omnius 1.0.196 → 1.0.197

package/dist/index.js

@@ -3871,6 +3871,18 @@ var init_system_deps = __esm({
         description: "Screenshot capture tool",
         group: "screenshot"
       },
+      {
+        command: "gnome-screenshot",
+        packages: { apt: "gnome-screenshot", dnf: "gnome-screenshot", pacman: "gnome-screenshot" },
+        description: "GNOME screenshot capture tool, useful on Wayland sessions",
+        group: "screenshot"
+      },
+      {
+        command: "grim",
+        packages: { apt: "grim", dnf: "grim", pacman: "grim" },
+        description: "Wayland screenshot capture tool for wlroots compositors",
+        group: "screenshot"
+      },
       {
         command: "xdotool",
         packages: { apt: "xdotool", dnf: "xdotool", pacman: "xdotool" },
@@ -3919,8 +3931,8 @@ import { existsSync as existsSync6, statSync as statSync3 } from "node:fs";
 function captureDesktopScreenshot(outputPath3) {
   const attempts = [];
   const out = quoteShell(outputPath3);
-  const tryCapture = (label, command) => {
-    const result = run(command, 1e4);
+  const tryCapture = (label, command, timeout2 = 1e4) => {
+    const result = run(command, timeout2);
     if (result.ok && existsSync6(outputPath3)) {
       const inspection = inspectScreenshot(outputPath3);
       if (inspection.ok)
@@ -3950,6 +3962,17 @@ $bitmap.Dispose()
     if (backend)
       return backend;
   } else if (process.platform === "linux") {
+    const desktop = `${process.env["XDG_CURRENT_DESKTOP"] || ""} ${process.env["DESKTOP_SESSION"] || ""}`;
+    if ((/wayland/i.test(process.env["XDG_SESSION_TYPE"] || "") || process.env["WAYLAND_DISPLAY"]) && hasCommand2("gdbus") && hasCommand2("dbus-monitor")) {
+      const backend = tryCapture("xdg-desktop-portal-screenshot", portalScreenshotCommand(outputPath3), 13e4);
+      if (backend)
+        return backend;
+    }
+    if (/gnome/i.test(desktop) && hasCommand2("gdbus")) {
+      const backend = tryCapture("gnome-shell-screenshot-dbus", `gdbus call --session --dest org.gnome.Shell.Screenshot --object-path /org/gnome/Shell/Screenshot --method org.gnome.Shell.Screenshot.Screenshot false false ${out}`);
+      if (backend)
+        return backend;
+    }
     const candidates = [
       ["gnome-screenshot", `gnome-screenshot -f ${out}`],
       ["grim", `grim ${out}`],
@@ -3965,6 +3988,20 @@ $bitmap.Dispose()
       if (backend)
         return backend;
     }
+    if (/wayland/i.test(process.env["XDG_SESSION_TYPE"] || "") || process.env["WAYLAND_DISPLAY"]) {
+      for (const binary of [/gnome/i.test(desktop) ? "gnome-screenshot" : "", "grim"].filter(Boolean)) {
+        const dep2 = ensureCommand(binary);
+        if (!dep2.available) {
+          if (dep2.error)
+            attempts.push({ label: `${binary} auto-install`, message: dep2.error });
+          continue;
+        }
+        const command = binary === "gnome-screenshot" ? `gnome-screenshot -f ${out}` : `grim ${out}`;
+        const backend = tryCapture(dep2.installed ? `${binary} (auto-installed)` : binary, command);
+        if (backend)
+          return backend;
+      }
+    }
     const dep = ensureCommand("scrot");
     if (dep.available) {
       const backend = tryCapture(dep.installed ? "scrot (auto-installed)" : "scrot", `scrot ${out}`);
@@ -3983,6 +4020,73 @@ $bitmap.Dispose()
   }
   throw new Error("No desktop screenshot backend succeeded.\n" + formatDesktopAttempts(attempts) + "\n" + desktopAutomationRecoveryMessage());
 }
+function portalScreenshotCommand(outputPath3) {
+  const script = `
+set -eu
+out=${quoteShell(outputPath3)}
+monitor_file="$(mktemp)"
+cleanup() {
+  if [ -n "\${monpid:-}" ]; then kill "$monpid" >/dev/null 2>&1 || true; fi
+  rm -f "$monitor_file"
+}
+trap cleanup EXIT
+
+call_out="$(gdbus call --session --dest org.freedesktop.portal.Desktop --object-path /org/freedesktop/portal/desktop --method org.freedesktop.portal.Screenshot.Screenshot "" "{'interactive': <true>, 'modal': <true>}" 2>&1)" || {
+  printf '%s\\n' "$call_out" >&2
+  exit 1
+}
+handle="$(printf '%s\\n' "$call_out" | sed -n "s/.*objectpath '\\([^']*\\)'.*/\\1/p")"
+if [ -z "$handle" ]; then
+  printf 'Portal screenshot did not return a request handle: %s\\n' "$call_out" >&2
+  exit 1
+fi
+
+dbus-monitor --session "type='signal',interface='org.freedesktop.portal.Request',member='Response',path='$handle'" > "$monitor_file" 2>&1 &
+monpid=$!
