nolo-cli 0.1.8 → 0.1.9

package/ai/token/stats.ts

@@ -0,0 +1,21 @@
+export function createInitialDayStats(
+  userId: string,
+  dateKey: string
+): DayStats {
+  const initial = {
+    userId,
+    period: "day",
+    timeKey: dateKey,
+    total: {
+      count: 0,
+      tokens: {
+        input: 0,
+        output: 0,
+      },
+      cost: 0,
+    },
+    models: {},
+    providers: {},
+  };
+  return initial;
+}

package/ai/token/tokenThunks.ts

@@ -0,0 +1,24 @@
+import { createAsyncThunk } from "@reduxjs/toolkit";
+import { AppThunkApi } from "app/store";
+import { getTokenStats, StatsParams, TokenStats } from "./query";
+import { queryUserTokens, QueryParams, QueryResult } from "./queryUserTokens";
+
+export const getTokenStatsThunk = createAsyncThunk<
+    TokenStats[],
+    StatsParams,
+    AppThunkApi
+>("token/getStats", async (params, { extra }) => {
+    const { db } = extra;
+    if (!db) throw new Error("Database not available");
+    return await getTokenStats(db, params);
+});
+
+export const queryUserTokensThunk = createAsyncThunk<
+    QueryResult,
+    QueryParams,
+    AppThunkApi
+>("token/queryUserTokens", async (params, { extra }) => {
+    const { db } = extra;
+    if (!db) throw new Error("Database not available");
+    return await queryUserTokens(db, params);
+});

package/ai/token/types.ts

@@ -0,0 +1,93 @@
+// ai/token/types.ts
+
+export const DEFAULT_QUERY_LIMIT = 100;
+
+export const TOKEN_PERIODS = {
+  DAY: "day",
+  WEEK: "week",
+  MONTH: "month",
+} as const;
+
+export const TOKEN_SCOPES = {
+  USER: "user",
+  CYBOT: "cybot",
+  SITE: "site",
+} as const;
+
+interface BillingUsageMetadata {
+  cost?: number;
+  billing_provider?: string;
+  billing_model?: string;
+  billing_service_tier?: string;
+  billing_estimated?: boolean;
+  image_generation_count?: number;
+}
+
+// 原始用量类型
+export interface RawUsageType1 extends BillingUsageMetadata {
+  output_tokens?: number;
+  input_tokens?: number;
+  cache_creation_input_tokens?: number;
+  cache_read_input_tokens?: number;
+}
+
+export interface RawUsageType2 extends BillingUsageMetadata {
+  prompt_tokens: number;
+  completion_tokens: number;
+  total_tokens: number;
+  prompt_cache_hit_tokens: number;
+  prompt_cache_miss_tokens: number;
+}
+
+export type RawUsage = RawUsageType1 | RawUsageType2;
+
+// 标准化后的用量数据
+export interface NormalizedUsage extends BillingUsageMetadata {
+  input_tokens: number;
+  output_tokens: number;
+  cache_creation_input_tokens: number;
+  cache_read_input_tokens: number;
+  cost: number;
+}
+
+// Token使用数据
+export interface TokenUsageData extends NormalizedUsage {
+  userId?: string;
+  cybotId: string;
+  model: string;
+  provider: string;
+  dialogId: string;
+  pay: any; // TODO: 明确支付数据类型
+}
+
+// Token记录
+export interface TokenRecord {
+  id: string;
+  userId: string;
+  username: string;
+  cybotId: string;
+  model: string;
+  provider: string;
+  dialogId: string;
+  cache_creation_input_tokens: number;
+  cache_read_input_tokens: number;
+  output_tokens: number;
+  input_tokens: number;
+  cost: number;
+  image_generation_count?: number;
+  pay: any;
+  createdAt: number; // UTC timestamp
+  type: string;
+}
+
+// Token统计数据
+export interface TokenStats {
+  total: number;
+  date: string; // YYYY-MM-DD in UTC
+  inputTokens: number;
+  outputTokens: number;
+  cost: number;
+  userId: string;
+  createdAt: number; // UTC timestamp
+  type: string;
+}

