@getpaseo/server 0.1.104-beta.2 → 0.1.104-beta.3

package/dist/server/server/browser-tools/tools.js

@@ -14,6 +14,16 @@ const BrowserToolOutputSchema = {
         retryable: z.boolean(),
     })
         .optional(),
+    dialogs: z
+        .array(z.object({
+        type: z.enum(["alert", "confirm", "prompt", "beforeunload"]),
+        message: z.string(),
+        defaultValue: z.string().optional(),
+        action: z.enum(["accepted", "dismissed"]),
+        promptText: z.string().optional(),
+        timestamp: z.number(),
+    }))
+        .optional(),
     context: z
         .object({
         agentId: z.string().optional(),
@@ -38,6 +48,8 @@ const BrowserHttpUrlInputSchema = z
     return normalized;
 });
 const BrowserRefInputSchema = z.string().regex(/^@e\d+$/);
+const BrowserClickButtonInputSchema = z.enum(["left", "right", "middle"]);
+const BrowserClickModifierInputSchema = z.enum(["Alt", "Control", "Meta", "Shift"]);
 const BrowserWaitInputSchema = z
     .object({
     text: z.string().min(1).optional(),
@@ -51,7 +63,7 @@ const BrowserWaitInputSchema = z
 export function registerBrowserTools(options) {
     options.registerTool("browser_list_tabs", {
         title: "List browser tabs",
-        description: "List open Paseo desktop browser tabs for this agent's workspace. Use returned browserId values with tab-scoped tools.",
+        description: "List open Paseo browser tabs for this agent's workspace across connected browser automation hosts. Use returned browserId values with tab-scoped tools.",
         inputSchema: {},
         outputSchema: BrowserToolOutputSchema,
     }, async () => {
@@ -73,7 +85,7 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_new_tab", {
         title: "Create browser tab",
-        description: "Create and focus a new Paseo desktop browser tab in this agent's workspace. Pass an http(s) URL or a scheme-less host URL, which is treated as http; the returned browserId is used by tab-scoped tools.",
+        description: "Create a new Paseo browser tab in this agent's workspace on the most recently connected browser automation host, opened in the background without switching the user's view. Pass an http(s) URL or a scheme-less host URL, which is treated as http; the returned browserId is used by tab-scoped tools.",
         inputSchema: {
             url: BrowserHttpUrlInputSchema.optional(),
         },
@@ -97,7 +109,7 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_snapshot", {
         title: "Snapshot browser page",
-        description: "Return a model-readable snapshot of a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
+        description: "Return a model-readable snapshot of a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
         inputSchema: {
             browserId: BrowserAutomationBrowserIdSchema,
         },
@@ -119,13 +131,16 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_click", {
         title: "Click browser element",
-        description: "Click an element in a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
+        description: "Click an element in a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
         inputSchema: {
             ref: BrowserRefInputSchema,
             browserId: BrowserAutomationBrowserIdSchema,
+            button: BrowserClickButtonInputSchema.optional(),
+            doubleClick: z.boolean().optional(),
+            modifiers: z.array(BrowserClickModifierInputSchema).optional(),
         },
         outputSchema: BrowserToolOutputSchema,
-    }, async ({ ref, browserId }) => {
+    }, async ({ ref, browserId, button, doubleClick, modifiers }) => {
         const context = resolveBrowserToolContext(options);
         const payload = await options.broker.execute({
             agentId: context.agentId,
@@ -136,6 +151,9 @@ export function registerBrowserTools(options) {
                 args: {
                     browserId,
                     ref,
+                    button: button ?? "left",
+                    doubleClick: doubleClick ?? false,
+                    modifiers: modifiers ?? [],
                 },
             },
         });
@@ -143,7 +161,7 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_fill", {
         title: "Fill browser element",
-        description: "Fill an input-like element in a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
+        description: "Fill an input-like element in a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
         inputSchema: {
             ref: BrowserRefInputSchema,
             value: z.string(),
@@ -169,7 +187,7 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_wait", {
         title: "Wait for browser condition",
-        description: "Wait until a Paseo desktop browser tab contains text or reaches a URL fragment. Use browserId from browser_new_tab or browser_list_tabs; waits up to 5s by default on the desktop side.",
