0agent 1.0.58 → 1.0.60

package/bin/chat.js

@@ -175,6 +175,89 @@ function renderMarkdown(text) {
   return out.join('\n');
 }
 
+// ─── Human-readable active task summary ───────────────────────────────────────
+// Translates raw AgentExecutor step labels into a plain-English 1-liner shown
+// as the live "what is happening right now" status line.
+function stepToHuman(step) {
+  // Tool invocations: "▶ tool_name(args...)"
+  const toolMatch = step.match(/^▶\s+(\w+)\((.{0,200})\)/);
+  if (toolMatch) {
+    const [, tool, args] = toolMatch;
+    // gui_automation — decode action
+    if (tool === 'gui_automation') {
+      const act  = (args.match(/action[=:]\s*["']?(\w+)["']?/i) || [])[1] ?? '';
+      const txt  = (args.match(/text[=:]\s*["']([^"']{0,40})["']/i) || [])[1] ?? '';
+      const url  = (args.match(/url[=:]\s*["']([^"']{0,60})["']/i) || [])[1] ?? '';
+      const secs = (args.match(/seconds[=:]\s*([\d.]+)/i) || [])[1] ?? '';
+      const x    = (args.match(/\bx[=:]\s*(\d+)/i) || [])[1] ?? '';
+      const y    = (args.match(/\by[=:]\s*(\d+)/i) || [])[1] ?? '';
+      const map = {
+        screenshot:    'Taking screenshot of screen',
+        click:         x ? `Clicking at (${x}, ${y})` : 'Clicking',
+        double_click:  x ? `Double-clicking at (${x}, ${y})` : 'Double-clicking',
+        right_click:   'Right-clicking',
+        move:          x ? `Moving cursor to (${x}, ${y})` : 'Moving cursor',
+        type:          txt ? `Typing: "${txt.slice(0,30)}"` : 'Typing text',
+        hotkey:        txt ? `Pressing ${txt}` : 'Pressing hotkey',
+        scroll:        'Scrolling',
+        drag:          'Dragging',
+        find_and_click: txt ? `Clicking "${txt}"` : 'Finding and clicking',
+        open_url:      url ? `Opening ${url}` : 'Opening URL in browser',
+        open_app:      txt ? `Opening ${txt}` : 'Opening app',
+        get_screen_size: 'Getting screen size',
+        get_cursor_pos:  'Getting cursor position',
+        wait:          secs ? `Waiting ${secs}s for UI to load` : 'Waiting for UI',
+      };
+      return map[act] ?? `GUI: ${act}`;
+    }
+    // shell_exec
+    if (tool === 'shell_exec') {
+      const cmd = args.replace(/^["']|["']$/g, '').replace(/\\n/g, ' ').trim().slice(0, 60);
+      return `Running: ${cmd}`;
+    }
+    // web_search
+    if (tool === 'web_search') {
+      const q = (args.match(/["']([^"']{1,60})["']/) || [])[1] ?? args.slice(0,50);
+      return `Searching: ${q}`;
+    }
+    // file_op
+    if (tool === 'file_op') {
+      const op   = (args.match(/op[=:]\s*["']?(\w+)["']?/i) || [])[1] ?? '';
+      const path = (args.match(/path[=:]\s*["']([^"']{1,50})["']/i) || [])[1] ?? '';
+      const opMap = { write: 'Writing', read: 'Reading', list: 'Listing', mkdir: 'Creating folder', delete: 'Deleting' };
+      return `${opMap[op] ?? op}: ${path}`;
+    }
+    // browser_open / scrape_url
+    if (tool === 'browser_open' || tool === 'scrape_url') {
+      const url = (args.match(/["']([^"']{1,60})["']/) || [])[1] ?? args.slice(0,50);
+      return `Reading page: ${url}`;
+    }
+    // memory_write
+    if (tool === 'memory_write') {
+      const label = (args.match(/label[=:]\s*["']([^"']{1,40})["']/i) || [])[1] ?? '';
+      return `Saving to memory: ${label}`;
+    }
+    // generic fallback
+    const cleanArgs = args.replace(/^["'](.*)["']$/, '$1').replace(/\\n/g, ' ').slice(0, 50);
+    return `${tool}: ${cleanArgs}`;
+  }
+
+  // Result line: "  ↳ text" — show briefly as status, not in history
+  if (/^\s*↳/.test(step)) {
+    return step.replace(/^\s*↳\s*/, '').slice(0, 80);
+  }
+
+  // Thinking / Continuing
+  if (/^Thinking/.test(step)) return 'Thinking…';
+  if (/^Continuing/.test(step)) return 'Working on it…';
+
+  // Self-heal, skill, misc
+  if (/^↺/.test(step)) return step.slice(0, 80);
+  if (/^✓/.test(step)) return step.slice(0, 80);
+
+  return step.slice(0, 80);
+}
+
 // ─── Step formatter ───────────────────────────────────────────────────────────
 // Converts raw step labels from AgentExecutor into icon + clean readable form.
 function formatStep(step) {
@@ -370,12 +453,29 @@ function handleWsEvent(event) {
     case 'session.step': {
       spinner.stop();
       if (streaming) { process.stdout.write('\n'); streaming = false; streamLineCount = 0; }
-      const formatted = formatStep(event.step);
-      if (formatted !== null) {
-        process.stdout.write('\r\x1b[2K');
-        console.log(formatted);
+
+      const step = event.step;
+      const isToolCall   = /^▶\s+\w+\(/.test(step);          // ▶ tool(args)
+      const isResult     = /^\s*↳/.test(step);                // ↳ result
+      const isThinking   = /^(Thinking|Continuing)/.test(step);
+      const isSummary    = /^(Done|Files|Commands|Matched|Fetching|Loaded|Selected|Extracting|Querying|↺|✓)/.test(step);
+
+      // Print tool invocations and summary milestones to scrolling history
+      if (isToolCall || isSummary) {
+        const formatted = formatStep(step);
+        if (formatted !== null) {
+          process.stdout.write('\r\x1b[2K');
+          console.log(formatted);
+        }
       }
-      spinner.startSession(event.step.slice(0, 50));
+
+      // Always update the live 1-liner status (overwrites current line)
+      const human = stepToHuman(step);
+      if (human) {
+        const icon = isThinking ? fmt(C.dim, '⠋') : isResult ? fmt(C.dim, '↳') : fmt(C.cyan, '⚡');
+        process.stdout.write(`\r\x1b[2K  ${icon} ${fmt(C.dim, human)}`);
+      }
+
       rl.prompt(true);
       break;
     }
@@ -653,6 +753,7 @@ async function runTask(input) {
       try {
         const r = await fetch(`${BASE_URL}/api/sessions/${sid}`, { signal: AbortSignal.timeout(2000) });
         const session = await r.json();
+        globalThis._daemonMisses = 0; // daemon responded — reset miss counter
 
