agentpage 0.0.41 → 0.0.43

package/dist/index.mjs

@@ -225,6 +225,32 @@ function isPotentialDomMutation(toolName, toolInput) {
 	].includes(action);
 }
 /**
+* 判定动作是否为"确定性推进"——比 isPotentialDomMutation 更窄。
+*
+* 仅包含必定产生可见状态变化的动作：
+* - 表单输入类：fill / type / select_option / clear / check / uncheck
+* - 导航类：navigate.*
+*
+* click 不在此列——因为 click 可能点了但完全没效果（如点击无 click listener 的元素）。
+*
+* 用途：协议缺失豁免。仅当本轮有"确定性推进"时才豁免协议缺失计数，
+* 避免模型反复点击无效目标导致死循环。
+*/
+function isConfirmedProgressAction(toolName, toolInput) {
+	if (toolName === "navigate") return true;
+	if (toolName !== "dom") return false;
+	const action = getToolAction(toolInput);
+	if (!action) return false;
+	return [
+		"fill",
+		"type",
+		"select_option",
+		"clear",
+		"check",
+		"uncheck"
+	].includes(action);
+}
+/**
 * 采集找不到元素任务。
 *
 * 返回 null 表示当前结果不属于“元素未找到”，
@@ -896,7 +922,7 @@ async function executeAgentLoop(params) {
 				}
 				pendingNotFoundRetry = void 0;
 			}
-			if (parsedInstructionState.hasRemainingProtocol) remainingInstruction = parsedInstructionState.nextInstruction;
+			if (parsedInstructionState.hasRemainingProtocol && parsedInstructionState.nextInstruction.trim().length === 0) remainingInstruction = "";
 			if (remainingInstruction.trim().length > 0 && round < maxRounds - 1) {
 				protocolViolationHint = [
 					"Protocol violation in previous round:",
@@ -949,6 +975,7 @@ async function executeAgentLoop(params) {
 		}
 		let roundHasError = false;
 		let roundHasPotentialDomMutation = false;
+		let roundHasConfirmedProgress = false;
 		const executedTaskCalls = [];
 		const roundMissingTasks = [];
 		for (const tc of response.toolCalls) {
@@ -980,6 +1007,7 @@ async function executeAgentLoop(params) {
 			if (missingTask) roundMissingTasks.push(missingTask);
 			if (result.details && typeof result.details === "object") roundHasError = roundHasError || Boolean(result.details.error);
 			if (!hasToolError(result) && isPotentialDomMutation(tc.name, tc.input)) roundHasPotentialDomMutation = true;
+			if (!hasToolError(result) && isConfirmedProgressAction(tc.name, tc.input)) roundHasConfirmedProgress = true;
 			if (tc.name === "page_info" && getToolAction(tc.input) === "snapshot") {
 				pageContext.latestSnapshot = toContentString(result.content);
 				recordSnapshotStats(pageContext.latestSnapshot);
@@ -1002,11 +1030,12 @@ async function executeAgentLoop(params) {
 				remainingInstruction = nextByHeuristic;
 				consecutiveNoProtocolRounds = 0;
 			} else if (executedTaskCalls.length > 0) {
-				if (!roundHasPotentialDomMutation || roundHasError) consecutiveNoProtocolRounds += 1;
+				if (!roundHasConfirmedProgress || roundHasError) consecutiveNoProtocolRounds += 1;
 			}
 		}
 		previousRoundModelOutput = parsedInstructionState.hasRemainingProtocol ? normalizeModelOutput(response.text) : `REMAINING: ${remainingInstruction || "DONE"}`;
 		lastRoundHadError = roundHasError;
+		if (executedTaskCalls.length === 0 && response.toolCalls.length > 0) lastRoundHadError = true;
 		previousRoundTasks = buildTaskArray(executedTaskCalls);
 		previousRoundPlannedTasks = plannedTasksCurrentRound;
 		if (parsedInstructionState.hasRemainingProtocol && remainingInstruction.trim().length === 0 && !roundHasError) {
@@ -1014,20 +1043,17 @@ async function executeAgentLoop(params) {
 			if (finalReply) callbacks?.onText?.(finalReply);
 			break;
 		}
-		if (consecutiveNoProtocolRounds >= 3) {
+		if (consecutiveNoProtocolRounds >= 5) {
 			finalReply = response.text?.trim() || "任务已完成。";
 			if (finalReply) callbacks?.onText?.(finalReply);
 			break;
 		}
-		if (consecutiveNoProtocolRounds >= 2) protocolViolationHint = [
-			"Protocol reminder: REMAINING protocol missing for 2+ rounds with tool calls.",
+		if (consecutiveNoProtocolRounds >= 3) protocolViolationHint = [
+			"Protocol reminder: REMAINING protocol missing for 3+ rounds with tool calls.",
 			"You MUST include REMAINING: <text> or REMAINING: DONE in every response.",
 			"If the task is fully complete, return REMAINING: DONE with no tool calls."
 		].join("\n");
-		const idleResult = detectIdleLoop(response.toolCalls.map((tc) => ({
-			name: tc.name,
-			input: tc.input
-		})), consecutiveReadOnlyRounds);
+		const idleResult = detectIdleLoop(executedTaskCalls, consecutiveReadOnlyRounds);
 		if (idleResult === -1) {
 			finalReply = response.text?.trim() || "任务已完成。";
 			if (finalReply) callbacks?.onText?.(finalReply);
@@ -1831,7 +1857,8 @@ function buildSystemPrompt(params = {}) {
 		"- Only interactive elements carry #hashID; others are context-only and cannot be targeted.",
 		"- Bracket tag may show ARIA role ([combobox], [slider]) as primary interaction hint.",
 		"- listeners=\"...\" = bound event handlers (abbrevs below). Prefer targets with matching listeners.",
-		"- Click priority: clk/pdn/mdn, onclick, native link/button, role=button/link. Avoid focus-only or hover-only signals.",
+		"- Click target MUST have click signal: listeners containing clk/pdn/mdn, or onclick attr, or native <a>/<button>, or role=button/link. NEVER click elements with only blr/fcs (focus/blur) — they are not click targets.",
+		"- If the text you want to click has no click signal, look at its parent row/container or nearby sibling that does have clk listener.",
 		"- No-effect fallback: try nearest actionable sibling/ancestor in same semantic group instead of repeating.",
 		"- Batch fill/type/check/select_option freely within one round. A click always ends the round — send at most ONE click as the LAST action in a batch.",
 		"- Input order (MANDATORY): focus/click → fill/type/select_option per target. Multi-field: focus A→fill A→focus B→fill B.",