+deadline=$(( $(date +%s) + 120 ))
+while :; do
+  if grep -q "member=Response" "$monitor_file"; then break; fi
+  if ! kill -0 "$monpid" >/dev/null 2>&1; then break; fi
+  if [ "$(date +%s)" -ge "$deadline" ]; then
+    printf 'Timed out waiting for the desktop screenshot permission prompt.\\n' >&2
+    exit 124
+  fi
+  sleep 0.2
+done
+kill "$monpid" >/dev/null 2>&1 || true
+
+response_code="$(awk '/member=Response/{seen=1; next} seen && /uint32/{print $2; exit}' "$monitor_file")"
+if [ "$response_code" != "0" ]; then
+  printf 'Screenshot permission was denied or cancelled by the desktop portal (response %s).\\n' "\${response_code:-unknown}" >&2
+  cat "$monitor_file" >&2
+  exit 1
+fi
+
+uri="$(sed -n 's/.*string "\\(file:[^"]*\\)".*/\\1/p' "$monitor_file" | tail -n 1)"
+if [ -z "$uri" ]; then
+  printf 'Desktop portal granted screenshot permission but did not return a file URI.\\n' >&2
+  cat "$monitor_file" >&2
+  exit 1
+fi
+
+if command -v gio >/dev/null 2>&1; then
+  gio copy -f "$uri" "$out"
+elif command -v python3 >/dev/null 2>&1; then
+  python3 - "$uri" "$out" <<'PY'
+import shutil, sys, urllib.parse
+uri, out = sys.argv[1], sys.argv[2]
+parsed = urllib.parse.urlparse(uri)
+if parsed.scheme != "file":
+    raise SystemExit(f"Unsupported screenshot URI scheme: {parsed.scheme}")
+shutil.copyfile(urllib.parse.unquote(parsed.path), out)
+PY
+else
+  printf 'Need gio or python3 to copy portal screenshot URI %s to %s.\\n' "$uri" "$out" >&2
+  exit 1
+fi
+`.trim();
+  return `bash -lc ${quoteShell(script)}`;
+}
 function moveDesktopPointer(x, y) {
   const result = performDesktopPointerAction({ x, y, moveOnly: true });
   if (!result.ok)
@@ -4111,10 +4215,13 @@ function desktopAutomationRecoveryMessage(command) {
     `DISPLAY=${display} WAYLAND_DISPLAY=${wayland}`,
     "For web pages and local dev servers, prefer browser_action or playwright_browser screenshots/clicks; they do not need OS desktop access.",
     "For OS-level desktop control, Omnius tries platform backends in order:",
-    "  Linux: xdotool/X11, ydotool, dotool, python-xlib",
+    "  Linux screenshots: xdg-desktop-portal permission prompt, GNOME Shell D-Bus/gnome-screenshot, grim, scrot, ImageMagick import, Python ImageGrab",
+    "  Linux pointer control: xdotool/X11, ydotool, dotool, python-xlib",
     "  macOS: cliclick, then System Events",
     "  Windows: PowerShell user32 input",
-    "On Wayland, install and enable ydotool or dotool when xdotool cannot open an X display."