         // Show any new steps not yet shown via WS
         const steps = session.steps ?? [];
@@ -689,7 +790,26 @@ async function runTask(input) {
           resolve_?.();
           drainQueue(); // auto-process queued messages
         }
-      } catch {}
+      } catch {
+        // Daemon not responding — track misses
+        if (typeof _daemonMisses === 'undefined') globalThis._daemonMisses = 0;
+        globalThis._daemonMisses = (globalThis._daemonMisses ?? 0) + 1;
+        if (globalThis._daemonMisses >= 4) {
+          globalThis._daemonMisses = 0;
+          clearInterval(pollTimer);
+          spinner.stop();
+          process.stdout.write(`\r\x1b[2K\n  ${fmt(C.yellow, '⚠')}  Daemon stopped — restarting…\n\n`);
+          const r = await _spawnDaemon();
+          if (r === 'ok') {
+            process.stdout.write(`  ${fmt(C.green, '✓')} Daemon restarted. Re-send your message.\n\n`);
+          } else {
+            process.stdout.write(`  ${fmt(C.red, '✗')} Could not restart daemon. Run: ${fmt(C.dim, '0agent start')}\n\n`);
+          }
+          if (pendingResolve) { pendingResolve(); pendingResolve = null; }
+          sessionId = null; streaming = false; streamLineCount = 0;
+          rl.prompt();
+        }
+      }
     }, 800);
 
     return new Promise(resolve => { pendingResolve = resolve; });

