fix agent video routing failures

lib/ai/auto-agent-graph.ts

@@ -9,6 +9,14 @@ import {
   type AgentProjectToolDefinition,
   type AgentProjectToolName,
 } from "@/lib/ai/tools-registry";
+import {
+  runHumanConfirmationGate,
+  type HumanConfirmationGateDecision,
+} from "@/lib/ai/runtime/middlewares/human-confirmation-gate";
+import {
+  normalizeAgentThreadState,
+  type AgentThreadState,
+} from "@/lib/ai/runtime/agent-thread-state";
 import {
   getAiModelAgentCodingModelId,
   isAiModelMultimodal,
@@ -140,6 +148,7 @@ export interface AutoAgentInput {
   historyMessages?: HistoryMessage[];
   context?: string;
   preferredModelId?: string;
+  threadState?: AgentThreadState | null;
 }
 
 export interface AutoAgentResult {
@@ -151,6 +160,8 @@ export interface AutoAgentResult {
   recommendedImageParams?: AgentRecommendedImageParams;
   videoPlan?: AgentVideoPlan;
   route?: RouteResult;
+  threadState?: AgentThreadState | null;
+  workflowDecision?: HumanConfirmationGateDecision;
 }
 
 const AutoAgentState = Annotation.Root({
@@ -168,6 +179,7 @@ const AutoAgentState = Annotation.Root({
   historyMessages: Annotation<HistoryMessage[]>,
   context: Annotation<string | null>,
   preferredModelId: Annotation<string | null>,
+  threadState: Annotation<AgentThreadState | null>,
   analysis: Annotation<AnalysisResult | null>,
   availableTools: Annotation<AgentProjectToolDefinition[]>,
   plan: Annotation<AgentExecutionPlan | null>,
@@ -177,6 +189,7 @@ const AutoAgentState = Annotation.Root({
   recommendedImageParams: Annotation<AgentRecommendedImageParams | null>,
   videoPlan: Annotation<AgentVideoPlan | null>,
   route: Annotation<RouteResult | null>,
+  workflowDecision: Annotation<HumanConfirmationGateDecision | null>,
 });
 
 type AutoAgentStateType = typeof AutoAgentState.State;
@@ -425,6 +438,53 @@ function prepareExecutionFromPlan(params: {
   };
 }
 
+function buildWorkflowGateBypassPayload(
+  state: AutoAgentStateType,
+  decision: HumanConfirmationGateDecision,
+) {
+  const analysis: AnalysisResult = {
+    intent: "chat",
+    confidence: 0.98,
+    needsClarification: decision.action === "awaiting_confirmation",
+    clarifyingQuestions:
+      decision.action === "awaiting_confirmation"
+        ? ["请先确认或修改当前待确认流程，我再继续执行。"]
+        : [],
+    conversationalReply: decision.friendlyMessage,
+  };
+
+  const plan: AgentExecutionPlan = {
+    goal: createPlanGoalSummary(state.text, state.hasImage),
+    nextAction: "respond",
+    intent: "chat",
+    targetSource: state.targetResolutionMode,
+    preferredCapability: "respond",
+    responseMode: "direct_answer",
+    shouldUseHistoryImage: false,
+    shouldUseStructuredReply: false,
+    confidence: 0.98,
+    toolCalls: [
+      {
+        name: "respondDirect",
+        reason: "当前工作流处于人工确认门控阶段，先处理用户确认或修改意图。",
+      },
+    ],
+  };
+
+  const execution = prepareExecutionFromPlan({
+    plan,
+    availableTools: [],
+  });
+
+  return {
+    analysis,
+    plan,
+    execution,
+    friendlyMessage: decision.friendlyMessage || null,
+    workflowDecision: decision,
+  };
+}
+
 function getEnvInt(name: string, fallback: number, min: number, max: number): number {
   const raw = process.env[name];
   const parsed = raw ? Number.parseInt(raw, 10) : Number.NaN;
@@ -1789,6 +1849,25 @@ const analyzeNode = async (state: AutoAgentStateType) => {
   return { analysis };
 };
 
+const confirmationGateNode = async (state: AutoAgentStateType) => {
+  const workflowDecision = runHumanConfirmationGate({
+    text: state.text,
+    threadState: state.threadState,
+  });
+
+  if (!workflowDecision.shouldBypassPlanner) {
+    return {
+      workflowDecision,
+    };
+  }
+
+  return buildWorkflowGateBypassPayload(state, workflowDecision);
+};
+
+const shouldBypassPlanner = (state: AutoAgentStateType) => {
+  return state.workflowDecision?.shouldBypassPlanner ? "route" : "analyze";
+};
+
 const planNode = async (state: AutoAgentStateType) => {
   if (!state.analysis) return {};
   const baseAvailableTools = await getEnabledAgentToolDefinitionsForContext({
@@ -1960,14 +2039,19 @@ let compiledGraph: ReturnType<typeof buildGraph> | null = null;
 
