@instaclaw/dispatch 0.3.1 → 0.5.0

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

@@ -3,8 +3,9 @@
  * @instaclaw/dispatch — Local relay for InstaClaw agents to control your computer.
  *
  * Usage:
- *   npx @instaclaw/dispatch                    # Interactive setup
- *   npx @instaclaw/dispatch --token XXX --vm HOST  # Direct connect
+ *   npx @instaclaw/dispatch                          # Interactive (pairing code)
+ *   npx @instaclaw/dispatch --pair ABCD-1234         # Pair with code
+ *   npx @instaclaw/dispatch --token XXX --vm HOST    # Direct connect (advanced)
  */
 import { loadConfig, saveConfig, getConfigPath } from "./config.js";
 import { checkPermissions, printPermissionGuide, runPermissionTest } from "./permissions.js";
@@ -13,6 +14,7 @@ import { closeSupervisor, setMode } from "./supervisor.js";
 import chalk from "chalk";
 import readline from "readline";
 import os from "os";
+const API_BASE = "https://instaclaw.io";
 async function main() {
     console.log(`
   ${chalk.bold("InstaClaw Dispatch")} ${chalk.dim("— Remote Computer Control")}
@@ -25,33 +27,37 @@ async function main() {
         printPermissionGuide(perms);
         if (perms.screenRecording === false || perms.accessibility === false) {
             await waitForEnter();
-            // Re-check after user grants
-            const recheck = checkPermissions();
-            if (recheck.screenRecording === false) {
-                console.log("  Warning: Screen Recording still not granted. Screenshots may fail.\n");
-            }
-            if (recheck.accessibility === false) {
-                console.log("  Warning: Accessibility still not granted. Click/type may fail.\n");
-            }
         }
         else {
-            // Run a functional test to be sure
             console.log("  Running permission test...");
             const test = await runPermissionTest();
             if (test.ok) {
-                console.log("  Permissions: all OK\n");
+                console.log(`  ${chalk.green("✓")} Permissions: all OK\n`);
             }
             else {
-                console.log(`  Warning: ${test.error}`);
+                console.log(`  ${chalk.yellow("⚠")} ${test.error}`);
                 console.log("  Continuing anyway — some commands may fail.\n");
             }
         }
     }
-    // 2. Load or create config
+    // 2. Resolve config: --pair, --token+--vm, saved config, or interactive
     let config = loadConfig();
     const args = parseArgs();
-    if (args.token && args.vm) {
-        // CLI args override config
+    // Path A: Pairing code (simplest)
+    if (args.pair) {
+        console.log(`  Redeeming pairing code ${chalk.bold(args.pair)}...`);
+        const result = await redeemPairingCode(args.pair);
+        if (result) {
+            config = { gatewayToken: result.token, vmAddress: result.vmAddress, port: 8765, mode: "supervised" };
+            saveConfig(config);
+            console.log(`  ${chalk.green("✓")} Paired successfully\n`);
+        }
+        else {
+            process.exit(1);
+        }
+    }
+    // Path B: Direct token + VM (advanced)
+    else if (args.token && args.vm) {
         config = {
             gatewayToken: args.token,
             vmAddress: args.vm,
@@ -60,40 +66,39 @@ async function main() {
         };
         saveConfig(config);
     }
-    if (!config?.gatewayToken || !config?.vmAddress) {
-        // Interactive setup
+    // Path C: No args — interactive setup
+    else if (!config?.gatewayToken || !config?.vmAddress) {
         config = await interactiveSetup(config);
         saveConfig(config);
         console.log(`  Config saved to ${getConfigPath()}\n`);
     }
     // Set supervisor mode
-    if (config.mode === "autonomous") {
+    if (config.mode === "autonomous" || args.autonomous) {
+        config.mode = "autonomous";
         setMode("autonomous");
     }
-    console.log(`  VM:   ${config.vmAddress}:${config.port}`);
-    console.log(`  Mode: ${config.mode}`);
-    console.log(`  Token: ${config.gatewayToken.substring(0, 8)}...\n`);
+    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