package/ai/tools/agent/agentTools.ts

@@ -0,0 +1,176 @@
+// 文件路径: ai/tools/agent/agentTools.ts
+
+import {
+    callAgentFunctionSchema,
+    callAgentFunc,
+} from "./callAgentTool";
+import {
+    createAgentToolFunctionSchema,
+    createAgentToolFunc,
+} from "./createAgentTool";
+import {
+    createSkillAgentToolFunctionSchema,
+    createSkillAgentToolFunc,
+} from "./createSkillAgentTool";
+import {
+    updateAgentToolFunctionSchema,
+    updateAgentToolFunc,
+} from "./updateAgentTool";
+import {
+    updateSelfToolFunctionSchema,
+    updateSelfToolFunc,
+} from "./updateSelfTool";
+import {
+    runStreamingAgentFunctionSchema,
+    runStreamingAgentFunc,
+} from "./runStreamingAgentTool";
+import {
+    streamParallelAgentsFunctionSchema,
+    streamParallelAgentsFunc,
+} from "./streamParallelAgentsTool";
+import {
+    createDialogFunctionSchema,
+    createDialogFunc,
+} from "./createDialogTool";
+import {
+    runLlmFunctionSchema,
+    runLlmToolFunc,
+} from "./runLlmTool";
+
+import type { ToolDefinition } from "../index";
+
+export const agentToolDefinitions: ToolDefinition[] = [
+    {
+        id: "createAgent",
+        schema: createAgentToolFunctionSchema,
+        executor: createAgentToolFunc,
+        description: {
+            name: "createAgent",
+            description:
+                "根据给定配置创建一个新的 Agent（智能体 / 应用），并返回其详细信息。",
+            category: "计划与编排",
+        },
+        behavior: "action",
+        uiGroup: "agent",
+        capability: "self_evolution",
+        riskLevel: "high",
+        costLevel: "low",
+        defaultConsent: "ask",
+    },
+    {
+        id: "createSkillAgent",
+        schema: createSkillAgentToolFunctionSchema,
+        executor: createSkillAgentToolFunc,
+        description: {
+            name: "createSkillAgent",
+            description:
+                "创建专门用于创建/评估 skill 文档协议的 Agent。",
+            category: "计划与编排",
+        },
+        behavior: "action",
+        uiGroup: "agent",
+        capability: "self_evolution",
+        riskLevel: "high",
+        costLevel: "low",
+        defaultConsent: "ask",
+    },
+    {
+        id: "updateSelf",
+        schema: updateSelfToolFunctionSchema,
+        executor: updateSelfToolFunc,
+        description: {
+            name: "updateSelf",
+            description:
+                "更新当前正在运行的 Agent 自己。低风险字段可直接更新，高影响字段会先请求确认。",
+            category: "计划与编排",
+        },
+        behavior: "action",
+        uiGroup: "agent",
+        capability: "self_evolution",
+        riskLevel: "high",
+        costLevel: "low",
+        defaultConsent: "ask",
+    },
+    {
+        id: "updateAgent",
+        schema: updateAgentToolFunctionSchema,
+        executor: updateAgentToolFunc,
+        description: {
+            name: "updateAgent",
+            description:
+                "根据给定配置更新已经存在的 Agent。只会修改参数中出现的字段，未提供的字段保持不变。",
+            category: "计划与编排",
+        },
+        behavior: "action",
+        uiGroup: "agent",
+        capability: "self_evolution",
+        riskLevel: "high",
+        costLevel: "low",
+        defaultConsent: "ask",
+    },
+    {
+        id: "callAgent",
+        schema: callAgentFunctionSchema,
+        executor: callAgentFunc,
+        description: {
+            name: "callAgent",
+            description:
+                "调用一个指定的 Agent 执行一次子任务，可用于多 Agent 编排、自动评测、自动对比等场景。",
+            category: "计划与编排",
+        },
+        behavior: "orchestrator",
+        uiGroup: "agent",
+    },
+    {
+        id: "runStreamingAgent",
+        schema: runStreamingAgentFunctionSchema,
+        executor: runStreamingAgentFunc,
+        description: {
+            name: "runStreamingAgent",
+            description:
+                "将本轮回答交由指定的 Agent 以流式方式输出，常用于从通用助手切换到某个专用应用。",
+            category: "计划与编排",
+        },
+        behavior: "orchestrator",
+        uiGroup: "agent",
+    },
+    {
+        id: "streamParallelAgents",
+        schema: streamParallelAgentsFunctionSchema,
+        executor: streamParallelAgentsFunc,
+        description: {
+            name: "streamParallelAgents",
+            description:
+                "并行调用多个 Agent，并让它们分别以流式方式输出结果。",
+            category: "计划与编排",
+        },
+        behavior: "orchestrator",
+        uiGroup: "agent",
+    },
+    {
+        id: "createDialog",
+        schema: createDialogFunctionSchema,
+        executor: createDialogFunc,
+        description: {
+            name: "createDialog",
+            description:
+                "创建一个新的对话，指定 Agent 响应，可选发送第一条消息触发执行。返回 dialogId 可随时查看结果。",
+            category: "计划与编排",
+        },
+        behavior: "action",
+        uiGroup: "agent",
+    },
+    {
+        id: "runLlm",
+        schema: runLlmFunctionSchema,
+        executor: runLlmToolFunc,
+        description: {
+            name: "runLlm",
+            description:
+                "对指定模型发起一次单轮 LLM 调用，不加载知识库也不调用工具。适合摘要、分类、格式化等纯文本处理任务。",
+            category: "计划与编排",
+        },
+        behavior: "action",
+        uiGroup: "agent",
+    },
+];

