dsh-vidi 0.3.1 → 0.4.1

package/README.md

@@ -8,16 +8,16 @@ A video tool for DeepSeek Harness: watch live streams, videos and podcasts with
 
 - **Opens anything**: YouTube, Bilibili and other sites yt-dlp supports, Xiaoyuzhou podcast episodes, m3u8 / rtmp / srt streams, local files (drop them in), or a stream pushed from OBS.
 - **Subtitles start by themselves** when a video opens; pause and resume any time. One or two languages (a main line and a second line); the site's own subtitles can be shown too, or used as a reference.
-- **Two paces**: real-time shows lines as they are heard; polished arrives a little later with misheard words fixed from context.
+- **Two paces**: real-time shows lines as they are heard; polished arrives a little later with misheard words fixed from context. In the subtitle list, polished lines carry a small tick; lines not polished yet are lighter.
 - **The subtitle model looks at the video**: polishing uses the page's title, description and comments, and — when the chosen model reads images — a frame of that stretch. It also keeps a glossary of names and terms for the video.
 - **Chat with DeepSeek beside the video**: grab this moment, cut this part, summarise, download the original — all in one conversation, with files saved in the video's folder.
 - **Restream**: push the video with its subtitles drawn into the picture to Bilibili, Douyin, Kuaishou, WeChat Channels, Douyu, Huya, YouTube, Twitch or any RTMP/SRT address — several at once; each platform connects and reconnects on its own.
-- **Export**: SRT / VTT subtitles, transcripts by line or paragraph, clips with subtitle tracks.
+- **Download and export**: the original or a video with subtitle tracks, just the sound (M4A), SRT / VTT subtitles, transcripts by line or paragraph.
 - English and Chinese interface; works on phones.
 
 ## Install
 
-In DSH, open **Plugins → Install third-party plugin** and enter `dsh-vidi`. A **Video** button appears in the sidebar; pasting a video link in a chat also offers to open it.
+In DSH, open **Plugins → Install third-party plugin** and enter `dsh-vidi`. A **Vidi video** button appears in the sidebar (the video page takes the main area like a chat, and the sidebar opens and closes as usual); pasting a video link in a chat also offers to open it.
 
 Needs DSH 0.2.2 or later.
 

package/README.zh-CN.md

@@ -8,16 +8,16 @@ DeepSeek Harness 里的视频工具：看直播、视频和播客，同时出任
 
 - **打开任何来源**：YouTube、B 站等 yt-dlp 支持的网站，小宇宙节目，m3u8 / rtmp / srt 直播流，本地文件（直接拖进来），也可以用 OBS 推流过来。
 - **自动出字幕**：视频一打开就开始，可以随时暂停、继续。单语、双语都行，主字幕和副字幕各选一种语言；网站自带的字幕也能直接显示，或者作为参考。
-- **两种节奏**：实时模式边播边出；精修模式晚一点出，但会结合上下文把听错的词改对。
+- **两种节奏**：实时模式边播边出；精修模式晚一点出，但会结合上下文把听错的词改对。字幕列表里精修过的句子带一个小勾，还没精修的颜色浅一些。
 - **字幕模型会看画面**：精修时结合视频页面的标题、简介、评论，所选模型能看图时还会带上那一段的画面，顺便整理这期视频的词表（人名、术语）。
 - **在视频旁边和 DeepSeek 对话**：截取此刻、剪下这段、总结要点、下载原片，都在同一个会话里完成，文件存在视频目录下。
 - **转播**：把视频连同画面里的字幕推到 B站、抖音、快手、视频号、斗鱼、虎牙、YouTube、Twitch 或任意 RTMP/SRT 地址，可以同时推多个平台；每个平台各自连接、断了各自重连。
-- **导出**：SRT / VTT 字幕，逐句或按段落的文字稿，带字幕轨的片段。
+- **下载和导出**：原片或带字幕轨的视频、只要声音的音频（M4A）、SRT / VTT 字幕，逐句或按段落的文字稿。
 - 中英文界面，手机上也能用。
 
 ## 安装
 
-在 DSH 里打开 **插件 → 安装第三方插件**，填 `dsh-vidi`。装好后侧栏会出现 **视频** 按钮；在对话里贴视频链接，输入框上方也会出现「在视频里打开」。
+在 DSH 里打开 **插件 → 安装第三方插件**，填 `dsh-vidi`。装好后侧栏会出现 **Vidi 视频** 按钮，视频页和对话一样占主区域，侧栏照常收起、展开；在对话里贴视频链接，输入框上方也会出现「在视频里打开」。
 
 需要 DSH 0.2.2 或更新的版本。
 

package/lib/client.js

@@ -247,6 +247,14 @@ const Icon = {
 		...P,
 		children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)("path", { d: "M3 7a2 2 0 0 1 2-2h4l2 2h8a2 2 0 0 1 2 2v8a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2z" })
 	}),
+	check: () => /* @__PURE__ */ (0, react_jsx_runtime.jsx)("svg", {
+		viewBox: "0 0 24 24",
+		width: "12",
+		height: "12",
+		...P,
+		strokeWidth: 2.6,
+		children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)("path", { d: "M5 12.5l4.5 4.5L19 7.5" })
+	}),
 	plus: () => /* @__PURE__ */ (0, react_jsx_runtime.jsx)("svg", {
 		viewBox: "0 0 24 24",
 		width: "14",
@@ -33299,9 +33307,12 @@ const Player = (0, react.forwardRef)(function Player(props, ref) {
 	const idleTimer = (0, react.useRef)();
 	/** Where playback was, so a new source (quality switch) continues from there. */
 	const lastTime = (0, react.useRef)(0);
-	(0, react.useImperativeHandle)(ref, () => ({ seek: (time) => {
-		if (video.current) video.current.currentTime = Math.max(0, time);
-	} }), []);
+	(0, react.useImperativeHandle)(ref, () => ({
+		seek: (time) => {
+			if (video.current) video.current.currentTime = Math.max(0, time);
+		},
+		closeMenu: () => setMenu(null)
+	}), []);
 	(0, react.useEffect)(() => {
 		const el = video.current;
 		if (!src || !el) return;
@@ -33836,6 +33847,11 @@ const Player = (0, react.forwardRef)(function Player(props, ref) {
 									className: "txt",
 									children: m.label
 								}),
+								m.id === "cc" && ccOn && current?.polished && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+									className: "dshl-okmark",
+									title: t("player.polished"),
+									children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.check, {})
+								}),
 								m.label && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", { className: "caret" })
 							]
 						}, m.id)),
@@ -33890,9 +33906,10 @@ function registry() {
 * itself stays in this package; companions only add.
 */
 const vidiUi = {
-	version: 1,
+	version: 2,
 	panels: registry(),
-	playerMenus: registry()
+	playerMenus: registry(),
+	benches: registry()
 };
 
