Browser tools: full accessibility snapshot, trusted input, dialogs, evaluate, and tab controls (#1881)

* feat(browser): replace flat snapshot with ARIA tree and page-side refs

* feat(browser): trusted CDP input with actionability waits

* feat(browser): auto-handle javascript dialogs

* feat(browser): add browser_evaluate

* feat(browser): add scroll, resize, and close_tab tools

* fix(browser): review fixes for snapshot/input/dialog seams

* fix(browser): address review findings on dialog capture, truncation, and keypress targets

* fix(harness): drop resize check that cannot run under parked compositor surfaces

* fix(browser): second review round — snapshot fidelity, key input, dialog scoping
This commit is contained in:
Mohamed Boudra
2026-07-04 01:38:07 +02:00
committed by GitHub
parent b28f291493
commit 9c220c2451
29 changed files with 5653 additions and 949 deletions
+155 -2
View File
@@ -1,9 +1,14 @@
import type { SessionInboundMessage, SessionOutboundMessage } from "@getpaseo/protocol/messages";
import { getDesktopHost, type DesktopHostBridge } from "@/desktop/host";
import { ensureResidentBrowserWebview as ensureResidentBrowserWebviewDefault } from "@/components/browser-webview-resident";
import { createWorkspaceBrowser } from "@/stores/browser-store";
import {
ensureResidentBrowserWebview as ensureResidentBrowserWebviewDefault,
removeResidentBrowserWebview,
resizeResidentBrowserWebview,
} from "@/components/browser-webview-resident";
import { createWorkspaceBrowser, getBrowserRecord, useBrowserStore } from "@/stores/browser-store";
import {
buildWorkspaceTabPersistenceKey,
collectAllTabs,
useWorkspaceLayoutStore,
} from "@/stores/workspace-layout-store";
@@ -114,6 +119,33 @@ async function handleBrowserAutomationRequest(params: {
return;
}
if (request.command.command === "resize") {
client.sendBrowserAutomationExecuteResponse({
type: "browser.automation.execute.response",
payload: resizeBrowserTabForRequest({ request, serverId }),
});
return;
}
if (request.command.command === "close_tab") {
try {
client.sendBrowserAutomationExecuteResponse({
type: "browser.automation.execute.response",
payload: await closeBrowserTabForRequest({
request,
serverId,
browserHost,
}),
});
} catch (error) {
client.sendBrowserAutomationExecuteResponse({
type: "browser.automation.execute.response",
payload: normalizeThrownBridgeError(request.requestId, error),
});
}
return;
}
if (!executeAutomationCommand) {
client.sendBrowserAutomationExecuteResponse({
type: "browser.automation.execute.response",
@@ -140,6 +172,127 @@ async function handleBrowserAutomationRequest(params: {
}
}
function resizeBrowserTabForRequest(params: {
request: BrowserAutomationExecuteRequest;
serverId?: string;
}): BrowserAutomationResponsePayload {
const { request, serverId } = params;
const command = request.command as Extract<
BrowserAutomationExecuteRequest["command"],
{ command: "resize" }
>;
const browserId = command.args.browserId;
if (!getBrowserRecord(browserId)) {
return browserAutomationFailure({
requestId: request.requestId,
code: "browser_tab_not_found",
message: `No browser tab found for ID: ${browserId}`,
});
}
const workspaceId = request.workspaceId;
if (serverId && workspaceId && !findWorkspaceBrowserTab({ serverId, workspaceId, browserId })) {
return browserAutomationFailure({
requestId: request.requestId,
code: "browser_tab_not_found",
message: `No browser tab found for ID: ${browserId}`,
});
}
const dimensions = resizeResidentBrowserWebview({
browserId,
width: command.args.width,
height: command.args.height,
});
if (!dimensions) {
return browserAutomationFailure({
requestId: request.requestId,
code: "browser_tab_not_found",
message: `No browser tab found for ID: ${browserId}`,
});
}
return {
requestId: request.requestId,
ok: true,
result: {
command: "resize",
browserId,
width: dimensions.width,
height: dimensions.height,
},
};
}
async function closeBrowserTabForRequest(params: {
request: BrowserAutomationExecuteRequest;
serverId?: string;
browserHost: DesktopHostBridge["browser"] | undefined;
}): Promise<BrowserAutomationResponsePayload> {
const { request, serverId, browserHost } = params;
const command = request.command as Extract<
BrowserAutomationExecuteRequest["command"],
{ command: "close_tab" }
>;
const browserId = command.args.browserId;
const workspaceId = request.workspaceId;
const workspaceTab = serverId
? findWorkspaceBrowserTab({ serverId, workspaceId, browserId })
: null;
if (!workspaceTab && (!serverId || !workspaceId)) {
return browserAutomationFailure({
requestId: request.requestId,
code: "browser_unsupported",
message: "Cannot close a browser tab without a workspace context.",
});
}
if (!workspaceTab || !getBrowserRecord(browserId)) {
return browserAutomationFailure({
requestId: request.requestId,
code: "browser_tab_not_found",
message: `No browser tab found for ID: ${browserId}`,
});
}
useWorkspaceLayoutStore.getState().closeTab(workspaceTab.workspaceKey, workspaceTab.tabId);
useBrowserStore.getState().removeBrowser(browserId);
removeResidentBrowserWebview(browserId);
await browserHost?.unregisterWorkspaceBrowser?.(browserId);
await browserHost?.clearPartition?.(browserId);
return {
requestId: request.requestId,
ok: true,
result: { command: "close_tab", browserId },
};
}
function findWorkspaceBrowserTab(input: {
serverId: string;
workspaceId: string | undefined;
browserId: string;
}): { workspaceKey: string; tabId: string } | null {
if (!input.workspaceId) {
return null;
}
const workspaceKey = buildWorkspaceTabPersistenceKey({
serverId: input.serverId,
workspaceId: input.workspaceId,
});
if (!workspaceKey) {
return null;
}
const layout = useWorkspaceLayoutStore.getState().layoutByWorkspace[workspaceKey];
const tab = layout
? collectAllTabs(layout.root).find((candidate) => {
return (
candidate.target.kind === "browser" && candidate.target.browserId === input.browserId
);
})
: null;
return tab ? { workspaceKey, tabId: tab.tabId } : null;
}
async function openBrowserTabForRequest(params: {
request: BrowserAutomationExecuteRequest;
serverId?: string;