Merge pull request #2248 from chan1103/feat/smart-paste-text-fallback

feat(coding-agent): fall back to text paste when clipboard holds no image

packages/coding-agent/src/config/keybindings.ts

@@ -137,7 +137,7 @@ export const KEYBINDINGS = {
 	},
 	"app.clipboard.pasteImage": {
 		defaultKeys: getDefaultPasteImageKeys(),
-		description: "Paste image from clipboard",
+		description: "Paste image or text from clipboard",
 	},
 	"app.clipboard.pasteTextRaw": {
 		defaultKeys: ["ctrl+shift+v", "alt+shift+v"],

packages/coding-agent/src/modes/controllers/input-controller.ts

@@ -49,7 +49,14 @@ const TINY_TITLE_PROGRESS_DONE_TTL_MS = 3_000;
 const TINY_TITLE_PROGRESS_REVEAL_DELAY_MS = 1_000;
 
 export class InputController {
-	constructor(private ctx: InteractiveModeContext) {}
+	constructor(
+		private ctx: InteractiveModeContext,
+		/** Injectable clipboard reads so tests can drive paste flows without a real clipboard. */
+		private clipboard: {
+			readImage: typeof readImageFromClipboard;
+			readText: typeof readTextFromClipboard;
+		} = { readImage: readImageFromClipboard, readText: readTextFromClipboard },
+	) {}
 
 	#enhancedPaste?: EnhancedPasteController;
 
@@ -876,10 +883,25 @@ export class InputController {
 
 	async handleImagePaste(): Promise<boolean> {
 		try {
-			const image = await readImageFromClipboard();
+			const image = await this.clipboard.readImage();
 			if (!image) {
-				this.ctx.showStatus("No image in clipboard (use terminal paste for text)");
-				return false;
+				// Smart paste (#1628): no image on the clipboard — fall back to
+				// pasting its text so the same chord covers both payload kinds.
+				// Hosts that pre-empt the terminal's own paste (VS Code's
+				// integrated terminal, Win+V clipboard history) deliver only
+				// this keypress, so a miss here must not dead-end.
+				const text = await this.clipboard.readText();
+				if (!text) {
+					this.ctx.showStatus("Clipboard is empty");
+					return false;
+				}
+				// Route to the focused component when it accepts pastes (modal
+				// Input prompts), matching the enhanced-paste text path (#2127).
+				const focused = this.ctx.ui.getFocused();
+				const target = focused && focused !== this.ctx.editor && hasPasteText(focused) ? focused : this.ctx.editor;
+				target.pasteText(text);
+				this.ctx.ui.requestRender();
+				return true;
 			}
 			return await this.#normalizeAndInsertPastedImage(
 				{
@@ -897,10 +919,11 @@ export class InputController {
 
 	async handleClipboardTextRawPaste(): Promise<void> {
 		try {
-			const text = await readTextFromClipboard();
+			const text = await this.clipboard.readText();
 			if (text) {
 				this.ctx.editor.insertText(text);
 				this.ctx.ui.requestRender();
+			} else {
 				this.ctx.showStatus("No text in clipboard to paste raw");
 			}
 		} catch {

packages/coding-agent/src/modes/utils/hotkeys-markdown.ts

@@ -48,7 +48,7 @@ export function buildHotkeysMarkdown(bindings: HotkeysMarkdownBindings): string
 		`| \`${appKey(bindings, "app.tools.expand")}\` | Toggle tool output expansion |`,
 		`| \`${appKey(bindings, "app.thinking.toggle")}\` | Toggle thinking block visibility |`,
 		`| \`${appKey(bindings, "app.editor.external")}\` | Edit message in external editor |`,
-		`| \`${appKey(bindings, "app.clipboard.pasteImage")}\` | Paste image from clipboard |`,
+		`| \`${appKey(bindings, "app.clipboard.pasteImage")}\` | Paste image or text from clipboard |`,
 		`| \`${appKey(bindings, "app.stt.toggle")}\` | Toggle speech-to-text recording |`,
 		`| \`${appKey(bindings, "app.agents.hub")}\` / \`${appKey(bindings, "app.session.observe")}\` / double-tap \`←\` (empty editor) | Open the agent hub |`,
 		"| `#` | Open prompt actions |",

packages/coding-agent/src/utils/clipboard.ts

@@ -125,6 +125,56 @@ async function readImageViaPowerShell(): Promise<ClipboardImage | null> {
 	}
 }
 
