@reconcrap/boss-recommend-mcp 2.0.37 → 2.0.39

package/src/domains/recommend/run-service.js

@@ -1,1424 +1,1431 @@
-import fs from "node:fs";
-import path from "node:path";
-import { createRunLifecycleManager } from "../../core/run/index.js";
-import {
-  addTiming,
-  imageEvidenceFilePath,
-  measureTiming
-} from "../../core/run/timing.js";
-import { captureScrolledNodeScreenshots } from "../../core/capture/index.js";
-import { waitForCvCaptureTarget } from "../../core/cv-capture-target/index.js";
-import { sleep } from "../../core/browser/index.js";
-import { GREET_CREDITS_EXHAUSTED_CODE } from "../../core/greet-quota/index.js";
-import {
-  compactCvAcquisitionState,
-  countParsedNetworkProfiles,
-  createCvAcquisitionState,
-  DEFAULT_MAX_IMAGE_PAGES,
-  getCvNetworkWaitPlan,
-  recordCvImageFallback,
-  recordCvNetworkHit,
-  recordCvNetworkMiss,
-  summarizeImageEvidence,
-  waitForCvNetworkEvents
-} from "../../core/cv-acquisition/index.js";
-import {
-  compactInfiniteListState,
-  createInfiniteListState,
-  detectInfiniteListBottomMarker,
-  getNextInfiniteListCandidate,
-  markInfiniteListCandidateProcessed,
-  resetInfiniteListForRefreshRound,
-  resolveInfiniteListFallbackPoint
-} from "../../core/infinite-list/index.js";
-import { createViewportRunGuard } from "../../core/self-heal/index.js";
-import {
-  callScreeningLlm,
-  compactScreeningLlmResult,
-  createFailedLlmScreeningResult,
-  llmResultToScreening,
-  screenCandidate
-} from "../../core/screening/index.js";
-import {
-  closeRecommendDetail,
-  createRecommendDetailNetworkRecorder,
-  extractRecommendDetailCandidate,
-  isStaleRecommendNodeError,
-  openRecommendCardDetailWithFreshRetry,
-  waitForRecommendDetailNetworkEvents
-} from "./detail.js";
-import {
-  readRecommendCardCandidate,
-  waitForRecommendCardNodeIds
-} from "./cards.js";
-import { selectAndConfirmFirstSafeFilter } from "./filters.js";
-import {
-  buildRecommendFilterSelectionOptions,
-  refreshRecommendListAtEnd
-} from "./refresh.js";
-import { selectRecommendJob } from "./jobs.js";
-import {
-  normalizeRecommendPageScope,
-  selectRecommendPageScope
-} from "./scopes.js";
-import {
-  RECOMMEND_BOTTOM_MARKER_SELECTORS,
-  RECOMMEND_CARD_SELECTOR,
-  RECOMMEND_END_REFRESH_SELECTOR,
-  RECOMMEND_LIST_CONTAINER_SELECTORS,
-  RECOMMEND_TARGET_URL
-} from "./constants.js";
-import {
-  clickRecommendActionControl,
-  normalizeRecommendPostAction,
-  resolveRecommendPostAction,
-  waitForRecommendDetailActionControls
-} from "./actions.js";
-import { getRecommendRoots } from "./roots.js";
-
-function normalizeLabels(labels = []) {
-  return labels.map((label) => String(label || "").trim()).filter(Boolean);
-}
-
-function isRefreshableListStall(reason = "") {
-  return new Set([
-    "stable_visible_signature",
-    "max_scrolls_exhausted",
-    "scroll_failed",
-    "scroll_anchor_unavailable"
-  ]).has(String(reason || ""));
-}
-
-function normalizeFilter(filter = {}) {
-  const filterGroups = Array.isArray(filter.filterGroups)
-    ? filter.filterGroups
-    : Array.isArray(filter.groups)
-      ? filter.groups
-      : [];
-  return {
-    enabled: filter.enabled !== false,
-    group: String(filter.group || ""),
-    labels: normalizeLabels(filter.labels || filter.filterLabels || []),
-    selectAllLabels: Boolean(filter.selectAllLabels),
-    filterGroups: filterGroups.map((group) => ({
-      group: String(group?.group || ""),
-      labels: normalizeLabels(group?.labels || group?.filterLabels || []),
-      selectAllLabels: group?.selectAllLabels !== false
-    })).filter((group) => group.group || group.labels.length)
-  };
-}
-
-function compactFilterResult(filterResult) {
-  if (!filterResult) return null;
-  return {
-    opened_panel: Boolean(filterResult.opened_panel),
-    selected_option: filterResult.selected_option
-      ? {
-          group: filterResult.selected_option.group,
-          label: filterResult.selected_option.label,
-          was_active: Boolean(filterResult.selected_option.was_active),
-          clicked: filterResult.selected_option.clicked !== false
-        }
-      : null,
-    selected_options: (filterResult.selected_options || []).map((option) => ({
-      group: option.group,
-      label: option.label,
-      was_active: Boolean(option.was_active),
-      clicked: option.clicked !== false
-    })),
-    confirmed: Boolean(filterResult.confirmed),
-    before_counts: filterResult.before_counts,
-    after_confirm_counts: filterResult.after_confirm_counts
-  };
-}
-
-function compactJobSelection(jobSelection) {
-  if (!jobSelection) return null;
-  return {
-    requested: jobSelection.requested || "",
-    selected: Boolean(jobSelection.selected),
-    already_current: Boolean(jobSelection.already_current),
-    reason: jobSelection.reason || null,
-    selected_option: jobSelection.selected_option || null,
-    options: (jobSelection.options || []).map((option) => ({
-      label: option.label,
-      label_without_salary: option.label_without_salary,
-      current: Boolean(option.current),
-      visible: Boolean(option.visible),
-      class_name: option.class_name
-    }))
-  };
-}
-
-function compactPageScopeSelection(pageScopeSelection) {
-  if (!pageScopeSelection) return null;
-  return {
-    requested_scope: pageScopeSelection.requested_scope || null,
-    effective_scope: pageScopeSelection.effective_scope || null,
-    fallback_scope: pageScopeSelection.fallback_scope || null,
-    fallback_applied: Boolean(pageScopeSelection.fallback_applied),
-    selected: Boolean(pageScopeSelection.selected),
-    already_current: Boolean(pageScopeSelection.already_current),
-    reason: pageScopeSelection.reason || null,
-    selected_tab: pageScopeSelection.selected_tab || null,
-    available_scopes: pageScopeSelection.available_scopes || [],
-    card_count: pageScopeSelection.after?.card_count || null
-  };
-}
-
-function compactScreening(screening) {
-  return {
-    status: screening.status,
-    passed: screening.passed,
-    score: screening.score,
-    reasons: screening.reasons,
-    candidate: {
-      domain: screening.candidate?.domain || "recommend",
-      source: screening.candidate?.source || "",
-      id: screening.candidate?.id || null,
-      identity: screening.candidate?.identity || {}
-    }
-  };
-}
-
-function compactCandidate(candidate) {
-  return {
-    id: candidate?.id || null,
-    identity: candidate?.identity || {},
-    text_length: candidate?.text?.raw?.length || 0,
-    tag_count: candidate?.tags?.length || 0
-  };
-}
-
-function compactDetail(detailResult) {
-  if (!detailResult) return null;
-  return {
-    popup_text_length: detailResult.detail?.popup_text?.length || 0,
-    resume_text_length: detailResult.detail?.resume_text?.length || 0,
-    network_body_count: detailResult.network_bodies?.filter((item) => item.body).length || 0,
-    parsed_network_profile_count: detailResult.parsed_network_profiles?.filter((item) => item.ok).length || 0,
-    cv_acquisition: detailResult.cv_acquisition || null,
-    image_evidence: summarizeImageEvidence(detailResult.image_evidence),
-    llm_screening: compactScreeningLlmResult(detailResult.llm_result),
-    close_result: detailResult.close_result
-  };
-}
-
-function normalizeScreeningMode(value) {
-  const normalized = String(value || "llm").trim().toLowerCase();
-  return ["deterministic", "local", "local_scorer"].includes(normalized)
-    ? "deterministic"
-    : "llm";
-}
-
-function createMissingLlmConfigResult() {
-  return createFailedLlmScreeningResult(new Error("LLM screening config is required for production recommend runs"));
-}
-
-function compactActionDiscovery(discovery) {
-  if (!discovery) return null;
-  return {
-    elapsed_ms: discovery.elapsed_ms,
-    timed_out: Boolean(discovery.timed_out),
-    detail_root_count: discovery.detail_root_count || 0,
-    summary: discovery.summary || null
-  };
-}
-
-async function runRecommendPostAction({
-  client,
-  screening,
-  actionDiscovery,
-  postAction = "none",
-  greetCount = 0,
-  maxGreetCount = null,
-  executePostAction = true,
-  afterClickDelayMs = 900
-} = {}) {
-  const plan = resolveRecommendPostAction({
-    postAction,
-    greetCount,
-    maxGreetCount
-  });
-  const result = {
-    requested: postAction,
-    execute_post_action: Boolean(executePostAction),
-    plan,
-    eligible: Boolean(screening?.passed),
-    action_attempted: false,
-    action_clicked: false,
-    counted_as_greet: false,
-    reason: ""
-  };
-
-  if (!screening?.passed) {
-    result.reason = "screening_not_passed";
-    return result;
-  }
-  if (plan.effective === "none") {
-    result.reason = "post_action_none";
-    return result;
-  }
-
-  const summary = actionDiscovery?.summary || {};
-  const control = plan.effective === "favorite" ? summary.favorite : summary.greet;
-  if (!control?.found) {
-    result.reason = `${plan.effective}_control_not_found`;
-    return result;
-  }
-  result.control = control;
-
-  if (plan.effective === "greet" && control.continue_chat) {
-    result.reason = "already_connected_continue_chat";
-    result.already_connected = true;
-    return result;
-  }
-  if (plan.effective === "greet" && control.greet_quota?.exhausted) {
-    result.reason = "greet_credits_exhausted";
-    result.out_of_greet_credits = true;
-    result.stop_run = true;
-    return result;
-  }
-  if (plan.effective === "greet" && control.available === false) {
-    result.reason = "greet_control_not_available";
-    return result;
-  }
-  if (plan.effective === "favorite" && control.active) {
-    result.reason = "already_favorited";
-    result.already_favorited = true;
-    return result;
-  }
-  if (control.disabled) {
-    result.reason = `${plan.effective}_control_disabled`;
-    return result;
-  }
-  if (!executePostAction) {
-    result.reason = "dry_run_post_action";
-    result.would_click = true;
-    return result;
-  }
-
-  result.action_attempted = true;
-  result.control_before = control;
-  let clickResult;
-  try {
-    clickResult = await clickRecommendActionControl(client, {
-      ...control,
-      kind: plan.effective
-    });
-  } catch (error) {
-    if (error?.code === GREET_CREDITS_EXHAUSTED_CODE) {
-      result.reason = "greet_credits_exhausted";
-      result.out_of_greet_credits = true;
-      result.stop_run = true;
-      result.greet_quota = error.greet_quota || control.greet_quota || null;
-      return result;
-    }
-    throw error;
-  }
-  result.click_result = clickResult;
-  result.action_clicked = true;
-  result.counted_as_greet = plan.effective === "greet";
-  result.reason = "clicked";
-  if (afterClickDelayMs > 0) await sleep(afterClickDelayMs);
-  try {
-    const afterDiscovery = await waitForRecommendDetailActionControls(client, {
-      timeoutMs: 2500,
-      intervalMs: 300,
-      requireAny: false
-    });
-    const afterSummary = afterDiscovery?.summary || {};
-    const afterControl = plan.effective === "favorite" ? afterSummary.favorite : afterSummary.greet;
-    result.action_discovery_after = compactActionDiscovery(afterDiscovery);
-    result.control_after = afterControl || null;
-    if (plan.effective === "greet") {
-      result.verified_after_click = Boolean(
-        afterControl?.continue_chat
-        || String(afterControl?.label || "").includes("继续沟通")
-      );
-    } else if (plan.effective === "favorite") {
-      result.verified_after_click = Boolean(
-        afterControl?.active
-        || String(afterControl?.label || "").includes("已")
-      );
-    }
-  } catch (error) {
-    result.verify_error = {
-      message: error?.message || String(error)
-    };
-  }
-  return result;
-}
-
-function compactRefreshAttempt(refreshAttempt) {
-  if (!refreshAttempt) return null;
-  return {
-    ok: Boolean(refreshAttempt.ok),
-    method: refreshAttempt.method || "",
-    reason: refreshAttempt.reason || null,
-    error: refreshAttempt.error || null,
-    forced_recent_not_view: Boolean(refreshAttempt.forced_recent_not_view),
-    target_url: refreshAttempt.target_url || null,
-    card_count: refreshAttempt.card_count || 0,
-    elapsed_ms: refreshAttempt.elapsed_ms || 0,
-    attempts: (refreshAttempt.attempts || []).map((attempt) => ({
-      ok: Boolean(attempt.ok),
-      method: attempt.method || "",
-      reason: attempt.reason || null,
-      error: attempt.error || null,
-      label: attempt.label || null,
+import fs from "node:fs";
+import path from "node:path";
+import { createRunLifecycleManager } from "../../core/run/index.js";
+import {
+  addTiming,
+  imageEvidenceFilePath,
+  measureTiming
+} from "../../core/run/timing.js";
+import { captureScrolledNodeScreenshots } from "../../core/capture/index.js";
+import { waitForCvCaptureTarget } from "../../core/cv-capture-target/index.js";
+import { sleep } from "../../core/browser/index.js";
+import { GREET_CREDITS_EXHAUSTED_CODE } from "../../core/greet-quota/index.js";
+import {
+  compactCvAcquisitionState,
+  countParsedNetworkProfiles,
+  createCvAcquisitionState,
+  DEFAULT_MAX_IMAGE_PAGES,
+  getCvNetworkWaitPlan,
+  recordCvImageFallback,
+  recordCvNetworkHit,
+  recordCvNetworkMiss,
+  summarizeImageEvidence,
+  waitForCvNetworkEvents
