open-agents-ai 0.187.573 → 0.187.574

package/dist/index.js

@@ -252600,7 +252600,9 @@ async function probeService() {
   try {
     const controller = new AbortController();
     const timeout2 = setTimeout(() => controller.abort(), 3e3);
-    const res = await fetch(`${BASE_URL}/health`, { signal: controller.signal });
+    const res = await fetch(`${BASE_URL}/health`, {
+      signal: controller.signal
+    });
     clearTimeout(timeout2);
     return res.ok;
   } catch {
@@ -252610,7 +252612,10 @@ async function probeService() {
 function findPython3() {
   for (const cmd of ["python3", "python"]) {
     try {
-      const ver = execSync19(`${cmd} --version 2>&1`, { stdio: "pipe", timeout: 5e3 }).toString().trim();
+      const ver = execSync19(`${cmd} --version 2>&1`, {
+        stdio: "pipe",
+        timeout: 5e3
+      }).toString().trim();
       if (ver.includes("Python 3"))
         return cmd;
     } catch {
@@ -252682,7 +252687,10 @@ async function ensureSession() {
   });
   const data = await res.json();
   if (!data.ok)
-    return { error: String(data.message ?? "Failed to start browser session"), sessionId: "" };
+    return {
+      error: String(data.message ?? "Failed to start browser session"),
+      sessionId: ""
+    };
   activeSessionId = data.session_id;
   return { sessionId: activeSessionId };
 }
@@ -252828,7 +252836,22 @@ var init_browser_action = __esm({
         properties: {
           action: {
             type: "string",
-            enum: ["navigate", "click", "click_xy", "type", "screenshot", "dom", "dom_summary", "vision_click", "scroll", "scroll_up", "scroll_down", "back", "forward", "close"],
+            enum: [
+              "navigate",
+              "click",
+              "click_xy",
+              "type",
+              "screenshot",
+              "dom",
+              "dom_summary",
+              "vision_click",
+              "scroll",
+              "scroll_up",
+              "scroll_down",
+              "back",
+              "forward",
+              "close"
+            ],
             description: "Browser action to perform. Key actions:\n- 'dom_summary': compact view of interactive elements (~1KB vs 200KB raw DOM)\n- 'vision_click': screenshot the page, use Moondream vision to find an element by description, then click it. Pass the element description in 'text' parameter (e.g. text='the login button'). This is the visual grounding loop from SeeAct.\n- 'click': click by CSS selector (fastest when you know the selector)\n- 'click_xy': click at pixel coordinates (when you have exact coords)"
           },
           url: {
@@ -252858,12 +252881,44 @@ var init_browser_action = __esm({
         },
         required: ["action"]
       };
+      /** TASK-CLEANUP: gracefully close the browser session when the task completes. */
+      async cleanup() {
+        if (activeSessionId) {
+          try {
+            const res = await fetch(`${BASE_URL}/session/close`, {
+              method: "POST",
+              headers: { "Content-Type": "application/json" },
+              body: JSON.stringify({ sid: activeSessionId }),
+              signal: AbortSignal.timeout(5e3)
+            });
+            await res.json();
+          } catch {
+          }
+          activeSessionId = null;
+        }
+        if (serviceProcess && serviceProcess.pid && !serviceProcess.killed) {
+          try {
+            process.kill(-serviceProcess.pid, "SIGKILL");
+          } catch {
+          }
+          try {
+            serviceProcess.kill("SIGKILL");
+          } catch {
+          }
+          serviceProcess = null;
+        }
+      }
       async execute(args) {
         const start2 = Date.now();
         const action = args.action;
         const launchErr = await launchService();
         if (launchErr) {
-          return { success: false, output: "", error: launchErr, durationMs: Date.now() - start2 };
+          return {
+            success: false,
+            output: "",
+            error: launchErr,
+            durationMs: Date.now() - start2
+          };
         }
         if (action === "close") {
           if (activeSessionId) {
@@ -252873,21 +252928,41 @@ var init_browser_action = __esm({
             }
             activeSessionId = null;
           }
-          return { success: true, output: "Browser session closed.", durationMs: Date.now() - start2 };
+          return {
+            success: true,
+            output: "Browser session closed.",
+            durationMs: Date.now() - start2
+          };
         }
         const session = await ensureSession();
         if (session.error) {
-          return { success: false, output: "", error: session.error, durationMs: Date.now() - start2 };
+          return {
+            success: false,
+            output: "",
+            error: session.error,
+            durationMs: Date.now() - start2
+          };
         }
         try {
           let result;
           switch (action) {
             case "navigate": {
               if (!args.url)
-                return { success: false, output: "", error: "url is required for navigate action", durationMs: Date.now() - start2 };
