plugin-ai-api 1.1.0 → 1.1.2

package/dist/locale/zh-CN.json

@@ -1,138 +1,177 @@
-{
-  "AI API Gateway": "AI API 网关",
-  "Configuration": "配置",
-  "Default AI Employee": "默认 AI 员工",
-  "Enabled LLM Services": "已启用 LLM 服务",
-  "Rate Limit": "速率限制",
-  "Save Configuration": "保存配置",
-  "Configuration saved": "配置已保存",
-  "Failed to save configuration": "保存配置失败",
-  "API mode": "API 模式",
-  "Direct LLM": "直接 LLM",
-  "AI Employee agent": "AI 员工代理",
-  "Default LLM service": "默认 LLM 服务",
-  "Enable user quotas": "启用用户配额",
-  "Default reserved output tokens": "默认预留输出令牌数",
-  "Refresh": "刷新",
-  "Model pricing": "模型定价",
-  "User quotas": "用户配额",
-  "Usage": "用量",
-  "LLM service": "LLM 服务",
-  "Model": "模型",
-  "Input price / 1M": "输入价格 / 百万令牌",
-  "Output price / 1M": "输出价格 / 百万令牌",
-  "Fixed request cost": "每次请求固定费用",
-  "Currency": "货币",
-  "Status": "状态",
-  "Enabled": "已启用",
-  "Disabled": "已禁用",
-  "Actions": "操作",
-  "Edit": "编辑",
-  "Delete": "删除",
-  "Delete this price?": "删除此价格？",
-  "Add price": "添加价格",
-  "Edit price": "编辑价格",
-  "Effective from": "生效时间",
-  "Effective to": "失效时间",
-  "Notes": "备注",
-  "Saved successfully": "保存成功",
-  "Deleted successfully": "删除成功",
-  "User": "用户",
-  "Period": "周期",
-  "Request limit": "请求限制",
-  "Token limit": "令牌限制",
-  "Cost limit": "费用限制",
-  "Timezone": "时区",
-  "Unlimited": "无限制",
-  "Add quota": "添加配额",
-  "Edit quota": "编辑配额",
-  "Delete this quota?": "删除此配额？",
-  "Daily": "每日",
-  "Monthly": "每月",
-  "Reject unpriced models": "拒绝未定价模型",
-  "Missing usage behavior": "缺少用量时的行为",
-  "Use reserved estimate": "使用预留估算",
-  "Allow without token charge": "允许且不计令牌",
-  "Context overflow behavior": "上下文超限处理",
-  "Reject request": "拒绝请求",
-  "Truncate oldest conversation turns": "截断最早的对话轮次",
-  "Started at": "开始时间",
-  "Requested model": "请求模型",
-  "Resolved service": "解析后的服务",
-  "Resolved model": "解析后的模型",
-  "Input tokens": "输入令牌",
-  "Output tokens": "输出令牌",
-  "Total tokens": "总令牌",
-  "Prompt cache tokens": "提示缓存令牌",
-  "Cost": "费用",
-  "Cost status": "费用状态",
-  "Request ID": "请求 ID",
-  "Failed to load models": "加载模型失败",
-  "Select a model": "选择模型",
-  "Select an AI Employee": "选择 AI 员工",
-  "Usage guide": "使用指南",
-  "OpenAI-compatible endpoint": "OpenAI 兼容端点",
-  "Base URL": "基础 URL",
-  "Use a NocoBase API key as the Bearer token.": "使用 NocoBase API 密钥作为 Bearer Token。",
-  "List available models": "列出可用模型",
-  "Send a chat completion": "发送聊天补全请求",
-  "Usage filters": "用量筛选",
-  "Time range": "时间范围",
-  "User ID": "用户 ID",
-  "Succeeded": "成功",
-  "Failed": "失败",
-  "Started": "已开始",
-  "Apply filters": "应用筛选",
-  "Reset": "重置",
-  "Requests": "请求数",
-  "Total cost": "总费用",
-  "Usage records": "用量记录",
-  "Model metadata": "模型元数据",
-  "Add override": "添加覆盖",
-  "Edit override": "编辑覆盖",
-  "Delete this override?": "确认删除此覆盖？",
-  "Context window": "上下文窗口",
-  "Max completion tokens": "最大输出 token 数",
-  "Owned by": "所属方",
-  "Display name": "显示名称",
-  "Description": "描述",
-  "Leave empty to not override": "留空则不覆盖",