+} from "../../core/cv-acquisition/index.js";
+import {
+  compactInfiniteListState,
+  createInfiniteListState,
+  detectInfiniteListBottomMarker,
+  getNextInfiniteListCandidate,
+  markInfiniteListCandidateProcessed,
+  resetInfiniteListForRefreshRound,
+  resolveInfiniteListFallbackPoint
+} from "../../core/infinite-list/index.js";
+import { createViewportRunGuard } from "../../core/self-heal/index.js";
+import {
+  callScreeningLlm,
+  compactScreeningLlmResult,
+  createFailedLlmScreeningResult,
+  llmResultToScreening,
+  screenCandidate
+} from "../../core/screening/index.js";
+import {
+  closeRecommendDetail,
+  createRecommendDetailNetworkRecorder,
+  extractRecommendDetailCandidate,
+  isStaleRecommendNodeError,
+  openRecommendCardDetailWithFreshRetry,
+  waitForRecommendDetailNetworkEvents
+} from "./detail.js";
+import {
+  readRecommendCardCandidate,
+  waitForRecommendCardNodeIds
+} from "./cards.js";
+import { selectAndConfirmFirstSafeFilter } from "./filters.js";
+import {
+  buildRecommendFilterSelectionOptions,
+  refreshRecommendListAtEnd
+} from "./refresh.js";
+import { selectRecommendJob } from "./jobs.js";
+import {
+  normalizeRecommendPageScope,
+  selectRecommendPageScope
+} from "./scopes.js";
+import {
+  RECOMMEND_BOTTOM_MARKER_SELECTORS,
+  RECOMMEND_CARD_SELECTOR,
+  RECOMMEND_END_REFRESH_SELECTOR,
+  RECOMMEND_LIST_CONTAINER_SELECTORS,
+  RECOMMEND_TARGET_URL
+} from "./constants.js";
+import {
+  clickRecommendActionControl,
+  normalizeRecommendPostAction,
+  resolveRecommendPostAction,
+  waitForRecommendDetailActionControls
+} from "./actions.js";
+import { getRecommendRoots } from "./roots.js";
+
+function normalizeLabels(labels = []) {
+  return labels.map((label) => String(label || "").trim()).filter(Boolean);
+}
+
+function isRefreshableListStall(reason = "") {
+  return new Set([
+    "stable_visible_signature",
+    "max_scrolls_exhausted",
+    "scroll_failed",
+    "scroll_anchor_unavailable"
+  ]).has(String(reason || ""));
+}
+
+function normalizeFilter(filter = {}) {
+  const filterGroups = Array.isArray(filter.filterGroups)
+    ? filter.filterGroups
+    : Array.isArray(filter.groups)
+      ? filter.groups
+      : [];
+  return {
+    enabled: filter.enabled !== false,
+    group: String(filter.group || ""),
+    labels: normalizeLabels(filter.labels || filter.filterLabels || []),
+    selectAllLabels: Boolean(filter.selectAllLabels),
+    filterGroups: filterGroups.map((group) => ({
+      group: String(group?.group || ""),
+      labels: normalizeLabels(group?.labels || group?.filterLabels || []),
+      selectAllLabels: group?.selectAllLabels !== false
+    })).filter((group) => group.group || group.labels.length)
+  };
+}
+
+function compactFilterResult(filterResult) {
+  if (!filterResult) return null;
+  return {
+    opened_panel: Boolean(filterResult.opened_panel),
+    selected_option: filterResult.selected_option
+      ? {
+          group: filterResult.selected_option.group,
+          label: filterResult.selected_option.label,
+          was_active: Boolean(filterResult.selected_option.was_active),
+          clicked: filterResult.selected_option.clicked !== false
+        }
+      : null,
+    selected_options: (filterResult.selected_options || []).map((option) => ({
+      group: option.group,
+      label: option.label,
+      was_active: Boolean(option.was_active),
+      clicked: option.clicked !== false
+    })),
+    confirmed: Boolean(filterResult.confirmed),
+    before_counts: filterResult.before_counts,
+    after_confirm_counts: filterResult.after_confirm_counts
+  };
+}
+
+function compactJobSelection(jobSelection) {
+  if (!jobSelection) return null;
+  return {
+    requested: jobSelection.requested || "",
+    selected: Boolean(jobSelection.selected),
+    already_current: Boolean(jobSelection.already_current),
+    reason: jobSelection.reason || null,
+    selected_option: jobSelection.selected_option || null,
+    options: (jobSelection.options || []).map((option) => ({
+      label: option.label,
+      label_without_salary: option.label_without_salary,
+      current: Boolean(option.current),
+      visible: Boolean(option.visible),
+      class_name: option.class_name
+    }))
+  };
+}
+
+function compactPageScopeSelection(pageScopeSelection) {
+  if (!pageScopeSelection) return null;
+  return {
+    requested_scope: pageScopeSelection.requested_scope || null,
+    effective_scope: pageScopeSelection.effective_scope || null,
+    fallback_scope: pageScopeSelection.fallback_scope || null,
+    fallback_applied: Boolean(pageScopeSelection.fallback_applied),
+    selected: Boolean(pageScopeSelection.selected),
+    already_current: Boolean(pageScopeSelection.already_current),
+    reason: pageScopeSelection.reason || null,
+    selected_tab: pageScopeSelection.selected_tab || null,
+    available_scopes: pageScopeSelection.available_scopes || [],
+    card_count: pageScopeSelection.after?.card_count || null
+  };
+}
+
+function compactScreening(screening) {
+  return {
+    status: screening.status,
+    passed: screening.passed,
+    score: screening.score,
+    reasons: screening.reasons,
+    candidate: {
+      domain: screening.candidate?.domain || "recommend",
+      source: screening.candidate?.source || "",
+      id: screening.candidate?.id || null,
+      identity: screening.candidate?.identity || {}
+    }
+  };
+}
+
+function compactCandidate(candidate) {
+  return {
+    id: candidate?.id || null,
+    identity: candidate?.identity || {},
+    text_length: candidate?.text?.raw?.length || 0,
+    tag_count: candidate?.tags?.length || 0
+  };
+}
+
+function compactDetail(detailResult) {
+  if (!detailResult) return null;
+  return {
+    popup_text_length: detailResult.detail?.popup_text?.length || 0,
+    resume_text_length: detailResult.detail?.resume_text?.length || 0,
+    network_body_count: detailResult.network_bodies?.filter((item) => item.body).length || 0,
+    parsed_network_profile_count: detailResult.parsed_network_profiles?.filter((item) => item.ok).length || 0,
+    cv_acquisition: detailResult.cv_acquisition || null,
+    image_evidence: summarizeImageEvidence(detailResult.image_evidence),
+    llm_screening: compactScreeningLlmResult(detailResult.llm_result),
+    close_result: detailResult.close_result
+  };
+}
+
+function normalizeScreeningMode(value) {
+  const normalized = String(value || "llm").trim().toLowerCase();
+  return ["deterministic", "local", "local_scorer"].includes(normalized)
+    ? "deterministic"
+    : "llm";
+}
+
+function createMissingLlmConfigResult() {
+  return createFailedLlmScreeningResult(new Error("LLM screening config is required for production recommend runs"));
+}
+
+function compactActionDiscovery(discovery) {
+  if (!discovery) return null;
+  return {
+    elapsed_ms: discovery.elapsed_ms,
+    timed_out: Boolean(discovery.timed_out),
+    detail_root_count: discovery.detail_root_count || 0,
+    summary: discovery.summary || null
+  };
+}
+
+async function runRecommendPostAction({
+  client,
+  screening,
+  actionDiscovery,
+  postAction = "none",
+  greetCount = 0,
+  maxGreetCount = null,
+  executePostAction = true,
+  afterClickDelayMs = 900
+} = {}) {
+  const plan = resolveRecommendPostAction({
+    postAction,
+    greetCount,
+    maxGreetCount
+  });
+  const result = {
+    requested: postAction,
+    execute_post_action: Boolean(executePostAction),
+    plan,
+    eligible: Boolean(screening?.passed),
+    action_attempted: false,
+    action_clicked: false,
+    counted_as_greet: false,
+    reason: ""
+  };
+
+  if (!screening?.passed) {
+    result.reason = "screening_not_passed";
+    return result;
+  }
+  if (plan.effective === "none") {
+    result.reason = "post_action_none";
+    return result;
+  }
+
+  const summary = actionDiscovery?.summary || {};
+  const control = plan.effective === "favorite" ? summary.favorite : summary.greet;
+  if (!control?.found) {
+    result.reason = `${plan.effective}_control_not_found`;
+    return result;
+  }
+  result.control = control;
+
+  if (plan.effective === "greet" && control.continue_chat) {
+    result.reason = "already_connected_continue_chat";
+    result.already_connected = true;
+    return result;
+  }
+  if (plan.effective === "greet" && control.greet_quota?.exhausted) {
+    result.reason = "greet_credits_exhausted";
+    result.out_of_greet_credits = true;
+    result.stop_run = true;
+    return result;
+  }
+  if (plan.effective === "greet" && control.available === false) {
+    result.reason = "greet_control_not_available";
+    return result;
+  }
+  if (plan.effective === "favorite" && control.active) {
+    result.reason = "already_favorited";
+    result.already_favorited = true;
+    return result;
+  }
+  if (control.disabled) {
+    result.reason = `${plan.effective}_control_disabled`;
+    return result;
+  }
+  if (!executePostAction) {
+    result.reason = "dry_run_post_action";
+    result.would_click = true;
+    return result;
+  }
+
+  result.action_attempted = true;
+  result.control_before = control;
+  let clickResult;
+  try {
+    clickResult = await clickRecommendActionControl(client, {
+      ...control,
+      kind: plan.effective
+    });
+  } catch (error) {
+    if (error?.code === GREET_CREDITS_EXHAUSTED_CODE) {
+      result.reason = "greet_credits_exhausted";
+      result.out_of_greet_credits = true;
+      result.stop_run = true;
+      result.greet_quota = error.greet_quota || control.greet_quota || null;
+      return result;
+    }
+    throw error;
+  }
+  result.click_result = clickResult;
+  result.action_clicked = true;
+  result.counted_as_greet = plan.effective === "greet";
+  result.reason = "clicked";
+  if (afterClickDelayMs > 0) await sleep(afterClickDelayMs);
+  try {
+    const afterDiscovery = await waitForRecommendDetailActionControls(client, {
+      timeoutMs: 2500,
+      intervalMs: 300,
+      requireAny: false
+    });
+    const afterSummary = afterDiscovery?.summary || {};
+    const afterControl = plan.effective === "favorite" ? afterSummary.favorite : afterSummary.greet;
+    result.action_discovery_after = compactActionDiscovery(afterDiscovery);
+    result.control_after = afterControl || null;
+    if (plan.effective === "greet") {
+      result.verified_after_click = Boolean(
+        afterControl?.continue_chat
+        || String(afterControl?.label || "").includes("继续沟通")
+      );
+    } else if (plan.effective === "favorite") {
+      result.verified_after_click = Boolean(
+        afterControl?.active
+        || String(afterControl?.label || "").includes("已")
+      );
+    }
+  } catch (error) {
+    result.verify_error = {
+      message: error?.message || String(error)
+    };
+  }
+  return result;
+}
+
+function compactRefreshAttempt(refreshAttempt) {
+  if (!refreshAttempt) return null;
+  return {
+    ok: Boolean(refreshAttempt.ok),
+    method: refreshAttempt.method || "",
+    reason: refreshAttempt.reason || null,
+    error: refreshAttempt.error || null,
+    forced_recent_not_view: Boolean(refreshAttempt.forced_recent_not_view),
+    target_url: refreshAttempt.target_url || null,
+    card_count: refreshAttempt.card_count || 0,
+    elapsed_ms: refreshAttempt.elapsed_ms || 0,
+    attempts: (refreshAttempt.attempts || []).map((attempt) => ({
+      ok: Boolean(attempt.ok),
+      method: attempt.method || "",
+      reason: attempt.reason || null,
+      error: attempt.error || null,
+      label: attempt.label || null,
       before_card_count: attempt.before_card_count || 0,
       after_card_count: attempt.after_card_count || 0,
       card_count: attempt.card_count || 0,
       elapsed_ms: attempt.elapsed_ms || 0
     })),
+    filter_reapply_attempts: (refreshAttempt.filter_reapply_attempts || []).map((attempt) => ({
+      ok: Boolean(attempt.ok),
+      method: attempt.method || "filter_reapply",
+      reason: attempt.reason || null,
+      error: attempt.error || null,
+      attempt: attempt.attempt || 0
+    })),
     job_selection: compactJobSelection(refreshAttempt.job_selection),
     page_scope: compactPageScopeSelection(refreshAttempt.page_scope),
     filter: compactFilterResult(refreshAttempt.filter)
   };
 }
-
-export function countRecommendResultStatuses(results = [], {
-  greetCount = 0
-} = {}) {
-  return {
-    processed: results.length,
-    screened: results.length,
-    detail_opened: results.filter((item) => item.detail).length,
-    passed: results.filter((item) => item.screening?.passed).length,
-    llm_screened: results.filter((item) => item.detail?.llm_screening || item.llm_screening).length,
-    greet_count: greetCount,
-    post_action_clicked: results.filter((item) => item.post_action?.action_clicked).length,
-    image_capture_failed: results.filter((item) => item.detail?.image_evidence?.ok === false).length,
-    detail_open_failed: results.filter((item) => (
-      item.error?.code === "DETAIL_STALE_NODE"
-      || item.error?.code === "DETAIL_OPEN_FAILED"
-    )).length,
-    transient_recovered: results.filter((item) => (
-      item.error?.code === "DETAIL_STALE_NODE"
-      || item.error?.code === "IMAGE_CAPTURE_STALE_NODE"
-      || item.error?.code === "IMAGE_CAPTURE_TIMEOUT"
