nolo-cli 0.1.30 → 0.1.32

package/ai/policy/runtimePolicy.ts

@@ -1,440 +0,0 @@
-import type { Agent } from "../../app/types";
-import type {
-  AgentBasePolicy,
-  DialogPolicyState,
-  KnowledgeCaptureLevel,
-  SelfEvolutionMode,
-  SpaceContextLevel,
-  TonePreset,
-  ToneResolutionMode,
-  UserPreferenceProfile,
-} from "./types";
-import {
-  DEFAULT_AGENT_BASE_POLICY,
-  DEFAULT_USER_PREFERENCE_PROFILE,
-} from "./types";
-
-export type CaptureIntent = "none" | "possible" | "strong";
-export type SpaceNeed = "none" | "light" | "deep";
-export type KnowledgeCaptureMode = "blocked" | "ask" | "suggest" | "auto";
-
-export interface ResolvedRuntimePolicy {
-  tonePreset: TonePreset;
-  toneResolutionMode: ToneResolutionMode;
-  knowledgeCapture: {
-    level: KnowledgeCaptureLevel;
-    mode: KnowledgeCaptureMode;
-    explicitRequest: boolean;
-    maxAutoCreatesPerDialog: number;
-    currentAutoCreates: number;
-    captureIntent: CaptureIntent;
-  };
-  spaceContext: {
-    level: SpaceContextLevel;
-    explicitRequest: boolean;
-    need: SpaceNeed;
-    maxReadCallsPerTurn: number;
-    maxReadCallsPerDialog: number;
-    currentAutoReads: number;
-    preloadSummaryCount: number;
-    preloadBudgetRatio: number;
-  };
-  selfEvolutionMode: SelfEvolutionMode;
-}
-
-const clampLevel = <T extends 1 | 2 | 3 | 4>(
-  value: unknown,
-  fallback: T,
-): T => {
-  const n = Number(value);
-  if (n === 1 || n === 2 || n === 3 || n === 4) {
-    return n as T;
-  }
-  return fallback;
-};
-
-const normalizeTonePreset = (value: unknown, fallback: TonePreset): TonePreset => {
-  switch (value) {
-    case "professional":
-    case "friendly":
-    case "direct":
-    case "pragmatic":
-    case "default":
-      return value;
-    default:
-      return fallback;
-  }
-};
-
-const normalizeToneResolutionMode = (
-  value: unknown,
-  fallback: ToneResolutionMode,
-): ToneResolutionMode => {
-  switch (value) {
-    case "agent_first":
-    case "user_first":
-    case "blend":
-      return value;
-    default:
-      return fallback;
-  }
-};
-
-const normalizeSelfEvolutionMode = (
-  value: unknown,
-  fallback: SelfEvolutionMode,
-): SelfEvolutionMode => {
-  switch (value) {
-    case "none":
-    case "knowledge_only":
-    case "prompt_and_refs":
-    case "full":
-      return value;
-    default:
-      return fallback;
-  }
-};
-
-export const resolveAgentBasePolicy = (agentConfig?: Agent | any): AgentBasePolicy => {
-  const raw = agentConfig?.basePolicy ?? {};
-
-  return {
-    version: 1,
-    tone: {
-      preset: normalizeTonePreset(
-        raw?.tone?.preset,
-        DEFAULT_AGENT_BASE_POLICY.tone?.preset ?? "default",
-      ),
-      resolutionMode: normalizeToneResolutionMode(
-        raw?.tone?.resolutionMode,
-        DEFAULT_AGENT_BASE_POLICY.tone?.resolutionMode ?? "blend",
-      ),
-    },
-    knowledgeCaptureMaxLevel: clampLevel(
-      raw?.knowledgeCaptureMaxLevel,
-      DEFAULT_AGENT_BASE_POLICY.knowledgeCaptureMaxLevel,
-    ),
-    spaceContextMaxLevel: clampLevel(
-      raw?.spaceContextMaxLevel,
-      DEFAULT_AGENT_BASE_POLICY.spaceContextMaxLevel,
-    ),
-    selfEvolutionMode: normalizeSelfEvolutionMode(
-      raw?.selfEvolutionMode,
-      DEFAULT_AGENT_BASE_POLICY.selfEvolutionMode,