     connect({
         vmAddress: config.vmAddress,
         port: config.port,
         gatewayToken: config.gatewayToken,
         mode: config.mode,
-        rejectUnauthorized: false, // Accept self-signed (TOFU)
+        rejectUnauthorized: false,
         certFingerprint: config.certFingerprint,
         onConnect: () => {
-            if (config.mode === "autonomous") {
-                console.log(`\n  Dispatch mode: AUTONOMOUS`);
-                console.log(`  All actions auto-approved (dangerous actions still require confirmation).`);
-            }
-            else {
-                console.log(`\n  Dispatch mode: SUPERVISED`);
-                console.log(`  Screenshots auto-approved. Actions require [Enter] to confirm.`);
-            }
-            console.log(`  Press 'a' to toggle autonomous mode. Press Ctrl+C to disconnect.\n`);
-            // Runtime mode switching via keypress
+            console.log("");
+            console.log(`  ${chalk.green("✓")} ${chalk.bold("Connected to your agent!")}`);
+            console.log(`  ${chalk.dim("Mode:")} ${config.mode === "autonomous" ? chalk.yellow("Autonomous") : chalk.green("Supervised")}`);
+            console.log(`  ${chalk.dim("Your agent can now control this computer.")}`);
+            console.log("");
+            console.log(`  ${chalk.dim("Press 'a' to toggle mode. Ctrl+C to disconnect.")}`);
+            console.log("");
+            // Runtime mode switching
             if (process.stdin.isTTY) {
-                process.stdin.setRawMode?.(false); // Ensure line mode for supervisor readline
+                process.stdin.setRawMode?.(false);
                 process.stdin.on("data", (chunk) => {
                     const key = chunk.toString().trim().toLowerCase();
                     if (key === "a") {
@@ -101,71 +106,101 @@ async function main() {
                         config.mode = newMode;
                         setMode(newMode);
                         saveConfig(config);
-                        console.log(`\n  Mode switched to: ${newMode.toUpperCase()}\n`);
+                        console.log(`\n  Mode: ${newMode === "autonomous" ? chalk.yellow("AUTONOMOUS") : chalk.green("SUPERVISED")}\n`);
                     }
                 });
             }
         },
-        onDisconnect: () => {
-            // Will auto-reconnect
-        },
+        onDisconnect: () => { },
         onCertFingerprint: (fp) => {
-            // TOFU: save fingerprint on first connection
             if (config && !config.certFingerprint) {
                 config.certFingerprint = fp;
                 saveConfig(config);
-                console.log(`  TLS fingerprint saved (TOFU): ${fp.substring(0, 20)}...`);
             }
         },
     });
-    // 4. Handle Ctrl+C
+    // 4. Ctrl+C
     process.on("SIGINT", () => {
-        console.log("\n  Disconnecting...");
+        console.log(`\n  ${chalk.dim("Disconnecting...")}`);
         disconnect();
         closeSupervisor();
         process.exit(0);
     });
 }
+// ── Pairing Code ──
+async function redeemPairingCode(code) {
+    try {
+        const res = await fetch(`${API_BASE}/api/vm/dispatch-pair/${encodeURIComponent(code)}`);
+        const data = await res.json();
+        if (data.error) {
+            console.log(`  ${chalk.red("✗")} ${data.error}`);
+            return null;
+        }
+        return { token: data.token, vmAddress: data.vmAddress };
+    }
+    catch (err) {
+        console.log(`  ${chalk.red("✗")} Failed to redeem pairing code: ${err}`);
+        return null;
+    }
+}
+// ── Interactive Setup ──
 async function interactiveSetup(existing) {
     const rl = readline.createInterface({ input: process.stdin, output: process.stdout });
     const ask = (q, def) => new Promise((resolve) => {
         const prompt = def ? `${q} [${def}]: ` : `${q}: `;
         rl.question(`  ${prompt}`, (ans) => resolve(ans.trim() || def || ""));
     });
-    console.log("  First-time setup:\n");
-    let token = await ask("Gateway token", existing?.gatewayToken);
-    let vm = await ask("VM address (IP or hostname)", existing?.vmAddress);
