ai-chatbot/lib/ai/runtime/intent-arbiter.ts

import "server-only";

import type { ConversationActionResolution } from "@/lib/ai/conversation-action-resolver";
import type { ResolvedTargetAssets } from "@/lib/ai/context-memory/types";
import type { AnalysisResult } from "@/lib/ai/smart-analyzer";
import type { TaskBoundaryDecision } from "@/lib/ai/runtime/task-boundary-classifier";
import { hasDirectAgentLocalEditIntent } from "@/lib/ai/runtime/local-edit-options";
import {
  hasCanvasAgentActionableImageEditIntent,
  inferCanvasAgentImageEditCapability,
} from "@/lib/ai/runtime/canvas-agent-kernel";

export type IntentArbiterMode =
  | "respond"
  | "clarify"
  | "analyze_image"
  | "generate_image"
  | "edit_image"
  | "generate_video";

export type IntentArbiterPreferredCapability =
  | "respond"
  | "analyze"
  | "generic_edit"
  | "remove_background"
  | "remove_watermark"
  | "enhance"
  | "extend"
  | "resize"
  | "angle_variant"
  | "local_edit"
  | "box_cutout"
  | "generate_image"
  | "generate_video"
  | "publish_canvas";

export interface IntentArbiterExecution {
  mode: IntentArbiterMode;
  routeIntent: AnalysisResult["intent"];
  routeSubIntent?: string;
  targetSource: ResolvedTargetAssets["resolutionMode"];
  preferredCapability: IntentArbiterPreferredCapability;
  shouldUseStructuredReply: boolean;
  shouldOptimizePrompt: boolean;
  shouldApplyImagePipeline: boolean;
  shouldApplyVideoPipeline: boolean;
  shouldStayConversational: boolean;
  shouldPublishToCanvas: boolean;
  toolCalls: Array<{ name: string; reason?: string; args?: Record<string, unknown> }>;
  availableTools: string[];
}

export interface IntentArbiterResult {
  analysis: AnalysisResult;
  execution: IntentArbiterExecution;
  overridden: boolean;
  reason: string;
  friendlyMessage?: string;
}

function normalizeCompactText(text: string): string {
  return String(text || "").toLowerCase().replace(/\s+/g, "").trim();
}

function isLikelyAdvisoryOrPlanningRequest(text: string): boolean {
  const normalizedText = String(text || "").toLowerCase();
  return /怎么|如何|为什么|是否|可不可以|能不能|行不行|建议|思路|方案|流程|步骤|教程|推荐|对比|区别|差别|优缺点|评估|分析|判断|解释|讲讲|科普|适合吗|可行吗|值不值得|要不要|帮我看看|帮我分析|帮我判断/.test(
    normalizedText,
  );
}

function isExplicitVideoGenerationRequest(text: string): boolean {
  const normalizedText = normalizeCompactText(text);
  if (!normalizedText) return false;
  const hasStrongVideoMarkers =
    /图生视频|文生视频|首帧|尾帧|分镜|运镜|视频生成/.test(normalizedText);
  if (hasStrongVideoMarkers) return true;
  if (isLikelyAdvisoryOrPlanningRequest(text)) return false;
  return (
    /让.+动起来/.test(normalizedText) ||
    (/(生成|制作|做|创建|输出|渲染|合成|出|做成|变成|转成)/.test(normalizedText) &&
      /(视频|动画|短片|片子|动起来|会动)/.test(normalizedText))
  );
}

function isExplicitImageGenerationRequest(text: string): boolean {
  const normalizedText = normalizeCompactText(text);
  if (!normalizedText || isLikelyAdvisoryOrPlanningRequest(text)) return false;
  if (/不要生图|先别生图|不需要生图|不要生成|别生成|只要提示词|仅给提示词|先给提示词/.test(normalizedText)) {
    return false;
  }

  const countToken = "(?:一|两|二|三|四|五|六|七|八|九|十|[1-9]\\d?)";
  const visualTarget =
    "(?:图|图片|海报|头像|插画|场景|主图|产品图|商品图|电商图|壁纸|封面|logo|图标)";
  const createVisualTarget = new RegExp(
    `(?:生成|创作|创建|制作|设计|做|出|画(?!布)|绘制)${countToken}?张?.{0,32}${visualTarget}`,
  );
  const drawSubject = new RegExp(
    `(?:画(?!布)|绘制|生成|创作|创建)${countToken}?(?:只|个|位|幅|张)?.{0,24}(?:小猫|猫咪|猫|狗|人物|角色|ip|头像|logo|图标|插画)`,
  );
  const continueVariants = new RegExp(
    `(?:再来|再出|再做|再给我|补|多来|继续生成|继续出)${countToken}?张`,
  );

