@kolisachint/hoocode-agent 0.4.103 → 0.4.105

package/dist/modes/interactive/interactive-mode.js

@@ -9,7 +9,7 @@ import * as path from "node:path";
 import { getProviders, } from "@kolisachint/hoocode-ai";
 import { CombinedAutocompleteProvider, Container, fuzzyFilter, getCapabilities, hyperlink, Loader, Markdown, matchesKey, ProcessTerminal, Spacer, setKeybindings, Text, TruncatedText, TUI, } from "@kolisachint/hoocode-tui";
 import { spawn, spawnSync } from "child_process";
-import { APP_NAME, APP_TITLE, getAgentDir, getAuthPath, getDocsPath, resolveVoicetoolsBin, VERSION, } from "../../config.js";
+import { APP_NAME, APP_TITLE, getAgentDir, getAuthPath, getDocsPath, VERSION } from "../../config.js";
 import { parseSkillBlock } from "../../core/agent-session.js";
 import { FooterDataProvider } from "../../core/footer-data-provider.js";
 import { KeybindingsManager } from "../../core/keybindings.js";
@@ -59,8 +59,9 @@ import { ToolExecutionComponent } from "./components/tool-execution.js";
 import { TreeSelectorComponent } from "./components/tree-selector.js";
 import { UserMessageComponent } from "./components/user-message.js";
 import { UserMessageSelectorComponent } from "./components/user-message-selector.js";
+import { VoicePanel } from "./components/voice-panel.js";
 import { getAvailableThemes, getAvailableThemesWithPaths, getEditorTheme, getMarkdownTheme, getThemeByName, initTheme, onThemeChange, setRegisteredThemes, setTheme, setThemeInstance, stopThemeWatcher, Theme, theme, } from "./theme/theme.js";
-import { startVoiceTranscribe } from "./voice-transcribe.js";
+import { startVoiceTranscribe, VoiceDaemon } from "./voice-transcribe.js";
 function isExpandable(obj) {
     return typeof obj === "object" && obj !== null && "setExpanded" in obj && typeof obj.setExpanded === "function";
 }
@@ -84,6 +85,12 @@ function isDeadTerminalError(error) {
     const code = error.code;
     return code !== undefined && DEAD_TERMINAL_ERROR_CODES.has(code);
 }
+// Mirrors voicetools' own default trailing-silence window (see `--silence-ms` /
+// VOICE_SILENCE_MS upstream). Only used to drive the cosmetic countdown in the
+// panel; the actual cutoff decision is made by the binary, not this timer.
+const VOICE_SILENCE_MS = 600;
+const VOICE_UNAVAILABLE_MESSAGE = "Voice input failed: voicetools binary unavailable and could not be downloaded. " +
+    "Install it, set VOICETOOLS_BIN, or ensure a published release exists for this platform.";
 const ANTHROPIC_SUBSCRIPTION_AUTH_WARNING = "Anthropic subscription auth: billed per token as extra usage, not plan limits.";
 function isAnthropicSubscriptionAuthKey(apiKey) {
     return typeof apiKey === "string" && apiKey.startsWith("sk-ant-oat");
@@ -116,6 +123,18 @@ export class InteractiveMode {
     editor;
     editorComponentFactory;
     voiceSession;
+    // Persistent `voicetools serve` process, once probed successfully. Stays alive
+    // (and warm) across captures for the life of the interactive session.
+    voiceDaemon;
+    // Sticky for the session: set once `serve` is confirmed unsupported (old binary)
+    // so later presses skip straight to the per-press `transcribe` fallback.
+    voiceDaemonUnsupported = false;
+    /** True while the voicetools binary is being resolved/downloaded before a session starts. */
+    voiceStarting = false;
+    /** True while a capture (daemon or legacy) is in flight. */
+    voiceActive = false;
+    /** The live multi-line status panel for the current capture (undefined when idle). */
+    voicePanel;
     autocompleteProvider;
     autocompleteProviderWrappers = [];
     fdPath;
@@ -2760,39 +2779,256 @@ export class InteractiveMode {
      * we update the previous status line instead of appending new ones to avoid log spam.
      */
     /**
-     * Toggle voice-to-text capture. First press spawns `voicetools`, streams
-     * decoded segments into the editor via bracketed paste, and reports state
-     * through the status line. Pressing the shortcut again stops early.
+     * Toggle voice-to-text capture.
+     *
+     * Prefers a persistent `voicetools serve` daemon: the first press probes for
+     * support and (if found) loads models once, showing a "warming up" spinner;
+     * every later press reuses the already-warm daemon and jumps straight to
+     * Listening. Binaries without `serve` support fall back to spawning
+     * `voicetools transcribe` per press, same as before. Pressing the shortcut
+     * again while listening (or while still warming up) cancels.
