open-agents-ai 0.23.0 → 0.23.2

package/dist/index.js

@@ -6961,20 +6961,45 @@ var init_desktop_click = __esm({
               durationMs: performance.now() - start
             };
           }
-          const { vl } = await import("moondream");
-          const apiKey = process.env["MOONDREAM_API_KEY"];
-          const endpoint = process.env["MOONDREAM_ENDPOINT"];
-          let client;
-          if (apiKey) {
-            client = new vl({ apiKey });
-          } else if (endpoint) {
-            client = new vl({ endpoint });
-          } else {
-            client = new vl({ endpoint: "http://localhost:2020/v1" });
+          let points = [];
+          let visionWorked = false;
+          try {
+            const { vl } = await import("moondream");
+            const apiKey = process.env["MOONDREAM_API_KEY"];
+            const endpoint = process.env["MOONDREAM_ENDPOINT"];
+            let client;
+            if (apiKey) {
+              client = new vl({ apiKey });
+            } else if (endpoint) {
+              client = new vl({ endpoint });
+            } else {
+              client = new vl({ endpoint: "http://localhost:2020/v1" });
+            }
+            const imageBuffer = readFileSync11(screenshotPath);
+            const pointResult = await client.point({ image: imageBuffer, object: target });
+            points = pointResult.points ?? [];
+            visionWorked = true;
+          } catch {
+          }
+          if (!visionWorked) {
+            const hints = [
+              `(Moondream vision not available \u2014 cannot locate "${target}" on screen)`,
+              `Screenshot saved: ${screenshotPath}`,
+              `Screen: ${dims.width}x${dims.height}`,
+              "",
+              "Use shell commands to interact with the desktop instead:",
+              "  xdotool search --name 'pattern' \u2014 find windows by title",
+              "  xdotool key 'ctrl+s' \u2014 send keyboard shortcuts",
+              "  xdotool mousemove X Y click 1 \u2014 click at known coordinates",
+              "  wmctrl -a 'window title' \u2014 activate a window by name",
+              "  xdg-open <url> \u2014 open a URL in the default browser"
+            ];
+            return {
+              success: true,
+              output: hints.join("\n"),
+              durationMs: performance.now() - start
+            };
           }
-          const imageBuffer = readFileSync11(screenshotPath);
-          const pointResult = await client.point({ image: imageBuffer, object: target });
-          const points = pointResult.points ?? [];
           if (points.length === 0) {
             return {
               success: false,
@@ -7066,37 +7091,62 @@ Screenshot: ${screenshotPath}`,
           captureScreenshot(screenshotPath);
           const dims = getImageDimensions2(screenshotPath);
           const imageBuffer = readFileSync11(screenshotPath);
-          const { vl } = await import("moondream");
-          const apiKey = process.env["MOONDREAM_API_KEY"];
-          const endpoint = process.env["MOONDREAM_ENDPOINT"];
-          let client;
-          if (apiKey) {
-            client = new vl({ apiKey });
-          } else if (endpoint) {
-            client = new vl({ endpoint });
-          } else {
-            client = new vl({ endpoint: "http://localhost:2020/v1" });
-          }
           const parts = [];
-          if (question) {
-            const result = await client.query({ image: imageBuffer, question });
-            const answer = typeof result.answer === "string" ? result.answer : "(streaming not supported)";
-            parts.push(`Q: ${question}`);
-            parts.push(`A: ${answer}`);
-          } else {
-            const result = await client.caption({ image: imageBuffer, length });
-            const caption = typeof result.caption === "string" ? result.caption : "(streaming not supported)";
-            parts.push(`Desktop description:
+          let visionWorked = false;
+          try {
+            const { vl } = await import("moondream");
+            const apiKey = process.env["MOONDREAM_API_KEY"];
+            const endpoint = process.env["MOONDREAM_ENDPOINT"];
+            let client;
+            if (apiKey) {
+              client = new vl({ apiKey });
+            } else if (endpoint) {
+              client = new vl({ endpoint });
+            } else {
+              client = new vl({ endpoint: "http://localhost:2020/v1" });
+            }
+            if (question) {
+              const result = await client.query({ image: imageBuffer, question });
+              const answer = typeof result.answer === "string" ? result.answer : "(streaming not supported)";
+              parts.push(`Q: ${question}`);
+              parts.push(`A: ${answer}`);
+            } else {
+              const result = await client.caption({ image: imageBuffer, length });
+              const caption = typeof result.caption === "string" ? result.caption : "(streaming not supported)";
+              parts.push(`Desktop description:
 ${caption}`);
+            }
+            visionWorked = true;
+          } catch {
+          }
+          if (!visionWorked) {
+            let ocrText = "";
+            try {
+              ocrText = execSync11(`tesseract ${JSON.stringify(screenshotPath)} stdout 2>/dev/null`, {
+                encoding: "utf8",
+                timeout: 15e3
+              }).trim();
+            } catch {
+            }
+            if (ocrText) {
+              parts.push("(Moondream vision not available \u2014 using OCR text extraction)");
+              parts.push(`
+Visible text on screen:
+${ocrText}`);
+            } else {
+              parts.push("(Moondream vision not available, OCR failed)");
+              parts.push("Screenshot captured but cannot describe contents.");
+              parts.push("Use shell commands to check window state instead:");
+              parts.push("  xdotool getactivewindow getwindowname  \u2014 get active window title");
+              parts.push("  wmctrl -l  \u2014 list all open windows");
+              parts.push("  xdg-open <url>  \u2014 open a URL in the default browser");
+            }
           }
           if (dims) {
             parts.push(`
 Screen: ${dims.width}x${dims.height}`);
           }
           parts.push(`Screenshot: ${screenshotPath}`);
-          const base64 = imageBuffer.toString("base64");
-          parts.push(`
-[IMAGE_BASE64:image/png:${base64}]`);
           return {
             success: true,
             output: parts.join("\n"),

