@xopcai/xopc 0.0.251 → 0.0.253

package/dist/src/agent/tools/browser/tool/browser-use-tool.js

@@ -1,186 +1,69 @@
-import { createLogger } from "../../../../utils/logger/index.js";
-import { init_logger } from "../../../../utils/logger.js";
-import { createBrowserActionRegistry } from "../../../../browser/actions/registry.js";
+import { formatBrowserObservation } from "../../../../browser/observation/format.js";
 import { BrowserUseSchema } from "./schemas.js";
 //#region src/agent/tools/browser/tool/browser-use-tool.ts
-init_logger();
-const log = createLogger("Agent:BrowserUse");
 function createBrowserUseTool(deps) {
-	const registry = createBrowserActionRegistry();
-	function buildContext(page, signal) {
-		return {
-			page,
-			manager: deps.getManager(),
-			config: deps.getConfig(),
-			taskId: deps.getTaskId(),
-			supervisor: deps.getSupervisor?.(),
-			signal
-		};
-	}
-	/**
-	* Render a {@link BrowserSetupHint} as a tool result. The text body is a
-	* JSON envelope (`kind: 'browser_setup_required'`) so the chat surface can
-	* detect it and render a Setup card; the embedded `message` keeps the
-	* payload human-readable for the LLM so it stops the browser attempt and
-	* tells the user to finish setup.
-	*/
-	function formatNotReady(hint) {
-		const message = `⚠ Browser is not ready (backend=${hint.backend}, reason=${hint.reason}). Do NOT retry. Tell the user to open Settings → Browser to finish setup, then ask them to confirm before running any browser action again.`;
-		const payload = {
-			kind: "browser_setup_required",
-			backend: hint.backend,
-			reason: hint.reason,
-			deepLink: hint.deepLink,
-			detail: hint.detail,
-			message
-		};
-		return {
-			content: [{
-				type: "text",
-				text: JSON.stringify(payload)
-			}],
-			details: {
-				ok: false,
-				kind: "browser_setup_required",
-				backend: hint.backend,
-				reason: hint.reason,
-				deepLink: hint.deepLink
-			}
-		};
-	}
-	function formatResult(result) {
-		if (result.ok) {
-			const parts = [];
-			if (result.text) parts.push(result.text);
-			if (result.data && !result.text) parts.push(JSON.stringify(result.data, null, 2));
-			return {
+	return {
+		name: "browser_use",
+		label: "🌐 Browser",
+		description: "Control a persistent browser through typed semantic actions. Start with observe or navigate, then use only refs and revision values from the latest observation. Screenshots are attached only when requested or semantic observation is insufficient.",
+		parameters: BrowserUseSchema,
+		async execute(_toolCallId, params, signal) {
+			const readiness = await deps.getReadiness?.();
+			if (readiness) return {
 				content: [{
 					type: "text",
-					text: parts.join("\n") || "OK"
+					text: `Browser is not ready: ${readiness.hint.detail || readiness.hint.reason}. Open Settings → Browser.`
 				}],
 				details: {
-					ok: true,
-					action: result.action,
-					...result.artifacts?.length ? { artifacts: result.artifacts.length } : {}
+					ok: false,
+					kind: "browser_setup_required",
+					hint: readiness.hint
 				}
 			};
+			return presentResult(await deps.getRuntime().execute(deps.getTaskId(), params, signal));
 		}
-		const parts = [];
-		if (result.error) parts.push(`[${result.error.code}] ${result.error.message}`);
-		if (result.diagnostics?.url) parts.push(`URL: ${result.diagnostics.url}`);
-		if (result.diagnostics?.snapshot) parts.push(`Snapshot: ${result.diagnostics.snapshot.slice(0, 2e3)}`);
+	};
+}
+function presentResult(result) {