 // src/client/RestreamPanel.tsx
@@ -34589,7 +34606,7 @@ const siteName = (url) => /youtu\.?be/.test(url) ? "YouTube" : /bilibili|b23\.tv
 const media = (dirId, file) => `/live/media/${encodeURIComponent(dirId)}/${encodeURIComponent(file)}`;
 function LiveView(props) {
 	const { sessionId, t, chatFace, chatSource, onClose, pending, onPendingDone } = props;
-	const pageMode = !!onClose;
+	const pageMode = !!onClose || !!props.page;
 	const [input, setInput] = (0, react.useState)("");
 	const [lines, setLinesRaw] = (0, react.useState)(() => store.get("lines", [browserLang(), "src"]));
 	const setLines = (l) => {
@@ -34632,6 +34649,8 @@ function LiveView(props) {
 	const [tab, setTab] = (0, react.useState)("transcript");
 	const [query, setQuery] = (0, react.useState)("");
 	const [trMode, setTrMode] = (0, react.useState)("lines");
+	/** Video page: the workbench shown in place of the chat (companion plugins), if any. */
+	const [bench, setBench] = (0, react.useState)(null);
 	/** Reading view: which languages to show ('src' = original); empty = follow the subtitle lines. */
 	const [readPick, setReadPick] = (0, react.useState)([]);
 	const [now, setNow] = (0, react.useState)(0);
@@ -34846,19 +34865,6 @@ function LiveView(props) {
 	const stop = async () => {
 		if (job) await post("stop", { jobId: job.id }).catch(() => {});
 	};
-	const downloadOnly = async () => {
-		const c = classify(input);
-		if (c?.source !== "url") return;
-		try {
-			await post("download", {
-				conversationId: sessionId,
-				url: c.url
-			});
-			refreshTasks();
-		} catch (e) {
-			setError(e.message);
-		}
-	};
 	const onFile = async (file) => {
 		setError("");
 		setUploading(0);
@@ -35036,7 +35042,6 @@ function LiveView(props) {
 	const statusDetail = job?.status === "running" && !job.message && !job.vod ? CLOCK(Math.max(0, (tick - job.startedAt) / 1e3)) : "";
 	const missing = deps.filter((d) => !d.ok && (d.name === "ffmpeg" || d.name === "yt-dlp" || (engine || "local") === "local" && d.name.startsWith("whisper")));
 	const needsModel = !hasTranslator && shownLines.some((l) => l !== "src");
-	const canDownloadLink = classify(input)?.source === "url" && !/^(rtmps?|srt|rtsp|udp):/i.test(classify(input).url);
 	const hasMain = !!job && !(job.status === "error" && !job.files.length);
 	const emptyBox = /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 		className: "dshl-drop",
@@ -35112,6 +35117,8 @@ function LiveView(props) {
 	const inline = pageMode && tab === "transcript" && !!job;
 	const extPanels = (0, react.useSyncExternalStore)(vidiUi.panels.subscribe, vidiUi.panels.get);
 	const extMenus = (0, react.useSyncExternalStore)(vidiUi.playerMenus.subscribe, vidiUi.playerMenus.get);
+	const extBenches = (0, react.useSyncExternalStore)(vidiUi.benches.subscribe, vidiUi.benches.get);
+	const activeBench = pageMode && job ? extBenches.find((b) => b.id === bench) : void 0;
 	const view = job ? {
 		job,
 		cues,
@@ -35120,7 +35127,9 @@ function LiveView(props) {
 		seek: (s) => player.current?.seek(s),
 		style: subStyle,
 		lines: shownLines,
-		ui
+		ui,
+		bench: activeBench?.id ?? null,
+		openBench: setBench
 	} : void 0;
 	/** Subtitle lines and glossary: under the player on the video page, beside it in a conversation. */
 	const transcriptBox = /* @__PURE__ */ (0, react_jsx_runtime.jsxs)(react_jsx_runtime.Fragment, { children: [
@@ -35188,21 +35197,30 @@ function LiveView(props) {
 						className: "dshl-muted pad",
 						children: t("tr.empty")
 					}), shown.map((c) => /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("button", {
-						className: `dshl-line ${c.id === activeId ? "active" : ""} ${c.polished ? "" : "raw"}`,
+						className: `dshl-line ${c.id === activeId ? "active" : ""} ${c.polished ? "ok" : "raw"}`,
 						onClick: () => player.current?.seek(c.start),
-						children: [/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-							className: "ts",
-							children: CLOCK(c.start)
-						}), /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-							className: "txt",
-							children: shownLines.map((k, i) => {
-								const v = lineText(c, k);
-								return v ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-									className: i === 0 ? "main" : "sub",
-									children: v
-								}, k) : null;
+						title: c.polished ? t("tr.polished") : t("tr.raw"),
+						children: [
+							/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+								className: "ts",
+								children: CLOCK(c.start)
+							}),
+							/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+								className: "txt",
+								children: shownLines.map((k, i) => {
+									const v = lineText(c, k);
+									return v ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+										className: i === 0 ? "main" : "sub",
+										children: v
+									}, k) : null;
+								})
+							}),
+							/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
+								className: "mark",
+								"aria-hidden": "true",
+								children: c.polished && /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.check, {})
 							})
-						})]
+						]
 					}, c.id))]
 				}) : /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 					className: "dshl-read",
@@ -35244,7 +35262,7 @@ function LiveView(props) {
 		})
 	] });
 	return /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
-		className: `dshl ${theater ? "theater" : ""} ${pageMode ? "page" : ""}`,
+		className: `dshl ${theater && !activeBench ? "theater" : ""} ${pageMode ? "page" : ""} ${activeBench ? "bench" : ""}`,
