {
         const recording = this.listenRecording;
         if (!recording)
             return;
@@ -810,7 +820,9 @@ export class SessionChatController {
             return;
         }
         const durationMs = Math.min(Math.max(Date.now() - recording.startedAt, 0), SPEECH_TO_TEXT_CLIENT_MAX_DURATION_MS);
-        await this.transcribeListenAudioFile(recording.audioPath, durationMs, recording.transcription);
+        const transcribed = await this.transcribeListenAudioFile(recording.audioPath, durationMs, recording.transcription);
+        if (mode && transcribed && !this.disposed)
+            this.view.submitEditor(mode);
     }
     async retryRetainedListenAudio() {
         if (this.pendingEditorMutations > 0) {
@@ -826,7 +838,9 @@ export class SessionChatController {
             this.view.showFooterNotice("no failed speech recording to retry", "default");
             return;
         }
-        await this.runListenTransition(() => this.transcribeListenAudioFile(retainedAudio.audioPath, retainedAudio.durationMs));
+        await this.runListenTransition(async () => {
+            await this.transcribeListenAudioFile(retainedAudio.audioPath, retainedAudio.durationMs);
+        });
     }
     async discardRetainedListenAudio() {
         if (this.listenRecording || this.listenTransition || this.listenActivityLabel) {
@@ -883,7 +897,7 @@ export class SessionChatController {
                 this.retainedListenAudio = undefined;
             }
             await cleanupListenTempFile(audioPath);
-            return;
+            return false;
         }
         if (!this.listenPreview)
             this.beginListenPreview();
@@ -897,7 +911,7 @@ export class SessionChatController {
                     `recording retained at ${audioPath}`,
                     "run /listen discard to delete it",
                 ]);
-                return;
+                return false;
             }
             let audio;
             try {
@@ -912,7 +926,7 @@ export class SessionChatController {
                     this.retainedListenAudio = undefined;
                 }
                 await cleanupListenTempFile(audioPath);
-                return;
+                return false;
             }
             if (audio.byteLength < LISTEN_RECORDING_MIN_BYTES) {
                 transcription?.abort();
@@ -921,7 +935,7 @@ export class SessionChatController {
                     this.retainedListenAudio = undefined;
                 }
                 await cleanupListenTempFile(audioPath);
-                return;
+                return false;
             }
             if (this.disposed) {
                 transcription?.abort();
@@ -929,7 +943,7 @@ export class SessionChatController {
                     this.retainedListenAudio = undefined;
                 }
                 await cleanupListenTempFile(audioPath);
-                return;
+                return false;
             }
             let activeTranscription = transcription;
             this.listenActivityLabel = "transcribing voice input";
@@ -942,7 +956,7 @@ export class SessionChatController {
                     mimeType: "audio/wav",
                 });
                 if (this.listenPreview !== preview)
-                    return;
+                    return false;
                 preview.commit(text);
                 if (text.trim() && !this.editorHiddenSystemMessages.includes(TRANSCRIPTION_GUIDANCE)) {
                     this.editorHiddenSystemMessages.push(TRANSCRIPTION_GUIDANCE);
@@ -957,6 +971,7 @@ export class SessionChatController {
                         `recording remains at ${audioPath}; delete it manually`,
                     ]);
                 }
+                return Boolean(text.trim());
             }
             catch (error) {
                 this.retainedListenAudio = { audioPath, durationMs };
@@ -965,6 +980,7 @@ export class SessionChatController {
                     `recording retained at ${audioPath}`,
                     "run /listen retry to try again, or /listen discard to delete it",
                 ]);
+                return false;
             }
             finally {
                 activeTranscription?.abort();
@@ -1117,6 +1133,9 @@ export class SessionChatController {
             this.ephemeralUnsubscribe?.();
             this.pendingUserMessagesUnsubscribe?.();
             this.subagentActivitiesUnsubscribe?.();
+            this.autoSpeak = false;
+            this.pendingAutoSpeak = undefined;
+            this.lastAutoSpeakMessageId = undefined;
             this.session = nextSession;
             this.snapshot = nextSnapshot;
             this.subagentActivities = nextSession.subagentActivities();
@@ -1218,9 +1237,48 @@ export class SessionChatController {
             this.snapshotRecoveryDeltas.push(delta);
             return;
         }
+        if (delta.toRevision <= this.snapshot.revision)
+            return;
+        const invalidatesSpeech = delta.cause.type === "rewind" ||
+            delta.cause.type === "compaction" ||
+            delta.cause.type === "resync";
+        if (invalidatesSpeech)
+            this.pendingAutoSpeak = undefined;
+        const previousLifecycle = this.snapshot.lifecycle;
+        const previousTurns = this.snapshot.turns;
         if (this.tryApplySdkDelta(delta)) {
+            if (this.autoSpeak && !invalidatesSpeech) {
+                if (previousLifecycle !== "running" && this.snapshot.lifecycle === "running") {
+                    this.pendingAutoSpeak = undefined;
+                    this.speakTask?.abortController.abort();
+                    this.lastAutoSpeakMessageId = this.snapshot.messages.findLast((entry) => isAssistantMessage(entry.message) && entry.state === "committed")?.id;
+                }
+                if (delta.delta.type === "snapshot.patch") {