+	if (result.ok) {
+		const observation = result.receipt.observation;
 		return {
-			content: [{
-				type: "text",
-				text: parts.join("\n") || "Failed"
-			}],
+			content: contentWithVisual(observation ? formatBrowserObservation(observation) : result.receipt.tabs ? formatTabs(result.receipt.tabs) : `${result.receipt.action} completed and verified.`, observation),
 			details: {
-				ok: false,
-				action: result.action,
-				error: result.error
+				ok: true,
+				receipt: stripVisual(result.receipt)
 			}
 		};
 	}
+	if (!("error" in result)) throw new Error("Invalid browser control result");
+	const observation = result.error.observation;
 	return {
-		name: "browser_use",
-		label: "🌐 Browser",
-		description: "Use a persistent browser for navigation, inspection, interaction, screenshots, and network capture. Use browser_recipe for saved repeatable browser work.",
-		parameters: BrowserUseSchema,
-		async execute(_toolCallId, params, signal, _onUpdate) {
-			const { mode, command, args: cmdArgs, options: _options } = params;
-			if (deps.getReadiness) try {
-				const notReady = await deps.getReadiness();
-				if (notReady) return formatNotReady(notReady.hint);
-			} catch (e) {
-				log.warn({ err: e }, "Readiness preflight threw; continuing with launch attempt");
-			}
-			if (mode === "inspect") {
-				const mgr = deps.getManager();
-				await mgr.ensureConnected();
-				const ext = mgr.getExtensionProvider();
-				if (ext) {
-					const timeoutMs = 3e4;
-					const urlRes = await ext.sendCommand("get_url", {}, { timeout: timeoutMs });
-					const titleRes = await ext.sendCommand("get_title", {}, { timeout: timeoutMs });
-					const snapRes = await ext.sendCommand("snapshot", {}, { timeout: timeoutMs });
-					const url = urlRes.ok && urlRes.data ? String(urlRes.data.url ?? "") : "";
-					const title = titleRes.ok && titleRes.data ? String(titleRes.data.title ?? "") : "";
-					let snapText = "";
-					if (snapRes.ok && snapRes.data) {
-						snapText = (snapRes.data.nodes ?? []).map((n) => `${n.role ?? "?"}: ${n.name ?? ""}`).join("\n");
-						if (snapText.length > 8e3) snapText = `${snapText.slice(0, 8e3)}\n... (truncated)`;
-					}
-					return {
-						content: [{
-							type: "text",
-							text: `URL: ${url}\nTitle: ${title}\n${snapText ? `--- Page Snapshot ---\n${snapText}` : ""}`
-						}],
-						details: {
-							ok: true,
-							mode: "inspect",
-							url,
-							title
-						}
-					};
-				}
-				const page = await deps.getPageForTask();
-				const ctx = buildContext(page, signal);
-				const result = await registry.execute("state", ctx, {});
-				const url = page.url();
-				const title = await page.title().catch(() => "");
-				return {
-					content: [{
-						type: "text",
-						text: `URL: ${url}\nTitle: ${title}\n${result.text ?? ""}`
-					}],
-					details: {
-						ok: true,
-						mode: "inspect",
-						url,
-						title
-					}
-				};
-			}
-			if (mode === "close") {
-				const taskId = deps.getTaskId();
-				await deps.getManager().ensureConnected();
-				await deps.getManager().closePage(taskId);
-				deps.notifyBrowserPageClosed?.(taskId);
-				return {
-					content: [{
-						type: "text",
-						text: "Browser page closed."
-					}],
-					details: {
-						ok: true,
-						mode: "close"
-					}
-				};
-			}
-			if (mode === "command") {
-				if (!command) return {
-					content: [{
-						type: "text",
-						text: "Missing `command` parameter for command mode."