 		"data-dsh-own-drop": "dsh-vidi",
 		onDragEnter: (e) => {
 			if (e.dataTransfer.types.includes("Files")) e.stopPropagation();
@@ -35315,12 +35333,6 @@ function LiveView(props) {
 								})
 							]
 						}),
-						canDownloadLink && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
-							className: "dshl-btn icon",
-							onClick: downloadOnly,
-							title: t("in.download.tip"),
-							children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.download, {})
-						}),
 						/* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
 							className: "dshl-btn primary",
 							onClick: () => start(),
@@ -35376,13 +35388,6 @@ function LiveView(props) {
 					children: t("st.retry")
 				})]
 			}),
-			tasks.length > 0 && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
-				className: "dshl-tasks",
-				children: tasks.filter((x) => x.kind === "original").slice(0, 3).map((x) => /* @__PURE__ */ (0, react_jsx_runtime.jsx)(TaskRow, {
-					task: x,
-					t
-				}, x.id))
-			}),
 			!hasMain && !pageMode ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("section", {
 				className: "dshl-empty",
 				children: emptyBox
@@ -35474,7 +35479,7 @@ function LiveView(props) {
 									id: `ext:${m.id}`,
 									icon: m.icon,
 									title: m.title(ui),
-									content: view ? m.render(view, () => {}) : null
+									content: view ? m.render(view, () => player.current?.closeMenu()) : null
 								})),
 								{
 									id: "download",
@@ -35528,7 +35533,10 @@ function LiveView(props) {
 					] })
 				}), /* @__PURE__ */ (0, react_jsx_runtime.jsx)("aside", {
 					className: "dshl-side",
-					children: pageMode ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)(ChatPanel, {
+					children: activeBench && view ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
+						className: "dshl-benchbox",
+						children: activeBench.render(view)
+					}) : pageMode ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)(ChatPanel, {
 						face: chatFace,
 						source: chatSource,
 						now,
@@ -35729,36 +35737,6 @@ function UsagePill({ usage, t, ui }) {
 		})]
 	});
 }
-function TaskRow({ task, t }) {
-	return /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
-		className: `dshl-task ${task.status}`,
-		children: [
-			/* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.download, {}),
-			/* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-				className: "name",
-				children: task.title ?? task.file
-			}),
-			task.status === "running" && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-				className: "bar",
-				children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)("i", { style: { width: `${task.pct}%` } })
-			}),
-			task.status === "running" && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-				className: "pct",
-				children: t("dl.pct").replace("{p}", String(task.pct))
-			}),
-			task.status === "done" && task.file && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("a", {
-				href: media(task.dirId, task.file),
-				download: true,
-				children: task.file
-			}),
-			task.status === "error" && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("span", {
-				className: "bad",
-				title: task.error,
-				children: t("dl.failed")
-			})
-		]
-	});
-}
 /** Download: choose what (video or subtitles), then one button that says exactly what you get. */
 function DownloadPanel(props) {
 	const { job, lines, tasks, sessionId, onTask, langName, siteChips, t } = props;
@@ -35781,6 +35759,13 @@ function DownloadPanel(props) {
 	].filter(Boolean).join(" · ");
 	const saveAs = `${title}（${[qLabel, withSubs ? subsLabel : ""].filter(Boolean).join("，")}）.mp4`;
 	const task = tasks.find((x) => x.kind === "video" && x.saveAs === saveAs);
+	const audioAs = `${title}（${t("dl.tab.audio")}）`;
+	const audioTask = tasks.find((x) => x.kind === "audio" && x.saveAs === audioAs);
+	const goAudio = () => post("audio", {
+		conversationId: sessionId,
+		jobId: job.id,
+		saveAs: audioAs
+	}).then(onTask).catch(() => onTask());
 	const go = () => post("video", {
 		conversationId: sessionId,
 		jobId: job.id,
@@ -35800,15 +35785,23 @@ function DownloadPanel(props) {
 		className: "dshl-dl",
 		children: [/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 			className: "dshl-seg wide",
-			children: [/* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
-				className: tab === "video" ? "on" : "",
-				onClick: () => setTab("video"),
-				children: t("dl.tab.video")
-			}), /* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
-				className: tab === "subs" ? "on" : "",
-				onClick: () => setTab("subs"),
-				children: t("dl.tab.subs")
-			})]
+			children: [
+				/* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
+					className: tab === "video" ? "on" : "",
+					onClick: () => setTab("video"),
+					children: t("dl.tab.video")
+				}),
+				/* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
+					className: tab === "audio" ? "on" : "",
+					onClick: () => setTab("audio"),
+					children: t("dl.tab.audio")
+				}),
+				/* @__PURE__ */ (0, react_jsx_runtime.jsx)("button", {
+					className: tab === "subs" ? "on" : "",
+					onClick: () => setTab("subs"),
+					children: t("dl.tab.subs")
+				})
+			]
 		}), tab === "video" ? /* @__PURE__ */ (0, react_jsx_runtime.jsxs)(react_jsx_runtime.Fragment, { children: [
 			/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 				className: "grp",
@@ -35856,6 +35849,37 @@ function DownloadPanel(props) {
 				className: "err",
 				children: task.error
