plugin-knowledge-base 1.1.2 → 1.1.4

package/dist/server/resources/ai-knowledge-base-documents.js

@@ -7,9 +7,11 @@
  * For more information, please refer to: https://www.nocobase.com/agreement.
  */
 
+var __create = Object.create;
 var __defProp = Object.defineProperty;
 var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
 var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
 var __hasOwnProp = Object.prototype.hasOwnProperty;
 var __export = (target, all) => {
   for (var name in all)
@@ -23,12 +25,28 @@ var __copyProps = (to, from, except, desc) => {
   }
   return to;
 };
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  // If the importer is in node compatibility mode or this is not an ESM
+  // file that has been converted to a CommonJS file using a Babel-
+  // compatible transform (i.e. "__esModule" has not been set), then set
+  // "default" to the CommonJS "module.exports" for node compatibility.
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
 var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
 var ai_knowledge_base_documents_exports = {};
 __export(ai_knowledge_base_documents_exports, {
   default: () => ai_knowledge_base_documents_default
 });
 module.exports = __toCommonJS(ai_knowledge_base_documents_exports);
+var import_plugin = __toESM(require("../plugin"));
+function getPlugin(ctx) {
+  try {
+    return ctx.app.pm.get(import_plugin.default);
+  } catch {
+    return null;
+  }
+}
 async function checkKBAccess(ctx, knowledgeBaseId) {
   var _a, _b, _c, _d;
   const userId = (_b = (_a = ctx.auth) == null ? void 0 : _a.user) == null ? void 0 : _b.id;
@@ -104,6 +122,10 @@ var ai_knowledge_base_documents_default = {
         const kb = await kbRepo.findOne({ filter: { id: values.knowledgeBaseId } });
         if (kb) {
           const kbData = kb.toJSON();
+          if (kbData.type === "EXTERNAL_RAG") {
+            ctx.throw(400, "Cannot upload documents to an external RAG knowledge base");
+            return;
+          }
           if (kbData.accessLevel === "BASIC" && kbData.ownerId !== userId) {
             ctx.throw(403, "Only the owner can upload documents to a personal knowledge base");
             return;
@@ -123,27 +145,19 @@ var ai_knowledge_base_documents_default = {
       }
       values.uploadedById = userId;
       const doc = await repo.create({ values });
-      const setClauses = [];
-      const bindValues = [];
-      let paramIdx = 1;
-      if (values.knowledgeBaseId) {
-        setClauses.push(`"knowledgeBaseId" = $${paramIdx++}`);
-        bindValues.push(values.knowledgeBaseId);
-      }
-      if (values.fileId) {
-        setClauses.push(`"fileId" = $${paramIdx++}`);
-        bindValues.push(values.fileId);
-      }
-      if (setClauses.length > 0) {
-        bindValues.push(doc.get("id"));
-        await ctx.db.sequelize.query(
-          `UPDATE "aiKnowledgeBaseDocuments" SET ${setClauses.join(", ")} WHERE id = $${paramIdx}`,
-          { bind: bindValues }
-        );
-        if (values.knowledgeBaseId) doc.set("knowledgeBaseId", values.knowledgeBaseId);
-        if (values.fileId) doc.set("fileId", values.fileId);
+      const fkUpdates = {};
+      if (values.knowledgeBaseId) fkUpdates.knowledgeBaseId = values.knowledgeBaseId;
+      if (values.fileId) fkUpdates.fileId = values.fileId;
+      if (Object.keys(fkUpdates).length > 0) {
+        await repo.update({
+          filterByTk: doc.get("id"),
+          values: fkUpdates
+        });
+        for (const [k, v] of Object.entries(fkUpdates)) {
+          doc.set(k, v);
+        }
       }
-      const plugin = ctx.app.pm.get("plugin-knowledge-base");
+      const plugin = getPlugin(ctx);
       if (plugin == null ? void 0 : plugin.vectorizationPipeline) {
         plugin.vectorizationPipeline.processDocument(doc.id).catch((err) => {
           ctx.logger.error(`Vectorization failed for document ${doc.id}:`, err);
@@ -185,11 +199,10 @@ var ai_knowledge_base_documents_default = {
       await next();
     },
     async reprocess(ctx, next) {
-      var _a, _b, _c, _d;
+      var _a, _b;
       const { filterByTk } = ctx.action.params;
       const repo = ctx.db.getRepository("aiKnowledgeBaseDocuments");
-      const userId = (_b = (_a = ctx.auth) == null ? void 0 : _a.user) == null ? void 0 : _b.id;
-      const roles = ((_c = ctx.state) == null ? void 0 : _c.currentRoles) ?? [];
+      const roles = ((_a = ctx.state) == null ? void 0 : _a.currentRoles) ?? [];
       const isAdmin = roles.includes("root") || roles.includes("admin");
       if (!isAdmin) {
         const doc = await repo.findOne({ filterByTk });
@@ -202,7 +215,7 @@ var ai_knowledge_base_documents_default = {
               return;
             }
             if ((kbData == null ? void 0 : kbData.accessLevel) === "SHARED") {
-              const canUpload = (_d = kbData.uploadRoles) == null ? void 0 : _d.some((r) => roles.includes(r));
+              const canUpload = (_b = kbData.uploadRoles) == null ? void 0 : _b.some((r) => roles.includes(r));
               if (!canUpload) {
                 ctx.throw(403, "You do not have permission to reprocess this document");
                 return;
@@ -215,7 +228,7 @@ var ai_knowledge_base_documents_default = {
         filterByTk,
         values: { status: "pending", error: null, chunkCount: 0 }
       });
-      const plugin = ctx.app.pm.get("plugin-knowledge-base");
+      const plugin = getPlugin(ctx);
       if (plugin == null ? void 0 : plugin.vectorizationPipeline) {
         plugin.vectorizationPipeline.processDocument(filterByTk).catch((err) => {
           ctx.logger.error(`Re-vectorization failed for document ${filterByTk}:`, err);