      */
     toggleVoiceTranscribe() {
-        if (this.voiceSession?.running) {
-            this.voiceSession.stop();
+        if (this.voiceActive) {
+            if (this.voiceDaemon?.isReady) {
+                this.voiceDaemon.cancel();
+            }
+            else {
+                this.voiceSession?.stop();
+            }
             this.voiceSession = undefined;
-            this.showStatus("Voice input cancelled");
+            this.voiceActive = false;
+            this.resetVoiceUI();
+            return;
+        }
+        if (this.voiceStarting) {
+            // A second press while resolving/warming up: honour the cancel. Any
+            // daemon that finishes loading afterwards is kept warm for next time.
+            this.voiceStarting = false;
+            this.resetVoiceUI();
+            return;
+        }
+        if (this.voiceDaemonUnsupported) {
+            this.voiceStarting = true;
+            this.showVoiceWarming("Starting voice input...");
+            void this.resolveVoiceBin()
+                .then((bin) => {
+                if (!this.voiceStarting)
+                    return;
+                this.voiceStarting = false;
+                if (!bin) {
+                    this.resetVoiceUI();
+                    this.showError(VOICE_UNAVAILABLE_MESSAGE);
+                    return;
+                }
+                this.beginLegacyVoiceCapture(bin);
+            })
+                .catch((err) => {
+                this.voiceStarting = false;
+                this.resetVoiceUI();
+                this.showError(`Voice input failed: ${err instanceof Error ? err.message : String(err)}`);
+            });
+            return;
+        }
+        if (this.voiceDaemon?.isReady) {
+            this.beginDaemonVoiceCapture();
             return;
         }
-        const statusLabels = {
-            recording: "Recording... speak now (press again to cancel)",
-            transcribing: "Transcribing...",
-            done: "Voice input done",
+        // No daemon yet: resolve the binary, then probe for `serve` support by
+        // spawning it. `VoiceDaemon.spawn` doubles as the probe: it resolves to a
+        // live daemon once READY arrives, to "unsupported" if the process exits
+        // with no output at all (an old binary rejecting the unrecognized `serve`
+        // subcommand), or to "error" if it printed a real ERROR first (e.g. no
+        // model installed yet) — already surfaced via onError, so that case skips
+        // the legacy fallback (it would just hit the same error) but leaves
+        // daemon mode available to retry on the next press.
+        this.voiceStarting = true;
+        this.showVoiceWarming("Warming up voice input...");
+        void this.resolveVoiceBin()
+            .then(async (bin) => {
+            if (!bin) {
+                this.voiceStarting = false;
+                this.resetVoiceUI();
+                this.showError(VOICE_UNAVAILABLE_MESSAGE);
+                return;
+            }
+            const result = await VoiceDaemon.spawn(bin, this.buildVoiceDaemonHandlers());
+            if (!result.ok) {
+                if (result.reason === "unsupported") {
+                    this.voiceDaemonUnsupported = true;
+                    if (!this.voiceStarting) {
+                        this.resetVoiceUI();
+                        return;
+                    }
+                    this.voiceStarting = false;
+                    this.beginLegacyVoiceCapture(bin);
+                    return;
+                }
+                this.voiceStarting = false;
+                this.resetVoiceUI();
+                return;
+            }
+            this.voiceDaemon = result.daemon;
+            if (!this.voiceStarting) {
+                // Cancelled while warming up: keep the loaded daemon warm for next time.
+                this.resetVoiceUI();
+                return;
+            }
+            this.voiceStarting = false;
+            this.beginDaemonVoiceCapture();
+        })
+            .catch((err) => {
+            this.voiceStarting = false;
+            this.resetVoiceUI();
+            this.showError(`Voice input failed: ${err instanceof Error ? err.message : String(err)}`);
+        });
+    }
+    /**
+     * Build the (stable, reused-across-captures) handlers for the daemon.
+     *
+     * Every handler bails out once `voiceActive` is false: CANCEL is a soft
+     * request (unlike the legacy path's `proc.kill()`, it doesn't sever the
+     * pipe), so the daemon can still have a trailing PARTIAL/FINAL/DONE for the
+     * just-cancelled capture in flight when the user presses cancel. Without
+     * this guard that stale text would land in the editor after the panel had
+     * already collapsed.
+     *
+     * v0.1.4 serve streams `PARTIAL <full growing hypothesis>` (live preview,
+     * never committed) and ends with a single `FINAL <complete text>` (the one
+     * commit to the editor). `SEGMENT` is still handled for the legacy
+     * transcribe path and any binary that streams committed chunks directly.