+    "On Wayland, Omnius requests screenshot permission through xdg-desktop-portal when available. Approve the system screenshot prompt to continue.",
+    "On GNOME Wayland, unattended screenshots may still be denied by compositor policy. Install gnome-screenshot or grant screenshot permission for the session if capture is blocked.",
+    "On Wayland pointer control, install and enable ydotool or dotool when xdotool cannot open an X display."
   ].join("\n");
 }
 function performDesktopPointerAction(options2) {
@@ -270549,6 +270656,9 @@ State: ${statePath}${clearArtifacts ? "\nArtifacts cleared: true" : ""}`,
           }
           const stamp = timestampSlug();
           const screenshotPath = join47(sessionDir2, `${stamp}-step-${step}-before.png`);
+          if (process.platform === "linux" && (process.env["WAYLAND_DISPLAY"] || /wayland/i.test(process.env["XDG_SESSION_TYPE"] || ""))) {
+            yield "Vision action loop: requesting desktop screenshot permission if the system prompts";
+          }
           yield `Vision action loop: capturing screenshot ${step}/${maxSteps}`;
           let screenshotBackend = "";
           try {
@@ -270641,6 +270751,9 @@ State: ${statePath}${clearArtifacts ? "\nArtifacts cleared: true" : ""}`,
                 actionTaken = operation === "move" ? `Moved pointer to (${pixelX}, ${pixelY}) via ${backend}` : `Clicked at (${pixelX}, ${pixelY}) via ${backend} [${button} ${clickType}]`;
                 outputLines.push(actionTaken);
                 afterScreenshotPath = join47(sessionDir2, `${timestampSlug()}-step-${step}-after.png`);
+                if (process.platform === "linux" && (process.env["WAYLAND_DISPLAY"] || /wayland/i.test(process.env["XDG_SESSION_TYPE"] || ""))) {
+                  yield "Vision action loop: requesting desktop screenshot permission for post-action verification if the system prompts";
+                }
                 yield `Vision action loop: capturing post-action screenshot ${step}/${maxSteps}`;
                 const afterBackend = captureDesktopScreenshot(afterScreenshotPath);
                 mutatedFiles.push(afterScreenshotPath);
@@ -528648,7 +528761,7 @@ var init_full_sub_agent = __esm({
     _activeSubProcesses = /* @__PURE__ */ new Map();
     FullSubAgentTool = class {
       name = "full_sub_agent";
-      description = "Spawn a COMPLETE Omnius sub-process with ALL tools and capabilities. Unlike sub_agent (shared process, limited tools), this spawns a separate omnius instance with its own clean context window, full tool set, memory, skills, and COHERE access. Use for complex multi-file tasks that benefit from a fresh context. The sub-process runs omnius --non-interactive and its output appears in the tab bar. Returns immediately with a process ID for monitoring.";
+      description = "Spawn a COMPLETE Omnius sub-process with ALL tools and capabilities. Unlike sub_agent (shared process, limited tools), this spawns a separate omnius instance with its own clean context window, full tool set, memory, skills, and COHERE access. Use for complex multi-file tasks that benefit from a fresh context. The sub-process runs omnius --non-interactive until task_complete, abort, or timeout. By default this tool returns immediately with a process ID for monitoring; set wait=true to block until the child exits.";
       parameters = {
         type: "object",
         properties: {
@@ -528668,6 +528781,14 @@ var init_full_sub_agent = __esm({
           id: {
             type: "string",
             description: "Sub-agent process ID (for status/output/stop)"
+          },
+          wait: {
+            type: "boolean",
+            description: "For action='spawn', wait for the subprocess to exit instead of returning immediately. The child omnius run uses unlimited turns by default and stops on task_complete, abort, or timeout."
+          },
+          timeout_ms: {
+            type: "number",
+            description: "Optional timeout passed through to the child omnius process."