+        description: "Wait until a Paseo browser tab contains text or reaches a URL fragment. Use browserId from browser_new_tab or browser_list_tabs; waits up to 5s by default on the browser host.",
         inputSchema: BrowserWaitInputSchema,
         outputSchema: BrowserToolOutputSchema,
     }, async ({ text, url, timeoutMs, browserId }) => {
@@ -245,7 +263,7 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_navigate", {
         title: "Navigate browser",
-        description: "Navigate a Paseo desktop browser tab to a URL. Use browserId from browser_new_tab or browser_list_tabs; pass an http(s) URL or a scheme-less host URL, which is treated as http.",
+        description: "Navigate a Paseo browser tab to a URL. Use browserId from browser_new_tab or browser_list_tabs; pass an http(s) URL or a scheme-less host URL, which is treated as http.",
         inputSchema: { url: BrowserHttpUrlInputSchema, browserId: BrowserAutomationBrowserIdSchema },
         outputSchema: BrowserToolOutputSchema,
     }, async ({ url, browserId }) => {
@@ -269,19 +287,19 @@ export function registerBrowserTools(options) {
             name: "browser_back",
             command: "back",
             title: "Browser back",
-            description: "Go back in a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs.",
+            description: "Go back in a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs.",
         },
         {
             name: "browser_forward",
             command: "forward",
             title: "Browser forward",
-            description: "Go forward in a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs.",
+            description: "Go forward in a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs.",
         },
         {
             name: "browser_reload",
             command: "reload",
             title: "Browser reload",
-            description: "Reload a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs.",
+            description: "Reload a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs.",
         },
     ]) {
         options.registerTool(toolConfig.name, {
@@ -307,7 +325,7 @@ export function registerBrowserTools(options) {
     }
     options.registerTool("browser_screenshot", {
         title: "Capture browser screenshot",
-        description: "Capture a PNG screenshot of a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs. Set fullPage to true to capture the full page.",
+        description: "Capture a PNG screenshot of a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs. Set fullPage to true to capture the full page.",
         inputSchema: {
             browserId: BrowserAutomationBrowserIdSchema,
             fullPage: z.boolean().default(false),
@@ -331,7 +349,7 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_upload", {
         title: "Upload files in browser",
-        description: "Set workspace files on a file input in a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
+        description: "Set workspace files on a file input in a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
         inputSchema: {
             ref: BrowserRefInputSchema,
             filePaths: z.array(z.string().min(1)).min(1),
@@ -360,7 +378,7 @@ export function registerBrowserTools(options) {
             name: "browser_hover",
             command: "hover",
             title: "Hover browser element",
-            description: "Hover an element in a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
+            description: "Hover an element in a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
         },
     ]) {
         options.registerTool(toolConfig.name, {
@@ -387,7 +405,7 @@ export function registerBrowserTools(options) {
     }
     options.registerTool("browser_select", {
         title: "Select browser option",
-        description: "Set a select element in a Paseo desktop browser tab to a value. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
+        description: "Set a select element in a Paseo browser tab to a value. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
         inputSchema: {
             ref: BrowserRefInputSchema,
             value: z.string(),
@@ -413,7 +431,7 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_drag", {
         title: "Drag browser element",
-        description: "Drag one element onto another in a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
+        description: "Drag one element onto another in a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs; refs come from the latest browser_snapshot of the same tab and expire when the page changes.",
         inputSchema: {
             sourceRef: BrowserRefInputSchema,
             targetRef: BrowserRefInputSchema,
@@ -439,7 +457,7 @@ export function registerBrowserTools(options) {
     });
     options.registerTool("browser_logs", {
         title: "Read browser logs",