 			})
+		] }) : tab === "audio" ? /* @__PURE__ */ (0, react_jsx_runtime.jsxs)(react_jsx_runtime.Fragment, { children: [
+			/* @__PURE__ */ (0, react_jsx_runtime.jsx)("p", {
+				className: "dshl-dl-note",
+				children: t("dl.audio.d")
+			}),
+			audioTask?.status === "running" ? /* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
+				className: "dshl-btn block progress",
+				style: { ["--p"]: `${audioTask.pct}%` },
+				children: audioTask.stage === "mux" ? t("dl.stage.audio") : t("dl.stage.audioDl").replace("{p}", String(audioTask.pct))
+			}) : audioTask?.status === "done" && audioTask.file ? /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("a", {
+				className: "dshl-btn block primary",
+				href: media(audioTask.dirId, audioTask.file),
+				download: `${audioAs}${audioTask.file.slice(audioTask.file.lastIndexOf("."))}`,
+				children: [
+					/* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.download, {}),
+					" ",
+					t("dl.save")
+				]
+			}) : /* @__PURE__ */ (0, react_jsx_runtime.jsxs)("button", {
+				className: "dshl-btn block primary",
+				onClick: () => void goAudio(),
+				children: [
+					/* @__PURE__ */ (0, react_jsx_runtime.jsx)(Icon.download, {}),
+					" ",
+					t("dl.go").replace("{what}", t("dl.audio.what"))
+				]
+			}),
+			audioTask?.status === "error" && /* @__PURE__ */ (0, react_jsx_runtime.jsx)("p", {
+				className: "err",
+				children: audioTask.error
+			})
 		] }) : /* @__PURE__ */ (0, react_jsx_runtime.jsxs)(react_jsx_runtime.Fragment, { children: [
 			/* @__PURE__ */ (0, react_jsx_runtime.jsxs)("div", {
 				className: "grp",
@@ -36137,8 +36161,6 @@ const en = {
 	"st.ready": "Ready",
 	"in.start": "Start",
 	"in.stop": "Stop",
-	"in.download": "Download",
-	"in.download.tip": "Download the video only",
 	"in.file": "File",
 	"in.file.tip": "Open a video or audio file",
 	"in.push": "Push",
@@ -36211,6 +36233,11 @@ const en = {
 	"subs.progressLive": "Generating subtitles · {t}",
 	"dl.tab.video": "Video",
 	"dl.tab.subs": "Subtitles",
+	"dl.tab.audio": "Audio",
+	"dl.audio.d": "Just the sound, for podcasts, listening on the go or other transcription tools.",
+	"dl.audio.what": "Audio",
+	"dl.stage.audioDl": "Downloading audio {p}%",
+	"dl.stage.audio": "Taking out the sound…",
 	"dl.quality": "Picture",
 	"dl.q.original": "Original",
 	"dl.q.original.d": "Best quality from the site",
@@ -36260,6 +36287,9 @@ const en = {
 	"chat.p.summary": "Summarise the key points of this video so far from the subtitles.",
 	"chat.p.download": "Download the original of this video.",
 	"page.back": "Back to chats",
+	"tr.polished": "Polished",
+	"tr.raw": "First pass; polishing comes a little later",
+	"player.polished": "This line is polished",
 	"page.open": "Vidi video",
 	"page.failed": "The video page could not open: {e}",
 	"side.subs": "Subtitles {p}%",
@@ -36453,8 +36483,6 @@ const zh = {
 	"st.ready": "已就绪",
 	"in.start": "开始",
 	"in.stop": "停止",
-	"in.download": "下载",
-	"in.download.tip": "只下载视频",
 	"in.file": "文件",
 	"in.file.tip": "打开视频或音频文件",
 	"in.push": "推流",
@@ -36527,6 +36555,11 @@ const zh = {
 	"subs.progressLive": "正在生成字幕 · 已到 {t}",
 	"dl.tab.video": "视频",
 	"dl.tab.subs": "字幕",
+	"dl.tab.audio": "音频",
+	"dl.audio.d": "只要声音，适合当播客听，或交给别的转写工具。",
+	"dl.audio.what": "音频",
+	"dl.stage.audioDl": "正在下载音频 {p}%",
+	"dl.stage.audio": "正在提取声音…",
 	"dl.quality": "画质",
 	"dl.q.original": "原片",
 	"dl.q.original.d": "网站最高画质",
@@ -36576,6 +36609,9 @@ const zh = {
 	"chat.p.summary": "根据字幕，总结这个视频到目前为止的要点。",
 	"chat.p.download": "下载这个视频的原片。",
 	"page.back": "返回对话",
+	"tr.polished": "已精修",
+	"tr.raw": "第一轮识别，稍后精修",
+	"player.polished": "这句字幕已精修",
 	"page.open": "Vidi 视频",
 	"page.failed": "视频页打不开：{e}",
 	"side.subs": "字幕 {p}%",
@@ -36836,9 +36872,22 @@ const CSS = `
    with them, and its empty space moves the window like the app's own top strip. Full screen and
    the web page need none of this. */
 .dshl-pagewrap { top: var(--dsh-frame-chrome-top, 0px); }
-html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top { padding: 4px 16px 8px 84px; -webkit-app-region: drag; app-region: drag; }
-html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top :is(button, input, label, a, select, .dshl-pick) { -webkit-app-region: no-drag; app-region: no-drag; }
-@media (max-width: 600px) { html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top { padding-left: 80px; } }
+html[data-platform='darwin']:not([data-fullscreen]) .dshl-pagewrap .dshl.page .dshl-top { padding: 4px 16px 8px 84px; }
+@media (max-width: 600px) { html[data-platform='darwin']:not([data-fullscreen]) .dshl-pagewrap .dshl.page .dshl-top { padding-left: 80px; } }
+/* Newer DS-Harness: the page is a main panel beside the app's sidebar. When the sidebar is closed
+   the app puts its own controls (traffic lights, sidebar toggle, new chat) at the top-left and
+   publishes how much room they take: the top bar starts after them. */
+.dshl-panelwrap { height: 100%; min-height: 0; background: var(--dsw-alias-bg-base, #fff); }
+.dshl-panelwrap .dshl.page .dshl-top { padding-left: max(20px, var(--dsh-frame-leading-clearance, 0px)); }
+/* Web page in a narrow window (sidebar as a drawer): the app gives its opener a row of its own above
+   the page, so the bar starts at the edge. The desktop app keeps the opener on the bar's row instead
+   (beside the traffic lights on macOS), and the clearance above leaves room for it. */
+html:not([data-platform]) [data-sidebar-drawer] .dshl-panelwrap .dshl.page .dshl-top { padding-left: 12px; }