package/dist/daemon.mjs

@@ -2367,6 +2367,8 @@ var init_ShellCapability = __esm({
       // These must never run autonomously — they survive uninstall and can
       // re-open apps (e.g. Brave) on every login or on a timer.
       static PERSISTENT_TASK_PATTERN = /crontab\s+-[eilr]|launchctl\s+load|launchctl\s+bootstrap|systemctl\s+enable|at\s+\d|make\s+login\s+item|LaunchAgents|LaunchDaemons|loginitems/i;
+      // Commands that make irreversible external state changes — require explicit user confirmation
+      static DESTRUCTIVE_PATTERN = /\bcurl\s+[^|&]*-[A-Za-z]*[XD]\s+(DELETE|POST|PUT|PATCH)\b|\bcurl\s+[^|&]*--(request|data)[=\s]+(DELETE|POST|PUT|PATCH)\b|rm\s+-[rf]{1,3}\s+[^|&;]{3}|DROP\s+TABLE|DELETE\s+FROM\s+\w/i;
       async execute(input, cwd, signal) {
         let command = String(input.command ?? "");
         const timeout = Number(input.timeout_ms ?? 3e4);
@@ -2378,6 +2380,13 @@ var init_ShellCapability = __esm({
             duration_ms: 0
           };
         }
+        if (_ShellCapability.DESTRUCTIVE_PATTERN.test(command)) {
+          return {
+            success: false,
+            output: `CONFIRM_REQUIRED: The command "${command.slice(0, 100)}" will make an irreversible change. Tell the user exactly what this will do and ask them to reply with explicit confirmation before you run it.`,
+            duration_ms: 0
+          };
+        }
         if (signal?.aborted) {
           return { success: false, output: "Cancelled.", duration_ms: 0 };
         }
@@ -2698,9 +2707,12 @@ var init_GUICapability = __esm({
             return { success: false, output: result.stderr.trim() || "Unknown error after install", duration_ms: Date.now() - start };
           }
           if (err.includes("accessibility") || err.includes("permission") || err.includes("AXIsProcessTrusted")) {
+            if (platform2() === "darwin") {
+              spawnSync4("open", ["x-apple.systempreferences:com.apple.preference.security?Privacy_Accessibility"], { timeout: 3e3 });
+            }
             return {
               success: false,
-              output: "macOS accessibility permission required. Go to: System Preferences \u2192 Privacy & Security \u2192 Accessibility \u2192 add Terminal (or the app running 0agent)",
+              output: "macOS Accessibility permission required for GUI automation.\n\u2192 System Settings has been opened automatically.\n\u2192 Go to: Privacy & Security \u2192 Accessibility \u2192 enable Terminal (or iTerm2 / the app running 0agent)\n\u2192 Then re-run your task.",
               duration_ms: Date.now() - start
             };
           }
@@ -2871,19 +2883,32 @@ for i, word in enumerate(data['text']):
         cy = data['top'][i] + data['height'][i] // 2
         found.append((cx, cy, word))
 