-        description: "Read recent console messages and browser performance network entries for a Paseo desktop browser tab. Use browserId from browser_new_tab or browser_list_tabs; maxEntries defaults to 50.",
+        description: "Read recent console messages and browser performance network entries for a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs; maxEntries defaults to 50.",
         inputSchema: {
             maxEntries: z.number().int().positive().max(200).optional(),
             browserId: BrowserAutomationBrowserIdSchema,
@@ -461,6 +479,108 @@ export function registerBrowserTools(options) {
         });
         return browserToolResult({ payload, context: { ...context, browserId } });
     });
+    options.registerTool("browser_evaluate", {
+        title: "Evaluate browser JavaScript",
+        description: "Evaluate a JavaScript function in a Paseo browser tab. Use browserId from browser_new_tab or browser_list_tabs; when ref is provided, refs come from the latest browser_snapshot and the resolved element is passed as the first argument.",
+        inputSchema: {
+            function: z.string().min(1),
+            ref: BrowserRefInputSchema.optional(),
+            browserId: BrowserAutomationBrowserIdSchema,
+        },
+        outputSchema: BrowserToolOutputSchema,
+    }, async ({ function: functionSource, ref, browserId }) => {
+        const context = resolveBrowserToolContext(options);
+        const payload = await options.broker.execute({
+            agentId: context.agentId,
+            cwd: context.cwd,
+            ...(context.workspaceId ? { workspaceId: context.workspaceId } : {}),
+            command: {
+                command: "evaluate",
+                args: {
+                    browserId,
+                    function: functionSource,
+                    ...(ref ? { ref } : {}),
+                },
+            },
+        });
+        return browserToolResult({ payload, context: { ...context, browserId } });
+    });
+    options.registerTool("browser_scroll", {
+        title: "Scroll browser",
+        description: "Scroll a Paseo browser tab by deltaX/deltaY CSS pixels. Use browserId from browser_new_tab or browser_list_tabs; optional ref comes from the latest browser_snapshot and centers the wheel input over that element.",
+        inputSchema: {
+            browserId: BrowserAutomationBrowserIdSchema,
+            ref: BrowserRefInputSchema.optional(),
+            deltaX: z.number(),
+            deltaY: z.number(),
+        },
+        outputSchema: BrowserToolOutputSchema,
+    }, async ({ browserId, ref, deltaX, deltaY }) => {
+        const context = resolveBrowserToolContext(options);
+        const payload = await options.broker.execute({
+            agentId: context.agentId,
+            cwd: context.cwd,
+            ...(context.workspaceId ? { workspaceId: context.workspaceId } : {}),
+            command: {
+                command: "scroll",
+                args: {
+                    browserId,
+                    ...(ref ? { ref } : {}),
+                    deltaX,
+                    deltaY,
+                },
+            },
+        });
+        return browserToolResult({ payload, context: { ...context, browserId } });
+    });
+    options.registerTool("browser_resize", {
+        title: "Resize browser viewport",
+        description: "Resize a Paseo browser tab's resident webview viewport. Use browserId from browser_new_tab or browser_list_tabs.",
+        inputSchema: {
+            browserId: BrowserAutomationBrowserIdSchema,
+            width: z.number().int().positive(),
+            height: z.number().int().positive(),
+        },
+        outputSchema: BrowserToolOutputSchema,
+    }, async ({ browserId, width, height }) => {
+        const context = resolveBrowserToolContext(options);
+        const payload = await options.broker.execute({
+            agentId: context.agentId,
+            cwd: context.cwd,
+            ...(context.workspaceId ? { workspaceId: context.workspaceId } : {}),
+            command: {
+                command: "resize",
+                args: {
+                    browserId,
+                    width,
+                    height,
+                },
+            },
+        });
+        return browserToolResult({ payload, context: { ...context, browserId } });
+    });
+    options.registerTool("browser_close_tab", {
+        title: "Close browser tab",
+        description: "Close a Paseo browser tab, remove its resident webview, and unregister it from the browser automation host. Use browserId from browser_new_tab or browser_list_tabs.",
+        inputSchema: {
+            browserId: BrowserAutomationBrowserIdSchema,
+        },
+        outputSchema: BrowserToolOutputSchema,
+    }, async ({ browserId }) => {
+        const context = resolveBrowserToolContext(options);
+        const payload = await options.broker.execute({