-					}],
-					details: {
-						ok: false,
-						mode: "command"
-					}
-				};
-				const ctx = buildContext(await deps.getPageForTask(), signal);
-				const args = cmdArgs ?? {};
-				return formatResult(await registry.execute(command, ctx, args));
-			}
-			return {
-				content: [{
-					type: "text",
-					text: `Unknown mode: ${mode}`
-				}],
-				details: { ok: false }
-			};
+		content: contentWithVisual([`[${result.error.code}] ${result.error.message}`, observation ? formatBrowserObservation(observation) : ""].filter(Boolean).join("\n\n"), observation),
+		details: {
+			ok: false,
+			kind: result.error.code === "APPROVAL_REQUIRED" ? "browser_approval_required" : "browser_error",
+			error: stripVisual(result.error)
 		}
 	};
 }
+function contentWithVisual(text, observation) {
+	const content = [{
+		type: "text",
+		text
+	}];
+	if (observation?.visual) content.push({
+		type: "image",
+		data: observation.visual.data,
+		mimeType: observation.visual.mimeType
+	});
+	return content;
+}
+function formatTabs(tabs) {
+	if (tabs.length === 0) return "No browser tabs are open.";
+	return tabs.map((tab) => `- ${tab.id}${tab.active ? " (active)" : ""}: ${tab.title}\n  ${tab.url}`).join("\n");
+}
+function stripVisual(value) {
+	return JSON.parse(JSON.stringify(value, (key, child) => key === "visual" ? void 0 : child));
+}
 //#endregion
 export { createBrowserUseTool };

package/dist/src/agent/tools/browser/tool/schemas.d.ts

@@ -1,16 +1,203 @@
-/**
- * BrowserUseSchema — Typebox schema for the unified `browser_use` AgentTool.
- */
-export declare const BrowserUseOptionsSchema: import("@sinclair/typebox").TObject<{
-    timeout: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TNumber>;
-    headless: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TBoolean>;
-}>;
-export declare const BrowserUseSchema: import("@sinclair/typebox").TObject<{
-    mode: import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"command">, import("@sinclair/typebox").TLiteral<"inspect">, import("@sinclair/typebox").TLiteral<"close">]>;
-    command: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
-    args: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TRecord<import("@sinclair/typebox").TString, import("@sinclair/typebox").TUnknown>>;
-    options: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
-        timeout: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TNumber>;
-        headless: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TBoolean>;
-    }>>;
-}>;
+import { type Static } from '@sinclair/typebox';
+export declare const BrowserUseSchema: import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"observe">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    visual: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"never">, import("@sinclair/typebox").TLiteral<"auto">, import("@sinclair/typebox").TLiteral<"always">]>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"navigate">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    url: import("@sinclair/typebox").TString;
+    expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+        urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+    }>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"click">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    revision: import("@sinclair/typebox").TNumber;
+    ref: import("@sinclair/typebox").TString;
+    expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+        urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+    }>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"fill">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    revision: import("@sinclair/typebox").TNumber;
+    ref: import("@sinclair/typebox").TString;
+    value: import("@sinclair/typebox").TString;
+    submit: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TBoolean>;
+    expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+        urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+    }>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"select">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    revision: import("@sinclair/typebox").TNumber;
+    ref: import("@sinclair/typebox").TString;
+    value: import("@sinclair/typebox").TString;
+    expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+        urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+    }>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"press">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    revision: import("@sinclair/typebox").TNumber;
+    ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    key: import("@sinclair/typebox").TString;
+    expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+        urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+    }>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"scroll">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    revision: import("@sinclair/typebox").TNumber;