-      || item.error?.code === "IMAGE_CAPTURE_TOTAL_TIMEOUT"
-    )).length
-  };
-}
-
-function countPassedResults(results = []) {
-  return countRecommendResultStatuses(results).passed;
-}
-
-function compactError(error, fallbackCode = "RECOMMEND_RUN_ERROR") {
-  if (!error) return null;
-  return {
-    code: error.code || fallbackCode,
-    message: error.message || String(error)
-  };
-}
-
-function createRecommendCloseFailureError(closeResult) {
-  const error = new Error(closeResult?.reason || "Recommend detail did not close before recovery");
-  error.code = "DETAIL_CLOSE_FAILED";
-  error.close_result = closeResult || null;
-  return error;
-}
-
-export function isRecoverableImageCaptureError(error) {
-  const code = String(error?.code || "");
-  if (code === "IMAGE_CAPTURE_TIMEOUT" || code === "IMAGE_CAPTURE_TOTAL_TIMEOUT") return true;
-  if (isStaleRecommendNodeError(error)) return true;
-  return /Image fallback capture timed out/i.test(String(error?.message || error || ""));
-}
-
-function collectPartialImageEvidencePaths(basePath = "", extension = "jpg", maxCount = 12) {
-  const resolved = String(basePath || "").trim();
-  if (!resolved) return [];
-  const parsed = path.parse(resolved);
-  const ext = parsed.ext || `.${String(extension || "jpg").replace(/^\./, "") || "jpg"}`;
-  const files = [];
-  for (let index = 0; index < Math.max(1, Number(maxCount) || 1); index += 1) {
-    const page = String(index + 1).padStart(2, "0");
-    const candidatePath = path.join(parsed.dir, `${parsed.name}-page-${page}${ext}`);
-    if (fs.existsSync(candidatePath)) files.push(candidatePath);
-  }
-  return files;
-}
-
-export function createRecoverableImageCaptureEvidence(error, {
-  elapsedMs = 0,
-  filePath = "",
-  extension = "jpg",
-  maxScreenshots = DEFAULT_MAX_IMAGE_PAGES
-} = {}) {
-  const filePaths = collectPartialImageEvidencePaths(filePath, extension, maxScreenshots);
-  return {
-    schema_version: 1,
-    ok: false,
-    source: "image-scroll-sequence",
-    elapsed_ms: Math.max(0, Math.round(Number(error?.elapsed_ms ?? elapsedMs) || 0)),
-    capture_count: filePaths.length,
-    screenshot_count: filePaths.length,
-    unique_screenshot_count: filePaths.length,
-    dropped_duplicate_count: 0,
-    total_byte_length: 0,
-    original_total_byte_length: 0,
-    llm_screenshot_count: 0,
-    llm_total_byte_length: 0,
-    llm_original_total_byte_length: 0,
-    llm_composition_error: null,
-    error_code: error?.code || (isStaleRecommendNodeError(error) ? "IMAGE_CAPTURE_STALE_NODE" : "IMAGE_CAPTURE_FAILED"),
-    error: error?.message || String(error || "Image capture failed"),
-    file_paths: filePaths,
-    llm_file_paths: []
-  };
-}
-
-function createImageCaptureFailureScreening(candidate, error) {
-  return {
-    status: "fail",
-    passed: false,
-    score: 0,
-    reasons: ["image_capture_failed"],
-    error: compactError(error, "IMAGE_CAPTURE_FAILED"),
-    candidate
-  };
-}
-
-export function isRecoverableRecommendDetailError(error) {
-  return isStaleRecommendNodeError(error);
-}
-
-function compactRecoverableDetailError(error) {
-  return compactError(error, isStaleRecommendNodeError(error) ? "DETAIL_STALE_NODE" : "DETAIL_OPEN_FAILED");
-}
-
-function createRecoverableDetailFailureScreening(candidate, error) {
-  return {
-    status: "fail",
-    passed: false,
-    score: 0,
-    reasons: isStaleRecommendNodeError(error)
-      ? ["detail_open_failed", "stale_node"]
-      : ["detail_open_failed"],
-    error: compactRecoverableDetailError(error),
-    candidate
-  };
-}
-
-export async function runRecommendWorkflow({
-  client,
-  targetUrl = "",
-  criteria = "",
-  jobLabel = "",
-  pageScope = "recommend",
-  fallbackPageScope = "recommend",
-  filter = {},
-  maxCandidates = 5,
-  detailLimit,
-  closeDetail = true,
-  delayMs = 0,
-  cardTimeoutMs = 10000,
-  maxImagePages = DEFAULT_MAX_IMAGE_PAGES,
-  imageWheelDeltaY = 650,
-  cvAcquisitionMode = "unknown",
-  listMaxScrolls = 20,
-  listStableSignatureLimit = 5,
-  listWheelDeltaY = 850,
-  listSettleMs = 2200,
-  listFallbackPoint = null,
-  refreshOnEnd = true,
-  maxRefreshRounds = 2,
-  refreshButtonSettleMs = 8000,
-  refreshReloadSettleMs = 8000,
-  postAction = "none",
-  maxGreetCount = null,
-  executePostAction = true,
-  actionTimeoutMs = 8000,
-  actionIntervalMs = 500,
-  actionAfterClickDelayMs = 900,
-  screeningMode = "llm",
-  llmConfig = null,
-  llmTimeoutMs = 120000,
-  llmImageLimit = 8,
-  llmImageDetail = "high",
-  imageOutputDir = ""
-} = {}, runControl) {
-  if (!client) throw new Error("runRecommendWorkflow requires a guarded CDP client");
-  const normalizedFilter = normalizeFilter(filter);
-  const normalizedPostAction = normalizeRecommendPostAction(postAction) || "none";
-  const requestedPageScope = normalizeRecommendPageScope(pageScope) || "recommend";
-  const normalizedFallbackPageScope = normalizeRecommendPageScope(fallbackPageScope) || "recommend";
-  const normalizedScreeningMode = normalizeScreeningMode(screeningMode);
-  const useLlmScreening = normalizedScreeningMode !== "deterministic";
-  const postActionEnabled = normalizedPostAction !== "none";
-  const targetPassCount = Math.max(1, Number(maxCandidates) || 1);
-  const detailCountLimit = detailLimit == null ? Number.POSITIVE_INFINITY : Math.max(0, Number(detailLimit) || 0);
-  const effectiveDetailLimit = postActionEnabled ? Number.POSITIVE_INFINITY : detailCountLimit;
-  const networkRecorder = effectiveDetailLimit > 0
-    ? createRecommendDetailNetworkRecorder(client)
-    : null;
-  const cvAcquisitionState = createCvAcquisitionState({ mode: cvAcquisitionMode });
-  const listState = createInfiniteListState({
-    domain: "recommend",
-    listName: "recommend-candidates"
-  });
-  const viewportGuard = createViewportRunGuard({
-    client,
-    domain: "recommend",
-    root: "frame",
-    frameOwnerRoot: "frameOwner",
-    runControl,
-    getRoots: getRecommendRoots
-  });
-  async function ensureRecommendViewport(rootState, phase) {
-    const result = await viewportGuard.ensure(rootState, { phase });
-    return result.rootState || rootState;
-  }
-  const results = [];
-  const refreshAttempts = [];
-  let refreshRounds = 0;
-  let contextRecoveryAttempts = 0;
-  let greetCount = 0;
-  const candidateRecoveryCounts = new Map();
-  let jobSelection = null;
-  let pageScopeSelection = null;
-  let filterResult = null;
-  let cardNodeIds = [];
-  let listEndReason = "";
-  const listFallbackResolver = listFallbackPoint || (async ({ items = [] } = {}) => resolveInfiniteListFallbackPoint(client, {
-    rootNodeId: rootState?.iframe?.documentNodeId,
-    containerSelectors: RECOMMEND_LIST_CONTAINER_SELECTORS,
-    itemNodeIds: items.map((item) => item.node_id).filter(Boolean),
-    itemSelectors: [RECOMMEND_CARD_SELECTOR],
-    viewportPoint: { xRatio: 0.28, yRatio: 0.5 },
-    validateViewportPoint: true
-  }));
-
-  function updateRecommendProgress(extra = {}) {
-    const counts = countRecommendResultStatuses(results, { greetCount });
-    const listSnapshot = compactInfiniteListState(listState);
-    runControl.updateProgress({
-      card_count: cardNodeIds.length,
-      target_count: targetPassCount,
-      target_count_semantics: "passed_candidates",
-      ...counts,
-      screening_mode: normalizedScreeningMode,
-      unique_seen: listSnapshot.seen_count,
-      scroll_count: listSnapshot.scroll_count,
-      refresh_rounds: refreshRounds,
-      refresh_attempts: refreshAttempts.length,
-      context_recoveries: contextRecoveryAttempts,
-      list_end_reason: listEndReason || null,
-      viewport_checks: viewportGuard.getStats().checks,
-      viewport_recoveries: viewportGuard.getStats().recoveries,
-      ...extra
-    });
-  }
-
-  function checkpointInProgressCandidate({
-    index = results.length,
-    candidateKey = "",
-    cardNodeId = null,
-    detailStep = "",
-    error = null
-  } = {}) {
-    runControl.checkpoint({
-      in_progress_candidate: {
-        index,
-        key: candidateKey,
-        card_node_id: cardNodeId,
-        detail_step: detailStep || null,
-        counters: countRecommendResultStatuses(results, { greetCount }),
-        error: compactError(error, "RECOMMEND_IN_PROGRESS_ERROR")
-      },
-      candidate_list: compactInfiniteListState(listState)
-    });
-  }
-
-  async function recoverAndReapplyRecommendContext(reason = "context_recovery", error = null, {
-    forceRecentNotView = true
-  } = {}) {
-    await runControl.waitIfPaused();
-    runControl.throwIfCanceled();
-    const started = Date.now();
-    runControl.setPhase("recommend:recover-context");
-    contextRecoveryAttempts += 1;
-    const refreshResult = await refreshRecommendListAtEnd(client, {
-      rootState,
-      jobLabel,
-      pageScope: pageScopeSelection?.effective_scope || requestedPageScope,
-      fallbackPageScope: normalizedFallbackPageScope,
-      filter: normalizedFilter,
-      preferEndRefreshButton: false,
-      forceNavigate: true,
-      targetUrl: targetUrl || RECOMMEND_TARGET_URL,
-      forceRecentNotView,
-      cardTimeoutMs,
-      buttonSettleMs: refreshButtonSettleMs,
-      reloadSettleMs: refreshReloadSettleMs
-    });
-    const compactRefresh = {
-      ...compactRefreshAttempt(refreshResult),
-      context_recovery: true,
-      recovery_reason: reason,
-      trigger_error: compactError(error, "RECOMMEND_CONTEXT_RECOVERY_TRIGGER"),
-      elapsed_ms: Date.now() - started
-    };
-    refreshAttempts.push(compactRefresh);
-    runControl.checkpoint({
-      context_recovery: {
-        attempt: contextRecoveryAttempts,
-        reason,
-        trigger_error: compactError(error, "RECOMMEND_CONTEXT_RECOVERY_TRIGGER"),
-        refresh: compactRefresh,
-        counters: countRecommendResultStatuses(results, { greetCount })
-      },
-      candidate_list: compactInfiniteListState(listState)
-    });
-    if (!refreshResult.ok) {
-      updateRecommendProgress({
-        refresh_method: refreshResult.method || null,
-        refresh_forced_recent_not_view: forceRecentNotView,
-        recovery_reason: reason
-      });
-      throw new Error(`Recommend context recovery failed after ${reason}: ${refreshResult.reason || refreshResult.error || "refresh returned no cards"}`);
-    }
-    rootState = refreshResult.root_state || await getRecommendRoots(client);
-    rootState = await ensureRecommendViewport(rootState, "recover_after");
-    cardNodeIds = await waitForRecommendCardNodeIds(client, rootState.iframe.documentNodeId, {
-      timeoutMs: cardTimeoutMs,
-      intervalMs: 300
-    });
-    resetInfiniteListForRefreshRound(listState, {
-      reason: `context_recovery:${reason}`,
-      round: contextRecoveryAttempts,
-      method: refreshResult.method,
-      metadata: {
-        card_count: cardNodeIds.length,
-        forced_recent_not_view: forceRecentNotView,
-        counters: countRecommendResultStatuses(results, { greetCount })
-      }
-    });
-    listEndReason = "";
-    updateRecommendProgress({
-      card_count: cardNodeIds.length,
-      refresh_method: refreshResult.method || null,
-      refresh_forced_recent_not_view: forceRecentNotView,
-      recovery_reason: reason
-    });
-    return refreshResult;
-  }
-
-  runControl.setPhase("recommend:cleanup");
-  await closeRecommendDetail(client, { attemptsLimit: 2 });
-
-  await runControl.waitIfPaused();
-  runControl.throwIfCanceled();
-  runControl.setPhase("recommend:roots");
-  let rootState = await getRecommendRoots(client);
-  rootState = await ensureRecommendViewport(rootState, "roots");
-  runControl.checkpoint({
-    iframe_selector: rootState.iframe.selector,
-    iframe_document_node_id: rootState.iframe.documentNodeId
-  });
-
-  if (jobLabel) {
-    await runControl.waitIfPaused();
-    runControl.throwIfCanceled();
-    runControl.setPhase("recommend:job");
-    jobSelection = await selectRecommendJob(client, rootState.iframe.documentNodeId, {
-      jobLabel,
-      settleMs: cardTimeoutMs > 45000 ? 12000 : 6000
-    });
-    if (!jobSelection.selected) {
-      throw new Error(`Requested recommend job was not selected: ${jobSelection.reason}`);
-    }
-    rootState = await getRecommendRoots(client);
-    rootState = await ensureRecommendViewport(rootState, "job");
-    runControl.checkpoint({
-      job_selection: compactJobSelection(jobSelection)
-    });
-  }
-
-  await runControl.waitIfPaused();
-  runControl.throwIfCanceled();
-  runControl.setPhase("recommend:page-scope");
-  pageScopeSelection = await selectRecommendPageScope(client, rootState.iframe.documentNodeId, {
-    pageScope: requestedPageScope,
-    fallbackScope: normalizedFallbackPageScope,
-    settleMs: cardTimeoutMs > 45000 ? 3000 : 1200,
-    timeoutMs: Math.min(Math.max(cardTimeoutMs, 10000), 60000)
-  });
-  if (!pageScopeSelection.selected) {
-    throw new Error(`Recommend page scope was not selected: ${pageScopeSelection.reason || pageScopeSelection.effective_scope || requestedPageScope}`);
-  }
-  rootState = await getRecommendRoots(client);
-  rootState = await ensureRecommendViewport(rootState, "page_scope");
-  runControl.checkpoint({
-    page_scope: compactPageScopeSelection(pageScopeSelection)