+/* macOS: the app's window controls (traffic lights, sidebar toggle, new chat) are centred at y 25 in
+   every state, full screen included; the bar's 40px row is centred on the same line. */
+html[data-platform='darwin'] .dshl-panelwrap .dshl.page .dshl-top { padding-top: 5px; padding-bottom: 8px; }
+html[data-platform='darwin'] .dshl.page .dshl-top { -webkit-app-region: drag; app-region: drag; }
+html[data-platform='darwin'] .dshl.page .dshl-top :is(button, input, label, a, select, .dshl-pick) { -webkit-app-region: no-drag; app-region: no-drag; }
 .dshl.page { display: flex; flex-direction: column; height: 100%; overflow: hidden; }
 .dshl.page .dshl-main { flex: 1; min-height: 0; grid-template-columns: minmax(0, 1fr) clamp(320px, 30vw, 420px); align-items: stretch; padding-bottom: 16px; }
 /* Video page: player on top, then the status row, then subtitles/glossary filling what is left; chat on the right. */
@@ -36846,6 +36895,9 @@ html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top :is(but
 .dshl.page .dshl-col > .dshl-drop { flex: 1; aspect-ratio: auto; min-height: 0; }
 .dshl.page .dshl-stage { flex: none; width: 100%; max-height: max(240px, calc(100cqh - 270px)); }
 .dshl.page .dshl-side { position: static; height: 100%; max-height: none; }
+/* Workbench open: the video and its subtitles keep a column on the left; the bench takes the rest. */
+.dshl.page.bench .dshl-main { grid-template-columns: clamp(340px, 32vw, 520px) minmax(0, 1fr); }
+.dshl-benchbox { flex: 1; min-height: 0; overflow: auto; overscroll-behavior: contain; }
 .dshl-under { display: flex; flex-direction: column; flex: 1; min-height: 210px; overflow: hidden; border: 1px solid var(--line); border-radius: 12px; background: var(--l1); }
 .dshl-under .dshl-sidetabs { align-items: center; padding: 0 12px 0 10px; }
 .dshl-under .dshl-sidetabs > button { flex: none; padding: 0 14px; }
@@ -36857,6 +36909,9 @@ html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top :is(but
 .dshl.page.theater .dshl-stage { max-height: 100cqh; }
 @media (max-width: 1000px) {
   .dshl.page { display: block; overflow: auto; }
+  .dshl.page.bench .dshl-main { grid-template-columns: minmax(0, 1fr); }
+  .dshl.page.bench .dshl-side { height: auto; min-height: 0; }
+  .dshl-benchbox { overflow: visible; }
   .dshl.page .dshl-main { grid-template-columns: minmax(0, 1fr); align-items: start; }
   .dshl.page .dshl-col { container-type: normal; }
   .dshl.page .dshl-col > .dshl-drop { min-height: 320px; aspect-ratio: 16 / 9; }
@@ -36865,6 +36920,10 @@ html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top :is(but
   .dshl.page .dshl-side { height: 70vh; min-height: 420px; }
 }
 @media (max-width: 600px) {
+  /* Tabs keep their row; the search, line/reading switch and copy move to one of their own. */
+  .dshl-under .dshl-sidetabs { flex-wrap: wrap; row-gap: 0; }
+  .dshl-tools.inline { flex: 1 0 100%; justify-content: flex-start; padding: 6px 0 8px; }
+  .dshl-tools.inline .dshl-field { flex: 1 1 auto; }
   .dshl-top { padding: 10px 12px; }
   .dshl-bar { flex-wrap: nowrap; gap: 6px; }
   .dshl-bar .dshl-field { flex: 1 1 0; height: 38px; padding: 0 4px 0 10px; gap: 6px; }
@@ -37056,6 +37115,7 @@ html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top :is(but
 .dshl-dl { display: grid; gap: 14px; padding: 0 6px; }
 .dshl-dl .grp { display: grid; gap: 8px; }
 .dshl-dl .cap { color: var(--muted); font-size: 12px; font-weight: 550; }
+.dshl-dl-note { margin: 0; color: var(--muted); font-size: 13px; line-height: 1.6; }
 .dshl-dl .cards { display: grid; grid-template-columns: 1fr 1fr; gap: 8px; }
 .dshl-card-opt { display: grid; align-content: start; gap: 2px; padding: 10px 12px; border: 1px solid var(--line); border-radius: 10px; background: var(--l1); color: var(--ink); text-align: left; cursor: pointer; }
 .dshl-card-opt b { font-size: 13px; font-weight: 600; }
@@ -37112,7 +37172,7 @@ html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top :is(but
 .dshl-lines { display: flex; flex-direction: column; gap: 2px; flex: 1; min-height: 0; margin: 0 -6px; overflow: auto; }
 .dshl:not(.page) .dshl-lines { max-height: calc(100vh - 330px); }
 @media (max-width: 1000px) { .dshl:not(.page) .dshl-lines { max-height: 60vh; } }
-.dshl-line { display: grid; grid-template-columns: 44px 1fr; gap: 6px; padding: 7px 6px; border: 0; border-radius: 8px; background: none; color: var(--ink); text-align: left; cursor: pointer; }
+.dshl-line { display: grid; grid-template-columns: 44px 1fr 14px; gap: 6px; padding: 7px 6px; border: 0; border-radius: 8px; background: none; color: var(--ink); text-align: left; cursor: pointer; }
 .dshl-lines > .pad { padding: 16px 6px; }
 .dshl-line:hover { background: var(--hover); }
 .dshl-line.active { background: color-mix(in srgb, var(--accent) 10%, transparent); }
@@ -37122,6 +37182,12 @@ html[data-platform='darwin']:not([data-fullscreen]) .dshl.page .dshl-top :is(but
 .dshl-line .main { font-weight: 550; }
 .dshl-line .sub { color: var(--muted); font-size: 13px; }
 .dshl-line.raw .main, .dshl-line.raw .sub { opacity: .62; }
+/* Polished lines: a quiet tick at the end; first-pass lines stay lighter until polished. */
+.dshl-line .mark { display: flex; padding-top: 4px; color: var(--faint); opacity: .7; }
+.dshl-line.active .mark { color: var(--accent); }
+/* Player: a faint tick on the subtitle button while the line on screen is polished (explained on hover). */
+.dshl-okmark { display: flex; margin: 0 -2px 0 1px; color: rgba(255,255,255,.45); }
+.dshl-okmark svg { width: 9px; height: 9px; }
 
