@instaclaw/dispatch 0.4.0 → 0.5.1

package/dist/connection.js

@@ -37,13 +37,17 @@ export function connect(opts) {
             const result = await executeCommand(command);
             commandsExecuted++;
             if (result.screenshotBuffer && result.screenshotMeta) {
-                // Screenshot: send metadata text frame, then binary frame
+                // Screenshot (or batch with screenshot): send metadata text frame, then binary frame
                 const meta = {
                     id: command.id,
                     type: "screenshot_result",
                     ...result.screenshotMeta,
                     size: result.screenshotBuffer.length,
                 };
+                // For batch commands, include action results in the metadata frame
+                if (result.data) {
+                    meta.data = result.data;
+                }
                 ws?.send(JSON.stringify(meta));
                 ws?.send(result.screenshotBuffer);
             }

package/dist/executor.js

@@ -32,6 +32,8 @@ export async function executeCommand(command) {
                 return await doDrag(params);
             case "windows":
                 return await listWindows();
+            case "batch":
+                return await executeBatch(params);
             default:
                 return { success: false, error: `Unknown command type: ${type}` };
         }
@@ -40,10 +42,87 @@ export async function executeCommand(command) {
         return { success: false, error: String(err) };
     }
 }
+/**
+ * Execute multiple actions sequentially with optional per-action wait times.
+ * Returns results for all actions; stops on first failure.
+ * Optionally takes a screenshot after the batch completes.
+ */
+async function executeBatch(params) {
+    const actions = params.actions;
+    if (!Array.isArray(actions) || actions.length === 0) {
+        return { success: false, error: "batch requires a non-empty 'actions' array" };
+    }
+    if (actions.length > 20) {
+        return { success: false, error: "batch limited to 20 actions max" };
+    }
+    const results = [];
+    for (const action of actions) {
+        const cmd = {
+            id: `batch_${Date.now()}`,
+            type: action.type,
+            params: action.params || {},
+        };
+        // Don't allow nested batches or screenshots inside a batch
+        if (cmd.type === "batch") {
+            results.push({ type: "batch", success: false, error: "nested batches not allowed" });
+            break;
+        }
+        if (cmd.type === "screenshot") {
+            results.push({ type: "screenshot", success: false, error: "use screenshotAfter param instead of screenshot in batch" });
+            break;
+        }
+        const result = await executeCommand(cmd);
+        results.push({
+            type: action.type,
+            success: result.success,
+            data: result.data,
+            error: result.error,
+        });
+        if (!result.success)
+            break; // Stop batch on first failure
+        // Per-action wait time (default: 50ms between actions)
+        const waitMs = action.waitAfterMs ?? 50;
+        if (waitMs > 0) {
+            await new Promise((r) => setTimeout(r, waitMs));
+        }
+    }
+    // Optionally take a screenshot after the batch completes
+    const screenshotAfter = params.screenshotAfter !== false; // default true
+    let screenshotResult;
+    if (screenshotAfter) {
+        // Wait for screen to settle before capturing
+        const settleMs = typeof params.settleMs === "number" ? params.settleMs : 300;
+        if (settleMs > 0) {
+            await new Promise((r) => setTimeout(r, settleMs));
+        }
+        screenshotResult = await takeScreenshot({
+            format: params.screenshotFormat || "webp",
+            quality: Number(params.screenshotQuality) || 55,
+        });
+    }
+    const batchResult = {
+        success: results.every((r) => r.success),
+        data: {
+            action: "batch",
+            actionsExecuted: results.length,
+            actionsTotal: actions.length,
+            results,
+        },
+    };
+    // If we took a screenshot, attach it
+    if (screenshotResult?.screenshotBuffer && screenshotResult.screenshotMeta) {
+        batchResult.screenshotBuffer = screenshotResult.screenshotBuffer;
+        batchResult.screenshotMeta = screenshotResult.screenshotMeta;
+        // Also include batch results in the screenshot meta so the agent gets both
+        batchResult.data.screenshotIncluded = true;
+    }
+    return batchResult;
+}
 async function takeScreenshot(params) {
     const capture = await captureScreenshot({
-        format: params.format || "jpeg",
-        quality: Number(params.quality) || 80,
+        format: params.format || "webp",
+        quality: Number(params.quality) || 55,
+        maxWidth: Number(params.maxWidth) || 1280,
     });
     // Cache coordMap for subsequent click/hover/drag commands
     lastCoordMap = capture.coordMap;