+// PowerShell one-liner that emits the clipboard text verbatim on stdout, or
+// nothing when the clipboard holds no text. `[Console]::Out.Write` avoids the
+// trailing newline Write-Output would add; output encoding is forced to UTF-8
+// so non-ASCII text survives the interop boundary regardless of console
+// codepage.
+const POWERSHELL_TEXT_SCRIPT = `
+$ErrorActionPreference = 'Stop'
+[Console]::OutputEncoding = [Text.Encoding]::UTF8
+[Console]::Out.Write([string](Get-Clipboard -Raw))
+`;
+
+/**
+ * Read clipboard text through Windows PowerShell — native win32 or the WSL
+ * host over interop.
+ *
+ * Same rationale as `readImageViaPowerShell`: under WSL, the WSLg Wayland
+ * clipboard only works when `wl-clipboard` happens to be installed in the
+ * distro, while `powershell.exe` is always reachable. Forcing UTF-8 output
+ * encoding keeps non-ASCII text intact regardless of the console codepage
+ * (the legacy win32 `Get-Clipboard` shell-out mangled it), and `Bun.spawn`
+ * keeps a cold PowerShell start off the TUI event loop.
+ *
+ * Returns null when the bridge fails (WSL callers fall through to
+ * wl-paste/xclip); an empty string is a successful "no text" read.
+ */
+async function readTextViaPowerShell(): Promise<string | null> {
+	try {
+		const proc = Bun.spawn(["powershell.exe", "-NoProfile", "-NonInteractive", "-Command", POWERSHELL_TEXT_SCRIPT], {
+			stdout: "pipe",
+			stderr: "ignore",
+			stdin: "ignore",
+		});
+		const timer = setTimeout(() => proc.kill(), POWERSHELL_TIMEOUT_MS);
+		let stdout = "";
+		try {
+			stdout = await new Response(proc.stdout).text();
+			await proc.exited;
+		} catch (err) {
+			logger.warn("clipboard: powershell text read failed", { error: String(err) });
+			return null;
+		} finally {
+			clearTimeout(timer);
+		}
+		if (proc.exitCode !== 0) return null;
+		return stdout.replaceAll("\r\n", "\n");
+	} catch {
+		return null;
+	}
+}
+
 /**
  * Read an image from the system clipboard.
  *
@@ -165,14 +215,16 @@ export async function readTextFromClipboard(): Promise<string> {
 			return execSync("pbpaste", { encoding: "utf8", timeout: 2000 }).toString();
 		}
 		if (p === "win32") {
-			return execSync('powershell.exe -NoProfile -Command "Get-Clipboard"', {
-				encoding: "utf8",
-				timeout: 2000,
-			}).toString();
+			return (await readTextViaPowerShell()) ?? "";
 		}
 		if (process.env.TERMUX_VERSION) {
 			return execSync("termux-clipboard-get", { encoding: "utf8", timeout: 2000 }).toString();
 		}
+		if (isWsl()) {
+			const text = await readTextViaPowerShell();
+			if (text !== null) return text;
+			// Bridge failed — fall through to the wl-paste/xclip paths below.
+		}
 		const hasWaylandDisplay = Boolean(process.env.WAYLAND_DISPLAY);
 		const hasX11Display = Boolean(process.env.DISPLAY);
 		if (hasWaylandDisplay) {