package/dist/server/resources/ai-knowledge-base.js

@@ -7,9 +7,11 @@
  * For more information, please refer to: https://www.nocobase.com/agreement.
  */
 
+var __create = Object.create;
 var __defProp = Object.defineProperty;
 var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
 var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
 var __hasOwnProp = Object.prototype.hasOwnProperty;
 var __export = (target, all) => {
   for (var name in all)
@@ -23,12 +25,24 @@ var __copyProps = (to, from, except, desc) => {
   }
   return to;
 };
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  // If the importer is in node compatibility mode or this is not an ESM
+  // file that has been converted to a CommonJS file using a Babel-
+  // compatible transform (i.e. "__esModule" has not been set), then set
+  // "default" to the CommonJS "module.exports" for node compatibility.
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
 var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
 var ai_knowledge_base_exports = {};
 __export(ai_knowledge_base_exports, {
   default: () => ai_knowledge_base_default
 });
 module.exports = __toCommonJS(ai_knowledge_base_exports);
+var import_plugin = __toESM(require("../plugin"));
+function getPlugin(ctx) {
+  return ctx.app.pm.get(import_plugin.default);
+}
 async function checkKBPermission(ctx, filterByTk, action) {
   var _a, _b;
   const repo = ctx.db.getRepository("aiKnowledgeBases");
@@ -89,17 +103,19 @@ var ai_knowledge_base_default = {
         filterByTk,
         appends: ["vectorStore", "documents"]
       });
-      if (record) {
-        const data = record.toJSON();
-        const userId = (_b = (_a = ctx.auth) == null ? void 0 : _a.user) == null ? void 0 : _b.id;
-        const roles = ctx.state.currentRoles ?? [];
-        const isAdmin = roles.includes("root") || roles.includes("admin");
-        if (!isAdmin) {
-          const hasAccess = data.accessLevel === "PUBLIC" || data.accessLevel === "BASIC" && data.ownerId === userId || data.accessLevel === "SHARED" && ((_c = data.allowedRoles) == null ? void 0 : _c.some((r) => roles.includes(r)));
-          if (!hasAccess) {
-            ctx.throw(403, "Access denied");
-            return;
-          }
+      if (!record) {
+        ctx.throw(404, "Knowledge base not found");
+        return;
+      }
+      const data = record.toJSON();
+      const userId = (_b = (_a = ctx.auth) == null ? void 0 : _a.user) == null ? void 0 : _b.id;
+      const roles = ctx.state.currentRoles ?? [];
+      const isAdmin = roles.includes("root") || roles.includes("admin");
+      if (!isAdmin) {
+        const hasAccess = data.accessLevel === "PUBLIC" || data.accessLevel === "BASIC" && data.ownerId === userId || data.accessLevel === "SHARED" && ((_c = data.allowedRoles) == null ? void 0 : _c.some((r) => roles.includes(r)));
+        if (!hasAccess) {
+          ctx.throw(403, "Access denied");
+          return;
         }
       }
       ctx.body = record;
@@ -120,13 +136,14 @@ var ai_knowledge_base_default = {
       if (values.accessLevel === "BASIC") {
         values.ownerId = userId;
       }
-      ctx.body = await repo.create({ values });
+      const record = await repo.create({ values });
       if (values.vectorStoreId) {
-        await ctx.db.sequelize.query(
-          `UPDATE "aiKnowledgeBases" SET "vectorStoreId" = $1 WHERE id = $2`,
-          { bind: [values.vectorStoreId, ctx.body.get("id")] }
-        );
+        await repo.update({
+          filterByTk: record.get("id"),
+          values: { vectorStoreId: values.vectorStoreId }
+        });
       }
+      ctx.body = record;
       await next();
     },
     async update(ctx, next) {
@@ -143,15 +160,8 @@ var ai_knowledge_base_default = {
         filterByTk,
         values
       });
-      if (values.vectorStoreId) {
-        await ctx.db.sequelize.query(
-          `UPDATE "aiKnowledgeBases" SET "vectorStoreId" = $1 WHERE id = $2`,
-          { bind: [values.vectorStoreId, filterByTk] }
-        );
-      }
       await next();
     },
-    // Fix #3: destroy with permission checks
     async destroy(ctx, next) {
       const { filterByTk } = ctx.action.params;
       const repo = ctx.db.getRepository("aiKnowledgeBases");