  return (
    /重新生成|生成新图|新建图片|创作一张|生成一张|画一张|给我画|帮我画|出一张|做一张图|做一张海报|创建一张图片|生成一只|生成一个|生成一幅|生成一版|生成一套|生成一组|生图|出图/.test(
      normalizedText,
    ) ||
    createVisualTarget.test(normalizedText) ||
    drawSubject.test(normalizedText) ||
    continueVariants.test(normalizedText) ||
    /参考.*生成|按这个风格生成|以这张图为灵感生成|仿照.*生成|同风格再来|同样风格再来/.test(
      normalizedText,
    ) ||
    /帮我生成.*(?:图|图片|海报|头像|插画|场景|小猫|猫咪|人物|产品图)|给我生成.*(?:图|图片|海报|头像|插画|场景|小猫|猫咪|人物|产品图)/.test(
      normalizedText,
    )
  );
}

function isLikelyImageEditRequest(text: string): boolean {
  return hasDirectAgentLocalEditIntent(text) || hasCanvasAgentActionableImageEditIntent(text);
}

function inferDefaultImageEditSubIntent(text: string): string {
  const inferredCapability = inferCanvasAgentImageEditCapability(text);
  if (
    inferredCapability &&
    inferredCapability !== "generic_edit" &&
    inferredCapability !== "generate_image" &&
    inferredCapability !== "generate_video" &&
    inferredCapability !== "analyze"
  ) {
    return inferredCapability;
  }
  if (hasDirectAgentLocalEditIntent(text)) return "local_edit";
  return "enhance";
}

function mapIntentToMode(
  intent: AnalysisResult["intent"],
  needsClarification?: boolean,
): IntentArbiterMode {
  if (needsClarification) return "clarify";
  switch (intent) {
    case "image_edit":
      return "edit_image";
    case "image_gen":
      return "generate_image";
    case "video_gen":
      return "generate_video";
    case "vision":
      return "analyze_image";
    case "chat":
    default:
      return "respond";
  }
}

function mapPreferredCapability(
  intent: AnalysisResult["intent"],
  subIntent?: string,
): IntentArbiterPreferredCapability {
  if (intent === "video_gen") return "generate_video";
  if (intent === "image_gen") return "generate_image";
  if (intent === "vision") return "analyze";
  if (intent === "chat") return "respond";

  switch (subIntent) {
    case "remove_background":
      return "remove_background";
    case "remove_watermark":
      return "remove_watermark";
    case "extend":
      return "extend";
    case "resize":
      return "resize";
    case "angle_variant":
    case "generate_angle_variant":
      return "angle_variant";
    case "local_edit":
    case "inpaint_region":
      return "local_edit";
    case "box_cutout":
      return "box_cutout";
    case "enhance":
      return "enhance";
    default:
      return "generic_edit";
  }
}

function buildExecutionFromAnalysis(params: {
  analysis: AnalysisResult;
  execution?: IntentArbiterExecution | null;
  targetSource: ResolvedTargetAssets["resolutionMode"];
}): IntentArbiterExecution {
  const mode = mapIntentToMode(params.analysis.intent, params.analysis.needsClarification);
  const shouldKeepPlannedToolCalls =
    params.execution?.mode === mode &&
    params.execution?.routeIntent === params.analysis.intent;

  return {
    mode,
    routeIntent: params.analysis.intent,
    routeSubIntent: params.analysis.subIntent,
    targetSource: params.targetSource,
    preferredCapability: mapPreferredCapability(params.analysis.intent, params.analysis.subIntent),
    shouldUseStructuredReply:
      params.analysis.intent === "image_edit" ||
      params.analysis.intent === "image_gen" ||
      params.analysis.intent === "video_gen",
    shouldOptimizePrompt:
      params.analysis.intent === "image_edit" ||
      params.analysis.intent === "image_gen" ||
      params.analysis.intent === "video_gen",
    shouldApplyImagePipeline:
      params.analysis.intent === "image_edit" || params.analysis.intent === "image_gen",
    shouldApplyVideoPipeline: params.analysis.intent === "video_gen",
    shouldStayConversational:
      mode === "respond" || mode === "analyze_image" || mode === "clarify",
    shouldPublishToCanvas: params.execution?.shouldPublishToCanvas || false,
    toolCalls: shouldKeepPlannedToolCalls ? params.execution?.toolCalls || [] : [],
    availableTools: params.execution?.availableTools || [],
  };
}

function buildDefaultFriendlyMessage(
  intent: AnalysisResult["intent"],
  params: { hasImage: boolean; text: string },
) {
  switch (intent) {
    case "image_gen":
      return params.hasImage ? "收到，我会结合你提供的参考图直接执行生图。" : "好的，我来直接帮你生成这张图。";
    case "image_edit":
      return params.hasImage ? "收到，我会按你的目标直接处理这张图。" : "收到，我来继续执行这轮图片处理。";
    case "video_gen":
      return params.hasImage ? "收到，我会结合这轮图片先规划并执行视频生成。" : "收到，我先按你的需求规划并执行视频生成。";
    case "vision":
      return params.hasImage ? "收到，我先分析这张图。" : "收到，我先按你的问题做分析。";
    case "chat":
    default:
      return undefined;
  }
}