 /* Reading (paragraphs by speaker) */
 .dshl-read { display: grid; align-content: start; gap: 16px; flex: 1; min-height: 0; overflow: auto; padding: 2px 2px 12px; }
@@ -37270,6 +37336,7 @@ let done = false;
 function ensureStyles() {
 	if (done || typeof document === "undefined") return;
 	done = true;
+	for (const old of document.querySelectorAll("style[data-dsh-live]")) old.remove();
 	const el = document.createElement("style");
 	el.dataset.dshLive = "";
 	el.textContent = CSS;
@@ -37278,6 +37345,26 @@ function ensureStyles() {
 
 // src/client/VideoPage.tsx
 const PAGE_SESSION = "dsh-live.pageSession";
+/** The id of the video page among DS-Harness's main panels (beside the conversation, Plugins…). */
+const PANEL_ID = "vidi";
+let panelHost;
+function setPanelHost(host) {
+	panelHost = host;
+	if (!host) return () => {};
+	const sync = () => {
+		const here = host.info.getSnapshot().activePanelId === PANEL_ID;
+		if (pageStore.get().open !== here) pageStore.set({
+			open: here,
+			...here ? {} : { pending: void 0 }
+		});
+	};
+	const off = host.info.subscribe(sync);
+	return () => {
+		off();
+		if (panelHost === host) panelHost = void 0;
+	};
+}
+const inPanel = () => !!panelHost;
 const samePath = (a, b) => !!a && !!b && a.replace(/\/+$/, "") === b.replace(/\/+$/, "");
 /**
 * The video page's own conversation: reused while it exists, created when not.
@@ -37327,6 +37414,16 @@ async function showPage(sessions, opts) {
 	const current = sessions.current();
 	const sid = opts.sessionId ?? await pageSession(sessions, opts.workspaces, opts.title ?? "Video");
 	if (sid !== current) sessions.open(sid);
+	if (panelHost) {
+		pageStore.set({
+			open: true,
+			sessionId: sid,
+			returnTo: void 0,
+			pending: opts.pending
+		});
+		panelHost.select(PANEL_ID);
+		return;
+	}
 	const prev = pageStore.get();
 	pageStore.set({
 		open: true,
@@ -37343,7 +37440,27 @@ function closePage(sessions) {
 	});
 	if (returnTo && sessions.has(returnTo)) sessions.open(returnTo);
 }
-/** Full-window video page (shell.overlay): player, subtitles, terms and the chat with DeepSeek. */
+/** The video page as a main panel (newer DS-Harness): the app's sidebar and its toggle stay. */
+function VideoPanel(props) {
+	const { sessions, t } = props;
+	const page = (0, react.useSyncExternalStore)(pageStore.subscribe, pageStore.get);
+	const face = (0, react.useSyncExternalStore)(sessions.subscribe, () => page.sessionId ? sessions.face(page.sessionId) : void 0);
+	const chat = (0, react.useSyncExternalStore)(sessions.subscribe, () => page.sessionId ? sessions.chat(page.sessionId) : void 0);
+	if (!page.sessionId) return null;
+	return /* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
+		className: "dshl-panelwrap",
+		children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)(LiveView, {
+			sessionId: page.sessionId,
+			t,
+			page: true,
+			chatFace: face,
+			chatSource: chat,
+			pending: page.pending,
+			onPendingDone: () => pageStore.set({ pending: void 0 })
+		}, page.sessionId)
+	});
+}
+/** Full-window video page (shell.overlay, older DS-Harness): player, subtitles, terms and the chat with DeepSeek. */
 function VideoPage(props) {
 	const { sessions, t } = props;
 	const page = (0, react.useSyncExternalStore)(pageStore.subscribe, pageStore.get);
@@ -37354,7 +37471,7 @@ function VideoPage(props) {
 		role: "alert",
 		children: t("page.failed").replace("{e}", page.error)
 	});
-	if (!page.open || !page.sessionId) return null;
+	if (inPanel() || !page.open || !page.sessionId) return null;
 	return /* @__PURE__ */ (0, react_jsx_runtime.jsx)("div", {
 		className: "dshl-pagewrap",
 		children: /* @__PURE__ */ (0, react_jsx_runtime.jsx)(LiveView, {
@@ -37459,7 +37576,8 @@ function PasteChip(props) {
 /**
 * dsh-vidi browser half:
 * - a "Video" tab beside Chat in every conversation (conversation.view);
-* - a "Watch video" button in the sidebar footer that opens the full video page (shell.overlay);
+* - a "Vidi video" button in the sidebar footer that opens the video page: one of the app's
+*   main panels where DS-Harness has them (ctx.layout), else a full-window page (shell.overlay);
 * - a prompt above the composer when a video link is pasted (conversation.input.dock);
 * - "show this video" requests from DeepSeek's live_open tool and the /video command.
 */
@@ -37555,6 +37673,24 @@ function apply(ctx) {
 			t
 		})
 	}, VideoPage));
+	ctx.inject(["layout"], (c) => {
+		c.effect(() => setPanelHost({
+			select: (id) => c.layout.selectPanel(id),
+			info: {
+				getSnapshot: () => c.layout.panelInfo.getSnapshot(),
+				subscribe: (fn) => c.layout.panelInfo.subscribe(fn)
+			}
+		}), "dsh-vidi: main panel host");
+		c.slots.inject("main", () => c.slots.register({
+			name: "main",
+			key: PANEL_ID,
+			locale: NS,
+			inject: () => ({
+				sessions: sessions(),
+				t
+			})
+		}, VideoPanel));
+	});
 	ctx.slots.inject("conversation.input.dock", () => ctx.slots.register({
 		name: "conversation.input.dock",
 		id: "dsh-vidi.paste",