 function buildGraph() {
   return new StateGraph(AutoAgentState)
+    .addNode("confirmationGate", confirmationGateNode)
     .addNode("analyze", analyzeNode)
-    .addNode("plan", planNode)
+    .addNode("planner", planNode)
     .addNode("prepareExecution", prepareExecutionNode)
     .addNode("optimize", optimizeNode)
     .addNode("routeModel", routeNode)
-    .addEdge(START, "analyze")
-    .addEdge("analyze", "plan")
-    .addEdge("plan", "prepareExecution")
+    .addEdge(START, "confirmationGate")
+    .addConditionalEdges("confirmationGate", shouldBypassPlanner, {
+      analyze: "analyze",
+      route: "routeModel",
+    })
+    .addEdge("analyze", "planner")
+    .addEdge("planner", "prepareExecution")
     .addConditionalEdges("prepareExecution", shouldOptimize, {
       optimize: "optimize",
       route: "routeModel",
@@ -2001,6 +2085,7 @@ export async function runAutoAgentGraph(params: AutoAgentInput): Promise<AutoAge
     historyMessages: params.historyMessages ?? [],
     context: params.context ?? null,
     preferredModelId: params.preferredModelId ?? null,
+    threadState: normalizeAgentThreadState(params.threadState),
     analysis: null,
     availableTools: [],
     plan: null,
@@ -2010,6 +2095,7 @@ export async function runAutoAgentGraph(params: AutoAgentInput): Promise<AutoAge
     recommendedImageParams: null,
     videoPlan: null,
     route: null,
+    workflowDecision: null,
   });
 
   return {
@@ -2021,5 +2107,7 @@ export async function runAutoAgentGraph(params: AutoAgentInput): Promise<AutoAge
     recommendedImageParams: result.recommendedImageParams ?? undefined,
     videoPlan: result.videoPlan ?? undefined,
     route: result.route ?? undefined,
+    threadState: result.threadState ?? undefined,
+    workflowDecision: result.workflowDecision ?? undefined,
   };
 }

lib/ai/tools/generate-video.ts

@@ -74,7 +74,35 @@ async function selectBestSubModel(
   const subModels = await getSubModelsByParentModelId(parentModelId);
 
   if (subModels.length === 0) {
-    throw new Error("没有可用的视频生成子模型");
+    const allModels = await getEnabledAiModelsWithProvider();
+    const directModel = allModels.find(
+      (item) =>
+        item.model.id === parentModelId &&
+        (item.provider.type === "volcano-video" || item.provider.type === "doubao-video"),
+    );
+
+    if (!directModel) {
+      throw new Error("没有可用的视频生成子模型");
+    }
+
+    const directCapabilities = directModel.model.capabilities as VideoModelCapabilities | undefined;
+    const supportedModes = getSupportedModes(directCapabilities, directModel.model.modelId);
+    if (!supportedModes.includes(mode)) {
+      throw new Error(`模型 ${directModel.model.displayName} 不支持 ${mode} 模式`);
+    }
+
+    return {
+      model: directModel,
+      capabilities:
+        directCapabilities ||
+        ({
+          supportedModes,
+          supportedResolutions: ["480p", "720p", "1080p"],
+          supportedRatios: ["16:9", "4:3", "1:1", "3:4", "9:16", "21:9", "adaptive"],
+          durationRange: { min: 2, max: 12 },
+          maxReferenceImages: 4,
+        } as VideoModelCapabilities),
+    };
   }
 
   console.log(`[Smart Model Selection] 找到 ${subModels.length} 个子模型`);
@@ -263,7 +291,7 @@ async function resolvePreferredVideoParentModelId(preferredModelRef?: string) {
   const models = await getEnabledAiModelsWithProvider();
   const matched = models.find(
     (item) =>
-      item.provider.type === "volcano-video" &&
+      (item.provider.type === "volcano-video" || item.provider.type === "doubao-video") &&
       (item.model.id === normalizedRef || item.model.modelId === normalizedRef),
   );
 
@@ -289,12 +317,14 @@ export async function executeDirectVideoGeneration(
   const fallbackParentModel =
     allModels.find(
       (item) =>
-        item.provider.type === "volcano-video" &&
+        (item.provider.type === "volcano-video" || item.provider.type === "doubao-video") &&
         !item.model.isSubModel &&
         item.model.modelId === "volcano-video-master",
     ) ||
     allModels.find(
-      (item) => item.provider.type === "volcano-video" && !item.model.isSubModel,
+      (item) =>
+        (item.provider.type === "volcano-video" || item.provider.type === "doubao-video") &&
+        !item.model.isSubModel,
     );
 
   const parentModelId = preferredParentModelId || fallbackParentModel?.model.id;