-  "Total input + output token capacity reported to clients.": "返回给客户端的输入+输出 token 总容量。",
-  "Maximum output tokens reported to clients.": "返回给客户端的最大输出 token 数。",
-  "Initial system prompt": "初始系统提示词",
-  "Prepended as the first system message, before any system prompt sent by the client. If the client sends no system prompt, this becomes the system prompt sent to the provider.": "作为第一条 system 消息插入到客户端发送的任何 system 提示词之前。如果客户端未发送 system 提示词，此提示词将作为发送给提供商的 system 提示词。",
-  "AI API": "AI API",
-  "Allow this role to use the AI API": "允许此角色使用 AI API",
-  "Allow all AI Employees": "允许所有 AI 员工",
-  "Select which AI Employees this role may use:": "选择此角色可使用的 AI 员工：",
-  "Select allowed AI Employees": "选择允许的 AI 员工",
-  "Max request body size (MB)": "请求体大小上限（MB）",
-  "Raise this to accept inline base64 images. Base64 adds about 33% to the original file size.": "调高此值以接收内联 base64 图片。base64 编码会使体积增加约 33%。",
-  "Usage groups": "用量组",
-  "Add group": "添加分组",
-  "Edit group": "编辑分组",
-  "Delete this group?": "删除此分组？",
-  "Mode": "模式",
-  "Share": "共享",
-  "Per user": "按用户",
-  "Rate limit per minute": "每分钟速率限制",
-  "Members": "成员",
-  "Add member": "添加成员",
-  "Member added": "成员已添加",
-  "Member removed": "成员已移除",
-  "Remove member?": "移除成员？",
-  "Remove": "移除",
-  "Search group by user": "按用户搜索分组",
-  "User belongs to": "用户属于",
-  "User not found": "未找到用户",
-  "Default": "默认",
-  "Allowed LLM services": "允许的 LLM 服务",
-  "Allow all models": "允许所有模型",
-  "Allowed models": "允许的模型",
-  "Model access": "模型访问",
-  "All models": "所有模型",
-  "All services": "所有服务",
-  "No models": "未允许任何模型",
-  "Leave empty to allow every service enabled in the general configuration.": "留空则允许通用配置中已启用的所有服务。",
-  "Users who do not belong to any other group automatically use this default group — no need to add members.": "不属于其他分组的用户会自动使用此默认分组，无需手动添加成员。"
-}
+{
+  "AI API Gateway": "AI API 网关",
+  "Configuration": "配置",
+  "Default AI Employee": "默认 AI 员工",
+  "Enabled LLM Services": "已启用 LLM 服务",
+  "Rate Limit": "速率限制",
+  "Save Configuration": "保存配置",
+  "Configuration saved": "配置已保存",
+  "Failed to save configuration": "保存配置失败",
+  "API mode": "API 模式",
+  "Direct LLM": "直接 LLM",
+  "AI Employee agent": "AI 员工代理",
+  "Default LLM service": "默认 LLM 服务",
+  "Enable user quotas": "启用用户配额",
+  "Default reserved output tokens": "默认预留输出令牌数",
+  "Refresh": "刷新",
+  "Model pricing": "模型定价",
+  "User quotas": "用户配额",
+  "Usage": "用量",
+  "LLM service": "LLM 服务",
+  "Model": "模型",
+  "Input price / 1M": "输入价格 / 百万令牌",
+  "Cache input price / 1M": "缓存输入价格 / 百万令牌",
+  "Output price / 1M": "输出价格 / 百万令牌",
+  "Fixed request cost": "每次请求固定费用",
+  "Currency": "货币",
+  "Status": "状态",
+  "Enabled": "已启用",
+  "Disabled": "已禁用",
+  "Actions": "操作",
+  "Edit": "编辑",
+  "Delete": "删除",
+  "Delete this price?": "删除此价格？",
+  "Add price": "添加价格",
+  "Edit price": "编辑价格",
+  "Effective from": "生效时间",
+  "Effective to": "失效时间",
+  "Notes": "备注",
+  "Saved successfully": "保存成功",
+  "Deleted successfully": "删除成功",