package/dist/index.js

@@ -8,27 +8,105 @@
  *   npx @instaclaw/dispatch --token XXX --vm HOST    # Direct connect (advanced)
  */
 import { loadConfig, saveConfig, getConfigPath } from "./config.js";
-import { checkPermissions, printPermissionGuide, runPermissionTest } from "./permissions.js";
+import { checkPermissions, runPermissionTest } from "./permissions.js";
 import { connect, disconnect } from "./connection.js";
 import { closeSupervisor, setMode } from "./supervisor.js";
 import chalk from "chalk";
 import readline from "readline";
 import os from "os";
+import fs from "fs";
+import path from "path";
 const API_BASE = "https://instaclaw.io";
+// Temp file to persist pairing code + args across Terminal restarts
+const RETRY_FILE = path.join(os.tmpdir(), "instaclaw-dispatch-retry.json");
+const RETRY_MAX_AGE_MS = 10 * 60 * 1000; // 10 minutes
 async function main() {
     console.log(`
   ${chalk.bold("InstaClaw Dispatch")} ${chalk.dim("— Remote Computer Control")}
   ${chalk.dim("─────────────────────────────────────────────")}
   `);
-    // 1. Check permissions (macOS only)
+    const args = parseArgs();
+    // ── Auto-retry: check if we're resuming after a Terminal restart ──
+    const retry = loadRetryState();
+    if (!args.pair && !args.token && retry) {
+        console.log(`  ${chalk.cyan("↻")} Resuming after Terminal restart...`);
+        console.log(`  ${chalk.dim("Pairing code:")} ${chalk.bold(retry.pair)}\n`);
+        args.pair = retry.pair;
+        if (retry.autonomous)
+            args.autonomous = true;
+        clearRetryState();
+    }
+    // ── 1. Check permissions FIRST (before consuming pairing code) ──
     if (os.platform() === "darwin") {
         console.log("  Checking permissions...");
         const perms = checkPermissions();
-        printPermissionGuide(perms);
-        if (perms.screenRecording === false || perms.accessibility === false) {
-            await waitForEnter();
+        if (perms.accessibility === false || perms.screenRecording === false) {
+            const terminal = getTerminalName();
+            // Save state so we can auto-resume after Terminal restart
+            if (args.pair) {
+                saveRetryState({ pair: args.pair, autonomous: args.autonomous });
+            }
+            console.log("");
+            if (perms.accessibility === false) {
+                console.log(`  ${chalk.red("✗")} Accessibility: ${chalk.bold("NOT GRANTED")}`);
+                console.log(`    → System Settings → Privacy & Security → Accessibility`);
+                console.log(`    → Enable "${terminal}"`);
+                console.log("");
+                console.log(chalk.yellow(`  ⚠  Enabling Accessibility will close ${terminal}.`));
+                console.log(chalk.yellow(`     Don't worry — just reopen ${terminal} and run the same command.`));
+                if (args.pair) {
+                    console.log(chalk.green(`     Your pairing code ${chalk.bold(args.pair)} will still work.`));
+                }
+                console.log("");
+            }
+            if (perms.screenRecording === false) {
+                console.log(`  ${chalk.red("✗")} Screen Recording: ${chalk.bold("NOT GRANTED")}`);
+                console.log(`    → System Settings → Privacy & Security → Screen Recording`);
+                console.log(`    → Enable "${terminal}"`);
+                console.log("");
+            }
+            if (perms.accessibility === true && perms.screenRecording === false) {
+                // Only Screen Recording is missing — Terminal won't be killed
+                console.log(`  Grant Screen Recording, then press ${chalk.bold("Enter")} to continue...`);
+                await waitForEnter();
+                // Re-check
+                const recheck = checkPermissions();
+                if (recheck.screenRecording === false) {
+                    console.log(`  ${chalk.yellow("⚠")} Screen Recording still not granted. Screenshots may not work.\n`);
+                }
+                else {
+                    console.log(`  ${chalk.green("✓")} Screen Recording: OK\n`);
+                }
+            }
+            else if (perms.accessibility === false) {
+                // Accessibility is missing — Terminal WILL be killed
+                console.log(`  After granting permissions and reopening ${terminal}, run:`);
+                console.log("");