-  });
-
-  if (normalizedFilter.enabled) {
-    await runControl.waitIfPaused();
-    runControl.throwIfCanceled();
-    runControl.setPhase("recommend:filter");
-    filterResult = await selectAndConfirmFirstSafeFilter(
-      client,
-      rootState.iframe.documentNodeId,
-      buildRecommendFilterSelectionOptions(normalizedFilter)
-    );
-    if (!filterResult.confirmed) {
-      throw new Error("Recommend run filter selection was not confirmed");
-    }
-    rootState = await getRecommendRoots(client);
-    rootState = await ensureRecommendViewport(rootState, "filter");
-    runControl.checkpoint({
-      filter: compactFilterResult(filterResult)
-    });
-  }
-
-  await runControl.waitIfPaused();
-  runControl.throwIfCanceled();
-  runControl.setPhase("recommend:cards");
-  rootState = await ensureRecommendViewport(rootState, "cards");
-  cardNodeIds = await waitForRecommendCardNodeIds(client, rootState.iframe.documentNodeId, {
-    timeoutMs: cardTimeoutMs,
-    intervalMs: 300
-  });
-  if (!cardNodeIds.length) {
-    throw new Error("No recommend candidate cards found for run service");
-  }
-
-  updateRecommendProgress({
-    list_end_reason: null
-  });
-
-  while (countPassedResults(results) < targetPassCount) {
-    const candidateStarted = Date.now();
-    const timings = {};
-    await runControl.waitIfPaused();
-    runControl.throwIfCanceled();
-    runControl.setPhase("recommend:candidate");
-    rootState = await ensureRecommendViewport(rootState, "candidate_loop");
-
-    const nextCandidateResult = await measureTiming(timings, "card_read_ms", () => getNextInfiniteListCandidate({
-      client,
-      state: listState,
-      maxScrolls: listMaxScrolls,
-      stableSignatureLimit: listStableSignatureLimit,
-      wheelDeltaY: listWheelDeltaY,
-      settleMs: listSettleMs,
-      fallbackPoint: listFallbackResolver,
-      findNodeIds: async () => {
-        let currentRootState = await getRecommendRoots(client);
-        currentRootState = await ensureRecommendViewport(currentRootState, "candidate_find_nodes");
-        rootState = currentRootState;
-        const currentCardNodeIds = await waitForRecommendCardNodeIds(client, currentRootState.iframe.documentNodeId, {
-          timeoutMs: Math.min(cardTimeoutMs, 5000),
-          intervalMs: 300
-        });
-        cardNodeIds = currentCardNodeIds;
-        return currentCardNodeIds;
-      },
-      readCandidate: async (nodeId, { visibleIndex }) => readRecommendCardCandidate(client, nodeId, {
-        targetUrl,
-        source: "recommend-run-card",
-        metadata: {
-          run_candidate_index: results.length,
-          visible_index: visibleIndex
-        }
-      }),
-      detectBottomMarker: async ({ scrollAttempt = 0, signature = {} } = {}) => detectInfiniteListBottomMarker(client, {
-        rootNodeId: rootState?.iframe?.documentNodeId,
-        markerSelectors: RECOMMEND_BOTTOM_MARKER_SELECTORS,
-        refreshSelectors: [RECOMMEND_END_REFRESH_SELECTOR],
-        textScanSelectors: scrollAttempt > 0 || (signature?.stable_signature_count || 0) >= 2 ? undefined : [],
-        maxTextScanNodes: 500
-      })
-    }));
-    if (!nextCandidateResult.ok) {
-      listEndReason = nextCandidateResult.reason || "list_exhausted";
-      if (
-          (nextCandidateResult.end_reached || isRefreshableListStall(nextCandidateResult.reason))
-          && refreshOnEnd
-          && countPassedResults(results) < targetPassCount
-          && refreshRounds < Math.max(0, Number(maxRefreshRounds) || 0)
-      ) {
-        await runControl.waitIfPaused();
-        runControl.throwIfCanceled();
-        runControl.setPhase("recommend:refresh");
-        refreshRounds += 1;
-        const refreshResult = await refreshRecommendListAtEnd(client, {
-          rootState,
-          jobLabel,
-          pageScope: pageScopeSelection?.effective_scope || requestedPageScope,
-          fallbackPageScope: normalizedFallbackPageScope,
-          filter: normalizedFilter,
-          forceRecentNotView: true,
-          cardTimeoutMs,
-          buttonSettleMs: refreshButtonSettleMs,
-          reloadSettleMs: refreshReloadSettleMs
-        });
-        const compactRefresh = compactRefreshAttempt(refreshResult);
-        refreshAttempts.push(compactRefresh);
-        runControl.checkpoint({
-          refresh_round: refreshRounds,
-          refresh: compactRefresh
-        });
-        updateRecommendProgress({
-          card_count: refreshResult.card_count || cardNodeIds.length,
-          refresh_method: refreshResult.method || null,
-          refresh_forced_recent_not_view: true,
-          list_end_reason: listEndReason
-        });
-        if (refreshResult.ok) {
-          rootState = refreshResult.root_state || await getRecommendRoots(client);
-          rootState = await ensureRecommendViewport(rootState, "refresh_after");
-          cardNodeIds = await waitForRecommendCardNodeIds(client, rootState.iframe.documentNodeId, {
-            timeoutMs: cardTimeoutMs,
-            intervalMs: 300
-          });
-          resetInfiniteListForRefreshRound(listState, {
-            reason: listEndReason,
-            round: refreshRounds,
-            method: refreshResult.method,
-            metadata: {
-              card_count: cardNodeIds.length,
-              forced_recent_not_view: true
-            }
-          });
-          listEndReason = "";
-          continue;
-        }
-      }
-      break;
-    }
-
-    const index = results.length;
-    let cardNodeId = nextCandidateResult.item.node_id;
-    const candidateKey = nextCandidateResult.item.key;
-    let cardCandidate = nextCandidateResult.item.candidate;
-
-    let screeningCandidate = cardCandidate;
-    let detailResult = null;
-    let recoverableDetailError = null;
-    let detailStep = "not_started";
-    if (index < effectiveDetailLimit) {
-      try {
-        await runControl.waitIfPaused();
-        runControl.throwIfCanceled();
-        runControl.setPhase("recommend:detail");
-        detailStep = "ensure_viewport";
-        rootState = await ensureRecommendViewport(rootState, "detail");
-        checkpointInProgressCandidate({ index, candidateKey, cardNodeId, detailStep });
-        detailStep = "open_detail";
-        networkRecorder.clear();
-        const openedDetail = await openRecommendCardDetailWithFreshRetry(client, {
-          cardNodeId,
-          candidateKey,
-          cardCandidate,
-          rootState,
-          targetUrl,
-          retryTimeoutMs: 8000,
-          maxAttempts: 3
-        });
-        addTiming(timings, "candidate_click_ms", openedDetail.timings?.candidate_click_ms);
-        addTiming(timings, "detail_open_ms", openedDetail.timings?.detail_open_ms);
-        cardNodeId = openedDetail.card_node_id || cardNodeId;
-        cardCandidate = openedDetail.card_candidate || cardCandidate;
-        screeningCandidate = cardCandidate;
-        const waitPlan = getCvNetworkWaitPlan(cvAcquisitionState);
-        detailStep = "wait_network";
-        const networkWait = await measureTiming(timings, "network_cv_wait_ms", () => waitForCvNetworkEvents(
-          waitForRecommendDetailNetworkEvents,
-          networkRecorder,
-          {
-            waitPlan,
-            minCount: 1,
-            requireLoaded: true,
-            intervalMs: 120
-          }
-        ));
-        if (networkWait?.elapsed_ms != null) {
-          timings.network_cv_wait_ms = Math.round(Number(networkWait.elapsed_ms) || 0);
-        }
-        detailStep = "extract_detail";
-        detailResult = await extractRecommendDetailCandidate(client, {
-          cardCandidate,
-          cardNodeId,
-          detailState: openedDetail.detail_state,
-          networkEvents: networkRecorder.events,
-          targetUrl,
-          closeDetail: false,
-          networkParseRetryMs: waitPlan.mode_before === "image" ? 500 : 2200,
-          networkParseIntervalMs: 250
-        });
-        addTiming(timings, "late_network_retry_ms", detailResult.network_parse_retry_elapsed_ms);
-
-        const parsedNetworkProfileCount = countParsedNetworkProfiles(detailResult);
-        let source = "network";
-        let imageEvidence = null;
-        let captureTarget = null;
-        let captureTargetWait = null;
-        if (parsedNetworkProfileCount > 0) {
-          recordCvNetworkHit(cvAcquisitionState, {
-            parsedNetworkProfileCount,
-            waitResult: networkWait
-          });
-        } else {
-          detailStep = "wait_capture_target";
-          captureTargetWait = await waitForCvCaptureTarget(client, openedDetail.detail_state, {
-            domain: "recommend",
-            timeoutMs: 6000,
-            intervalMs: 250
-          });
-          captureTarget = captureTargetWait.target || null;
-          const captureNodeId = captureTarget?.node_id || null;
-          if (captureNodeId) {
-            const imageEvidencePath = imageEvidenceFilePath({
-              imageOutputDir,
-              domain: "recommend",
-              runId: runControl?.runId,
-              index,
-              extension: "jpg"
-            });
-            try {
-              detailStep = "capture_image";
-              imageEvidence = await measureTiming(timings, "screenshot_capture_ms", () => captureScrolledNodeScreenshots(client, captureNodeId, {
-                filePath: imageEvidencePath,
-                format: "jpeg",
-                quality: 72,
-                optimize: true,
-                resizeMaxWidth: 1100,
-                captureViewport: false,
-                padding: 0,
-                maxScreenshots: maxImagePages,
-                wheelDeltaY: imageWheelDeltaY,
-                settleMs: 350,
-                scrollMethod: "dom-anchor-fallback-input",
-                stepTimeoutMs: 45000,
-                totalTimeoutMs: 90000,
-                duplicateStopCount: 1,
-                skipDuplicateScreenshots: true,
-                composeForLlm: true,
-                llmPagesPerImage: 3,
-                llmResizeMaxWidth: 1100,
-                llmQuality: 72,
-                metadata: {
-                  domain: "recommend",
-                  capture_mode: "scroll_sequence",
-                  acquisition_reason: "network_miss_image_fallback",
-                  run_candidate_index: index,
-                  candidate_key: candidateKey,
-                  capture_target: captureTarget,
-                  capture_target_wait: captureTargetWait
-                }
-              }));
-              source = "image";
-            } catch (error) {
-              if (!isRecoverableImageCaptureError(error)) throw error;
-              const recoveryCount = candidateRecoveryCounts.get(candidateKey) || 0;
-              if (recoveryCount < 1) {
-                candidateRecoveryCounts.set(candidateKey, recoveryCount + 1);
-                timings.image_capture_recovery_trigger = compactError(error, "IMAGE_CAPTURE_FAILED");
-                checkpointInProgressCandidate({ index, candidateKey, cardNodeId, detailStep, error });
-                await closeRecommendDetail(client, { attemptsLimit: 2 }).catch(() => null);
-                await recoverAndReapplyRecommendContext(`image_capture:${detailStep}`, error, {
-                  forceRecentNotView: true
-                });
-                continue;
-              }
-              imageEvidence = createRecoverableImageCaptureEvidence(error, {
-                elapsedMs: timings.screenshot_capture_ms,
-                filePath: imageEvidencePath,
-                extension: "jpg",
-                maxScreenshots: maxImagePages
-              });
-              source = "image_capture_failed";
-            }
-            recordCvImageFallback(cvAcquisitionState, {
-              reason: source === "image_capture_failed"
-                ? "network_miss_image_capture_failed"
-                : "network_miss_image_fallback",
-              parsedNetworkProfileCount,
-              waitResult: networkWait,
-              imageEvidence
-            });
-          } else {
-            source = "missing_capture_node";
-            recordCvNetworkMiss(cvAcquisitionState, {
-              reason: "network_miss_no_capture_node",
-              parsedNetworkProfileCount,
-              waitResult: networkWait
-            });
-          }
-        }
-
-        detailResult.image_evidence = imageEvidence;
-        detailResult.cv_acquisition = {
-          source,
-          mode_after: compactCvAcquisitionState(cvAcquisitionState).mode,
-          wait_plan: waitPlan,
-          network_wait: networkWait,
-          parsed_network_profile_count: parsedNetworkProfileCount,
-          image_evidence: summarizeImageEvidence(imageEvidence),
-          capture_target: captureTarget || null,
-          capture_target_wait: captureTargetWait
-        };
-        screeningCandidate = detailResult.candidate;
-      } catch (error) {
-        if (!isRecoverableRecommendDetailError(error)) throw error;
-        const recoveryCount = candidateRecoveryCounts.get(candidateKey) || 0;
-        if (recoveryCount < 1) {
-          candidateRecoveryCounts.set(candidateKey, recoveryCount + 1);