-    ),
-  };
-};
-
-export const resolveUserPreferenceProfile = (
-  settingsRecord?: Record<string, any> | null,
-): UserPreferenceProfile => {
-  const fallbackSpaceLevel =
-    settingsRecord?.enableReadCurrentSpace === false
-      ? 1
-      : DEFAULT_USER_PREFERENCE_PROFILE.spaceContextLevel;
-
-  return {
-    version: 1,
-    tone: {
-      preset: normalizeTonePreset(
-        settingsRecord?.userTonePreset,
-        DEFAULT_USER_PREFERENCE_PROFILE.tone?.preset ?? "default",
-      ),
-    },
-    knowledgeCaptureLevel: clampLevel(
-      settingsRecord?.knowledgeCaptureLevel,
-      DEFAULT_USER_PREFERENCE_PROFILE.knowledgeCaptureLevel,
-    ),
-    spaceContextLevel: clampLevel(
-      settingsRecord?.spaceContextLevel,
-      fallbackSpaceLevel,
-    ),
-  };
-};
-
-const containsAny = (text: string, patterns: string[]): boolean =>
-  patterns.some((pattern) => text.includes(pattern));
-
-export const inferCaptureIntent = (userInput: string): CaptureIntent => {
-  const normalized = userInput.toLowerCase();
-
-  if (
-    containsAny(normalized, [
-      "保存",
-      "存一下",
-      "沉淀",
-      "记住",
-      "写入文档",
-      "建文档",
-      "创建文档",
-      "导入 skill",
-      "导入这个 skill",
-      "确认导入",
-      "确认导入这个 skill",
-      "导入这个技能",
-      "确认导入这个技能",
-      "创建 skill",
-      "创建这个 skill",
-      "skill 文档",
-      "创建表格",
-      "建表",
-      "建个表",
-      "做个表",
-      "做一张表",
-      "整理成表",
-      "整理成一张表",
-      "建立一个表",
-      "建立一张表",
-      "建立成表",
-      "建立成一张表",
-      "创建这个表格",
-      "确认创建",
-      "确认创建这张表格",
-      "create doc",
-      "import skill",
-      "confirm import",
-      "create skill",
-      "create table",
-      "save this",
-      "remember this",
-      "document this",
-    ])
-  ) {
-    return "strong";
-  }
-
-  if (
-    containsAny(normalized, [
-      "总结",
-      "方案",
-      "调研",
-      "规范",
-      "会议纪要",
-      "总结一下",
-      "research",
-      "proposal",
-      "spec",
-      "summary",
-    ])
-  ) {
-    return "possible";
-  }
-
-  return "none";
-};
-
-export const inferSpaceNeed = (userInput: string): SpaceNeed => {
-  const normalized = userInput.toLowerCase();
-
-  if (
-    containsAny(normalized, [
-      "当前空间",
-      "这个空间",
-      "这些文档",
-      "根据空间",
-      "workspace",
-      "current space",
-      "these docs",
-      "based on the docs",
-    ])
-  ) {
-    return "deep";
-  }
-
-  if (
-    containsAny(normalized, [
-      "文档",
-      "文件",
-      "页面",
-      "目录",
-      "page",
-      "pages",
-      "docs",
-      "files",
-      "目录结构",
-    ])
-  ) {
-    return "light";
-  }
-
-  return "none";
-};
-
-const resolveTonePreset = (
-  agentTone: TonePreset,
-  userTone: TonePreset,
-  mode: ToneResolutionMode,
-): TonePreset => {
-  if (mode === "agent_first") {
-    return agentTone === "default" ? userTone : agentTone;
-  }
-  if (mode === "user_first") {
-    return userTone === "default" ? agentTone : userTone;
-  }
-  if (userTone !== "default") return userTone;
-  return agentTone;
-};
-
-const maxReadCallsPerTurnByLevel = (
-  level: SpaceContextLevel,
-  need: SpaceNeed,
-): number => {
-  if (need === "deep") {
-    return { 1: 2, 2: 3, 3: 4, 4: 6 }[level];
-  }
-  if (need === "light") {
-    return { 1: 1, 2: 2, 3: 3, 4: 4 }[level];