+                if (args.pair) {
+                    console.log(`  ${chalk.cyan(`npx @instaclaw/dispatch@0.5.0 --pair ${args.pair}`)}`);
+                }
+                else {
+                    console.log(`  ${chalk.cyan("npx @instaclaw/dispatch")}`);
+                }
+                console.log("");
+                console.log(`  ${chalk.dim("Or just reopen Terminal — we'll auto-resume if possible.")}`);
+                console.log("");
+                console.log(`  Press ${chalk.bold("Enter")} when ready (or grant Accessibility and Terminal will restart)...`);
+                await waitForEnter();
+                // If we get here, user pressed Enter without Terminal being killed
+                const recheck = checkPermissions();
+                if (recheck.accessibility === false) {
+                    console.log(`\n  ${chalk.yellow("⚠")} Accessibility still not granted. Click/type commands will fail.`);
+                    console.log(`  ${chalk.dim("Continuing anyway — screenshots may still work.\n")}`);
+                }
+                else {
+                    console.log(`\n  ${chalk.green("✓")} Permissions: all OK\n`);
+                    clearRetryState();
+                }
+            }
         }
         else {
+            // All permissions OK — run functional test
             console.log("  Running permission test...");
             const test = await runPermissionTest();
             if (test.ok) {
@@ -38,11 +116,11 @@ async function main() {
                 console.log(`  ${chalk.yellow("⚠")} ${test.error}`);
                 console.log("  Continuing anyway — some commands may fail.\n");
             }
+            clearRetryState();
         }
     }
-    // 2. Resolve config: --pair, --token+--vm, saved config, or interactive
+    // ── 2. Resolve config: --pair, --token+--vm, saved config, or interactive ──
     let config = loadConfig();
-    const args = parseArgs();
     // Path A: Pairing code (simplest)
     if (args.pair) {
         console.log(`  Redeeming pairing code ${chalk.bold(args.pair)}...`);
@@ -51,6 +129,7 @@ async function main() {
             config = { gatewayToken: result.token, vmAddress: result.vmAddress, port: 8765, mode: "supervised" };
             saveConfig(config);
             console.log(`  ${chalk.green("✓")} Paired successfully\n`);
+            clearRetryState();
         }
         else {
             process.exit(1);
@@ -80,7 +159,7 @@ async function main() {
     console.log(`  ${chalk.dim("VM:")}   ${config.vmAddress}:${config.port}`);
     console.log(`  ${chalk.dim("Mode:")} ${config.mode}`);
     console.log(`  ${chalk.dim("Token:")} ${config.gatewayToken.substring(0, 8)}...\n`);
-    // 3. Connect to VM
+    // ── 3. Connect to VM ──
     connect({
         vmAddress: config.vmAddress,
         port: config.port,
@@ -119,7 +198,7 @@ async function main() {
             }
         },
     });
-    // 4. Ctrl+C
+    // ── 4. Ctrl+C ──
     process.on("SIGINT", () => {
         console.log(`\n  ${chalk.dim("Disconnecting...")}`);
         disconnect();
@@ -127,6 +206,46 @@ async function main() {
         process.exit(0);
     });
 }
+// ── Retry State (persists across Terminal restarts) ──
+function saveRetryState(state) {
+    try {
+        fs.writeFileSync(RETRY_FILE, JSON.stringify({ ...state, ts: Date.now() }));
+    }
+    catch { }
+}
+function loadRetryState() {
+    try {
+        if (!fs.existsSync(RETRY_FILE))
+            return null;
+        const data = JSON.parse(fs.readFileSync(RETRY_FILE, "utf-8"));
+        // Only use if recent (within 10 min)
+        if (Date.now() - data.ts > RETRY_MAX_AGE_MS) {
+            clearRetryState();
+            return null;
+        }
+        if (!data.pair)
+            return null;
+        return { pair: data.pair, autonomous: data.autonomous };
+    }
+    catch {
+        return null;
+    }
+}
+function clearRetryState() {
+    try {
+        fs.unlinkSync(RETRY_FILE);
+    }
+    catch { }
+}
+function getTerminalName() {
+    const terminal = process.env.TERM_PROGRAM || "your terminal app";
+    return {
+        "Apple_Terminal": "Terminal",
+        "iTerm.app": "iTerm2",
+        "vscode": "Visual Studio Code",
+        "WarpTerminal": "Warp",
+    }[terminal] || terminal;
+}
 // ── Pairing Code ──
 async function redeemPairingCode(code) {
     try {