-          timings.detail_recovery_trigger = compactRecoverableDetailError(error);
-          checkpointInProgressCandidate({ index, candidateKey, cardNodeId, detailStep, error });
-          await closeRecommendDetail(client, { attemptsLimit: 2 }).catch(() => null);
-          await recoverAndReapplyRecommendContext(`detail:${detailStep}`, error, {
-            forceRecentNotView: true
-          });
-          continue;
-        }
-        recoverableDetailError = error;
-        detailResult = null;
-        timings.detail_recovered_error = compactRecoverableDetailError(error);
-        await closeRecommendDetail(client, { attemptsLimit: 2 }).catch(() => null);
-      }
-    }
-
-    await runControl.waitIfPaused();
-    runControl.throwIfCanceled();
-    runControl.setPhase("recommend:screening");
-    let llmResult = null;
-    if (useLlmScreening) {
-      if (recoverableDetailError || detailResult?.image_evidence?.ok === false) {
-        llmResult = null;
-      } else if (!llmConfig) {
-        llmResult = createMissingLlmConfigResult();
-      } else {
-        try {
-          const llmTimingKey = detailResult?.image_evidence?.file_paths?.length
-            ? "vision_model_ms"
-            : "text_model_ms";
-          llmResult = await measureTiming(timings, llmTimingKey, () => callScreeningLlm({
-            candidate: screeningCandidate,
-            criteria,
-            config: llmConfig,
-            timeoutMs: llmTimeoutMs,
-            imageEvidence: detailResult?.image_evidence || null,
-            maxImages: llmImageLimit,
-            imageDetail: llmImageDetail
-          }));
-        } catch (error) {
-          llmResult = createFailedLlmScreeningResult(error);
-        }
-      }
-      if (detailResult) detailResult.llm_result = llmResult;
-    }
-    const screening = recoverableDetailError
-      ? createRecoverableDetailFailureScreening(screeningCandidate, recoverableDetailError)
-      : detailResult?.image_evidence?.ok === false
-      ? createImageCaptureFailureScreening(screeningCandidate, {
-        code: detailResult.image_evidence.error_code,
-        message: detailResult.image_evidence.error
-      })
-      : useLlmScreening
-        ? llmResultToScreening(llmResult, screeningCandidate)
-        : screenCandidate(screeningCandidate, { criteria });
-    let actionDiscovery = null;
-    let postActionResult = null;
-    if (postActionEnabled && detailResult) {
-      const postActionStarted = Date.now();
-      await runControl.waitIfPaused();
-      runControl.throwIfCanceled();
-      runControl.setPhase("recommend:post-action");
-      actionDiscovery = await waitForRecommendDetailActionControls(client, {
-        timeoutMs: actionTimeoutMs,
-        intervalMs: actionIntervalMs,
-        requireAny: true
-      });
-      postActionResult = await runRecommendPostAction({
-        client,
-        screening,
-        actionDiscovery,
-        postAction: normalizedPostAction,
-        greetCount,
-        maxGreetCount: Number.isInteger(maxGreetCount) ? maxGreetCount : null,
-        executePostAction,
-        afterClickDelayMs: actionAfterClickDelayMs
-      });
-      if (postActionResult.counted_as_greet && postActionResult.action_clicked) {
-        greetCount += 1;
-      }
-      addTiming(timings, "post_action_ms", Date.now() - postActionStarted);
-    }
-    if (detailResult && closeDetail) {
-      detailResult.close_result = await measureTiming(timings, "close_detail_ms", () => closeRecommendDetail(client));
-      if (!detailResult.close_result?.closed) {
-        const closeError = createRecommendCloseFailureError(detailResult.close_result);
-        const recovery = await recoverAndReapplyRecommendContext("detail_close_failed", closeError, {
-          forceRecentNotView: true
-        });
-        detailResult.cv_acquisition = {
-          ...(detailResult.cv_acquisition || {}),
-          close_recovery: {
-            ok: Boolean(recovery.ok),
-            method: recovery.method || "",
-            forced_recent_not_view: Boolean(recovery.forced_recent_not_view),
-            card_count: recovery.card_count || 0
-          }
-        };
-      }
-    }
-    timings.total_ms = Date.now() - candidateStarted;
-    const compactResult = {
-      index,
-      candidate_key: candidateKey,
-      card_node_id: cardNodeId,
-      candidate: compactCandidate(screeningCandidate),
-      detail: compactDetail(detailResult),
-      llm_screening: detailResult ? null : compactScreeningLlmResult(llmResult),
-      screening: compactScreening(screening),
-      action_discovery: compactActionDiscovery(actionDiscovery),
-      post_action: postActionResult,
-      error: recoverableDetailError
-        ? compactRecoverableDetailError(recoverableDetailError)
-        : detailResult?.image_evidence?.ok === false
-        ? compactError({
-          code: detailResult.image_evidence.error_code,
-          message: detailResult.image_evidence.error
-        }, "IMAGE_CAPTURE_FAILED")
-        : null,
-      timings
-    };
-    results.push(compactResult);
-    markInfiniteListCandidateProcessed(listState, candidateKey, {
-      metadata: {
-        result_index: index,
-        candidate_id: screeningCandidate.id || null
-      }
-    });
-
-    updateRecommendProgress({
-      last_candidate_id: screeningCandidate.id || null,
-      last_candidate_key: candidateKey,
-      last_score: screening.score
-    });
-    const checkpointStarted = Date.now();
-    runControl.checkpoint({
-      results,
-      last_candidate: {
-        id: screeningCandidate.id || null,
-        key: candidateKey,
-        identity: screeningCandidate.identity || {},
-        screening: {
-          status: screening.status,
-          passed: screening.passed,
-          score: screening.score
-        },
-        llm_screening: compactScreeningLlmResult(llmResult),
-        error: compactResult.error,
-        post_action: postActionResult
-      }
-    });
-    addTiming(compactResult.timings, "checkpoint_save_ms", Date.now() - checkpointStarted);
-
-    if (postActionResult?.stop_run) {
-      listEndReason = postActionResult.reason || "post_action_stop";
-      break;
-    }
-
-    if (delayMs > 0) {
-      const sleepStarted = Date.now();
-      await runControl.sleep(delayMs);
-      addTiming(compactResult.timings, "sleep_ms", Date.now() - sleepStarted);
-      compactResult.timings.total_ms = Date.now() - candidateStarted;
-    }
-  }
-
-  runControl.setPhase("recommend:done");
-  return {
-    domain: "recommend",
-    target_url: targetUrl,
-    job_selection: compactJobSelection(jobSelection),
-    page_scope: compactPageScopeSelection(pageScopeSelection),
-    filter: compactFilterResult(filterResult),
-    card_count: cardNodeIds.length,
-    candidate_list: compactInfiniteListState(listState),
-    viewport_health: {
-      stats: viewportGuard.getStats(),
-      events: viewportGuard.getEvents()
-    },
-    list_end_reason: listEndReason || null,
-    refresh_rounds: refreshRounds,
-    refresh_attempts: refreshAttempts,
-    context_recoveries: contextRecoveryAttempts,
-    ...countRecommendResultStatuses(results, { greetCount }),
-    results
-  };
-}
-
-export function createRecommendRunService({
-  lifecycle,
-  idPrefix = "recommend",
-  workflow = runRecommendWorkflow,
-  onSnapshot = null
-} = {}) {
-  const manager = lifecycle || createRunLifecycleManager({ idPrefix, onSnapshot });
-
-  function startRecommendRun({
-    runId = "",
-    pid = process.pid,
-    client,
-    targetUrl = "",
-    criteria = "",
-    jobLabel = "",
-    pageScope = "recommend",
-    fallbackPageScope = "recommend",
-    filter = {},
-    maxCandidates = 5,
-    detailLimit,
-    closeDetail = true,
-    delayMs = 0,
-    cardTimeoutMs = 10000,
-    maxImagePages = DEFAULT_MAX_IMAGE_PAGES,
-    imageWheelDeltaY = 650,
-    cvAcquisitionMode = "unknown",
-    listMaxScrolls = 20,
-    listStableSignatureLimit = 5,
-    listWheelDeltaY = 850,
-    listSettleMs = 2200,
-    listFallbackPoint = null,
-    refreshOnEnd = true,
-    maxRefreshRounds = 2,
-    refreshButtonSettleMs = 8000,
-    refreshReloadSettleMs = 8000,
-    postAction = "none",
-    maxGreetCount = null,
-    executePostAction = true,
-    actionTimeoutMs = 8000,
-    actionIntervalMs = 500,
-    actionAfterClickDelayMs = 900,
-    screeningMode = "llm",
-    llmConfig = null,
-    llmTimeoutMs = 120000,
-    llmImageLimit = 8,
-    llmImageDetail = "high",
-    imageOutputDir = "",
-    name = "recommend-domain-run"
-  } = {}) {
-    if (!client) throw new Error("startRecommendRun requires a guarded CDP client");
-    const normalizedFilter = normalizeFilter(filter);
-    const normalizedPostAction = normalizeRecommendPostAction(postAction) || "none";
-    const requestedPageScope = normalizeRecommendPageScope(pageScope) || "recommend";
-    const normalizedFallbackPageScope = normalizeRecommendPageScope(fallbackPageScope) || "recommend";
-    const normalizedScreeningMode = normalizeScreeningMode(screeningMode);
-    const candidateLimit = Math.max(1, Number(maxCandidates) || 1);
-    const normalizedDetailLimit = detailLimit == null ? null : Math.max(0, Number(detailLimit) || 0);
-    return manager.startRun({
-      runId,
-      name,
-      pid,
-      context: {
-        domain: "recommend",
-        target_url: targetUrl,
-        criteria_present: Boolean(criteria),
-        job_label: jobLabel || "",
-        requested_page_scope: requestedPageScope,
-        fallback_page_scope: normalizedFallbackPageScope,
-        filter: normalizedFilter,
-        max_candidates: maxCandidates,
-        max_candidates_semantics: "passed_candidates",
-        detail_limit: normalizedDetailLimit,
-        close_detail: closeDetail,
-        cv_acquisition_mode: cvAcquisitionMode,
-        max_image_pages: maxImagePages,
-        image_wheel_delta_y: imageWheelDeltaY,
-        list_max_scrolls: listMaxScrolls,
-        list_stable_signature_limit: listStableSignatureLimit,
-        list_wheel_delta_y: listWheelDeltaY,
-        list_settle_ms: listSettleMs,
-        list_fallback_point: listFallbackPoint,
-        refresh_on_end: refreshOnEnd,
-        max_refresh_rounds: maxRefreshRounds,
-        refresh_button_settle_ms: refreshButtonSettleMs,
-        refresh_reload_settle_ms: refreshReloadSettleMs,
-        post_action: normalizedPostAction,
-        max_greet_count: Number.isInteger(maxGreetCount) ? maxGreetCount : null,
-        execute_post_action: Boolean(executePostAction),
-        action_timeout_ms: actionTimeoutMs,
-        screening_mode: normalizedScreeningMode,
-        llm_configured: Boolean(llmConfig),
-        llm_timeout_ms: llmTimeoutMs,
-        llm_image_limit: llmImageLimit,
-        llm_image_detail: llmImageDetail,
-        image_output_dir: imageOutputDir || ""
-      },
-      progress: {
-        card_count: 0,
-        target_count: candidateLimit,
-        target_count_semantics: "passed_candidates",
-        processed: 0,
-        screened: 0,
-        detail_opened: 0,
-        llm_screened: 0,
-        passed: 0,
-        greet_count: 0,
-        post_action_clicked: 0,
-        image_capture_failed: 0,
-        detail_open_failed: 0,
-        transient_recovered: 0,
-        context_recoveries: 0
-      },
-      checkpoint: {},
-      task: (runControl) => workflow({
-        client,
-        targetUrl,
-        criteria,
-        jobLabel,
-        pageScope: requestedPageScope,
-        fallbackPageScope: normalizedFallbackPageScope,
-        filter: normalizedFilter,
-        maxCandidates,
-        detailLimit: normalizedDetailLimit,
-        closeDetail,
-        delayMs,
-        cardTimeoutMs,
-        maxImagePages,
-        imageWheelDeltaY,
-        cvAcquisitionMode,
-        listMaxScrolls,
-        listStableSignatureLimit,
-        listWheelDeltaY,
-        listSettleMs,
-        listFallbackPoint,
-        refreshOnEnd,
-        maxRefreshRounds,
-        refreshButtonSettleMs,
-        refreshReloadSettleMs,
-        postAction: normalizedPostAction,
-        maxGreetCount,
-        executePostAction,
-        actionTimeoutMs,
-        actionIntervalMs,
-        actionAfterClickDelayMs,
-        screeningMode: normalizedScreeningMode,
-        llmConfig,
-        llmTimeoutMs,
-        llmImageLimit,
-        llmImageDetail,
-        imageOutputDir
-      }, runControl)
-    });
-  }
-
-  return {
-    startRecommendRun,
-    getRecommendRun: manager.getRun,
-    pauseRecommendRun: manager.pauseRun,
-    resumeRecommendRun: manager.resumeRun,
-    cancelRecommendRun: manager.cancelRun,
-    waitForRecommendRun: manager.waitForRun,
-    listRecommendRuns: manager.listRuns,
-    manager
-  };
-}
+
+export function countRecommendResultStatuses(results = [], {
+  greetCount = 0
+} = {}) {
+  return {
+    processed: results.length,
+    screened: results.length,
+    detail_opened: results.filter((item) => item.detail).length,
+    passed: results.filter((item) => item.screening?.passed).length,
+    llm_screened: results.filter((item) => item.detail?.llm_screening || item.llm_screening).length,
+    greet_count: greetCount,
+    post_action_clicked: results.filter((item) => item.post_action?.action_clicked).length,