-  }
-  return { 1: 0, 2: 0, 3: 1, 4: 2 }[level];
-};
-
-const maxReadCallsPerDialogByLevel = (
-  level: SpaceContextLevel,
-  need: SpaceNeed,
-): number => {
-  if (need === "deep") {
-    return { 1: 3, 2: 4, 3: 6, 4: 8 }[level];
-  }
-  if (need === "light") {
-    return { 1: 2, 2: 3, 3: 4, 4: 6 }[level];
-  }
-  return { 1: 0, 2: 0, 3: 2, 4: 4 }[level];
-};
-
-const preloadSummaryCountByLevel = (level: SpaceContextLevel): number =>
-  ({ 1: 0, 2: 0, 3: 8, 4: 16 }[level]);
-
-const preloadBudgetRatioByLevel = (level: SpaceContextLevel): number =>
-  ({ 1: 0, 2: 0.01, 3: 0.04, 4: 0.08 }[level]);
-
-export const resolveRuntimePolicy = (params: {
-  agentConfig?: Agent | any;
-  settingsRecord?: Record<string, any> | null;
-  userInput: string;
-  dialogPolicyState?: DialogPolicyState | null;
-}): ResolvedRuntimePolicy => {
-  const { agentConfig, settingsRecord, userInput, dialogPolicyState } = params;
-
-  const agentBasePolicy = resolveAgentBasePolicy(agentConfig);
-  const userPreferenceProfile = resolveUserPreferenceProfile(settingsRecord);
-
-  const knowledgeLevel = Math.min(
-    agentBasePolicy.knowledgeCaptureMaxLevel,
-    userPreferenceProfile.knowledgeCaptureLevel,
-  ) as KnowledgeCaptureLevel;
-  const spaceLevel = Math.min(
-    agentBasePolicy.spaceContextMaxLevel,
-    userPreferenceProfile.spaceContextLevel,
-  ) as SpaceContextLevel;
-
-  const captureIntent = inferCaptureIntent(userInput);
-  const spaceNeed = inferSpaceNeed(userInput);
-  const explicitCaptureRequest = captureIntent === "strong";
-  const explicitSpaceRequest = spaceNeed !== "none";
-
-  const currentAutoCreates = Math.max(
-    0,
-    Number(dialogPolicyState?.autoKnowledgeCaptureCount ?? 0),
-  );
-  const currentAutoReads = Math.max(
-    0,
-    Number(dialogPolicyState?.autoSpaceReadCount ?? 0),
-  );
-
-  let knowledgeMode: KnowledgeCaptureMode = "blocked";
-  if (explicitCaptureRequest) {
-    knowledgeMode = "auto";
-  } else if (knowledgeLevel === 2) {
-    knowledgeMode = "ask";
-  } else if (knowledgeLevel === 3) {
-    knowledgeMode = "suggest";
-  } else if (knowledgeLevel >= 4 && captureIntent !== "none") {
-    knowledgeMode = "auto";
-  }
-
-  return {
-    tonePreset: resolveTonePreset(
-      agentBasePolicy.tone?.preset ?? "default",
-      userPreferenceProfile.tone?.preset ?? "default",
-      agentBasePolicy.tone?.resolutionMode ?? "blend",
-    ),
-    toneResolutionMode: agentBasePolicy.tone?.resolutionMode ?? "blend",
-    knowledgeCapture: {
-      level: knowledgeLevel,
-      mode: knowledgeMode,
-      explicitRequest: explicitCaptureRequest,
-      maxAutoCreatesPerDialog: explicitCaptureRequest ? 99 : 1,
-      currentAutoCreates,
-      captureIntent,
-    },
-    spaceContext: {
-      level: spaceLevel,
-      explicitRequest: explicitSpaceRequest,
-      need: spaceNeed,
-      maxReadCallsPerTurn: maxReadCallsPerTurnByLevel(spaceLevel, spaceNeed),
-      maxReadCallsPerDialog: maxReadCallsPerDialogByLevel(spaceLevel, spaceNeed),
-      currentAutoReads,
-      preloadSummaryCount: preloadSummaryCountByLevel(spaceLevel),
-      preloadBudgetRatio: preloadBudgetRatioByLevel(spaceLevel),
-    },
-    selfEvolutionMode: agentBasePolicy.selfEvolutionMode,
-  };
-};
-