-try:
-    if found:
-        cx, cy, word = found[0]
-        pyautogui.click(cx, cy, duration=${duration})
-        print(f"Found '{word}' at ({cx},{cy}) \u2014 clicked")
+if found:
+    cx, cy, word = found[0]
+    pyautogui.click(cx, cy, duration=${duration})
+    print(f"Found '{word}' at ({cx},{cy}) \u2014 clicked")
+else:
+    # Retry once after a brief wait (element may still be loading)
+    time.sleep(1.5)
+    img2 = pyautogui.screenshot()
+    data2 = pytesseract.image_to_data(img2, output_type=pytesseract.Output.DICT)
+    found2 = []
+    for i, word in enumerate(data2['text']):
+        if target in word.lower() and int(data2['conf'][i]) > 40:
+            cx2 = data2['left'][i] + data2['width'][i] // 2
+            cy2 = data2['top'][i] + data2['height'][i] // 2
+            found2.append((cx2, cy2, word))
+    if found2:
+        cx2, cy2, word2 = found2[0]
+        pyautogui.click(cx2, cy2, duration=${duration})
+        print(f"Found '{word2}' at ({cx2},{cy2}) after retry \u2014 clicked")
     else:
-        print(f"Text '${safeText}' not found on screen. Take a screenshot to see current state.")
+        print(f"Text '${safeText}' not found on screen after retry. Take a screenshot to see what changed.")
         sys.exit(1)
-finally:
-    try:
-        os.remove(shot_path)
-    except Exception:
-        pass
+try:
+    os.remove(shot_path)
+except Exception:
+    pass
 `;
           }
           case "open_url": {
@@ -3221,6 +3246,7 @@ var init_AgentExecutor = __esm({
             break;
           }
           this.onStep(i === 0 ? "Thinking\u2026" : "Continuing\u2026");
+          if (messages.length > 28) this._compressHistory(messages);
           let response;
           let llmFailed = false;
           {
@@ -3241,6 +3267,14 @@ var init_AgentExecutor = __esm({
                 break;
               } catch (err) {
                 const msg = err instanceof Error ? err.message : String(err);
+                const isRateLimit = /RateLimit:\d+/.test(msg);
+                if (isRateLimit) {
+                  const waitSec = parseInt(msg.split(":")[1] ?? "30", 10);
+                  const waitMs = Math.min(waitSec * 1e3, 12e4);
+                  this.onStep(`Rate limited \u2014 waiting ${waitSec}s before retry\u2026`);
+                  await new Promise((r) => setTimeout(r, waitMs));
+                  continue;
+                }
                 const isTimeout = /timeout|AbortError|aborted/i.test(msg);
                 if (isTimeout && llmRetry < 2) {
                   llmRetry++;
@@ -3274,6 +3308,11 @@ var init_AgentExecutor = __esm({
             try {
               const capResult = await this.registry.execute(tc.name, tc.input, this.cwd, signal);
               result = capResult.output;
+              const MAX_TOOL_OUTPUT = 4e3;
+              if (result.length > MAX_TOOL_OUTPUT) {
+                result = result.slice(0, MAX_TOOL_OUTPUT) + `
+[...${result.length - MAX_TOOL_OUTPUT} chars truncated]`;
+              }
               if (capResult.fallback_used) {
                 this.onStep(`  (used fallback: ${capResult.fallback_used})`);
               }
@@ -3503,6 +3542,9 @@ content = element.text if element else page.get_all_text()` : `content = page.ge
           `- For research tasks: use web_search first, then scrape_url for full page content`,
           `- Use relative paths from the working directory`,
           `- Be concise in your final response: state what was done and where to find it`,
+          `- For tasks with 3+ distinct steps or multiple apps/services, BRIEFLY LIST the steps first, then execute one at a time`,
+          `- CONFIRM BEFORE SENDING: Before sending any message (WhatsApp, email, Slack, SMS, tweet), show the user the exact text and recipient and wait for explicit confirmation`,
+          `- CONFIRM BEFORE DELETING: Before deleting files, database records, or any data, state what will be deleted and confirm with the user`,
           ``,
           `\u2550\u2550\u2550 EXECUTION DISCIPLINE \u2014 follow strictly \u2550\u2550\u2550`,
           `- SEQUENTIAL: complete each step fully before starting the next. Never start step 2 while step 1 is still in progress.`,
@@ -3571,6 +3613,19 @@ content = element.text if element else page.get_all_text()` : `content = page.ge
         if (extra) lines.push(``, `Context:`, extra);
         return lines.join("\n");
       }
+      _compressHistory(messages) {
+        const KEEP_TAIL = 14;
+        if (messages.length <= KEEP_TAIL + 2) return;
+        const head = messages.slice(0, 1);
+        const tail = messages.slice(-KEEP_TAIL);
+        const middle = messages.slice(1, -KEEP_TAIL);
+        const toolResults = middle.filter((m) => m.role === "tool").map((m) => String(m.content).slice(0, 120).replace(/\n/g, " ")).join(" | ");
+        const summary = {
+          role: "user",
+          content: `[Earlier context compressed \u2014 ${middle.length} messages. Key tool results: ${toolResults.slice(0, 600)}]`
+        };
+        messages.splice(0, messages.length, ...head, summary, ...tail);
+      }
       /** Returns true if task is a self-modification request. Self-mod tasks get longer LLM timeouts. */
       isSelfModTask(task) {
         return SELF_MOD_PATTERN.test(task);
@@ -4471,6 +4526,10 @@ var LLMExecutor = class {
       signal: signal ? AbortSignal.any([signal, AbortSignal.timeout(12e4)]) : AbortSignal.timeout(12e4)
     });
     if (!res.ok) {
+      if (res.status === 429) {
+        const retryAfter = parseInt(res.headers.get("retry-after") ?? res.headers.get("x-ratelimit-reset-requests") ?? "30", 10);
+        throw new Error(`RateLimit:${Math.min(retryAfter, 120)}`);
+      }
       const err = await res.text();
       throw new Error(`Anthropic ${res.status}: ${err}`);
     }
@@ -4595,6 +4654,10 @@ var LLMExecutor = class {
       signal: signal ? AbortSignal.any([signal, AbortSignal.timeout(12e4)]) : AbortSignal.timeout(12e4)
     });
     if (!res.ok) {
+      if (res.status === 429) {
+        const retryAfter = parseInt(res.headers.get("retry-after") ?? "30", 10);
+        throw new Error(`RateLimit:${Math.min(retryAfter, 120)}`);
+      }
       const err = await res.text();
       throw new Error(`OpenAI ${res.status}: ${err}`);
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "0agent",
-  "version": "1.0.58",
+  "version": "1.0.60",
   "description": "A persistent, learning AI agent that runs on your machine. An agent that learns.",
   "private": false,
   "license": "Apache-2.0",