+            agentId: context.agentId,
+            cwd: context.cwd,
+            ...(context.workspaceId ? { workspaceId: context.workspaceId } : {}),
+            command: {
+                command: "close_tab",
+                args: {
+                    browserId,
+                },
+            },
+        });
+        return browserToolResult({ payload, context: { ...context, browserId } });
+    });
 }
 function resolveBrowserToolContext(options) {
     const callerAgent = options.resolveCallerAgent();
@@ -518,15 +638,22 @@ function browserToolResult(params) {
             structuredContent: {
                 ok: true,
                 result: browserToolStructuredResult(payload.result),
+                ...(payload.dialogs ? { dialogs: payload.dialogs } : {}),
                 context,
             },
         };
     }
     return {
-        content: [{ type: "text", text: summarizeBrowserError(payload.error) }],
+        content: [
+            {
+                type: "text",
+                text: appendDialogSummary(summarizeBrowserError(payload.error), payload.dialogs),
+            },
+        ],
         structuredContent: {
             ok: false,
             error: payload.error,
+            ...(payload.dialogs ? { dialogs: payload.dialogs } : {}),
             context,
         },
     };
@@ -554,29 +681,30 @@ function browserToolImageContent(result) {
     };
 }
 function summarizeBrowserSuccess(payload) {
+    const withDialogs = (summary) => appendDialogSummary(summary, payload.dialogs);
     const controlSummary = summarizeBrowserControlSuccess(payload.result);
     if (controlSummary) {
-        return controlSummary;
+        return withDialogs(controlSummary);
     }
     const refActionSummary = summarizeBrowserRefActionSuccess(payload.result);
     if (refActionSummary) {
-        return refActionSummary;
+        return withDialogs(refActionSummary);
     }
     const diagnosticsSummary = summarizeBrowserDiagnosticsSuccess(payload.result);
     if (diagnosticsSummary) {
-        return diagnosticsSummary;
+        return withDialogs(diagnosticsSummary);
     }
     const keyboardSummary = summarizeBrowserKeyboardSuccess(payload.result);
     if (keyboardSummary) {
-        return keyboardSummary;
+        return withDialogs(keyboardSummary);
     }
     const navigationSummary = summarizeBrowserNavigationSuccess(payload.result);
     if (navigationSummary) {
-        return navigationSummary;
+        return withDialogs(navigationSummary);
     }
     const mediaSummary = summarizeBrowserMediaSuccess(payload.result);
     if (mediaSummary) {
-        return mediaSummary;
+        return withDialogs(mediaSummary);
     }
     if (payload.result.command === "list_tabs") {
         const count = payload.result.tabs.length;
@@ -587,22 +715,35 @@ function summarizeBrowserSuccess(payload) {
             const active = tab.isActive ? " active" : "";
             return `- browserId=${tab.browserId}${active} title=${JSON.stringify(tab.title || "Untitled")} url=${tab.url}`;
         });
-        return [
+        return withDialogs([
             `Found ${count} Paseo browser tab${count === 1 ? "" : "s"}. Use these browserId values for tab-scoped browser tools.`,
             ...tabLines,
-        ].join("\n");
+        ].join("\n"));
     }
     if (payload.result.command === "new_tab") {
-        return `Created browser tab browserId=${payload.result.browserId} url=${payload.result.url}. Use this browserId for tab-scoped browser tools.`;
+        return withDialogs(`Created browser tab browserId=${payload.result.browserId} url=${payload.result.url}. Use this browserId for tab-scoped browser tools.`);
     }
     if (payload.result.command === "snapshot") {
-        const count = payload.result.elements.length;
-        return `Snapshot captured ${count} element${count === 1 ? "" : "s"}.`;
+        return withDialogs([
+            `Snapshot captured ${payload.result.stats.nodeCount} node${payload.result.stats.nodeCount === 1 ? "" : "s"} with ${payload.result.stats.refCount} ref${payload.result.stats.refCount === 1 ? "" : "s"}.`,
+            `Title: ${payload.result.title || "Untitled"}`,
+            `URL: ${payload.result.url}`,
+            "",
+            payload.result.snapshot,
+        ].join("\n"));
     }
     if (payload.result.command === "wait") {
-        return `Browser wait matched ${payload.result.matched}.`;
+        return withDialogs(`Browser wait matched ${payload.result.matched}.`);
     }
-    return `Browser ${payload.result.command} complete.`;
+    return withDialogs(`Browser ${payload.result.command} complete.`);
+}
+function appendDialogSummary(summary, dialogs) {
+    if (!dialogs || dialogs.length === 0) {
+        return summary;
+    }
+    return `${summary}\nHandled browser dialog${dialogs.length === 1 ? "" : "s"}: ${dialogs
+        .map((dialog) => `${dialog.action} ${dialog.type} ${JSON.stringify(dialog.message)}`)