+  "User": "用户",
+  "Period": "周期",
+  "Request limit": "请求限制",
+  "Token limit": "令牌限制",
+  "Cost limit": "费用限制",
+  "Timezone": "时区",
+  "Unlimited": "无限制",
+  "Add quota": "添加配额",
+  "Edit quota": "编辑配额",
+  "Delete this quota?": "删除此配额？",
+  "Daily": "每日",
+  "Monthly": "每月",
+  "Reject unpriced models": "拒绝未定价模型",
+  "Missing usage behavior": "缺少用量时的行为",
+  "Use reserved estimate": "使用预留估算",
+  "Allow without token charge": "允许且不计令牌",
+  "Context overflow behavior": "上下文超限处理",
+  "Reject request": "拒绝请求",
+  "Truncate oldest conversation turns": "截断最早的对话轮次",
+  "Started at": "开始时间",
+  "Requested model": "请求模型",
+  "Resolved service": "解析后的服务",
+  "Resolved model": "解析后的模型",
+  "Input tokens": "输入令牌",
+  "Output tokens": "输出令牌",
+  "Total tokens": "总令牌",
+  "Prompt cache tokens": "提示缓存令牌",
+  "Cost": "费用",
+  "Cost status": "费用状态",
+  "Request ID": "请求 ID",
+  "Failed to load models": "加载模型失败",
+  "Select a model": "选择模型",
+  "Select an AI Employee": "选择 AI 员工",
+  "Usage guide": "使用指南",
+  "OpenAI-compatible endpoint": "OpenAI 兼容端点",
+  "Base URL": "基础 URL",
+  "Use a NocoBase API key as the Bearer token.": "使用 NocoBase API 密钥作为 Bearer Token。",
+  "List available models": "列出可用模型",
+  "Send a chat completion": "发送聊天补全请求",
+  "Usage filters": "用量筛选",
+  "Time range": "时间范围",
+  "User ID": "用户 ID",
+  "Succeeded": "成功",
+  "Failed": "失败",
+  "Started": "已开始",
+  "Apply filters": "应用筛选",
+  "Reset": "重置",
+  "Requests": "请求数",
+  "Total cost": "总费用",
+  "Usage records": "用量记录",
+  "Model metadata": "模型元数据",
+  "Add override": "添加覆盖",
+  "Edit override": "编辑覆盖",
+  "Delete this override?": "确认删除此覆盖？",
+  "Context window": "上下文窗口",
+  "Max completion tokens": "最大输出 token 数",
+  "Owned by": "所属方",
+  "Display name": "显示名称",
+  "Description": "描述",
+  "Leave empty to not override": "留空则不覆盖",
+  "Total input + output token capacity reported to clients.": "返回给客户端的输入+输出 token 总容量。",
+  "Maximum output tokens reported to clients.": "返回给客户端的最大输出 token 数。",
+  "Initial system prompt": "初始系统提示词",
+  "Prepended as the first system message, before any system prompt sent by the client. If the client sends no system prompt, this becomes the system prompt sent to the provider.": "作为第一条 system 消息插入到客户端发送的任何 system 提示词之前。如果客户端未发送 system 提示词，此提示词将作为发送给提供商的 system 提示词。",
+  "AI API": "AI API",
+  "Allow this role to use the AI API": "允许此角色使用 AI API",
+  "Allow all AI Employees": "允许所有 AI 员工",
+  "Select which AI Employees this role may use:": "选择此角色可使用的 AI 员工：",
+  "Select allowed AI Employees": "选择允许的 AI 员工",
+  "Max request body size (MB)": "请求体大小上限（MB）",
+  "Raise this to accept inline base64 images. Base64 adds about 33% to the original file size.": "调高此值以接收内联 base64 图片。base64 编码会使体积增加约 33%。",
+  "Usage groups": "用量组",
+  "Add group": "添加分组",
+  "Edit group": "编辑分组",
+  "Delete this group?": "删除此分组？",
+  "Mode": "模式",
+  "Share": "共享",
+  "Per user": "按用户",
+  "Rate limit per minute": "每分钟速率限制",
+  "Members": "成员",
+  "Add member": "添加成员",
+  "Member added": "成员已添加",