package/dist/launcher.cjs

@@ -1,29 +1,32 @@
 #!/usr/bin/env node
 // Node version gate — uses only ES5 syntax so it parses on any Node version.
-// On old Node: offers to install Node 20 via nvm with y/n prompts.
+// On old Node: walks through nvm install, then Node 20, then reinstall.
 var nodeVersion = parseInt(process.versions.node, 10);
 
 if (nodeVersion < 18) {
   var os = require("os");
   var path = require("path");
+  var fs = require("fs");
   var childProcess = require("child_process");
   var readline = require("readline");
 
   var platform = os.platform();
   var arch = os.arch();
 
-  console.log(
-    "\n  open-agents requires Node.js >= 18 (you have " + process.version + ")" +
-    "\n  Platform: " + platform + "/" + arch + "\n"
-  );
+  console.log("");
+  console.log("  ┌─────────────────────────────────────────────────┐");
+  console.log("  │         open-agents — Node.js Upgrade           │");
+  console.log("  └─────────────────────────────────────────────────┘");
+  console.log("");
+  console.log("  Your Node.js:  " + process.version);
+  console.log("  Required:      >= 18.0.0");
+  console.log("  Platform:      " + platform + "/" + arch);
+  console.log("");
 
-  // Check if nvm is already installed
+  // Detect if nvm is already installed
   var nvmDir = process.env.NVM_DIR || path.join(os.homedir(), ".nvm");
   var hasNvm = false;
-  try {
-    require("fs").statSync(path.join(nvmDir, "nvm.sh"));
-    hasNvm = true;
-  } catch (e) {}
+  try { fs.statSync(path.join(nvmDir, "nvm.sh")); hasNvm = true; } catch (e) {}
 
   var rl = readline.createInterface({ input: process.stdin, output: process.stdout });
 
@@ -33,10 +36,10 @@ if (nodeVersion < 18) {
     });
   }
 
-  function run(cmd, opts) {
+  function run(cmd) {
     console.log("  $ " + cmd);
     try {
-      childProcess.execSync(cmd, Object.assign({ stdio: "inherit" }, opts || {}));
+      childProcess.execSync(cmd, { stdio: "inherit" });
       return true;
     } catch (e) {
       console.error("  Command failed: " + (e.message || e));
@@ -44,70 +47,123 @@ if (nodeVersion < 18) {
     }
   }
 
-  function installNode() {
-    if (!hasNvm) {
-      console.log("\n  nvm not found. Installing nvm first...\n");
+  function bail(msg) {
+    console.log(msg);
+    rl.close();
+    process.exit(1);
+  }
+
+  function doInstallNvm(next) {
+    if (hasNvm) {
+      console.log("  nvm found at " + nvmDir);
+      console.log("");
+      next();
+      return;
+    }
+    ask("  nvm is not installed. Install nvm now? [Y/n] ", function(a) {
+      if (a === "n" || a === "no") {
+        bail(
+          "\n  Install nvm manually:\n" +
+          "    curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.1/install.sh | bash\n" +
+          "    source ~/.bashrc\n" +
+          "    nvm install 20\n" +
+          "    npm i -g open-agents-ai\n"
+        );
+        return;
+      }
+      console.log("");
       var ok = run("curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.1/install.sh | bash");
       if (!ok) {
-        console.error("\n  nvm install failed. Install manually:");
-        console.error("    https://github.com/nvm-sh/nvm#installing-and-updating\n");
-        rl.close();
-        process.exit(1);
+        bail("\n  nvm install failed. See https://github.com/nvm-sh/nvm#installing-and-updating\n");
+        return;
       }
-      // Re-source nvm for this session
       nvmDir = process.env.NVM_DIR || path.join(os.homedir(), ".nvm");
-    } else {
-      console.log("\n  nvm found at " + nvmDir + "\n");
-    }
-
-    // nvm is a shell function, so we run through bash
-    var shell = process.env.SHELL || "/bin/bash";
-    var nvmCmd = 'export NVM_DIR="' + nvmDir + '" && ' +
-      '[ -s "$NVM_DIR/nvm.sh" ] && . "$NVM_DIR/nvm.sh" && ' +
-      "nvm install 20 && nvm use 20 && nvm alias default 20";
+      hasNvm = true;
+      console.log("");
+      next();
+    });
+  }
 