package/lib/index.mjs

@@ -1,6 +1,6 @@
 import { appendFileSync, chmodSync, createReadStream, createWriteStream, existsSync, mkdirSync, readFileSync, readdirSync, rmSync, statSync, writeFileSync } from "node:fs";
 import { homedir, networkInterfaces } from "node:os";
-import { basename, extname, join, normalize, resolve, sep } from "node:path";
+import { basename, dirname, extname, join, normalize, resolve, sep } from "node:path";
 import z from "@deepseek-ai/schemastery";
 import { execFile, spawn } from "node:child_process";
 import { promisify } from "node:util";
@@ -161,6 +161,16 @@ function defaultExports(targets) {
 }
 // src/glossary.ts
 const MAX = 80;
+/** Does the text mention the term (its correct form or a known wrong one)? Latin words match whole words only. */
+function mentions(text, t) {
+	return [t.src, ...t.aliases ?? []].some((w) => {
+		const k = w.trim().toLowerCase();
+		if (!k) return false;
+		if (!/^[\x00-\x7f]+$/.test(k)) return text.includes(k);
+		const esc = k.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
+		return new RegExp(`(^|[^a-z0-9])${esc}($|[^a-z0-9])`).test(text);
+	});
+}
 function readTerms(file) {
 	if (!file || !existsSync(file)) return [];
 	try {
@@ -174,7 +184,9 @@ function readTerms(file) {
 * holds terms confirmed by polished text, and gets corrected as polishing
 * goes on). Used by recognition (correct spellings as hints), translation
 * and the next polishing round. Users can correct or remove entries; their
-* entries win and are kept in a shared library for later videos.
+* entries win and are kept in a shared library for later videos. A library
+* entry joins a video's glossary only once that video mentions it (title,
+* description or subtitles), so one video's names never show up in another.
 */
 var Glossary = class {
 	file;
@@ -183,20 +195,24 @@ var Glossary = class {
 	log;
 	library;
 	terms = [];
+	/** Library entries this video has not mentioned yet: not shown, not in any prompt. */
+	shelf = [];
 	constructor(file, targets, onChange, log, seed, library) {
 		this.file = file;
 		this.targets = targets;
 		this.onChange = onChange;
 		this.log = log;
 		this.library = library;
-		this.terms = readTerms(file).map((t) => ({
+		const lib = readTerms(library).map((t) => ({
 			...t,
-			origin: t.origin ?? "auto"
+			origin: "user"
 		}));
-		for (const t of readTerms(library)) if (!this.find(t.src)) this.terms.push({
+		const inLib = (t) => lib.some((x) => x.src.toLowerCase() === t.src.toLowerCase());
+		this.terms = readTerms(file).map((t) => ({
 			...t,
-			origin: "user"
-		});
+			origin: t.origin ?? "auto"
+		})).filter((t) => t.origin !== "user" || !inLib(t));
+		this.shelf = lib.filter((t) => !this.find(t.src));
 		for (const line of seed.split("\n").map((l) => l.trim()).filter(Boolean)) {
 			const [src, rest] = line.split(/\s*=\s*/);
 			if (src && !this.find(src)) this.terms.push({
@@ -206,6 +222,17 @@ var Glossary = class {
 			});
 		}
 	}
+	/** New text from this video (title, description, subtitle lines): bring in the library entries it mentions. */
+	notice(text) {
+		if (!this.shelf.length || !text) return;
+		const low = text.toLowerCase();
+		const hit = this.shelf.filter((t) => mentions(low, t) && !this.find(t.src));
+		if (!hit.length) return;
+		this.shelf = this.shelf.filter((t) => !hit.includes(t));
+		this.terms.push(...hit);
+		this.log(`glossary: ${hit.map((t) => t.src).join(", ")} from the library`);
+		this.save();
+	}
 	setTargets(targets) {
 		this.targets = targets;
 	}
@@ -304,6 +331,7 @@ var Glossary = class {
 		});
 		this.save();
 		const t = this.find(src);
+		this.shelf = this.shelf.filter((x) => x.src.toLowerCase() !== src.toLowerCase());
 		this.saveLibrary((lib) => [...lib.filter((x) => x.src.toLowerCase() !== src.toLowerCase()), {
 			...t,
 			origin: "user"
@@ -313,6 +341,7 @@ var Glossary = class {
 		const t = this.find(src);
 		if (!t) return;
 		this.terms = this.terms.filter((x) => x !== t);
+		this.shelf = this.shelf.filter((x) => x.src.toLowerCase() !== t.src.toLowerCase());
 		this.save();
 		this.saveLibrary((lib) => lib.filter((x) => x.src.toLowerCase() !== t.src.toLowerCase()));
 	}
@@ -1082,6 +1111,46 @@ function downloadOriginal(url, dir, browser, onProgress, signal) {
 		});
 	});
 }
+/** Download only the sound of a page URL (best audio stream), next to the job files. Resolves to the saved file path. */
+function downloadAudio(url, dir, browser, onProgress) {
+	return new Promise((resolve, reject) => {
+		const args = [
+			"--no-warnings",
+			"--no-playlist",
+			...cookieArgs(browser),
+			"-f",
+			"bestaudio[ext=m4a]/bestaudio/best",
+			"--newline",
+			"--progress-template",
+			"download:%(progress._percent_str)s",
+			"-o",
+			join(dir, "audio-src.%(ext)s"),
+			"--print",
+			"after_move:filepath",
+			url
+		];
+		const p = spawn("yt-dlp", args, { stdio: [
+			"ignore",
+			"pipe",
+			"pipe"
+		] });
+		let out = "", err = "";
+		p.stdout.on("data", (d) => {
+			out += d.toString();
+			const m = d.toString().match(/([\d.]+)%\s*$/m);
+			if (m) onProgress(Math.min(99, Math.round(Number(m[1]))));
+		});
+		p.stderr.on("data", (d) => {
+			err += d.toString();
+		});
+		p.on("error", reject);
+		p.on("exit", (code) => {
+			const file = out.split("\n").map((l) => l.trim()).filter((l) => l.startsWith("/")).at(-1);
+			if (code === 0 && file && existsSync(file)) resolve(file);
+			else reject(new Error(ytdlpError({ stderr: err })));