+    image_capture_failed: results.filter((item) => item.detail?.image_evidence?.ok === false).length,
+    detail_open_failed: results.filter((item) => (
+      item.error?.code === "DETAIL_STALE_NODE"
+      || item.error?.code === "DETAIL_OPEN_FAILED"
+    )).length,
+    transient_recovered: results.filter((item) => (
+      item.error?.code === "DETAIL_STALE_NODE"
+      || item.error?.code === "IMAGE_CAPTURE_STALE_NODE"
+      || item.error?.code === "IMAGE_CAPTURE_TIMEOUT"
+      || item.error?.code === "IMAGE_CAPTURE_TOTAL_TIMEOUT"
+    )).length
+  };
+}
+
+function countPassedResults(results = []) {
+  return countRecommendResultStatuses(results).passed;
+}
+
+function compactError(error, fallbackCode = "RECOMMEND_RUN_ERROR") {
+  if (!error) return null;
+  return {
+    code: error.code || fallbackCode,
+    message: error.message || String(error)
+  };
+}
+
+function createRecommendCloseFailureError(closeResult) {
+  const error = new Error(closeResult?.reason || "Recommend detail did not close before recovery");
+  error.code = "DETAIL_CLOSE_FAILED";
+  error.close_result = closeResult || null;
+  return error;
+}
+
+export function isRecoverableImageCaptureError(error) {
+  const code = String(error?.code || "");
+  if (code === "IMAGE_CAPTURE_TIMEOUT" || code === "IMAGE_CAPTURE_TOTAL_TIMEOUT") return true;
+  if (isStaleRecommendNodeError(error)) return true;
+  return /Image fallback capture timed out/i.test(String(error?.message || error || ""));
+}
+
+function collectPartialImageEvidencePaths(basePath = "", extension = "jpg", maxCount = 12) {
+  const resolved = String(basePath || "").trim();
+  if (!resolved) return [];
+  const parsed = path.parse(resolved);
+  const ext = parsed.ext || `.${String(extension || "jpg").replace(/^\./, "") || "jpg"}`;
+  const files = [];
+  for (let index = 0; index < Math.max(1, Number(maxCount) || 1); index += 1) {
+    const page = String(index + 1).padStart(2, "0");
+    const candidatePath = path.join(parsed.dir, `${parsed.name}-page-${page}${ext}`);
+    if (fs.existsSync(candidatePath)) files.push(candidatePath);
+  }
+  return files;
+}
+
+export function createRecoverableImageCaptureEvidence(error, {
+  elapsedMs = 0,
+  filePath = "",
+  extension = "jpg",
+  maxScreenshots = DEFAULT_MAX_IMAGE_PAGES
+} = {}) {
+  const filePaths = collectPartialImageEvidencePaths(filePath, extension, maxScreenshots);
+  return {
+    schema_version: 1,
+    ok: false,
+    source: "image-scroll-sequence",
+    elapsed_ms: Math.max(0, Math.round(Number(error?.elapsed_ms ?? elapsedMs) || 0)),
+    capture_count: filePaths.length,
+    screenshot_count: filePaths.length,
+    unique_screenshot_count: filePaths.length,
+    dropped_duplicate_count: 0,
+    total_byte_length: 0,
+    original_total_byte_length: 0,
+    llm_screenshot_count: 0,
+    llm_total_byte_length: 0,
+    llm_original_total_byte_length: 0,
+    llm_composition_error: null,
+    error_code: error?.code || (isStaleRecommendNodeError(error) ? "IMAGE_CAPTURE_STALE_NODE" : "IMAGE_CAPTURE_FAILED"),
+    error: error?.message || String(error || "Image capture failed"),
+    file_paths: filePaths,
+    llm_file_paths: []
+  };
+}
+
+function createImageCaptureFailureScreening(candidate, error) {
+  return {
+    status: "fail",
+    passed: false,
+    score: 0,
+    reasons: ["image_capture_failed"],
+    error: compactError(error, "IMAGE_CAPTURE_FAILED"),
+    candidate
+  };
+}
+
+export function isRecoverableRecommendDetailError(error) {
+  return isStaleRecommendNodeError(error);
+}
+
+function compactRecoverableDetailError(error) {
+  return compactError(error, isStaleRecommendNodeError(error) ? "DETAIL_STALE_NODE" : "DETAIL_OPEN_FAILED");
+}
+
+function createRecoverableDetailFailureScreening(candidate, error) {
+  return {
+    status: "fail",
+    passed: false,
+    score: 0,
+    reasons: isStaleRecommendNodeError(error)
+      ? ["detail_open_failed", "stale_node"]
+      : ["detail_open_failed"],
+    error: compactRecoverableDetailError(error),
+    candidate
+  };
+}
+
+export async function runRecommendWorkflow({
+  client,
+  targetUrl = "",
+  criteria = "",
+  jobLabel = "",
+  pageScope = "recommend",
+  fallbackPageScope = "recommend",
+  filter = {},
+  maxCandidates = 5,
+  detailLimit,
+  closeDetail = true,
+  delayMs = 0,
+  cardTimeoutMs = 10000,
+  maxImagePages = DEFAULT_MAX_IMAGE_PAGES,
+  imageWheelDeltaY = 650,
+  cvAcquisitionMode = "unknown",
+  listMaxScrolls = 20,
+  listStableSignatureLimit = 5,
+  listWheelDeltaY = 850,
+  listSettleMs = 2200,
+  listFallbackPoint = null,
+  refreshOnEnd = true,
+  maxRefreshRounds = 2,
+  refreshButtonSettleMs = 8000,
+  refreshReloadSettleMs = 8000,
+  postAction = "none",
+  maxGreetCount = null,
+  executePostAction = true,
+  actionTimeoutMs = 8000,
+  actionIntervalMs = 500,
+  actionAfterClickDelayMs = 900,
+  screeningMode = "llm",
+  llmConfig = null,
+  llmTimeoutMs = 120000,
+  llmImageLimit = 8,
+  llmImageDetail = "high",
+  imageOutputDir = ""
+} = {}, runControl) {
+  if (!client) throw new Error("runRecommendWorkflow requires a guarded CDP client");
+  const normalizedFilter = normalizeFilter(filter);
+  const normalizedPostAction = normalizeRecommendPostAction(postAction) || "none";
+  const requestedPageScope = normalizeRecommendPageScope(pageScope) || "recommend";
+  const normalizedFallbackPageScope = normalizeRecommendPageScope(fallbackPageScope) || "recommend";
+  const normalizedScreeningMode = normalizeScreeningMode(screeningMode);
+  const useLlmScreening = normalizedScreeningMode !== "deterministic";
+  const postActionEnabled = normalizedPostAction !== "none";
+  const targetPassCount = Math.max(1, Number(maxCandidates) || 1);
+  const detailCountLimit = detailLimit == null ? Number.POSITIVE_INFINITY : Math.max(0, Number(detailLimit) || 0);
+  const effectiveDetailLimit = postActionEnabled ? Number.POSITIVE_INFINITY : detailCountLimit;
+  const networkRecorder = effectiveDetailLimit > 0
+    ? createRecommendDetailNetworkRecorder(client)
+    : null;
+  const cvAcquisitionState = createCvAcquisitionState({ mode: cvAcquisitionMode });
+  const listState = createInfiniteListState({
+    domain: "recommend",
+    listName: "recommend-candidates"
+  });
+  const viewportGuard = createViewportRunGuard({
+    client,
+    domain: "recommend",
+    root: "frame",
+    frameOwnerRoot: "frameOwner",
+    runControl,
+    getRoots: getRecommendRoots
+  });
+  async function ensureRecommendViewport(rootState, phase) {
+    const result = await viewportGuard.ensure(rootState, { phase });
+    return result.rootState || rootState;
+  }
+  const results = [];
+  const refreshAttempts = [];
+  let refreshRounds = 0;
+  let contextRecoveryAttempts = 0;
+  let greetCount = 0;
+  const candidateRecoveryCounts = new Map();
+  let jobSelection = null;
+  let pageScopeSelection = null;
+  let filterResult = null;
+  let cardNodeIds = [];
+  let listEndReason = "";
+  const listFallbackResolver = listFallbackPoint || (async ({ items = [] } = {}) => resolveInfiniteListFallbackPoint(client, {
+    rootNodeId: rootState?.iframe?.documentNodeId,
+    containerSelectors: RECOMMEND_LIST_CONTAINER_SELECTORS,
+    itemNodeIds: items.map((item) => item.node_id).filter(Boolean),
+    itemSelectors: [RECOMMEND_CARD_SELECTOR],
+    viewportPoint: { xRatio: 0.28, yRatio: 0.5 },
+    validateViewportPoint: true
+  }));
+
+  function updateRecommendProgress(extra = {}) {
+    const counts = countRecommendResultStatuses(results, { greetCount });
+    const listSnapshot = compactInfiniteListState(listState);
+    runControl.updateProgress({
+      card_count: cardNodeIds.length,
+      target_count: targetPassCount,
+      target_count_semantics: "passed_candidates",
+      ...counts,
+      screening_mode: normalizedScreeningMode,
+      unique_seen: listSnapshot.seen_count,
+      scroll_count: listSnapshot.scroll_count,
+      refresh_rounds: refreshRounds,
+      refresh_attempts: refreshAttempts.length,
+      context_recoveries: contextRecoveryAttempts,
+      list_end_reason: listEndReason || null,
+      viewport_checks: viewportGuard.getStats().checks,
+      viewport_recoveries: viewportGuard.getStats().recoveries,
+      ...extra
+    });
+  }
+
+  function checkpointInProgressCandidate({
+    index = results.length,
+    candidateKey = "",
+    cardNodeId = null,
+    detailStep = "",
+    error = null
+  } = {}) {
+    runControl.checkpoint({
+      in_progress_candidate: {
+        index,
+        key: candidateKey,
+        card_node_id: cardNodeId,
+        detail_step: detailStep || null,
+        counters: countRecommendResultStatuses(results, { greetCount }),
+        error: compactError(error, "RECOMMEND_IN_PROGRESS_ERROR")
+      },
+      candidate_list: compactInfiniteListState(listState)
+    });
+  }
+
+  async function recoverAndReapplyRecommendContext(reason = "context_recovery", error = null, {
+    forceRecentNotView = true
+  } = {}) {
+    await runControl.waitIfPaused();
+    runControl.throwIfCanceled();
+    const started = Date.now();
+    runControl.setPhase("recommend:recover-context");
+    contextRecoveryAttempts += 1;
+    const refreshResult = await refreshRecommendListAtEnd(client, {
+      rootState,
+      jobLabel,
+      pageScope: pageScopeSelection?.effective_scope || requestedPageScope,
+      fallbackPageScope: normalizedFallbackPageScope,
+      filter: normalizedFilter,
+      preferEndRefreshButton: false,
+      forceNavigate: true,
+      targetUrl: targetUrl || RECOMMEND_TARGET_URL,
+      forceRecentNotView,
+      cardTimeoutMs,
+      buttonSettleMs: refreshButtonSettleMs,
+      reloadSettleMs: refreshReloadSettleMs
+    });
+    const compactRefresh = {
+      ...compactRefreshAttempt(refreshResult),
+      context_recovery: true,
+      recovery_reason: reason,
+      trigger_error: compactError(error, "RECOMMEND_CONTEXT_RECOVERY_TRIGGER"),
+      elapsed_ms: Date.now() - started
+    };
+    refreshAttempts.push(compactRefresh);
+    runControl.checkpoint({
+      context_recovery: {
+        attempt: contextRecoveryAttempts,
+        reason,
+        trigger_error: compactError(error, "RECOMMEND_CONTEXT_RECOVERY_TRIGGER"),
+        refresh: compactRefresh,
+        counters: countRecommendResultStatuses(results, { greetCount })
+      },
+      candidate_list: compactInfiniteListState(listState)
+    });
+    if (!refreshResult.ok) {
+      updateRecommendProgress({
+        refresh_method: refreshResult.method || null,
+        refresh_forced_recent_not_view: forceRecentNotView,
+        recovery_reason: reason
+      });
+      throw new Error(`Recommend context recovery failed after ${reason}: ${refreshResult.reason || refreshResult.error || "refresh returned no cards"}`);
+    }
+    rootState = refreshResult.root_state || await getRecommendRoots(client);
+    rootState = await ensureRecommendViewport(rootState, "recover_after");
+    cardNodeIds = await waitForRecommendCardNodeIds(client, rootState.iframe.documentNodeId, {
+      timeoutMs: cardTimeoutMs,
+      intervalMs: 300
+    });
+    resetInfiniteListForRefreshRound(listState, {
+      reason: `context_recovery:${reason}`,
+      round: contextRecoveryAttempts,
+      method: refreshResult.method,
+      metadata: {
+        card_count: cardNodeIds.length,
+        forced_recent_not_view: forceRecentNotView,
+        counters: countRecommendResultStatuses(results, { greetCount })
+      }
+    });
+    listEndReason = "";
+    updateRecommendProgress({
+      card_count: cardNodeIds.length,
+      refresh_method: refreshResult.method || null,
+      refresh_forced_recent_not_view: forceRecentNotView,
+      recovery_reason: reason
+    });
+    return refreshResult;
+  }
+
+  runControl.setPhase("recommend:cleanup");
+  await closeRecommendDetail(client, { attemptsLimit: 2 });
+
+  await runControl.waitIfPaused();
+  runControl.throwIfCanceled();
+  runControl.setPhase("recommend:roots");
+  let rootState = await getRecommendRoots(client);
+  rootState = await ensureRecommendViewport(rootState, "roots");
+  runControl.checkpoint({
+    iframe_selector: rootState.iframe.selector,
+    iframe_document_node_id: rootState.iframe.documentNodeId
+  });
+
+  if (jobLabel) {
+    await runControl.waitIfPaused();
+    runControl.throwIfCanceled();
+    runControl.setPhase("recommend:job");
+    jobSelection = await selectRecommendJob(client, rootState.iframe.documentNodeId, {
+      jobLabel,
+      settleMs: cardTimeoutMs > 45000 ? 12000 : 6000
+    });
+    if (!jobSelection.selected) {
+      throw new Error(`Requested recommend job was not selected: ${jobSelection.reason}`);
+    }
+    rootState = await getRecommendRoots(client);
+    rootState = await ensureRecommendViewport(rootState, "job");
+    runControl.checkpoint({
+      job_selection: compactJobSelection(jobSelection)
+    });
+  }
+
+  await runControl.waitIfPaused();
+  runControl.throwIfCanceled();
+  runControl.setPhase("recommend:page-scope");
+  pageScopeSelection = await selectRecommendPageScope(client, rootState.iframe.documentNodeId, {
+    pageScope: requestedPageScope,
+    fallbackScope: normalizedFallbackPageScope,