-              result = await apiCall("/navigate", "POST", { url: args.url });
+                return {
+                  success: false,
+                  output: "",
+                  error: "url is required for navigate action",
+                  durationMs: Date.now() - start2
+                };
+              result = await apiCall("/navigate", "POST", {
+                url: args.url
+              });
               if (result.ok) {
-                return { success: true, output: `Navigated to ${args.url}`, durationMs: Date.now() - start2 };
+                return {
+                  success: true,
+                  output: `Navigated to ${args.url}`,
+                  durationMs: Date.now() - start2
+                };
               }
               const navMsg = String(result.message ?? "Navigation failed");
               const navHint = navMsg.toLowerCase().includes("connection") || navMsg.toLowerCase().includes("refused") || navMsg.toLowerCase().includes("err_connection") ? " (the URL appears unreachable — check if the target server is running and accepting connections)" : navMsg.toLowerCase().includes("timeout") ? " (page load timed out — try again or use a different URL)" : "";
@@ -252900,10 +252975,21 @@ var init_browser_action = __esm({
             }
             case "click": {
               if (!args.selector)
-                return { success: false, output: "", error: "selector is required for click action", durationMs: Date.now() - start2 };
-              result = await apiCall("/click", "POST", { selector: args.selector });
+                return {
+                  success: false,
+                  output: "",
+                  error: "selector is required for click action",
+                  durationMs: Date.now() - start2
+                };
+              result = await apiCall("/click", "POST", {
+                selector: args.selector
+              });
               if (result.ok) {
-                return { success: true, output: `Clicked element: ${args.selector}`, durationMs: Date.now() - start2 };
+                return {
+                  success: true,
+                  output: `Clicked element: ${args.selector}`,
+                  durationMs: Date.now() - start2
+                };
               }
               const clickMsg = String(result.message ?? "Click failed");
               return {
@@ -252915,10 +253001,19 @@ var init_browser_action = __esm({
             }
             case "click_xy": {
               if (args.x == null || args.y == null)
-                return { success: false, output: "", error: "x and y are required for click_xy action", durationMs: Date.now() - start2 };
+                return {
+                  success: false,
+                  output: "",
+                  error: "x and y are required for click_xy action",
+                  durationMs: Date.now() - start2
+                };
               result = await apiCall("/click_xy", "POST", { x: args.x, y: args.y });
               if (result.ok) {
-                return { success: true, output: `Clicked at (${args.x}, ${args.y})`, durationMs: Date.now() - start2 };
+                return {
+                  success: true,
+                  output: `Clicked at (${args.x}, ${args.y})`,
+                  durationMs: Date.now() - start2
+                };
               }
               const xyMsg = String(result.message ?? "Click failed");
               return {
@@ -252930,10 +253025,22 @@ var init_browser_action = __esm({
             }
             case "type": {
               if (!args.selector || !args.text)
-                return { success: false, output: "", error: "selector and text are required for type action", durationMs: Date.now() - start2 };
-              result = await apiCall("/type", "POST", { selector: args.selector, text: args.text });
+                return {
+                  success: false,
+                  output: "",
+                  error: "selector and text are required for type action",
+                  durationMs: Date.now() - start2
+                };
+              result = await apiCall("/type", "POST", {
+                selector: args.selector,
+                text: args.text
+              });
               if (result.ok) {
-                return { success: true, output: `Typed "${args.text.slice(0, 50)}" into ${args.selector}`, durationMs: Date.now() - start2 };
+                return {
+                  success: true,
+                  output: `Typed "${args.text.slice(0, 50)}" into ${args.selector}`,
+                  durationMs: Date.now() - start2
+                };
               }
               const typeMsg = String(result.message ?? "Type failed");
               return {
@@ -252981,16 +253088,30 @@ var init_browser_action = __esm({
                   durationMs: Date.now() - start2
                 };
               }
-              return { success: false, output: "", error: "Screenshot failed", durationMs: Date.now() - start2 };
+              return {
+                success: false,
+                output: "",
+                error: "Screenshot failed",
+                durationMs: Date.now() - start2
+              };
             }