-    console.log("  Installing Node.js 20...\n");
-    var ok = run(shell + ' -c \'' + nvmCmd + "'");
-    if (!ok) {
-      console.error("\n  Node 20 install failed.\n");
-      rl.close();
-      process.exit(1);
-    }
+  function doInstallNode(next) {
+    ask("  Install Node.js 20 via nvm? [Y/n] ", function(a) {
+      if (a === "n" || a === "no") {
+        bail(
+          "\n  Install manually:\n" +
+          "    nvm install 20 && nvm alias default 20\n" +
+          "    npm i -g open-agents-ai\n"
+        );
+        return;
+      }
+      console.log("");
+      var shell = process.env.SHELL || "/bin/bash";
+      var cmd = 'export NVM_DIR="' + nvmDir + '" && ' +
+        '[ -s "$NVM_DIR/nvm.sh" ] && . "$NVM_DIR/nvm.sh" && ' +
+        "nvm install 20 && nvm use 20 && nvm alias default 20";
+      var ok = run(shell + " -c '" + cmd + "'");
+      if (!ok) {
+        bail("\n  Node 20 install failed.\n");
+        return;
+      }
+      console.log("");
+      next();
+    });
+  }
 
-    // Now reinstall open-agents under the new Node
-    console.log("\n  Node 20 installed. Reinstalling open-agents...\n");
-    var reinstallCmd = 'export NVM_DIR="' + nvmDir + '" && ' +
-      '[ -s "$NVM_DIR/nvm.sh" ] && . "$NVM_DIR/nvm.sh" && ' +
-      "npm i -g open-agents-ai";
-    var ok2 = run(shell + ' -c \'' + reinstallCmd + "'");
-    if (!ok2) {
-      console.error("\n  Reinstall failed. After opening a new terminal, run:");
-      console.error("    npm i -g open-agents-ai\n");
+  function doReinstall() {
+    ask("  Reinstall open-agents under Node 20? [Y/n] ", function(a) {
+      if (a === "n" || a === "no") {
+        console.log(
+          "\n  Open a new terminal, then run:\n" +
+          "    npm i -g open-agents-ai\n"
+        );
+        rl.close();
+        process.exit(0);
+        return;
+      }
+      console.log("");
+      var shell = process.env.SHELL || "/bin/bash";
+      var cmd = 'export NVM_DIR="' + nvmDir + '" && ' +
+        '[ -s "$NVM_DIR/nvm.sh" ] && . "$NVM_DIR/nvm.sh" && ' +
+        "nvm use 20 && npm i -g open-agents-ai";
+      var ok = run(shell + " -c '" + cmd + "'");
+      if (!ok) {
+        console.log(
+          "\n  Reinstall failed. Open a new terminal, then run:\n" +
+          "    npm i -g open-agents-ai\n"
+        );
+        rl.close();
+        process.exit(1);
+        return;
+      }
+      console.log("");
+      console.log("  ┌─────────────────────────────────────────────────┐");
+      console.log("  │  Done! Open a new terminal, then run:  oa       │");
+      console.log("  └─────────────────────────────────────────────────┘");
+      console.log("");
       rl.close();
-      process.exit(1);
-    }
-
-    console.log(
-      "\n  Done! Open a new terminal (or run: source ~/.bashrc) then run:" +
-      "\n    oa\n"
-    );
-    rl.close();
-    process.exit(0);
+      process.exit(0);
+    });
   }
 
-  ask("  Install Node.js 20 now? [Y/n] ", function(answer) {
-    if (answer === "n" || answer === "no") {
-      console.log(
+  // Walk through the steps
+  ask("  Upgrade to Node.js 20 now? [Y/n] ", function(a) {
+    if (a === "n" || a === "no") {
+      bail(
         "\n  To install manually:\n" +
         "    curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.1/install.sh | bash\n" +
         "    source ~/.bashrc\n" +
         "    nvm install 20\n" +
+        "    nvm alias default 20\n" +
         "    npm i -g open-agents-ai\n"
       );
-      rl.close();
-      process.exit(1);
+      return;
     }
-    installNode();
+    console.log("");
+    doInstallNvm(function() {
+      doInstallNode(function() {
+        doReinstall();
+      });
+    });
   });
 } else {
   // Node >= 18 — load the ESM bundle

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.23.0",
+  "version": "0.23.2",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",