+  "Member removed": "成员已移除",
+  "Remove member?": "移除成员？",
+  "Remove": "移除",
+  "Search group by user": "按用户搜索分组",
+  "User belongs to": "用户属于",
+  "User not found": "未找到用户",
+  "Default": "默认",
+  "Allowed LLM services": "允许的 LLM 服务",
+  "Allow all models": "允许所有模型",
+  "Allowed models": "允许的模型",
+  "Model access": "模型访问",
+  "All models": "所有模型",
+  "All services": "所有服务",
+  "No models": "未允许任何模型",
+  "Leave empty to allow every service enabled in the general configuration.": "留空则允许通用配置中已启用的所有服务。",
+  "Users who do not belong to any other group automatically use this default group — no need to add members.": "不属于其他分组的用户会自动使用此默认分组，无需手动添加成员。",
+  "Membership": "默认成员",
+  "Model routing": "模型路由",
+  "Model routing (virtual models)": "模型路由（虚拟模型）",
+  "Alias": "别名",
+  "Fallback model": "兜底模型",
+  "A fallback model is required": "必须填写兜底模型",
+  "Used when no capability bucket candidate is usable and the fallback is permitted for the caller.": "当能力桶中没有可用模型且调用者有权使用回退模型时使用。",
+  "Vision models (in order)": "视觉模型（按顺序）",
+  "Requests with an image or file block use the first permitted model here.": "包含图片或文件块的请求会使用此处第一个有权限的模型。",
+  "Tool-calling models (in order)": "工具调用模型（按顺序）",
+  "Requests with tools/tool_choice use the first permitted model here.": "包含 tools/tool_choice 的请求会使用此处第一个有权限的模型。",
+  "Reasoning models (in order)": "推理模型（按顺序）",
+  "Requests with an explicit reasoning or reasoning_effort parameter use the first permitted model here.": "显式包含 reasoning 或 reasoning_effort 参数的请求会使用此处第一个有权限的模型。",
+  "Cheap models (in order)": "低成本模型（按顺序）",
+  "Optional. When set, cheap-eligible requests use the first permitted model here.": "可选。设置后，低复杂度请求会使用此处第一个有权限的模型。",
+  "General models (in order)": "通用模型（按顺序）",
+  "Default bucket when no capability rule matched. Leave empty to derive from all enabled models ordered by Model metadata sortOrder.": "没有任何能力规则命中时的默认桶。留空则按模型元数据 sortOrder 对所有已启用模型排序后派生。",
+  "Type service/model and press Enter — earlier entries are preferred": "输入 service/模型 后回车——越靠前优先级越高",
+  "An empty bucket is derived automatically from Model metadata (capability flags + sortOrder).": "留空的桶会自动从“模型元数据”页的能力标志 + sortOrder 派生。",
+  "Save": "保存",
+  "Supports vision": "支持视觉输入",
+  "Used by virtual-model routing for image/file requests.": "用于对包含图片或文件的请求进行虚拟模型路由。",
+  "Supports tool calling": "支持工具调用",
+  "Used by virtual-model routing for requests with tools/tool_choice.": "用于对包含 tools/tool_choice 的请求进行虚拟模型路由。",
+  "Reasoning tier": "推理层级",
+  "cheap | general | reasoning. Used by virtual-model routing buckets.": "cheap | general | reasoning。用于虚拟模型路由桶。",
+  "Cheap": "低成本",
+  "General": "通用",
+  "Reasoning": "推理",
+  "Routing priority": "路由优先级",
+  "Ascending — lower is preferred when a bucket is derived from metadata.": "升序 — 从元数据派生路由桶时，数值越小优先级越高。",
+  "Select models — the order shown is the routing priority": "选择模型 — 显示顺序即路由优先级",
+  "Select a fallback model": "选择兜底模型",
+  "Name": "名称",
+  "Rate limit/min": "每分钟限制",
+
+
+  "unavailable": "(不可用)"
+}