           }
         },
         required: []
@@ -528706,6 +528827,8 @@ var init_full_sub_agent = __esm({
             if (!task)
               return { success: false, output: "", error: "task is required", durationMs: performance.now() - start2 };
             const model = String(args["model"] ?? this.model);
+            const wait = args["wait"] === true;
+            const timeoutMs = typeof args["timeout_ms"] === "number" && Number.isFinite(args["timeout_ms"]) ? Math.max(1, Math.floor(args["timeout_ms"])) : void 0;
             const broker = getModelBroker();
             const decision2 = await broker.ensureModelLoadable({
               name: model || "default",
@@ -528729,7 +528852,7 @@ var init_full_sub_agent = __esm({
                 durationMs: performance.now() - start2
               };
             }
-            const entry = spawnFullSubAgent(task, { model, backendUrl: this.backendUrl, workingDir: this.workingDir }, (text) => this.onViewWrite?.(entry.id, text), (id, exitCode, output) => {
+            const entry = spawnFullSubAgent(task, { model, backendUrl: this.backendUrl, workingDir: this.workingDir, timeoutMs }, (text) => this.onViewWrite?.(entry.id, text), (id, exitCode, output) => {
               this.onViewStatus?.(id, exitCode === 0 ? "completed" : "failed");
               broker.unregisterLoaded("subprocess", id, "sub-agent-exited");
               this.onComplete?.(id, task, exitCode, output);
@@ -528746,6 +528869,33 @@ var init_full_sub_agent = __esm({
               priority: 1
             });
             this.onViewRegister?.(entry.id, entry.id, "full");
+            if (wait) {
+              const exitCode = await this.waitForExit(entry);
+              const output = entry.outputBuffer.join("\n");
+              const tail = entry.outputBuffer.slice(-120).join("\n");
+              const durationMs = performance.now() - start2;
+              if (exitCode === 0) {
+                return {
+                  success: true,
+                  output: `Full Omnius sub-agent completed: ${entry.id}
+  PID: ${entry.pid}
+  Model: ${model}
+  Duration: ${(durationMs / 1e3).toFixed(1)}s
+  Output lines: ${entry.outputBuffer.length}` + (tail ? `
+
+${tail}` : ""),
+                  durationMs
+                };
+              }
+              return {
+                success: false,
+                output: output ? `Full Omnius sub-agent failed: ${entry.id}
+
+${tail || output}` : "",
+                error: `Full Omnius sub-agent exited before successful completion (exit code ${exitCode ?? "unknown"}).`,
+                durationMs
+              };
+            }
             return {
               success: true,
               output: `Full Omnius sub-agent spawned: ${entry.id}
@@ -528805,6 +528955,14 @@ ${tail}`, durationMs: performance.now() - start2 };
             return { success: false, output: "", error: `Unknown action: ${action}`, durationMs: performance.now() - start2 };
         }
       }
+      waitForExit(entry) {
+        if (entry.status !== "running")
+          return Promise.resolve(entry.exitCode);
+        return new Promise((resolvePromise) => {
+          entry.process.once("exit", (code8) => resolvePromise(code8));
+          entry.process.once("error", () => resolvePromise(-1));
+        });
+      }
     };
   }
 });
@@ -672837,7 +672995,11 @@ function createSubAgentTool(config, repoRoot, ctxWindowSize) {
         },
         max_turns: {
           type: "number",
-          description: "Maximum turns for the sub-agent (default: 15)"
+          description: "Maximum turns for the sub-agent (default: 15). Use 0 to run until task_complete or timeout."
+        },
+        until_task_complete: {
+          type: "boolean",
+          description: "If true, run with max_turns=0 so the sub-agent stops only at task_complete, abort, or timeout."
         }
       },
       required: ["task"]
@@ -672845,7 +673007,7 @@ function createSubAgentTool(config, repoRoot, ctxWindowSize) {
     async execute(args) {
       const task = String(args["task"] ?? "");
       const background = Boolean(args["background"]);
-      const maxTurns = typeof args["max_turns"] === "number" ? args["max_turns"] : 15;
+      const maxTurns = args["until_task_complete"] === true ? 0 : typeof args["max_turns"] === "number" ? args["max_turns"] : 15;
       if (!task) {
         return { success: false, output: "", error: "task is required" };
       }

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.196",
+  "version": "1.0.197",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.196",
+      "version": "1.0.197",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.196",
+  "version": "1.0.197",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",