package/dist/screenshot.d.ts

@@ -1,6 +1,8 @@
 export interface ScreenshotOptions {
-    format?: "jpeg" | "png";
+    format?: "jpeg" | "png" | "webp";
     quality?: number;
+    /** Max width — downscale if screen is larger (default 1280) */
+    maxWidth?: number;
 }
 export interface ScreenshotCapture {
     buffer: Buffer;

package/dist/screenshot.js

@@ -1,38 +1,70 @@
 /**
- * Screenshot — captures screen via usecomputer, converts PNG→JPEG via sharp.
- * Separated per PRD Section 5.1 package structure.
+ * Screenshot — captures screen via usecomputer, compresses via sharp.
+ * Optimized for speed: WebP format, quality 55, resolution capped at 1280x720.
  */
 import fs from "fs";
 import path from "path";
+// Pre-warm sharp at module load (avoid dynamic import per screenshot)
+// eslint-disable-next-line @typescript-eslint/no-explicit-any
+let sharpFn = null;
+const sharpReady = import("sharp").then((m) => { sharpFn = m.default || m; }).catch(() => { });
+// Pre-warm usecomputer at module load
+let ucMod = null;
+const ucReady = import("usecomputer").then((m) => { ucMod = m; }).catch(() => { });
 /**
  * Capture a screenshot and return it as a buffer with metadata.
  */
 export async function captureScreenshot(opts = {}) {
-    const uc = await import("usecomputer");
+    // Ensure modules are loaded
+    if (!ucMod)
+        await ucReady;
+    if (!sharpFn)
+        await sharpReady;
+    const uc = ucMod;
     const tmpPath = path.join("/tmp", `dispatch-ss-${Date.now()}.png`);
-    const format = opts.format || "jpeg";
-    const quality = opts.quality || 80;
+    const format = opts.format || "webp";
+    const quality = opts.quality || 55;
+    const maxWidth = opts.maxWidth || 1280;
     try {
         const result = await uc.screenshot({ path: tmpPath, display: null, window: null, region: null, annotate: null });
         const pngBuffer = fs.readFileSync(tmpPath);
         let outputBuffer;
-        if (format === "jpeg") {
+        let outWidth = result.imageWidth;
+        let outHeight = result.imageHeight;
+        if (sharpFn) {
             try {
-                const sharp = (await import("sharp")).default;
-                outputBuffer = await sharp(pngBuffer).jpeg({ quality }).toBuffer();
+                let pipeline = sharpFn(pngBuffer);
+                // Downscale if wider than maxWidth (HiDPI screens capture at 2x+)
+                if (result.imageWidth > maxWidth) {
+                    pipeline = pipeline.resize(maxWidth, null, { fit: "inside", withoutEnlargement: true });
+                    const ratio = maxWidth / result.imageWidth;
+                    outWidth = maxWidth;
+                    outHeight = Math.round(result.imageHeight * ratio);
+                }
+                // Encode to target format
+                if (format === "webp") {
+                    outputBuffer = await pipeline.webp({ quality, effort: 2 }).toBuffer();
+                }
+                else if (format === "jpeg") {
+                    outputBuffer = await pipeline.jpeg({ quality }).toBuffer();
+                }
+                else {
+                    outputBuffer = await pipeline.png().toBuffer();
+                }
             }
             catch {
-                // sharp not available — fall back to PNG
+                // sharp failed — fall back to raw PNG
                 outputBuffer = pngBuffer;
             }
         }
         else {
+            // No sharp available — return raw PNG
             outputBuffer = pngBuffer;
         }
         return {
             buffer: outputBuffer,
-            width: result.imageWidth,
-            height: result.imageHeight,
+            width: outWidth,
+            height: outHeight,
             format,
             coordMap: result.coordMap,
         };

package/dist/supervisor.js

@@ -85,6 +85,13 @@ function isDangerous(cmd) {
         const key = String(p.key || "").toLowerCase();
         return DANGEROUS_KEY_COMBOS.has(key);
     }
+    // Check all sub-actions in a batch
+    if (cmd.type === "batch") {
+        const actions = p.actions;
+        if (Array.isArray(actions)) {
+            return actions.some((a) => isDangerous({ id: "", type: a.type, params: a.params || {} }));
+        }
+    }
     return false;
 }
 function formatCommand(cmd) {
@@ -100,6 +107,13 @@ function formatCommand(cmd) {
             return `Scroll ${p.direction} by ${p.amount || 3}`;
         case "drag":
             return `Drag from (${p.fromX}, ${p.fromY}) to (${p.toX}, ${p.toY})`;
+        case "batch": {
+            const actions = p.actions;
+            if (!Array.isArray(actions))
+                return "Batch (empty)";
+            const summary = actions.map((a) => formatCommand({ id: "", type: a.type, params: a.params || {} })).join(" → ");
+            return `Batch (${actions.length} actions): ${summary}`;
+        }
         default:
             return `${cmd.type}: ${JSON.stringify(p).substring(0, 80)}`;
     }

package/dist/types.d.ts

@@ -1,7 +1,7 @@
 /** Command sent from VM agent → dispatch server → local relay */
 export interface DispatchCommand {
     id: string;
-    type: "screenshot" | "click" | "type" | "press" | "scroll" | "drag" | "windows" | "status";
+    type: "screenshot" | "click" | "type" | "press" | "scroll" | "drag" | "windows" | "status" | "batch";
     params: Record<string, unknown>;
     description?: string;
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@instaclaw/dispatch",
-  "version": "0.4.0",
+  "version": "0.5.1",
   "description": "Let your InstaClaw AI agent control your computer — screenshots, clicks, typing, and more via a secure WebSocket relay",
   "bin": {
     "instaclaw-dispatch": "./dist/index.js"