package/dist/server/billing.js

@@ -86,9 +86,15 @@ function formatUnits(value, scale) {
 function divideRounded(value, divisor) {
   return (value + divisor / 2n) / divisor;
 }
-function calculateCostUnits(inputTokens, outputTokens, price) {
+function calculateCostUnits(inputTokens, outputTokens, promptCacheTokens, price) {
+  const cachedInputTokens = Math.min(Math.max(promptCacheTokens, 0), inputTokens);
+  const uncachedInputTokens = inputTokens - cachedInputTokens;
   const input = divideRounded(
-    BigInt(inputTokens) * decimalUnits(price.inputPricePerMillionTokens, PRICE_SCALE),
+    BigInt(uncachedInputTokens) * decimalUnits(price.inputPricePerMillionTokens, PRICE_SCALE),
+    PRICE_TO_COST_DIVISOR
+  );
+  const cacheInput = divideRounded(
+    BigInt(cachedInputTokens) * decimalUnits(price.cacheInputPricePerMillionTokens, PRICE_SCALE),
     PRICE_TO_COST_DIVISOR
   );
   const output = divideRounded(
@@ -96,10 +102,10 @@ function calculateCostUnits(inputTokens, outputTokens, price) {
     PRICE_TO_COST_DIVISOR
   );
   const fixed = divideRounded(decimalUnits(price.fixedCostPerRequest, PRICE_SCALE), 100n);
-  return input + output + fixed;
+  return input + cacheInput + output + fixed;
 }
-function calculateLlmCost(inputTokens, outputTokens, price) {
-  return formatUnits(calculateCostUnits(inputTokens, outputTokens, price), COST_SCALE);
+function calculateLlmCost(inputTokens, outputTokens, price, promptCacheTokens = 0) {
+  return formatUnits(calculateCostUnits(inputTokens, outputTokens, promptCacheTokens, price), COST_SCALE);
 }
 function normalizePositiveInteger(value, fallback) {
   const parsed = Number(value);
@@ -145,6 +151,7 @@ async function findPrice(ctx, service, modelId) {
     id: valueOf(price, "id"),
     currency: valueOf(price, "currency"),
     inputPricePerMillionTokens: decimalString(valueOf(price, "inputPricePerMillionTokens"), PRICE_SCALE),
+    cacheInputPricePerMillionTokens: decimalString(valueOf(price, "cacheInputPricePerMillionTokens"), PRICE_SCALE),
     outputPricePerMillionTokens: decimalString(valueOf(price, "outputPricePerMillionTokens"), PRICE_SCALE),
     fixedCostPerRequest: decimalString(valueOf(price, "fixedCostPerRequest"), PRICE_SCALE)
   };
@@ -252,7 +259,7 @@ function usageNumbers(usage) {
   };
 }
 async function finalizeLlmBilling(ctx, providerUsage, succeeded) {
-  var _a, _b, _c, _d, _e, _f;
+  var _a, _b, _c, _d, _e, _f, _g;
   const billing = stateOf(ctx).aiApiLlmBilling;
   if (!billing) return {};
   let numbers = usageNumbers(providerUsage);
@@ -269,7 +276,7 @@ async function finalizeLlmBilling(ctx, providerUsage, succeeded) {
   } else {
     costStatus = billing.price ? "usage_unavailable" : "unpriced";
   }
-  const cost = numbers && billing.price ? calculateLlmCost(numbers.input, numbers.output, billing.price) : void 0;
+  const cost = numbers && billing.price ? calculateLlmCost(numbers.input, numbers.output, billing.price, (providerUsage == null ? void 0 : providerUsage.prompt_cache_tokens) ?? 0) : void 0;
   const reservation = billing.reservation;
   if (reservation) {
     const Bucket = ctx.db.getModel("aiApiGroupQuotaBuckets");
@@ -315,8 +322,9 @@ async function finalizeLlmBilling(ctx, providerUsage, succeeded) {
     groupId: reservation == null ? void 0 : reservation.groupId,
     quotaMode: reservation == null ? void 0 : reservation.quotaMode,
     inputPricePerMillionTokens: (_d = billing.price) == null ? void 0 : _d.inputPricePerMillionTokens,
-    outputPricePerMillionTokens: (_e = billing.price) == null ? void 0 : _e.outputPricePerMillionTokens,
-    fixedCostPerRequest: (_f = billing.price) == null ? void 0 : _f.fixedCostPerRequest
+    cacheInputPricePerMillionTokens: (_e = billing.price) == null ? void 0 : _e.cacheInputPricePerMillionTokens,
+    outputPricePerMillionTokens: (_f = billing.price) == null ? void 0 : _f.outputPricePerMillionTokens,
+    fixedCostPerRequest: (_g = billing.price) == null ? void 0 : _g.fixedCostPerRequest
   };
 }
 // Annotate the CommonJS export names for ESM import in node:

package/dist/server/collections/ai-api-model-metadata.js

@@ -72,6 +72,32 @@ var ai_api_model_metadata_default = (0, import_database.defineCollection)({
       allowNull: true,
       comment: "Initial system prompt prepended as the first system message of every request for this model. Never replaces the client system prompt."
     },
+    {
+      name: "supportsVision",
+      type: "boolean",
+      defaultValue: false,
+      index: true,
+      comment: "Whether the model accepts image content blocks. Used by virtual-model routing."
+    },
+    {
+      name: "supportsToolCalling",
+      type: "boolean",
+      defaultValue: true,
+      index: true,
+      comment: "Whether the model accepts tools / tool_choice. Used by virtual-model routing."
+    },
+    {
+      name: "reasoningTier",
+      type: "string",
+      defaultValue: "general",
+      comment: "Cost/capability tier for routing: cheap | general | reasoning."
+    },
+    {
+      name: "sortOrder",
+      type: "integer",
+      defaultValue: 0,
+      comment: "Ascending priority used when deriving a routing bucket from metadata (lower = preferred)."
+    },
     {
       name: "enabled",
       type: "boolean",

package/dist/server/collections/ai-api-model-prices.js

@@ -39,6 +39,14 @@ var ai_api_model_prices_default = (0, import_database.defineCollection)({
     { name: "model", type: "string", allowNull: false, index: true },
     { name: "currency", type: "string", allowNull: false, defaultValue: "USD" },
     { name: "inputPricePerMillionTokens", type: "decimal", precision: 20, scale: 10, allowNull: false },
+    {
+      name: "cacheInputPricePerMillionTokens",
+      type: "decimal",
+      precision: 20,
+      scale: 10,
+      allowNull: false,
+      defaultValue: 0
+    },
     { name: "outputPricePerMillionTokens", type: "decimal", precision: 20, scale: 10, allowNull: false },
     { name: "fixedCostPerRequest", type: "decimal", precision: 20, scale: 10, allowNull: false, defaultValue: 0 },
     { name: "effectiveFrom", type: "datetimeTz", allowNull: false, index: true },

package/dist/server/collections/ai-api-response-records.js

@@ -0,0 +1,101 @@
+/**
+ * This file is part of the NocoBase (R) project.
+ * Copyright (c) 2020-2024 NocoBase Co., Ltd.
+ * Authors: NocoBase Team.
+ *
+ * This project is dual-licensed under AGPL-3.0 and NocoBase Commercial License.
+ * For more information, please refer to: https://www.nocobase.com/agreement.
+ */
+
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var ai_api_response_records_exports = {};
+__export(ai_api_response_records_exports, {
+  default: () => ai_api_response_records_default
+});
+module.exports = __toCommonJS(ai_api_response_records_exports);
+var import_database = require("@nocobase/database");
+var ai_api_response_records_default = (0, import_database.defineCollection)({
+  name: "aiApiResponseRecords",
+  dataCategory: "runtime",
+  hidden: true,
+  autoGenId: true,
+  fields: [
+    {
+      name: "responseId",
+      type: "string",
+      unique: true,
+      index: true,
+      comment: "Public response ID (resp_xxx) exposed to clients"
+    },
+    {
+      name: "userId",
+      type: "bigInt",
+      allowNull: false,
+      index: true,
+      comment: "Owner of this response chain"
+    },
+    {
+      name: "user",
+      type: "belongsTo",
+      target: "users",
+      targetKey: "id",
+      foreignKey: "userId",
+      constraints: false
+    },
+    {
+      name: "model",
+      type: "string",
+      allowNull: false,
+      index: true,
+      comment: "Model used for this response (service/model format)"
+    },
+    {
+      name: "input",
+      type: "jsonb",
+      allowNull: false,
+      comment: "Original Responses API input (string or array)"
+    },
+    {
+      name: "output",
+      type: "jsonb",
+      allowNull: false,
+      comment: "Complete ResponseObject as returned to the client"
+    },
+    {
+      name: "previousResponseId",
+      type: "string",
+      allowNull: true,
+      index: true,
+      comment: "Links to the previous response in this conversation chain"
+    },
+    {
+      name: "metadata",
+      type: "jsonb",
+      allowNull: true,
+      comment: "Client-provided metadata from the request"
+    },
+    {
+      name: "expiresAt",
+      type: "datetimeTz",
+      allowNull: false,
+      index: true,
+      comment: "Auto-delete after 30 days (OpenAI retention policy)"
+    }
+  ]
+});