             case "dom": {
               result = await apiCall("/dom", "GET");
               const dom = result.dom;
               if (dom) {
                 const truncated = dom.length > 5e4 ? dom.slice(0, 5e4) + "\n... (truncated)" : dom;
-                return { success: true, output: truncated, durationMs: Date.now() - start2 };
+                return {
+                  success: true,
+                  output: truncated,
+                  durationMs: Date.now() - start2
+                };
               }
-              return { success: false, output: "", error: "DOM capture failed", durationMs: Date.now() - start2 };
+              return {
+                success: false,
+                output: "",
+                error: "DOM capture failed",
+                durationMs: Date.now() - start2
+              };
             }
             // dom_summary: Research-grounded DOM downsampling
             // Paper: AgentOccam (arXiv:2410.13825, ICLR 2025) — pivotal node extraction
@@ -253004,9 +253125,18 @@ var init_browser_action = __esm({
               result = await apiCall("/dom", "GET");
               const rawDom = result.dom;
               if (!rawDom)
-                return { success: false, output: "", error: "DOM capture failed", durationMs: Date.now() - start2 };
+                return {
+                  success: false,
+                  output: "",
+                  error: "DOM capture failed",
+                  durationMs: Date.now() - start2
+                };
               const summary = downsampleDom(rawDom);
-              return { success: true, output: summary, durationMs: Date.now() - start2 };
+              return {
+                success: true,
+                output: summary,
+                durationMs: Date.now() - start2
+              };
             }
             // vision_click: Screenshot → Moondream point detection → Click
             // Paper: SeeAct (arXiv:2401.01614) — visual grounding for web agents
@@ -253019,14 +253149,24 @@ var init_browser_action = __esm({
             case "vision_click": {
               const target = args.text;
               if (!target)
-                return { success: false, output: "", error: "text parameter is required for vision_click — describe what to click (e.g. 'the login button')", durationMs: Date.now() - start2 };
+                return {
+                  success: false,
+                  output: "",
+                  error: "text parameter is required for vision_click — describe what to click (e.g. 'the login button')",
+                  durationMs: Date.now() - start2
+                };
               const ssResult = await apiCall("/screenshot", "GET");
               const ssB64 = ssResult.b64;
               const ssWidth = ssResult.width || 1280;
               const ssHeight = ssResult.height || 720;
               const ssFile = ssResult.file;
               if (!ssB64 && !ssFile) {
-                return { success: false, output: "", error: "Screenshot failed — cannot perform vision click", durationMs: Date.now() - start2 };
+                return {
+                  success: false,
+                  output: "",
+                  error: "Screenshot failed — cannot perform vision click",
+                  durationMs: Date.now() - start2
+                };
               }
               let imagePath = "";
               if (ssFile) {
@@ -253038,7 +253178,12 @@ var init_browser_action = __esm({
                   wfs(tmpPath, fileBuffer);
                   imagePath = tmpPath;
                 } catch (e2) {
-                  return { success: false, output: "", error: `Failed to save screenshot: ${e2}`, durationMs: Date.now() - start2 };
+                  return {
+                    success: false,
+                    output: "",
+                    error: `Failed to save screenshot: ${e2}`,
+                    durationMs: Date.now() - start2
+                  };
                 }
               } else if (ssB64) {
                 const tmpPath = join41(process.env["TMPDIR"] || "/tmp", `oa-vision-click-${Date.now()}.png`);
@@ -253056,7 +253201,12 @@ var init_browser_action = __esm({
                   prompt: target
                 });
                 if (!visionResult.success) {
-                  return { success: false, output: `Vision could not find "${target}" on the page. Try using dom_summary to find the CSS selector instead.`, error: visionResult.error, durationMs: Date.now() - start2 };
+                  return {
+                    success: false,
+                    output: `Vision could not find "${target}" on the page. Try using dom_summary to find the CSS selector instead.`,
+                    error: visionResult.error,
+                    durationMs: Date.now() - start2
+                  };
                 }
                 const coordMatch = visionResult.output.match(/\((\d+\.?\d*),\s*(\d+\.?\d*)\)/);
                 if (coordMatch) {
@@ -253066,10 +253216,19 @@ var init_browser_action = __esm({
                   pointY = Math.round(normY * ssHeight);
                 }
               } catch (e2) {
-                return { success: false, output: "", error: `Vision detection failed: ${e2}`, durationMs: Date.now() - start2 };