-    // Input validation: detect swapped fields
-    const isHexLong = (s) => /^[0-9a-f]{32,}$/i.test(s);
-    const isIPish = (s) => /^\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}$/.test(s);
-    if (isHexLong(vm)) {
-        console.log(chalk.yellow("\n  That VM address looks like a gateway token (long hex string)."));
-        console.log(chalk.yellow("  The VM address should be an IP like 104.237.145.128"));
-        const swap = await ask("Did you swap the fields? Enter your VM IP now");
-        if (swap) {
-            token = vm;
-            vm = swap;
+    console.log(`  ${chalk.bold("Welcome!")} Let's connect to your agent.\n`);
+    console.log(`  Got a pairing code from ${chalk.underline("instaclaw.io/settings")}?`);
+    const input = await ask("Enter pairing code (or 'a' for advanced setup)");
+    // Check if it's a pairing code (format: XXXX-YYYY)
+    if (input.toLowerCase() === "a") {
+        // Advanced: manual token + IP
+        console.log("");
+        let token = await ask("Gateway token", existing?.gatewayToken);
+        let vm = await ask("VM address (IP)", existing?.vmAddress);
+        // Swap detection
+        const isHexLong = (s) => /^[0-9a-f]{32,}$/i.test(s);
+        const isIPish = (s) => /^\d{1,3}\.\d{1,3}\.\d{1,3}\.\d{1,3}$/.test(s);
+        if (isHexLong(vm)) {
+            console.log(chalk.yellow("\n  That looks like a gateway token, not an IP address."));
+            const swap = await ask("Enter your VM IP");
+            if (swap) {
+                token = vm;
+                vm = swap;
+            }
         }
-    }
-    else if (isIPish(token)) {
-        console.log(chalk.yellow("\n  That gateway token looks like an IP address."));
-        console.log(chalk.yellow("  The gateway token should be a long hex string."));
-        const swap = await ask("Did you swap the fields? Enter your gateway token now");
-        if (swap) {
-            vm = token;
-            token = swap;
+        else if (isIPish(token)) {
+            console.log(chalk.yellow("\n  That looks like an IP, not a gateway token."));
+            const swap = await ask("Enter your gateway token");
+            if (swap) {
+                vm = token;
+                token = swap;
+            }
         }
+        rl.close();
+        return { gatewayToken: token, vmAddress: vm, port: 8765, mode: "supervised" };
+    }
+    // It's a pairing code or something else
+    const isHexLong = (s) => /^[0-9a-f]{32,}$/i.test(s);
+    if (isHexLong(input)) {
+        console.log(chalk.yellow("\n  That looks like a gateway token, not a pairing code."));
+        console.log(`  Get your pairing code at ${chalk.underline("instaclaw.io/settings")}`);
+        console.log(`  Or press 'a' for advanced setup.\n`);
+        rl.close();
+        return interactiveSetup(existing); // Retry
     }
-    const portStr = await ask("Port", String(existing?.port || 8765));
+    // Try to redeem as pairing code
     rl.close();
-    return {
-        gatewayToken: token,
-        vmAddress: vm,
-        port: parseInt(portStr, 10) || 8765,
-        mode: "supervised",
-        certFingerprint: existing?.certFingerprint,
-    };
+    const result = await redeemPairingCode(input);
+    if (result) {
+        return { gatewayToken: result.token, vmAddress: result.vmAddress, port: 8765, mode: "supervised" };
+    }
+    // Failed — fall back to advanced
+    console.log("  Falling back to advanced setup...\n");
+    return interactiveSetup(existing);
 }
+// ── Arg Parsing ──
 function parseArgs() {
     const args = process.argv.slice(2);
     const result = {};
@@ -182,6 +217,9 @@ function parseArgs() {
         else if (args[i] === "--autonomous" || args[i] === "--auto") {
             result.autonomous = true;
         }
+        else if ((args[i] === "--pair" || args[i] === "--code") && args[i + 1]) {
+            result.pair = args[++i];
+        }
     }
     return result;
 }

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": "0.3.1",
+  "version": "0.5.0",
   "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"