+    ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    deltaY: import("@sinclair/typebox").TNumber;
+    expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+        urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+    }>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"wait">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    revision: import("@sinclair/typebox").TNumber;
+    condition: import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"page_idle">, import("@sinclair/typebox").TLiteral<"text">, import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>;
+    value: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    timeoutMs: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TNumber>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"upload">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    revision: import("@sinclair/typebox").TNumber;
+    ref: import("@sinclair/typebox").TString;
+    paths: import("@sinclair/typebox").TArray<import("@sinclair/typebox").TString>;
+    expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+        urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+    }>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"tabs">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    operation: import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"list">, import("@sinclair/typebox").TLiteral<"create">, import("@sinclair/typebox").TLiteral<"activate">, import("@sinclair/typebox").TLiteral<"close">]>;
+    tabId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    url: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"sequence">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+} & {
+    revision: import("@sinclair/typebox").TNumber;
+    steps: import("@sinclair/typebox").TArray<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TObject<{
+        action: import("@sinclair/typebox").TLiteral<"click">;
+        ref: import("@sinclair/typebox").TString;
+        expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+            urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+        }>>;
+    }>, import("@sinclair/typebox").TObject<{
+        action: import("@sinclair/typebox").TLiteral<"fill">;
+        ref: import("@sinclair/typebox").TString;
+        value: import("@sinclair/typebox").TString;
+        submit: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TBoolean>;
+        expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+            urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+        }>>;
+    }>, import("@sinclair/typebox").TObject<{
+        action: import("@sinclair/typebox").TLiteral<"select">;
+        ref: import("@sinclair/typebox").TString;
+        value: import("@sinclair/typebox").TString;
+        expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+            urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+        }>>;
+    }>, import("@sinclair/typebox").TObject<{
+        action: import("@sinclair/typebox").TLiteral<"press">;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        key: import("@sinclair/typebox").TString;
+        expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+            urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+        }>>;
+    }>, import("@sinclair/typebox").TObject<{
+        action: import("@sinclair/typebox").TLiteral<"scroll">;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        deltaY: import("@sinclair/typebox").TNumber;
+        expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+            urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+            state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+        }>>;
+    }>]>>;
+    expect: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TObject<{
+        urlIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        titleIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        textIncludes: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        ref: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+        state: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TUnion<[import("@sinclair/typebox").TLiteral<"visible">, import("@sinclair/typebox").TLiteral<"hidden">]>>;
+    }>>;
+}>, import("@sinclair/typebox").TObject<{
+    action: import("@sinclair/typebox").TLiteral<"close">;
+    sessionId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+    approvalId: import("@sinclair/typebox").TOptional<import("@sinclair/typebox").TString>;
+}>]>;
+export type BrowserUseInput = Static<typeof BrowserUseSchema>;

package/dist/src/agent/tools/browser/tool/schemas.js

@@ -1,21 +1,151 @@
 import { Type } from "@sinclair/typebox";