+		});
+	});
+}
 /** Keep the most recent downloaded original of a job dir, if any. */
 function findOriginal(dir) {
 	if (!existsSync(dir)) return void 0;
@@ -1682,6 +1751,7 @@ var LiveJob = class LiveJob extends EventEmitter {
 			still: (at) => this.still(at),
 			onLine: (cue, src, tr, speaker) => {
 				if (!this.cues.includes(cue)) return;
+				this.deps.glossary.notice(src);
 				cue.src = src;
 				if (speaker) cue.speaker = speaker;
 				if (Object.keys(tr).length) cue.tr = {
@@ -1776,6 +1846,11 @@ var LiveJob = class LiveJob extends EventEmitter {
 			job.cues.push(...[...seen.values()].sort((a, b) => a.start - b.start));
 			job.nextId = Math.max(0, ...job.cues.map((c) => c.id)) + 1;
 			if (existsSync(join(dir, "brief.txt"))) job.brief = readFileSync(join(dir, "brief.txt"), "utf8");
+			deps.glossary.notice([
+				job.title ?? "",
+				job.brief,
+				...job.cues.map((c) => c.src)
+			].join("\n"));
 			job.siteLang = j.siteLang;
 			const q = j;
 			if (q.height) job.height = q.height;
@@ -2317,6 +2392,7 @@ var LiveJob = class LiveJob extends EventEmitter {
 		if (cue.final && key !== void 0) this.drafts.delete(key);
 		this.emit("cue", cue);
 		if (cue.final) {
+			this.deps.glossary.notice(cue.src);
 			if (this.opts.mode === "realtime" && !this.provider?.translates && cue.tr === void 0 && this.targets().length) this.translate(cue);
 			else {
 				this.record(cue);
@@ -2762,6 +2838,7 @@ var LiveJob = class LiveJob extends EventEmitter {
 		try {
 			writeFileSync(join(this.deps.dir, "brief.txt"), this.brief);
 		} catch {}
+		this.deps.glossary.notice(`${this.title ?? ""}\n${this.brief}`);
 	}
 	/** The site's subtitle text around [from, to] (media seconds), per track — a reference, not ground truth. */
 	siteText(from, to) {
@@ -3037,6 +3114,86 @@ var Downloads = class {
 		});
 		return task;
 	}
+	/**
+	* Just the sound, as a file people can play anywhere (M4A, or the site's own
+	* MP3/M4A as is). Uses what is already here (original, recording, source
+	* file); otherwise fetches only the audio stream, never the whole video.
+	*/
+	audio(opts) {
+		const dir = join(this.jobsDir, opts.dirId);
+		const task = this.add({
+			kind: "audio",
+			conversationId: opts.conversationId,
+			url: opts.url,
+			dirId: opts.dirId,
+			title: opts.title,
+			saveAs: opts.saveAs
+		});
+		(async () => {
+			const recording = join(dir, "recording.mp4");
+			let source = opts.localFile ?? findOriginal(dir) ?? (existsSync(recording) ? recording : void 0);
+			if (!source) {
+				if (!opts.url) throw new Error("nothing to take the sound from yet");
+				task.stage = "download";
+				source = await downloadAudio(opts.url, dir, opts.browser, (pct) => {
+					task.pct = Math.round(pct * .95);
+				});
+			}
+			if (dirname(source) === dir && /^\.(m4a|mp3)$/i.test(extname(source))) {
+				task.file = basename(source);
+				return;
+			}
+			task.stage = "mux";
+			task.pct = Math.max(task.pct, 95);
+			const out = join(dir, "audio.m4a");
+			try {
+				await run("ffmpeg", [
+					"-v",
+					"error",
+					"-y",
+					"-i",
+					source,
+					"-vn",
+					"-map",
+					"0:a:0",
+					"-c:a",
+					"copy",
+					"-movflags",
+					"+faststart",
+					out
+				], { timeout: 9e5 });
+			} catch {
+				await run("ffmpeg", [
+					"-v",
+					"error",
+					"-y",
+					"-i",
+					source,
+					"-vn",
+					"-map",
+					"0:a:0",
+					"-c:a",
+					"aac",
+					"-b:a",
+					"192k",
+					"-movflags",
+					"+faststart",
+					out
+				], { timeout: 18e5 });
+			}
+			task.file = basename(out);
+		})().then(() => {
+			task.status = "done";
+			task.pct = 100;
+			task.stage = void 0;
+		}).catch((err) => {
+			const msg = String(err.stderr || err.message);
+			task.status = "error";
+			task.error = /HTTP Error (412|429)/.test(msg) ? "The site is refusing requests for now (rate limit). Try again in a minute." : msg.slice(-300);
+			this.log(`audio download failed: ${task.error}`);
+		});
+		return task;
+	}
 };
 // src/keys.ts
 /** Online speech services; each reads its key from the names DSH and other tools already use. */
@@ -5125,6 +5282,21 @@ function apply(ctx, config) {
 					saveAs: typeof b.saveAs === "string" ? b.saveAs.replace(/[\\/:*?"<>|]+/g, " ").slice(0, 150) : "video.mp4"
 				}));
 			}
+			if (route === "audio" && req.method === "POST") {
+				const b = await readJson(req);
+				const job = jobs.get(String(b.jobId));
+				if (!job) return send(res, 404, { error: "no such job" });
+				const s = job.summary();
+				return send(res, 200, downloads.audio({
+					conversationId: String(b.conversationId ?? s.conversationId),
+					dirId: s.id,
+					title: s.title,
+					url: s.source === "url" ? s.url : void 0,
+					browser: s.loginBrowser,
+					localFile: s.source === "file" ? s.url : void 0,
+					saveAs: typeof b.saveAs === "string" ? b.saveAs.replace(/[\\/:*?"<>|]+/g, " ").slice(0, 150) : "audio"
+				}));
+			}
 			if (route === "job") {
 				const jid = url.searchParams.get("jobId");
 				const job = jid ? jobs.get(jid) : jobForConversation(url.searchParams.get("conversationId") ?? "");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-vidi",
-  "version": "0.3.1",
+  "version": "0.4.1",
   "description": "A video tool for DeepSeek Harness: watch live streams, videos and podcasts with real-time subtitles in any language, then clip and download",
   "type": "module",
   "license": "PolyForm-Noncommercial-1.0.0",