+                    for (const change of delta.delta.changes) {
+                        if (change.type !== "turn.set")
+                            continue;
+                        const turn = change.turn;
+                        if (turn.state === "running") {
+                            this.pendingAutoSpeak = undefined;
+                            this.speakTask?.abortController.abort();
+                            continue;
+                        }
+                        if (previousTurns[turn.userHistoryEntryId]?.state === "running" &&
+                            turn.outcome.status === "completed") {
+                            this.queueAutoSpeak();
+                        }
+                    }
+                }
+                if (previousLifecycle === "running" &&
+                    this.snapshot.lifecycle === "idle" &&
+                    !Object.values(previousTurns).some((turn) => turn.state === "running")) {
+                    this.queueAutoSpeak();
+                }
+                this.drainAutoSpeak();
+            }
             return;
         }
+        this.pendingAutoSpeak = undefined;
         this.snapshotRecoveryDeltas.push(delta);
         void this.recoverFromRevisionGap();
     }
@@ -1988,7 +2046,50 @@ export class SessionChatController {
             this.view.addTranscriptNotice("failed to clear goal", "error", [error.message]);
         }
     }
-    async speakLastAssistantMessage() {
+    setAutoSpeak(enabled) {
+        if ((enabled ?? !this.autoSpeak) && this.deps.env.platform() !== "darwin") {
+            this.view.showFooterNotice("auto-speak is currently supported only on macOS", "default");
+            return;
+        }
+        const next = enabled ?? !this.autoSpeak;
+        if (next && !this.autoSpeak) {
+            this.lastAutoSpeakMessageId = this.snapshot.messages.findLast((entry) => isAssistantMessage(entry.message))?.id;
+        }
+        this.autoSpeak = next;
+        if (!this.autoSpeak) {
+            this.pendingAutoSpeak = undefined;
+            this.speakTask?.abortController.abort();
+        }
+        this.view.showFooterNotice(`auto-speak ${this.autoSpeak ? "on" : "off"}`, "default");
+        this.refreshStatus();
+    }
+    queueAutoSpeak() {
+        if (!this.autoSpeak)
+            return;
+        const entry = this.snapshot.messages.findLast((entry) => isAssistantMessage(entry.message));
+        if (!entry || entry.id === this.lastAutoSpeakMessageId || !isAssistantMessage(entry.message))
+            return;
+        this.lastAutoSpeakMessageId = entry.id;
+        if (entry.message.stopReason === "stop" || entry.message.stopReason === "length") {
+            this.pendingAutoSpeak = entry.message;
+        }
+    }
+    drainAutoSpeak() {
+        if (!this.pendingAutoSpeak || this.isSessionOperationActive())
+            return;
+        const message = this.pendingAutoSpeak;
+        this.pendingAutoSpeak = undefined;
+        if (!this.autoSpeak ||
+            this.disposed ||
+            this.listenRecording ||
+            this.listenTransition ||
+            this.listenActivityLabel ||
+            this.speakTask ||
+            this.hasPendingUserMessages)
+            return;
+        void this.speakLastAssistantMessage(message);
+    }
+    async speakLastAssistantMessage(message) {
         if (this.speakTask) {
             this.view.showFooterNotice("speech playback already in progress", "default");
             return;
@@ -2001,7 +2102,7 @@ export class SessionChatController {
             this.view.showFooterNotice("/speak is currently supported only on macOS.", "default");
             return;
         }
-        const lastAssistant = this.getLastAssistantMessage();
+        const lastAssistant = message ?? this.getLastAssistantMessage();
         if (!lastAssistant) {
             this.view.showFooterNotice("no assistant message to speak yet.", "default");
             return;
@@ -2067,6 +2168,7 @@ export class SessionChatController {
         });
     }
     refreshStatus() {
+        this.drainAutoSpeak();
         const activityLabel = this.getActivityLabel();
         this.view.updateStatus({
             footer: activityLabel
@@ -2078,6 +2180,7 @@ export class SessionChatController {
                     sessionCost: this.getSessionCostString(),
                     duration: this.getTurnDurationString(),
                     pursuingGoal: this.snapshot.goal?.status === "active",
+                    autoSpeak: this.autoSpeak,
                 },
             editor: {
                 mode: this.getInputMode(),
@@ -2278,9 +2381,6 @@ export class SessionChatController {
             !this.listenActivityLabel &&
             !this.speakTask);
     }
-    resolveDiffToolConfig() {
-        return this.config.diffTool ?? this.defaultDiffTool;
-    }
     async startDiffReviewBridge(args, session, sessionSnapshot, executionEnvironment = session) {
         const cwd = sessionSnapshot.executionEnvironment.cwd;
         const backend = createSdkToolExecutionBackend({ executionEnvironment, cwd });
@@ -2305,7 +2405,6 @@ export class SessionChatController {
                 ...(options.reasoning !== undefined ? { reasoning: options.reasoning } : {}),
             }),
             deps: this.deps,
-            ...(this.diffToolLauncher ? { toolLauncher: this.diffToolLauncher } : {}),
         });
         const unsubscribeEphemeral = session.onEphemeral((message) => {
             const event = message.event;
@@ -2326,7 +2425,7 @@ export class SessionChatController {
             if (args.signal.aborted) {
                 throw new Error("diff review start aborted");
             }
-            await bridge.launchTool(args.diffTool);
+            await bridge.launchTool(this.diffToolLauncher ?? launchBuiltInDiffTool);
         }
         catch (error) {
             unsubscribeEphemeral();