+    settleMs: cardTimeoutMs > 45000 ? 3000 : 1200,
+    timeoutMs: Math.min(Math.max(cardTimeoutMs, 10000), 60000)
+  });
+  if (!pageScopeSelection.selected) {
+    throw new Error(`Recommend page scope was not selected: ${pageScopeSelection.reason || pageScopeSelection.effective_scope || requestedPageScope}`);
+  }
+  rootState = await getRecommendRoots(client);
+  rootState = await ensureRecommendViewport(rootState, "page_scope");
+  runControl.checkpoint({
+    page_scope: compactPageScopeSelection(pageScopeSelection)
+  });
+
+  if (normalizedFilter.enabled) {
+    await runControl.waitIfPaused();
+    runControl.throwIfCanceled();
+    runControl.setPhase("recommend:filter");
+    filterResult = await selectAndConfirmFirstSafeFilter(
+      client,
+      rootState.iframe.documentNodeId,
+      buildRecommendFilterSelectionOptions(normalizedFilter)
+    );
+    if (!filterResult.confirmed) {
+      throw new Error("Recommend run filter selection was not confirmed");
+    }
+    rootState = await getRecommendRoots(client);
+    rootState = await ensureRecommendViewport(rootState, "filter");
+    runControl.checkpoint({
+      filter: compactFilterResult(filterResult)
+    });
+  }
+
+  await runControl.waitIfPaused();
+  runControl.throwIfCanceled();
+  runControl.setPhase("recommend:cards");
+  rootState = await ensureRecommendViewport(rootState, "cards");
+  cardNodeIds = await waitForRecommendCardNodeIds(client, rootState.iframe.documentNodeId, {
+    timeoutMs: cardTimeoutMs,
+    intervalMs: 300
+  });
+  if (!cardNodeIds.length) {
+    throw new Error("No recommend candidate cards found for run service");
+  }
+
+  updateRecommendProgress({
+    list_end_reason: null
+  });
+
+  while (countPassedResults(results) < targetPassCount) {
+    const candidateStarted = Date.now();
+    const timings = {};
+    await runControl.waitIfPaused();
+    runControl.throwIfCanceled();
+    runControl.setPhase("recommend:candidate");
+    rootState = await ensureRecommendViewport(rootState, "candidate_loop");
+
+    const nextCandidateResult = await measureTiming(timings, "card_read_ms", () => getNextInfiniteListCandidate({
+      client,
+      state: listState,
+      maxScrolls: listMaxScrolls,
+      stableSignatureLimit: listStableSignatureLimit,
+      wheelDeltaY: listWheelDeltaY,
+      settleMs: listSettleMs,
+      fallbackPoint: listFallbackResolver,
+      findNodeIds: async () => {
+        let currentRootState = await getRecommendRoots(client);
+        currentRootState = await ensureRecommendViewport(currentRootState, "candidate_find_nodes");
+        rootState = currentRootState;
+        const currentCardNodeIds = await waitForRecommendCardNodeIds(client, currentRootState.iframe.documentNodeId, {
+          timeoutMs: Math.min(cardTimeoutMs, 5000),
+          intervalMs: 300
+        });
+        cardNodeIds = currentCardNodeIds;
+        return currentCardNodeIds;
+      },
+      readCandidate: async (nodeId, { visibleIndex }) => readRecommendCardCandidate(client, nodeId, {
+        targetUrl,
+        source: "recommend-run-card",
+        metadata: {
+          run_candidate_index: results.length,
+          visible_index: visibleIndex
+        }
+      }),
+      detectBottomMarker: async ({ scrollAttempt = 0, signature = {} } = {}) => detectInfiniteListBottomMarker(client, {
+        rootNodeId: rootState?.iframe?.documentNodeId,
+        markerSelectors: RECOMMEND_BOTTOM_MARKER_SELECTORS,
+        refreshSelectors: [RECOMMEND_END_REFRESH_SELECTOR],
+        textScanSelectors: scrollAttempt > 0 || (signature?.stable_signature_count || 0) >= 2 ? undefined : [],
+        maxTextScanNodes: 500
+      })
+    }));
+    if (!nextCandidateResult.ok) {
+      listEndReason = nextCandidateResult.reason || "list_exhausted";
+      if (
+          (nextCandidateResult.end_reached || isRefreshableListStall(nextCandidateResult.reason))
+          && refreshOnEnd
+          && countPassedResults(results) < targetPassCount
+          && refreshRounds < Math.max(0, Number(maxRefreshRounds) || 0)
+      ) {
+        await runControl.waitIfPaused();
+        runControl.throwIfCanceled();
+        runControl.setPhase("recommend:refresh");
+        refreshRounds += 1;
+        const refreshResult = await refreshRecommendListAtEnd(client, {
+          rootState,
+          jobLabel,
+          pageScope: pageScopeSelection?.effective_scope || requestedPageScope,
+          fallbackPageScope: normalizedFallbackPageScope,
+          filter: normalizedFilter,
+          forceRecentNotView: true,
+          cardTimeoutMs,
+          buttonSettleMs: refreshButtonSettleMs,
+          reloadSettleMs: refreshReloadSettleMs
+        });
+        const compactRefresh = compactRefreshAttempt(refreshResult);
+        refreshAttempts.push(compactRefresh);
+        runControl.checkpoint({
+          refresh_round: refreshRounds,
+          refresh: compactRefresh
+        });
+        updateRecommendProgress({
+          card_count: refreshResult.card_count || cardNodeIds.length,
+          refresh_method: refreshResult.method || null,
+          refresh_forced_recent_not_view: true,
+          list_end_reason: listEndReason
+        });
+        if (refreshResult.ok) {
+          rootState = refreshResult.root_state || await getRecommendRoots(client);
+          rootState = await ensureRecommendViewport(rootState, "refresh_after");
+          cardNodeIds = await waitForRecommendCardNodeIds(client, rootState.iframe.documentNodeId, {
+            timeoutMs: cardTimeoutMs,
+            intervalMs: 300
+          });
+          resetInfiniteListForRefreshRound(listState, {
+            reason: listEndReason,
+            round: refreshRounds,
+            method: refreshResult.method,
+            metadata: {
+              card_count: cardNodeIds.length,
+              forced_recent_not_view: true
+            }
+          });
+          listEndReason = "";
+          continue;
+        }
+      }
+      break;
+    }
+
+    const index = results.length;
+    let cardNodeId = nextCandidateResult.item.node_id;
+    const candidateKey = nextCandidateResult.item.key;
+    let cardCandidate = nextCandidateResult.item.candidate;
+
+    let screeningCandidate = cardCandidate;
+    let detailResult = null;
+    let recoverableDetailError = null;
+    let detailStep = "not_started";
+    if (index < effectiveDetailLimit) {
+      try {
+        await runControl.waitIfPaused();
+        runControl.throwIfCanceled();
+        runControl.setPhase("recommend:detail");
+        detailStep = "ensure_viewport";
+        rootState = await ensureRecommendViewport(rootState, "detail");
+        checkpointInProgressCandidate({ index, candidateKey, cardNodeId, detailStep });
+        detailStep = "open_detail";
+        networkRecorder.clear();
+        const openedDetail = await openRecommendCardDetailWithFreshRetry(client, {
+          cardNodeId,
+          candidateKey,
+          cardCandidate,
+          rootState,
+          targetUrl,
+          retryTimeoutMs: 8000,
+          maxAttempts: 3
+        });
+        addTiming(timings, "candidate_click_ms", openedDetail.timings?.candidate_click_ms);
+        addTiming(timings, "detail_open_ms", openedDetail.timings?.detail_open_ms);
+        cardNodeId = openedDetail.card_node_id || cardNodeId;
+        cardCandidate = openedDetail.card_candidate || cardCandidate;
+        screeningCandidate = cardCandidate;
+        const waitPlan = getCvNetworkWaitPlan(cvAcquisitionState);
+        detailStep = "wait_network";
+        const networkWait = await measureTiming(timings, "network_cv_wait_ms", () => waitForCvNetworkEvents(
+          waitForRecommendDetailNetworkEvents,
+          networkRecorder,
+          {
+            waitPlan,
+            minCount: 1,
+            requireLoaded: true,
+            intervalMs: 120
+          }
+        ));
+        if (networkWait?.elapsed_ms != null) {
+          timings.network_cv_wait_ms = Math.round(Number(networkWait.elapsed_ms) || 0);
+        }
+        detailStep = "extract_detail";
+        detailResult = await extractRecommendDetailCandidate(client, {
+          cardCandidate,
+          cardNodeId,
+          detailState: openedDetail.detail_state,
+          networkEvents: networkRecorder.events,
+          targetUrl,
+          closeDetail: false,
+          networkParseRetryMs: waitPlan.mode_before === "image" ? 500 : 2200,
+          networkParseIntervalMs: 250
+        });
+        addTiming(timings, "late_network_retry_ms", detailResult.network_parse_retry_elapsed_ms);
+
+        const parsedNetworkProfileCount = countParsedNetworkProfiles(detailResult);
+        let source = "network";
+        let imageEvidence = null;
+        let captureTarget = null;
+        let captureTargetWait = null;
+        if (parsedNetworkProfileCount > 0) {
+          recordCvNetworkHit(cvAcquisitionState, {
+            parsedNetworkProfileCount,
+            waitResult: networkWait
+          });
+        } else {
+          detailStep = "wait_capture_target";
+          captureTargetWait = await waitForCvCaptureTarget(client, openedDetail.detail_state, {
+            domain: "recommend",
+            timeoutMs: 6000,
+            intervalMs: 250
+          });
+          captureTarget = captureTargetWait.target || null;
+          const captureNodeId = captureTarget?.node_id || null;
+          if (captureNodeId) {
+            const imageEvidencePath = imageEvidenceFilePath({
+              imageOutputDir,
+              domain: "recommend",
+              runId: runControl?.runId,
+              index,
+              extension: "jpg"
+            });
+            try {
+              detailStep = "capture_image";
+              imageEvidence = await measureTiming(timings, "screenshot_capture_ms", () => captureScrolledNodeScreenshots(client, captureNodeId, {
+                filePath: imageEvidencePath,
+                format: "jpeg",
+                quality: 72,
+                optimize: true,
+                resizeMaxWidth: 1100,
+                captureViewport: false,
+                padding: 0,
+                maxScreenshots: maxImagePages,
+                wheelDeltaY: imageWheelDeltaY,
+                settleMs: 350,
+                scrollMethod: "dom-anchor-fallback-input",
+                stepTimeoutMs: 45000,
+                totalTimeoutMs: 90000,
+                duplicateStopCount: 1,
+                skipDuplicateScreenshots: true,
+                composeForLlm: true,
+                llmPagesPerImage: 3,
+                llmResizeMaxWidth: 1100,
+                llmQuality: 72,
+                metadata: {
+                  domain: "recommend",
+                  capture_mode: "scroll_sequence",
+                  acquisition_reason: "network_miss_image_fallback",
+                  run_candidate_index: index,
+                  candidate_key: candidateKey,
+                  capture_target: captureTarget,
+                  capture_target_wait: captureTargetWait
+                }
+              }));
+              source = "image";
+            } catch (error) {
+              if (!isRecoverableImageCaptureError(error)) throw error;
+              const recoveryCount = candidateRecoveryCounts.get(candidateKey) || 0;
+              if (recoveryCount < 1) {
+                candidateRecoveryCounts.set(candidateKey, recoveryCount + 1);
+                timings.image_capture_recovery_trigger = compactError(error, "IMAGE_CAPTURE_FAILED");
+                checkpointInProgressCandidate({ index, candidateKey, cardNodeId, detailStep, error });
+                await closeRecommendDetail(client, { attemptsLimit: 2 }).catch(() => null);
+                await recoverAndReapplyRecommendContext(`image_capture:${detailStep}`, error, {
+                  forceRecentNotView: true
+                });
+                continue;
+              }
+              imageEvidence = createRecoverableImageCaptureEvidence(error, {
+                elapsedMs: timings.screenshot_capture_ms,
+                filePath: imageEvidencePath,
+                extension: "jpg",
+                maxScreenshots: maxImagePages
+              });
+              source = "image_capture_failed";
+            }
+            recordCvImageFallback(cvAcquisitionState, {
+              reason: source === "image_capture_failed"
+                ? "network_miss_image_capture_failed"
+                : "network_miss_image_fallback",
+              parsedNetworkProfileCount,
+              waitResult: networkWait,
+              imageEvidence
+            });
+          } else {
+            source = "missing_capture_node";
+            recordCvNetworkMiss(cvAcquisitionState, {
+              reason: "network_miss_no_capture_node",
+              parsedNetworkProfileCount,
+              waitResult: networkWait
+            });
+          }
+        }
+
+        detailResult.image_evidence = imageEvidence;
+        detailResult.cv_acquisition = {
+          source,
+          mode_after: compactCvAcquisitionState(cvAcquisitionState).mode,
+          wait_plan: waitPlan,
+          network_wait: networkWait,
+          parsed_network_profile_count: parsedNetworkProfileCount,