-export const buildUserPolicyContext = (
-  policy: ResolvedRuntimePolicy,
-): string => {
-  const lines = [
-    `用户语气偏好：${policy.tonePreset}`,
-    `知识沉淀级别：${policy.knowledgeCapture.level}（当前模式：${policy.knowledgeCapture.mode}）`,
-    `空间上下文级别：${policy.spaceContext.level}（本轮最多自动 read ${policy.spaceContext.maxReadCallsPerTurn} 次）`,
-    "如果工具返回 policy limit，不要重试同一路径，应改为解释限制、先给答案或向用户确认。",
-  ];
-
-  if (policy.knowledgeCapture.mode === "ask") {
-    lines.push("当前用户偏好要求：创建知识文档前先询问用户。");
-  } else if (policy.knowledgeCapture.mode === "suggest") {
-    lines.push("当前用户偏好要求：先完成回答，再建议是否沉淀成文档/表格。");
-  }
-
-  return lines.join("\n");
-};
-
-export const buildStaticUserPolicyContext = (params: {
-  agentConfig?: Agent | any;
-  settingsRecord?: Record<string, any> | null;
-}): string => {
-  const agentBasePolicy = resolveAgentBasePolicy(params.agentConfig);
-  const userPreferenceProfile = resolveUserPreferenceProfile(params.settingsRecord);
-  const tonePreset = resolveTonePreset(
-    agentBasePolicy.tone?.preset ?? "default",
-    userPreferenceProfile.tone?.preset ?? "default",
-    agentBasePolicy.tone?.resolutionMode ?? "blend",
-  );
-
-  return [
-    `用户语气偏好：${tonePreset}`,
-    `知识沉淀级别：${Math.min(agentBasePolicy.knowledgeCaptureMaxLevel, userPreferenceProfile.knowledgeCaptureLevel)}`,
-    `空间上下文级别：${Math.min(agentBasePolicy.spaceContextMaxLevel, userPreferenceProfile.spaceContextLevel)}`,
-    "请尽量保留 agent 自身角色设定，同时按用户偏好调整表达和自动化程度。",
-  ].join("\n");
-};
-
-export const resolveSpaceContextPreloadPlan = (level: SpaceContextLevel) => ({
-  preloadSummaryCount: preloadSummaryCountByLevel(level),
-  preloadBudgetRatio: preloadBudgetRatioByLevel(level),
-  includeRecentContent: level >= 3,
-});
-
-export const isBudgetedWorkspaceReadKey = (dbKey: string): boolean => {
-  const normalized = dbKey.toLowerCase();
-  return (
-    normalized.startsWith("page-") ||
-    normalized.startsWith("dialog-") ||
-    normalized.startsWith("space-") ||
-    normalized.startsWith("table-") ||
-    normalized.startsWith("table_row-") ||
-    normalized.startsWith("row-") ||
-    normalized.startsWith("meta-")
-  );
-};

package/ai/policy/selfUpdateFields.ts

@@ -1,48 +0,0 @@
-export const AGENT_UPDATE_FIELD_NAMES = [
-  "name",
-  "model",
-  "provider",
-  "prompt",
-  "introduction",
-  "greeting",
-  "isPublic",
-  "tags",
-  "tools",
-  "references",
-  "temperature",
-  "top_p",
-  "frequency_penalty",
-  "presence_penalty",
-  "max_tokens",
-  "reasoning_effort",
-] as const;
-
-export type AgentUpdateField = (typeof AGENT_UPDATE_FIELD_NAMES)[number];
-
-export const DEFAULT_AUTO_APPROVED_SELF_UPDATE_FIELDS: AgentUpdateField[] = [
-  "greeting",
-  "introduction",
-  "tags",
-];
-
-const VALID_AGENT_UPDATE_FIELDS = new Set<string>(AGENT_UPDATE_FIELD_NAMES);
-
-export const normalizeAgentUpdateFieldList = (
-  value: unknown,
-  fallback: readonly AgentUpdateField[] = DEFAULT_AUTO_APPROVED_SELF_UPDATE_FIELDS,
-): AgentUpdateField[] => {
-  if (!Array.isArray(value)) {
-    return [...fallback];
-  }
-
-  const normalized = Array.from(
-    new Set(
-      value.filter(
-        (item): item is AgentUpdateField =>
-          typeof item === "string" && VALID_AGENT_UPDATE_FIELDS.has(item),
-      ),
-    ),
-  );
-
-  return normalized;
-};