+                return {
+                  success: false,
+                  output: "",
+                  error: `Vision detection failed: ${e2}`,
+                  durationMs: Date.now() - start2
+                };
               }
               if (pointX < 0 || pointY < 0) {
-                return { success: false, output: `Could not determine click coordinates for "${target}". Vision returned no valid points.`, durationMs: Date.now() - start2 };
+                return {
+                  success: false,
+                  output: `Could not determine click coordinates for "${target}". Vision returned no valid points.`,
+                  durationMs: Date.now() - start2
+                };
               }
               const clickResult = await apiCall("/click_xy", "POST", {
                 x: pointX,
@@ -253092,22 +253251,49 @@ var init_browser_action = __esm({
               };
             }
             case "scroll":
-              result = await apiCall("/scroll", "POST", { amount: args.amount ?? 600 });
-              return { success: !!result.ok, output: `Scrolled ${args.amount ?? 600}px`, durationMs: Date.now() - start2 };
+              result = await apiCall("/scroll", "POST", {
+                amount: args.amount ?? 600
+              });
+              return {
+                success: !!result.ok,
+                output: `Scrolled ${args.amount ?? 600}px`,
+                durationMs: Date.now() - start2
+              };
             case "scroll_up":
               result = await apiCall("/scroll/up", "POST");
-              return { success: !!result.ok, output: "Scrolled up", durationMs: Date.now() - start2 };
+              return {
+                success: !!result.ok,
+                output: "Scrolled up",
+                durationMs: Date.now() - start2
+              };
             case "scroll_down":
               result = await apiCall("/scroll/down", "POST");
-              return { success: !!result.ok, output: "Scrolled down", durationMs: Date.now() - start2 };
+              return {
+                success: !!result.ok,
+                output: "Scrolled down",
+                durationMs: Date.now() - start2
+              };
             case "back":
               result = await apiCall("/history/back", "POST");
-              return { success: !!result.ok, output: "Navigated back", durationMs: Date.now() - start2 };
+              return {
+                success: !!result.ok,
+                output: "Navigated back",
+                durationMs: Date.now() - start2
+              };
             case "forward":
               result = await apiCall("/history/forward", "POST");
-              return { success: !!result.ok, output: "Navigated forward", durationMs: Date.now() - start2 };
+              return {
+                success: !!result.ok,
+                output: "Navigated forward",
+                durationMs: Date.now() - start2
+              };
             default:
-              return { success: false, output: "", error: `Unknown action: ${action}. Available: navigate, click, click_xy, type, screenshot, dom, scroll, scroll_up, scroll_down, back, forward, close`, durationMs: Date.now() - start2 };
+              return {
+                success: false,
+                output: "",
+                error: `Unknown action: ${action}. Available: navigate, click, click_xy, type, screenshot, dom, scroll, scroll_up, scroll_down, back, forward, close`,
+                durationMs: Date.now() - start2
+              };
           }
         } catch (err) {
           return {
@@ -534893,6 +535079,14 @@ ${sr.result.output}`;
                         } else {
                           completed = true;
                           summary = extractTaskCompleteSummary(r2.tc.arguments);
+                          for (const tool of this.tools.values()) {
+                            if (tool.cleanup) {
+                              try {
+                                await tool.cleanup();
+                              } catch {
+                              }
+                            }
+                          }
                           if (summary && !this._assistantTextEmitted) {
                             this.emit({
                               type: "assistant_text",
@@ -613853,7 +614047,9 @@ function adaptTool6(tool) {
         output: result.output,
         error: result.error
       };
-    }
+    },
+    // Pass through lifecycle hooks from the underlying Tool implementation
+    cleanup: tool.cleanup
   };
 }
 function scanForSessionSignals(toolOutput) {

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "open-agents-ai",
-  "version": "0.187.573",
+  "version": "0.187.574",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "open-agents-ai",
-      "version": "0.187.573",
+      "version": "0.187.574",
       "hasInstallScript": true,
       "license": "CC-BY-NC-4.0",
       "dependencies": {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "open-agents-ai",
-  "version": "0.187.573",
+  "version": "0.187.574",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",