package/ai/tools/agent/agentUpdateShared.ts

@@ -0,0 +1,311 @@
+import type { Agent } from "app/types";
+import type { FormData as AgentFormData } from "ai/agent/createAgentSchema";
+import {
+  type AgentUpdateField,
+  AGENT_UPDATE_FIELD_NAMES,
+} from "ai/policy/selfUpdateFields";
+import { ToolResultError } from "ai/tools/toolResultError";
+
+type ReasoningEffort = "low" | "medium" | "high";
+
+export type ReferenceArg = {
+  dbKey: string;
+  title?: string;
+  type?: "knowledge" | "instruction" | "page";
+};
+
+export type GreetingMenuItemArg = {
+  id: string;
+  label: string;
+  userMessage?: string;
+};
+
+export type GreetingConfigArg = {
+  text?: string;
+  menu?: GreetingMenuItemArg[];
+};
+
+export type AgentUpdateArgsShape = {
+  __confirmedSelfEvolution?: boolean;
+  name?: string;
+  model?: string;
+  provider?: string;
+  prompt?: string;
+  introduction?: string;
+  greeting?: string | GreetingConfigArg;
+  isPublic?: boolean;
+  tags?: string[] | string;
+  tools?: string[];
+  references?: ReferenceArg[];
+  temperature?: number;
+  top_p?: number;
+  frequency_penalty?: number;
+  presence_penalty?: number;
+  max_tokens?: number;
+  reasoning_effort?: ReasoningEffort;
+};
+
+export type UpdateAgentToolArgs = AgentUpdateArgsShape & {
+  agentId: string;
+};
+
+export type UpdateSelfToolArgs = AgentUpdateArgsShape;
+
+export const agentUpdateFieldSchemaProperties = {
+  name: { type: "string", description: "Agent 的新名称。" },
+  model: { type: "string", description: "模型名称，例如 'gpt-4o-mini'。" },
+  provider: { type: "string", description: "模型提供方，例如 'openai'。若不确定可留空。" },
+  prompt: { type: "string", description: "系统提示词，描述 Agent 的角色和行为。" },
+  introduction: { type: "string", description: "面向终端用户的简介文案。" },
+  greeting: {
+    description: "欢迎配置：纯文本字符串，或包含欢迎语 + 菜单的对象。",
+    anyOf: [
+      { type: "string", description: "纯文本欢迎语。" },
+      {
+        type: "object",
+        description: "结构化欢迎配置。",
+        properties: {
+          text: { type: "string", description: "欢迎语文本。" },
+          menu: {
+            type: "array",
+            description: "快捷菜单项。",
+            items: {
+              type: "object",
+              properties: {
+                id: { type: "string", description: "菜单项唯一标识。" },
+                label: { type: "string", description: "按钮显示文案。" },
+                userMessage: { type: "string", description: "点击后发送给 Agent 的消息。" },
+              },
+              required: ["id", "label"],
+            },
+          },
+        },
+      },
+    ],
+  },
+  isPublic: { type: "boolean", description: "是否公开到应用市场。" },
+  tags: {
+    type: "array",
+    items: { type: "string" },
+    description: "标签列表；传空数组 [] 表示清空。",
+  },
+  tools: {
+    type: "array",
+    items: { type: "string" },
+    description: "允许调用的工具名称数组。",
+  },
+  references: {
+    type: "array",
+    description: "引用集合；传空数组 [] 表示清空。",
+    items: {
+      type: "object",
+      properties: {
+        dbKey: { type: "string", description: "引用条目的数据库键。" },
+        title: { type: "string", description: "引用在 UI 中展示的标题。" },
+        type: {
+          type: "string",
+          enum: ["knowledge", "instruction", "page"],
+          description: "引用类型。",