+          image_evidence: summarizeImageEvidence(imageEvidence),
+          capture_target: captureTarget || null,
+          capture_target_wait: captureTargetWait
+        };
+        screeningCandidate = detailResult.candidate;
+      } catch (error) {
+        if (!isRecoverableRecommendDetailError(error)) throw error;
+        const recoveryCount = candidateRecoveryCounts.get(candidateKey) || 0;
+        if (recoveryCount < 1) {
+          candidateRecoveryCounts.set(candidateKey, recoveryCount + 1);
+          timings.detail_recovery_trigger = compactRecoverableDetailError(error);
+          checkpointInProgressCandidate({ index, candidateKey, cardNodeId, detailStep, error });
+          await closeRecommendDetail(client, { attemptsLimit: 2 }).catch(() => null);
+          await recoverAndReapplyRecommendContext(`detail:${detailStep}`, error, {
+            forceRecentNotView: true
+          });
+          continue;
+        }
+        recoverableDetailError = error;
+        detailResult = null;
+        timings.detail_recovered_error = compactRecoverableDetailError(error);
+        await closeRecommendDetail(client, { attemptsLimit: 2 }).catch(() => null);
+      }
+    }
+
+    await runControl.waitIfPaused();
+    runControl.throwIfCanceled();
+    runControl.setPhase("recommend:screening");
+    let llmResult = null;
+    if (useLlmScreening) {
+      if (recoverableDetailError || detailResult?.image_evidence?.ok === false) {
+        llmResult = null;
+      } else if (!llmConfig) {
+        llmResult = createMissingLlmConfigResult();
+      } else {
+        try {
+          const llmTimingKey = detailResult?.image_evidence?.file_paths?.length
+            ? "vision_model_ms"
+            : "text_model_ms";
+          llmResult = await measureTiming(timings, llmTimingKey, () => callScreeningLlm({
+            candidate: screeningCandidate,
+            criteria,
+            config: llmConfig,
+            timeoutMs: llmTimeoutMs,
+            imageEvidence: detailResult?.image_evidence || null,
+            maxImages: llmImageLimit,
+            imageDetail: llmImageDetail
+          }));
+        } catch (error) {
+          llmResult = createFailedLlmScreeningResult(error);
+        }
+      }
+      if (detailResult) detailResult.llm_result = llmResult;
+    }
+    const screening = recoverableDetailError
+      ? createRecoverableDetailFailureScreening(screeningCandidate, recoverableDetailError)
+      : detailResult?.image_evidence?.ok === false
+      ? createImageCaptureFailureScreening(screeningCandidate, {
+        code: detailResult.image_evidence.error_code,
+        message: detailResult.image_evidence.error
+      })
+      : useLlmScreening
+        ? llmResultToScreening(llmResult, screeningCandidate)
+        : screenCandidate(screeningCandidate, { criteria });
+    let actionDiscovery = null;
+    let postActionResult = null;
+    if (postActionEnabled && detailResult) {
+      const postActionStarted = Date.now();
+      await runControl.waitIfPaused();
+      runControl.throwIfCanceled();
+      runControl.setPhase("recommend:post-action");
+      actionDiscovery = await waitForRecommendDetailActionControls(client, {
+        timeoutMs: actionTimeoutMs,
+        intervalMs: actionIntervalMs,
+        requireAny: true
+      });
+      postActionResult = await runRecommendPostAction({
+        client,
+        screening,
+        actionDiscovery,
+        postAction: normalizedPostAction,
+        greetCount,
+        maxGreetCount: Number.isInteger(maxGreetCount) ? maxGreetCount : null,
+        executePostAction,
+        afterClickDelayMs: actionAfterClickDelayMs
+      });
+      if (postActionResult.counted_as_greet && postActionResult.action_clicked) {
+        greetCount += 1;
+      }
+      addTiming(timings, "post_action_ms", Date.now() - postActionStarted);
+    }
+    if (detailResult && closeDetail) {
+      detailResult.close_result = await measureTiming(timings, "close_detail_ms", () => closeRecommendDetail(client));
+      if (!detailResult.close_result?.closed) {
+        const closeError = createRecommendCloseFailureError(detailResult.close_result);
+        const recovery = await recoverAndReapplyRecommendContext("detail_close_failed", closeError, {
+          forceRecentNotView: true
+        });
+        detailResult.cv_acquisition = {
+          ...(detailResult.cv_acquisition || {}),
+          close_recovery: {
+            ok: Boolean(recovery.ok),
+            method: recovery.method || "",
+            forced_recent_not_view: Boolean(recovery.forced_recent_not_view),
+            card_count: recovery.card_count || 0
+          }
+        };
+      }
+    }
+    timings.total_ms = Date.now() - candidateStarted;
+    const compactResult = {
+      index,
+      candidate_key: candidateKey,
+      card_node_id: cardNodeId,
+      candidate: compactCandidate(screeningCandidate),
+      detail: compactDetail(detailResult),
+      llm_screening: detailResult ? null : compactScreeningLlmResult(llmResult),
+      screening: compactScreening(screening),
+      action_discovery: compactActionDiscovery(actionDiscovery),
+      post_action: postActionResult,
+      error: recoverableDetailError
+        ? compactRecoverableDetailError(recoverableDetailError)
+        : detailResult?.image_evidence?.ok === false
+        ? compactError({
+          code: detailResult.image_evidence.error_code,
+          message: detailResult.image_evidence.error
+        }, "IMAGE_CAPTURE_FAILED")
+        : null,
+      timings
+    };
+    results.push(compactResult);
+    markInfiniteListCandidateProcessed(listState, candidateKey, {
+      metadata: {
+        result_index: index,
+        candidate_id: screeningCandidate.id || null
+      }
+    });
+
+    updateRecommendProgress({
+      last_candidate_id: screeningCandidate.id || null,
+      last_candidate_key: candidateKey,
+      last_score: screening.score
+    });
+    const checkpointStarted = Date.now();
+    runControl.checkpoint({
+      results,
+      last_candidate: {
+        id: screeningCandidate.id || null,
+        key: candidateKey,
+        identity: screeningCandidate.identity || {},
+        screening: {
+          status: screening.status,
+          passed: screening.passed,
+          score: screening.score
+        },
+        llm_screening: compactScreeningLlmResult(llmResult),
+        error: compactResult.error,
+        post_action: postActionResult
+      }
+    });
+    addTiming(compactResult.timings, "checkpoint_save_ms", Date.now() - checkpointStarted);
+
+    if (postActionResult?.stop_run) {
+      listEndReason = postActionResult.reason || "post_action_stop";
+      break;
+    }
+
+    if (delayMs > 0) {
+      const sleepStarted = Date.now();
+      await runControl.sleep(delayMs);
+      addTiming(compactResult.timings, "sleep_ms", Date.now() - sleepStarted);
+      compactResult.timings.total_ms = Date.now() - candidateStarted;
+    }
+  }
+
+  runControl.setPhase("recommend:done");
+  return {
+    domain: "recommend",
+    target_url: targetUrl,
+    job_selection: compactJobSelection(jobSelection),
+    page_scope: compactPageScopeSelection(pageScopeSelection),
+    filter: compactFilterResult(filterResult),
+    card_count: cardNodeIds.length,
+    candidate_list: compactInfiniteListState(listState),
+    viewport_health: {
+      stats: viewportGuard.getStats(),
+      events: viewportGuard.getEvents()
+    },
+    list_end_reason: listEndReason || null,
+    refresh_rounds: refreshRounds,
+    refresh_attempts: refreshAttempts,
+    context_recoveries: contextRecoveryAttempts,
+    ...countRecommendResultStatuses(results, { greetCount }),
+    results
+  };
+}
+
+export function createRecommendRunService({
+  lifecycle,
+  idPrefix = "recommend",
+  workflow = runRecommendWorkflow,
+  onSnapshot = null
+} = {}) {
+  const manager = lifecycle || createRunLifecycleManager({ idPrefix, onSnapshot });
+
+  function startRecommendRun({
+    runId = "",
+    pid = process.pid,
+    client,
+    targetUrl = "",
+    criteria = "",
+    jobLabel = "",
+    pageScope = "recommend",
+    fallbackPageScope = "recommend",
+    filter = {},
+    maxCandidates = 5,
+    detailLimit,
+    closeDetail = true,
+    delayMs = 0,
+    cardTimeoutMs = 10000,
+    maxImagePages = DEFAULT_MAX_IMAGE_PAGES,
+    imageWheelDeltaY = 650,
+    cvAcquisitionMode = "unknown",
+    listMaxScrolls = 20,
+    listStableSignatureLimit = 5,
+    listWheelDeltaY = 850,
+    listSettleMs = 2200,
+    listFallbackPoint = null,
+    refreshOnEnd = true,
+    maxRefreshRounds = 2,
+    refreshButtonSettleMs = 8000,
+    refreshReloadSettleMs = 8000,
+    postAction = "none",
+    maxGreetCount = null,
+    executePostAction = true,
+    actionTimeoutMs = 8000,
+    actionIntervalMs = 500,
+    actionAfterClickDelayMs = 900,
+    screeningMode = "llm",
+    llmConfig = null,
+    llmTimeoutMs = 120000,
+    llmImageLimit = 8,
+    llmImageDetail = "high",
+    imageOutputDir = "",
+    name = "recommend-domain-run"
+  } = {}) {
+    if (!client) throw new Error("startRecommendRun requires a guarded CDP client");
+    const normalizedFilter = normalizeFilter(filter);
+    const normalizedPostAction = normalizeRecommendPostAction(postAction) || "none";
+    const requestedPageScope = normalizeRecommendPageScope(pageScope) || "recommend";
+    const normalizedFallbackPageScope = normalizeRecommendPageScope(fallbackPageScope) || "recommend";
+    const normalizedScreeningMode = normalizeScreeningMode(screeningMode);
+    const candidateLimit = Math.max(1, Number(maxCandidates) || 1);
+    const normalizedDetailLimit = detailLimit == null ? null : Math.max(0, Number(detailLimit) || 0);
+    return manager.startRun({
+      runId,
+      name,
+      pid,
+      context: {
+        domain: "recommend",
+        target_url: targetUrl,
+        criteria_present: Boolean(criteria),
+        job_label: jobLabel || "",
+        requested_page_scope: requestedPageScope,
+        fallback_page_scope: normalizedFallbackPageScope,
+        filter: normalizedFilter,
+        max_candidates: maxCandidates,
+        max_candidates_semantics: "passed_candidates",
+        detail_limit: normalizedDetailLimit,
+        close_detail: closeDetail,
+        cv_acquisition_mode: cvAcquisitionMode,
+        max_image_pages: maxImagePages,
+        image_wheel_delta_y: imageWheelDeltaY,
+        list_max_scrolls: listMaxScrolls,
+        list_stable_signature_limit: listStableSignatureLimit,
+        list_wheel_delta_y: listWheelDeltaY,
+        list_settle_ms: listSettleMs,
+        list_fallback_point: listFallbackPoint,
+        refresh_on_end: refreshOnEnd,
+        max_refresh_rounds: maxRefreshRounds,
+        refresh_button_settle_ms: refreshButtonSettleMs,
+        refresh_reload_settle_ms: refreshReloadSettleMs,
+        post_action: normalizedPostAction,
+        max_greet_count: Number.isInteger(maxGreetCount) ? maxGreetCount : null,
+        execute_post_action: Boolean(executePostAction),
+        action_timeout_ms: actionTimeoutMs,
+        screening_mode: normalizedScreeningMode,
+        llm_configured: Boolean(llmConfig),
+        llm_timeout_ms: llmTimeoutMs,
+        llm_image_limit: llmImageLimit,
+        llm_image_detail: llmImageDetail,
+        image_output_dir: imageOutputDir || ""
+      },
+      progress: {
+        card_count: 0,
+        target_count: candidateLimit,
+        target_count_semantics: "passed_candidates",
+        processed: 0,
+        screened: 0,
+        detail_opened: 0,
+        llm_screened: 0,
+        passed: 0,
+        greet_count: 0,
+        post_action_clicked: 0,
+        image_capture_failed: 0,
+        detail_open_failed: 0,
+        transient_recovered: 0,
+        context_recoveries: 0
+      },
+      checkpoint: {},
+      task: (runControl) => workflow({
+        client,
+        targetUrl,
+        criteria,
+        jobLabel,
+        pageScope: requestedPageScope,
+        fallbackPageScope: normalizedFallbackPageScope,
+        filter: normalizedFilter,
+        maxCandidates,
+        detailLimit: normalizedDetailLimit,
+        closeDetail,
+        delayMs,
+        cardTimeoutMs,
+        maxImagePages,
+        imageWheelDeltaY,
+        cvAcquisitionMode,
+        listMaxScrolls,
+        listStableSignatureLimit,
+        listWheelDeltaY,
+        listSettleMs,
+        listFallbackPoint,
+        refreshOnEnd,
+        maxRefreshRounds,
+        refreshButtonSettleMs,
+        refreshReloadSettleMs,
+        postAction: normalizedPostAction,
+        maxGreetCount,
+        executePostAction,
+        actionTimeoutMs,
+        actionIntervalMs,
+        actionAfterClickDelayMs,
+        screeningMode: normalizedScreeningMode,
+        llmConfig,
+        llmTimeoutMs,
+        llmImageLimit,
+        llmImageDetail,
+        imageOutputDir
+      }, runControl)
+    });
+  }
+
+  return {
+    startRecommendRun,
+    getRecommendRun: manager.getRun,
+    pauseRecommendRun: manager.pauseRun,
+    resumeRecommendRun: manager.resumeRun,
+    cancelRecommendRun: manager.cancelRun,
+    waitForRecommendRun: manager.waitForRun,
+    listRecommendRuns: manager.listRuns,
+    manager
+  };
+}