 //#region src/agent/tools/browser/tool/schemas.ts
-/**
-* BrowserUseSchema — Typebox schema for the unified `browser_use` AgentTool.
-*/
-const BrowserUseOptionsSchema = Type.Object({
-	timeout: Type.Optional(Type.Number({ description: "Action timeout override (ms)." })),
-	headless: Type.Optional(Type.Boolean({ description: "Override headless mode for this call." }))
-});
-const BrowserUseSchema = Type.Object({
-	mode: Type.Union([
-		Type.Literal("command"),
-		Type.Literal("inspect"),
+const ApprovalId = Type.Optional(Type.String({ description: "One-time local-owner approval id for this exact action." }));
+const SessionId = Type.Optional(Type.String({ description: "Browser session id returned by the previous observation." }));
+const Revision = Type.Number({ description: "Observation revision used to resolve element refs." });
+const Ref = Type.String({ description: "Ephemeral element ref from the latest observation." });
+const Expectation = Type.Optional(Type.Object({
+	urlIncludes: Type.Optional(Type.String()),
+	titleIncludes: Type.Optional(Type.String()),
+	textIncludes: Type.Optional(Type.String()),
+	ref: Type.Optional(Type.String()),
+	state: Type.Optional(Type.Union([Type.Literal("visible"), Type.Literal("hidden")]))
+}, { additionalProperties: false }));
+function action(name, properties) {
+	return Type.Object({
+		action: Type.Literal(name),
+		sessionId: SessionId,
+		approvalId: ApprovalId,
+		...properties
+	}, { additionalProperties: false });
+}
+const Observe = action("observe", { visual: Type.Optional(Type.Union([
+	Type.Literal("never"),
+	Type.Literal("auto"),
+	Type.Literal("always")
+])) });
+const Navigate = action("navigate", {
+	url: Type.String({ minLength: 1 }),
+	expect: Expectation
+});
+const Click = action("click", {
+	revision: Revision,
+	ref: Ref,
+	expect: Expectation
+});
+const Fill = action("fill", {
+	revision: Revision,
+	ref: Ref,
+	value: Type.String(),
+	submit: Type.Optional(Type.Boolean()),
+	expect: Expectation
+});
+const Select = action("select", {
+	revision: Revision,
+	ref: Ref,
+	value: Type.String(),
+	expect: Expectation
+});
+const Press = action("press", {
+	revision: Revision,
+	ref: Type.Optional(Ref),
+	key: Type.String({ minLength: 1 }),
+	expect: Expectation
+});
+const Scroll = action("scroll", {
+	revision: Revision,
+	ref: Type.Optional(Ref),
+	deltaY: Type.Number(),
+	expect: Expectation
+});
+const Wait = action("wait", {
+	revision: Revision,
+	condition: Type.Union([
+		Type.Literal("page_idle"),
+		Type.Literal("text"),
+		Type.Literal("visible"),
+		Type.Literal("hidden")
+	]),
+	value: Type.Optional(Type.String()),
+	ref: Type.Optional(Ref),
+	timeoutMs: Type.Optional(Type.Number({
+		minimum: 100,
+		maximum: 12e4
+	}))
+});
+const Upload = action("upload", {
+	revision: Revision,
+	ref: Ref,
+	paths: Type.Array(Type.String({ minLength: 1 }), {
+		minItems: 1,
+		maxItems: 20
+	}),
+	expect: Expectation
+});
+const Tabs = action("tabs", {
+	operation: Type.Union([
+		Type.Literal("list"),
+		Type.Literal("create"),
+		Type.Literal("activate"),
 		Type.Literal("close")
-	], { description: "Execution mode." }),
-	command: Type.Optional(Type.String({ description: "Browser action name (command mode). E.g. open, click, type, screenshot." })),
-	args: Type.Optional(Type.Record(Type.String(), Type.Unknown(), { description: "Action arguments (command mode)." })),
-	options: Type.Optional(BrowserUseOptionsSchema)
+	]),
+	tabId: Type.Optional(Type.String()),
+	url: Type.Optional(Type.String())
+});
+const SequenceStep = Type.Union([
+	Type.Object({
+		action: Type.Literal("click"),
+		ref: Ref,
+		expect: Expectation
+	}, { additionalProperties: false }),
+	Type.Object({
+		action: Type.Literal("fill"),
+		ref: Ref,
+		value: Type.String(),
+		submit: Type.Optional(Type.Boolean()),
+		expect: Expectation
+	}, { additionalProperties: false }),
+	Type.Object({
+		action: Type.Literal("select"),
+		ref: Ref,
+		value: Type.String(),
+		expect: Expectation
+	}, { additionalProperties: false }),
+	Type.Object({
+		action: Type.Literal("press"),
+		ref: Type.Optional(Ref),
+		key: Type.String(),
+		expect: Expectation
+	}, { additionalProperties: false }),
+	Type.Object({
+		action: Type.Literal("scroll"),
+		ref: Type.Optional(Ref),
+		deltaY: Type.Number(),
+		expect: Expectation
+	}, { additionalProperties: false })
+]);
+const Sequence = action("sequence", {
+	revision: Revision,
+	steps: Type.Array(SequenceStep, {
+		minItems: 1,
+		maxItems: 10
+	}),
+	expect: Expectation
 });
+const Close = action("close", {});
+const BrowserUseSchema = Type.Union([
+	Observe,
+	Navigate,
+	Click,
+	Fill,
+	Select,
+	Press,
+	Scroll,
+	Wait,
+	Upload,
+	Tabs,
+	Sequence,
+	Close
+]);
 //#endregion
-export { BrowserUseOptionsSchema, BrowserUseSchema };
+export { BrowserUseSchema };