package/dist/server/collections/ai-api-usage-records.js

@@ -70,6 +70,7 @@ var ai_api_usage_records_default = (0, import_database.defineCollection)({
     { name: "quotaPolicyId", type: "bigInt", allowNull: true, index: true },
     { name: "groupId", type: "bigInt", allowNull: true, index: true },
     { name: "inputPricePerMillionTokens", type: "decimal", allowNull: true, precision: 20, scale: 10 },
+    { name: "cacheInputPricePerMillionTokens", type: "decimal", allowNull: true, precision: 20, scale: 10 },
     { name: "outputPricePerMillionTokens", type: "decimal", allowNull: true, precision: 20, scale: 10 },
     { name: "fixedCostPerRequest", type: "decimal", allowNull: true, precision: 20, scale: 10 },
     { name: "providerRequestId", type: "string", allowNull: true },

package/dist/server/collections/ai-api-virtual-models.js

@@ -0,0 +1,68 @@
+/**
+ * This file is part of the NocoBase (R) project.
+ * Copyright (c) 2020-2024 NocoBase Co., Ltd.
+ * Authors: NocoBase Team.
+ *
+ * This project is dual-licensed under AGPL-3.0 and NocoBase Commercial License.
+ * For more information, please refer to: https://www.nocobase.com/agreement.
+ */
+
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var ai_api_virtual_models_exports = {};
+__export(ai_api_virtual_models_exports, {
+  default: () => ai_api_virtual_models_default
+});
+module.exports = __toCommonJS(ai_api_virtual_models_exports);
+var import_database = require("@nocobase/database");
+var ai_api_virtual_models_default = (0, import_database.defineCollection)({
+  name: "aiApiVirtualModels",
+  autoGenId: true,
+  fields: [
+    {
+      name: "name",
+      type: "string",
+      unique: true,
+      allowNull: false,
+      comment: 'Alias the client passes as the model id, e.g. "auto".'
+    },
+    {
+      name: "mode",
+      type: "string",
+      defaultValue: "chat",
+      comment: "Endpoint family this alias serves: chat | embedding."
+    },
+    {
+      name: "fallbackModel",
+      type: "string",
+      allowNull: false,
+      comment: 'Concrete "service/modelId" used when no capability bucket candidate is usable and fallback is permitted.'
+    },
+    {
+      name: "visionModels",
+      type: "json",
+      defaultValue: [],
+      comment: 'Ordered "service/modelId" list for vision requests.'
+    },
+    { name: "toolModels", type: "json", defaultValue: [], comment: "Ordered list for tool-calling requests." },
+    { name: "reasoningModels", type: "json", defaultValue: [], comment: "Ordered list for reasoning-tier requests." },
+    { name: "cheapModels", type: "json", defaultValue: [], comment: "Ordered list for cheap-tier requests." },
+    { name: "generalModels", type: "json", defaultValue: [], comment: "Ordered list for general requests." },
+    { name: "enabled", type: "boolean", defaultValue: true, index: true }
+  ]
+});