function resolvePolicyIntent(params: {
  text: string;
  hasImage: boolean;
  hasHistoryImage: boolean;
  taskBoundaryDecision: TaskBoundaryDecision;
  actionResolution: ConversationActionResolution;
  analysis: AnalysisResult;
}): { intent: AnalysisResult["intent"]; subIntent?: string; reason: string } {
  const explicitImageGen = isExplicitImageGenerationRequest(params.text);
  const explicitVideoGen = isExplicitVideoGenerationRequest(params.text);
  const likelyEdit = isLikelyImageEditRequest(params.text);
  const hasUsableImage = params.hasImage || params.hasHistoryImage;

  if (explicitVideoGen) {
    return {
      intent: "video_gen",
      reason: "explicit_video_generation_request",
    };
  }

  if (hasUsableImage && likelyEdit) {
    return {
      intent: "image_edit",
      subIntent: params.analysis.subIntent || inferDefaultImageEditSubIntent(params.text),
      reason: "image_edit_policy_override",
    };
  }

  switch (params.actionResolution.kind) {
    case "image_edit_followup":
      return {
        intent: "image_edit",
        subIntent: params.analysis.subIntent || inferDefaultImageEditSubIntent(params.text),
        reason: "conversation_action_image_edit_followup",
      };
    case "image_generation_followup":
    case "prompt_carry_forward":
      return {
        intent: "image_gen",
        reason: "conversation_action_image_generation_followup",
      };
    default:
      break;
  }

  switch (params.taskBoundaryDecision.taskType) {
    case "prompt_only":
    case "analysis_only":
      return {
        intent: "vision",
        reason: `task_boundary_${params.taskBoundaryDecision.taskType}`,
      };
    case "execution_video":
      return {
        intent: "video_gen",
        reason: "task_boundary_execution_video",
      };
    case "execution_image":
      return {
        intent: "image_gen",
        reason: "task_boundary_execution_image_generate",
      };
    case "workflow_followup":
      if (hasUsableImage && likelyEdit) {
        return {
          intent: "image_edit",
          subIntent: params.analysis.subIntent || inferDefaultImageEditSubIntent(params.text),
          reason: "task_boundary_workflow_followup_edit",
        };
      }
      if (explicitImageGen) {
        return {
          intent: "image_gen",
          reason: "task_boundary_workflow_followup_generate",
        };
      }
      break;
    default:
      break;
  }

  if (explicitImageGen) {
    return {
      intent: "image_gen",
      reason: "explicit_image_generation_request",
    };
  }

  return {
    intent: params.analysis.intent,
    subIntent:
      params.analysis.intent === "image_edit"
        ? params.analysis.subIntent || inferDefaultImageEditSubIntent(params.text)
        : params.analysis.subIntent,
    reason: "keep_ai_analysis",
  };
}

export function arbitrateAgentIntent(params: {
  text: string;
  hasImage: boolean;
  hasHistoryImage: boolean;
  actionResolution: ConversationActionResolution;
  taskBoundaryDecision: TaskBoundaryDecision;
  analysis: AnalysisResult;
  execution?: IntentArbiterExecution | null;
  targetSource: ResolvedTargetAssets["resolutionMode"];
}): IntentArbiterResult {
  const resolved = resolvePolicyIntent(params);
  const nextAnalysis: AnalysisResult = {
    ...params.analysis,
    intent: resolved.intent,
    subIntent:
      resolved.intent === "image_edit"
        ? resolved.subIntent || inferDefaultImageEditSubIntent(params.text)
        : undefined,
    confidence:
      resolved.reason === "keep_ai_analysis"
        ? params.analysis.confidence
        : Math.max(params.analysis.confidence, 0.84),
    needsClarification:
      resolved.reason === "keep_ai_analysis"
        ? Boolean(params.analysis.needsClarification)
        : false,
    clarifyingQuestions:
      resolved.reason === "keep_ai_analysis" && params.analysis.needsClarification
        ? params.analysis.clarifyingQuestions || []
        : [],
  };
  const nextExecution = buildExecutionFromAnalysis({
    analysis: nextAnalysis,
    execution: params.execution,
    targetSource: params.targetSource,
  });
  const overridden =
    nextAnalysis.intent !== params.analysis.intent ||
    nextAnalysis.subIntent !== params.analysis.subIntent ||
    nextExecution.mode !== params.execution?.mode ||
    nextExecution.routeIntent !== params.execution?.routeIntent;
  const friendlyMessage =
    overridden || !params.analysis.conversationalReply?.trim()
      ? buildDefaultFriendlyMessage(nextAnalysis.intent, {
          hasImage: params.hasImage || params.hasHistoryImage,
          text: params.text,
        })
      : undefined;

  return {
    analysis: friendlyMessage
      ? {
          ...nextAnalysis,
          conversationalReply: friendlyMessage,
        }
      : nextAnalysis,
    execution: nextExecution,
    overridden,
    reason: resolved.reason,
    friendlyMessage,
  };
}