+     */
+    buildVoiceDaemonHandlers() {
+        return {
+            onSegment: (text) => {
+                if (!this.voiceActive)
+                    return;
+                this.commitVoiceText(text);
+            },
+            onPartial: (text) => {
+                if (!this.voiceActive)
+                    return;
+                this.voicePanel?.setPartial(text);
+            },
+            onFinal: (text) => {
+                if (!this.voiceActive)
+                    return;
+                this.commitVoiceText(text);
+            },
+            onStatus: (status) => {
+                if (!this.voiceActive)
+                    return;
+                if (status === "done") {
+                    this.voiceActive = false;
+                    this.resetVoiceUI();
+                    return;
+                }
+                if (status === "transcribing") {
+                    this.voicePanel?.setTranscribing();
+                }
+                else if (status === "listening") {
+                    this.voicePanel?.startListening();
+                }
+            },
+            onLevel: (rms) => {
+                if (!this.voiceActive)
+                    return;
+                this.voicePanel?.pushLevel(rms);
+            },
+            onPhase: (phase) => {
+                if (!this.voiceActive)
+                    return;
+                if (phase === "silence") {
+                    this.voicePanel?.beginSilence(VOICE_SILENCE_MS);
+                }
+                else {
+                    this.voicePanel?.endSilence();
+                }
+            },
+            onError: (message) => {
+                this.voiceActive = false;
+                this.resetVoiceUI();
+                this.showError(`Voice input failed: ${message}`);
+            },
+            onCrash: (message) => {
+                this.voiceActive = false;
+                this.voiceDaemon = undefined;
+                this.resetVoiceUI();
+                this.showError(`Voice input daemon crashed: ${message}. It will restart on next use.`);
+            },
         };
-        this.voiceSession = startVoiceTranscribe(resolveVoicetoolsBin(), {
+    }
+    /** Inject decoded text into the editor via bracketed paste (with a trailing space). */
+    commitVoiceText(text) {
+        this.editor.handleInput(`\x1b[200~${text} \x1b[201~`);
+    }
+    beginDaemonVoiceCapture() {
+        if (!this.voiceDaemon?.isReady)
+            return;
+        this.voiceActive = true;
+        this.showVoicePanel().startListening();
+        this.voiceDaemon.startCapture();
+    }
+    beginLegacyVoiceCapture(bin) {
+        this.voiceActive = true;
+        const panel = this.showVoicePanel();
+        panel.startListening();
+        this.voiceSession = startVoiceTranscribe(bin, {
             onStatus: (status) => {
-                this.showStatus(statusLabels[status] ?? status);
                 if (status === "done") {
+                    this.voiceActive = false;
                     this.voiceSession = undefined;
+                    this.resetVoiceUI();
+                    return;
                 }
+                // Old binaries emit no LEVEL/PARTIAL, so the panel shows a spinner
+                // for the batch phases; committed words still stream into the editor.
+                if (status === "transcribing")
+                    panel.setTranscribing();
             },
             onSegment: (text) => {
-                // Inject via bracketed paste so the editor treats it as pasted text.
-                this.editor.handleInput(`\x1b[200~${text} \x1b[201~`);
+                this.commitVoiceText(text);
             },
             onError: (message) => {
+                this.voiceActive = false;
                 this.voiceSession = undefined;
+                this.resetVoiceUI();
                 this.showError(`Voice input failed: ${message}`);
             },
         });
     }
+    /** Resolve the `voicetools` binary path. Prefers an explicit VOICETOOLS_BIN
+     * override, otherwise resolves via the managed tools manager (bin dir / PATH /
+     * download from the published release). Returns undefined when unavailable.
+     */
+    async resolveVoiceBin() {
+        const override = process.env.VOICETOOLS_BIN?.trim();
+        if (override)
+            return override;
+        return ensureTool("voicetools", true);
+    }
+    /** Create (or reuse) the voice panel and mount it in the status container. */
+    showVoicePanel() {
+        if (!this.voicePanel) {
+            this.statusContainer.clear();
+            this.voicePanel = new VoicePanel(this.ui, keyHint("app.input.voiceTranscribe", "cancel"));
+            this.statusContainer.addChild(this.voicePanel);
+        }
+        this.ui.requestRender();
+        return this.voicePanel;
+    }
+    showVoiceWarming(message) {
+        this.showVoicePanel().setWarming(message);
+    }
+    /** Collapse the voice panel back to nothing (idle) and stop its animation. */
+    resetVoiceUI() {
+        if (this.voicePanel) {
+            this.voicePanel.dispose();
+            this.voicePanel = undefined;
+        }
+        this.statusContainer.clear();
+        this.ui.requestRender();
+    }
     showStatus(message) {
         const children = this.chatContainer.children;
         const last = children.length > 0 ? children[children.length - 1] : undefined;
@@ -4471,6 +4707,14 @@ export class InteractiveMode {
             this.voiceSession.stop();
             this.voiceSession = undefined;
         }
+        this.voiceDaemon?.shutdown();
+        this.voiceDaemon = undefined;
+        if (this.voicePanel) {
+            this.voicePanel.dispose();
+            this.voicePanel = undefined;
+        }
+        this.voiceActive = false;
+        this.voiceStarting = false;
         if (this.settingsManager.getShowTerminalProgress()) {
             this.ui.terminal.setProgress(false);
         }