+        .join("; ")}.`;
 }
 function summarizeBrowserMediaSuccess(result) {
     if (result.command === "screenshot") {
@@ -637,6 +778,13 @@ function summarizeBrowserNavigationSuccess(result) {
     return null;
 }
 function summarizeBrowserDiagnosticsSuccess(result) {
+    if (result.command === "evaluate") {
+        return [
+            "Browser evaluate returned:",
+            result.resultJson,
+            ...(result.truncated ? ["Result was truncated."] : []),
+        ].join("\n");
+    }
     if (result.command !== "logs") {
         return null;
     }
@@ -663,20 +811,31 @@ function summarizeBrowserControlSuccess(result) {
     if (result.command === "drag") {
         return `Dragged browser element ${result.sourceRef} to ${result.targetRef}.`;
     }
+    if (result.command === "scroll") {
+        return result.ref
+            ? `Scrolled browser element ${result.ref} by ${result.deltaX}, ${result.deltaY}.`
+            : `Scrolled browser by ${result.deltaX}, ${result.deltaY}.`;
+    }
+    if (result.command === "resize") {
+        return `Resized browser viewport to ${result.width}x${result.height}.`;
+    }
+    if (result.command === "close_tab") {
+        return `Closed browser tab ${result.browserId}.`;
+    }
     return null;
 }
 function summarizeBrowserError(error) {
     switch (error.code) {
         case "browser_disabled":
-            return "Browser tools are disabled. Enable desktop browser tools on the host, then try again.";
-        case "browser_no_desktop":
-            return "No desktop browser automation client is connected. Open the Paseo desktop app and try again.";
+            return "Browser tools are disabled. Enable browser tools on the host, then try again.";
+        case "browser_no_host":
+            return error.message;
         case "browser_timeout":
-            return "The browser did not respond before the timeout. Try again or check the desktop app.";
+            return "The browser did not respond before the timeout. Try again or check the browser host.";
         case "screenshot_no_frame":
             return error.message;
         case "browser_unsupported":
-            return "This desktop build does not support that browser automation request yet.";
+            return error.message;
         case "browser_stale_ref":
             return "That browser element reference is stale. Take a new browser snapshot and try again.";
         default:

package/dist/server/server/session.d.ts

@@ -28,7 +28,7 @@ import { LoopService } from "./loop-service.js";
 import { ScheduleService } from "./schedule/service.js";
 import { type GitHubService } from "../services/github-service.js";
 import type { ProviderUsageService } from "../services/quota-fetcher/service.js";
-import { generateBranchNameFromFirstAgentContext } from "./worktree-branch-name-generator.js";
+import { WorkspaceAutoName } from "./workspace-auto-name.js";
 export declare function resolveWaitForFinishError(options: {
     status: "permission" | "error" | "idle";
     final: AgentSnapshotPayload | null;
@@ -68,8 +68,8 @@ export interface SessionOptions {
     github?: GitHubService;
     createAgentMcpTransport?: AgentMcpTransportFactory;
     renameCurrentBranch?: typeof renameCurrentBranchDefault;
-    generateWorkspaceName?: typeof generateBranchNameFromFirstAgentContext;
     workspaceGitService: WorkspaceGitService;
+    workspaceAutoName: WorkspaceAutoName;
     daemonConfigStore: DaemonConfigStore;
     mcpBaseUrl?: string | null;
     stt: Resolvable<SpeechToTextProvider | null>;
@@ -141,8 +141,8 @@ export declare class Session {
     private readonly filesystem;
     private readonly github;
     private readonly renameCurrentBranch;
-    private readonly generateWorkspaceName;
     private readonly workspaceGitService;
+    private readonly workspaceAutoName;
     private readonly gitMutation;
     private readonly workspaceProvisioning;
     private readonly daemonConfigStore;
@@ -275,12 +275,6 @@ export declare class Session {
     private handleCancelAgentRequest;
     private handleAgentRewindRequest;
     private buildAgentSessionConfig;
-    private scheduleAutoNameWorkspaceBranchForFirstAgent;
-    private maybeAutoNameWorkspaceBranchForFirstAgent;
-    private applyGeneratedWorkspaceTitle;
-    private generateWorkspaceTitleFromContext;
-    private maybeAutoNameDirectoryWorkspaceTitle;
-    private scheduleAutoNameLocalWorkspaceTitleForFirstAgent;
     private isPathWithinRoot;
     /**
      * Handle clearing agent attention flag
@@ -355,7 +349,6 @@ export declare class Session {
     private registerWorkspaceForImportedAgent;
     private handleWorkspaceCreateRequest;
     private handleWorkspaceCreateLocal;
-    private scheduleWorkspaceNaming;
     private handleWorkspaceCreateWorktree;
     private resolveWorktreeSourceCwd;
     private handleOpenProjectRequest;