+        },
+      },
+      required: ["dbKey"],
+    },
+  },
+  temperature: { type: "number", description: "采样温度，0~2。" },
+  top_p: { type: "number", description: "nucleus sampling，0~1。" },
+  frequency_penalty: { type: "number", description: "重复惩罚，-2~2。" },
+  presence_penalty: { type: "number", description: "新话题激励，-2~2。" },
+  max_tokens: { type: "number", description: "单次回答最大 token 数。" },
+  reasoning_effort: {
+    type: "string",
+    enum: ["low", "medium", "high"],
+    description: "推理强度。",
+  },
+} as const;
+
+type AgentPatch = Partial<AgentFormData> & {
+  greeting?: string | GreetingConfigArg;
+  tags?: string[] | string;
+  references?: ReferenceArg[];
+};
+
+export const validateUpdateArgs = (
+  userId: string | undefined,
+  options?: { requireAgentId?: boolean; agentId?: string },
+): void => {
+  if (!userId) throw new Error("更新 Agent 失败：当前未登录或缺少 userId。");
+  if (options?.requireAgentId && !options?.agentId?.trim()) {
+    throw new Error("更新 Agent 失败：必须提供非空的 agentId。");
+  }
+};
+
+export const fetchAgentByDbKey = async (
+  dbKey: string,
+  db: any,
+): Promise<Agent | undefined> => {
+  if (!db) return undefined;
+  return db.get(dbKey).catch(() => undefined);
+};
+
+export const extractAgentId = (dbKey: string): string => {
+  const parts = dbKey.trim().split("-");
+  return parts.length >= 3 ? parts[parts.length - 1] : dbKey.trim();
+};
+
+export const buildPatch = (args: AgentUpdateArgsShape): AgentPatch => {
+  const patch: AgentPatch = {};
+  const {
+    name,
+    model,
+    provider,
+    prompt,
+    introduction,
+    greeting,
+    isPublic,
+    tags,
+    tools,
+    references,
+    temperature,
+    top_p,
+    frequency_penalty,
+    presence_penalty,
+    max_tokens,
+    reasoning_effort,
+  } = args;
+
+  if (name !== undefined) patch.name = String(name).trim();
+  if (model !== undefined) patch.model = String(model).trim();
+  if (provider !== undefined) patch.provider = String(provider).trim();
+  if (prompt !== undefined) patch.prompt = prompt;
+  if (introduction !== undefined) patch.introduction = introduction;
+  if (greeting !== undefined) patch.greeting = greeting;
+  if (isPublic !== undefined) patch.isPublic = isPublic;
+  if (tags !== undefined) patch.tags = tags as any;
+  if (tools !== undefined) patch.tools = tools ?? [];
+  if (references !== undefined) patch.references = references as any;
+  if (temperature !== undefined) patch.temperature = temperature;
+  if (top_p !== undefined) patch.top_p = top_p;
+  if (frequency_penalty !== undefined) patch.frequency_penalty = frequency_penalty;
+  if (presence_penalty !== undefined) patch.presence_penalty = presence_penalty;
+  if (max_tokens !== undefined) patch.max_tokens = max_tokens;
+  if (reasoning_effort !== undefined) patch.reasoning_effort = reasoning_effort;
+
+  return patch;
+};
+
+const AGENT_UPDATE_FIELD_NAME_SET = new Set<string>(AGENT_UPDATE_FIELD_NAMES);
+
+export const listRequestedFields = (
+  args: Record<string, unknown>,
+): AgentUpdateField[] =>
+  Object.keys(args).filter(
+    (key): key is AgentUpdateField =>
+      key !== "agentId" &&
+      key !== "__confirmedSelfEvolution" &&
+      AGENT_UPDATE_FIELD_NAME_SET.has(key) &&
+      args[key] !== undefined,
+  );
+
+export const assertAgentUpdateConfirmation = ({
+  scope,
+  requestedFields,
+  confirmed,
+  autoApprovedFields = [],
+}: {
+  scope: "self" | "generic";
+  requestedFields: AgentUpdateField[];
+  confirmed?: boolean;
+  autoApprovedFields?: AgentUpdateField[];
+}): void => {
+  if (requestedFields.length === 0 || confirmed) return;
+
+  const allowedFields =
+    scope === "self"
+      ? requestedFields.filter((field) => autoApprovedFields.includes(field))
+      : [];
+  const blockedFields =
+    scope === "self"
+      ? requestedFields.filter((field) => !autoApprovedFields.includes(field))
+      : requestedFields;
+
+  if (blockedFields.length === 0) return;
+
+  const displayPrefix = scope === "self" ? "updateSelf" : "updateAgent";
+  throw new ToolResultError(
+    scope === "self"
+      ? `当前自我更新字段需要先确认：${blockedFields.join(", ")}`
+      : `当前通用 Agent 更新需要先确认：${blockedFields.join(", ")}`,
+    {
+      code: "agent_update_requires_confirmation",
+      retryable: false,
+      displayData: `${displayPrefix} 需要先确认：${blockedFields.join(", ")}`,
+      rawData: {
+        error: "agent_update_requires_confirmation",
+        message:
+          scope === "self"
+            ? "当前自我更新包含需要用户确认的字段。"
+            : "当前通用 Agent 更新默认要求用户确认。",
+        policy: {
+          capability: scope === "self" ? "self_update" : "agent_update",
+          scope,
+          decision: "ask",
+          requestedFields,
+          allowedFields,
+          blockedFields,
+          autoApprovedFields,
+        },
+      },
+    },
+  );
+};
+
+export const buildRawDataWithUpdateInfo = (
+  agent: Agent,
+  previousAgent: Agent,
+  requestedFields: AgentUpdateField[],
+) => {
+  const changes: Record<string, { o: any; n: any }> = {};
+  for (const key of requestedFields) {
+    const newVal = (agent as any)[key];
+    const oldVal = (previousAgent as any)[key];
+    if (JSON.stringify(newVal) !== JSON.stringify(oldVal)) {
+      changes[key] = { o: oldVal, n: newVal };
+    }
+  }
+
+  return {
+    ...agent,
+    _isUpdate: true,
+    _changes: Object.keys(changes).length > 0 ? changes : undefined,
+  };
+};
+
+export const formatUpdatedAgentOutput = (agent: Agent): string =>
+  [
+    `✅ 已更新 Agent：${agent.name ?? "(名称未变)"}`,
+    `- ID: ${agent.id}`,
+    `- 是否公开: ${agent.isPublic ? "是" : "否"}`,
+    agent.model && `- 模型: ${agent.model}`,
+    (agent as any).provider && `- Provider: ${(agent as any).provider}`,
+    agent.tags?.length && `- 标签: ${agent.tags.join(", ")}`,
+    typeof agent.temperature === "number" && `- temperature: ${agent.temperature}`,
+  ]
+    .filter(Boolean)
+    .join("\n");
+
+export const buildUpdateThunkPreviousAgent = (
+  previousAgent: Agent,
+  userId: string,
+): Agent => {
+  const ownerUserId = String(previousAgent.userId ?? "").trim();
+  if (previousAgent.isPublic && ownerUserId !== userId) {
+    return {
+      ...previousAgent,
+      isPublic: false,
+    };
+  }
+  return previousAgent;
+};
+
+// TODO: Persist lightweight audit events for updateSelf/updateAgent so version history
+// and future rollback can inspect who changed what without rebuilding it from dialogs.