package/dist/server/resources/ai-vector-databases.js

@@ -29,6 +29,7 @@ __export(ai_vector_databases_exports, {
   default: () => ai_vector_databases_default
 });
 module.exports = __toCommonJS(ai_vector_databases_exports);
+var import_pg = require("pg");
 var ai_vector_databases_default = {
   name: "aiVectorDatabase",
   actions: {
@@ -85,8 +86,7 @@ var ai_vector_databases_default = {
         return;
       }
       try {
-        const { Client } = require("pg");
-        const client = new Client({
+        const client = new import_pg.Client({
           host: connectParams.host,
           port: connectParams.port || 5432,
           user: connectParams.username,

package/dist/server/resources/ai-vector-stores.js

@@ -58,13 +58,14 @@ var ai_vector_stores_default = {
       const rawValues = ctx.action.params.values || {};
       const values = rawValues.values || rawValues;
       const repo = ctx.db.getRepository("aiVectorStores");
-      ctx.body = await repo.create({ values });
+      const record = await repo.create({ values });
       if (values.vectorDatabaseId) {
-        await ctx.db.sequelize.query(
-          `UPDATE "aiVectorStores" SET "vectorDatabaseId" = $1 WHERE id = $2`,
-          { bind: [values.vectorDatabaseId, ctx.body.get("id")] }
-        );
+        await repo.update({
+          filterByTk: record.get("id"),
+          values: { vectorDatabaseId: values.vectorDatabaseId }
+        });
       }
+      ctx.body = record;
       await next();
     },
     async update(ctx, next) {
@@ -76,12 +77,6 @@ var ai_vector_stores_default = {
         filterByTk,
         values
       });
-      if (values.vectorDatabaseId) {
-        await ctx.db.sequelize.query(
-          `UPDATE "aiVectorStores" SET "vectorDatabaseId" = $1 WHERE id = $2`,
-          { bind: [values.vectorDatabaseId, filterByTk] }
-        );
-      }
       await next();
     },
     async destroy(ctx, next) {

package/dist/server/services/docpixie-extractor.js

@@ -0,0 +1,139 @@
+/**
+ * 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 docpixie_extractor_exports = {};
+__export(docpixie_extractor_exports, {
+  DocPixieExtractor: () => DocPixieExtractor
+});
+module.exports = __toCommonJS(docpixie_extractor_exports);
+var import_os = require("os");
+var import_path = require("path");
+var import_promises = require("fs/promises");
+var import_fs = require("fs");
+class DocPixieExtractor {
+  constructor(db, getDocpixiePlugin) {
+    this.db = db;
+    this.getDocpixiePlugin = getDocpixiePlugin;
+  }
+  /** Returns true if plugin-docpixie is loaded and ready (configured + LLM provider set) */
+  isAvailable() {
+    var _a;
+    const plugin = this.getDocpixiePlugin();
+    return !!((_a = plugin == null ? void 0 : plugin.service) == null ? void 0 : _a.isReady());
+  }
+  /**
+   * Extract text from a local file path.
+   * processDocument() is synchronous — it waits for full OCR + summarization before returning.
+   * Returns null on any failure so callers can fall through to the next extractor.
+   */
+  async extractFromPath(filePath, filename, userId) {
+    var _a;
+    const plugin = this.getDocpixiePlugin();
+    if (!((_a = plugin == null ? void 0 : plugin.service) == null ? void 0 : _a.isReady())) return null;
+    try {
+      const documentId = await plugin.service.processDocument(filePath, {
+        name: filename,
+        userId
+      });
+      const text = await this.getPageTexts(documentId);
+      return { text, documentId };
+    } catch {
+      return null;
+    }
+  }
+  /**
+   * Extract text from an in-memory buffer (for files already downloaded, e.g. from S3).
+   * Writes to a temp file, runs DocPixie extraction, then cleans up.
+   */
+  async extractFromBuffer(buffer, filename, extname, userId) {
+    var _a;
+    const plugin = this.getDocpixiePlugin();
+    if (!((_a = plugin == null ? void 0 : plugin.service) == null ? void 0 : _a.isReady())) return null;
+    const ext = extname.startsWith(".") ? extname : `.${extname}`;
+    const tempPath = (0, import_path.join)(
+      (0, import_os.tmpdir)(),
+      `kb-docpixie-${Date.now()}-${Math.random().toString(36).slice(2)}${ext}`
+    );
+    try {
+      await (0, import_promises.writeFile)(tempPath, buffer);
+      return await this.extractFromPath(tempPath, filename, userId);
+    } finally {
+      if ((0, import_fs.existsSync)(tempPath)) {
+        (0, import_promises.unlink)(tempPath).catch(() => {
+        });
+      }
+    }
+  }
+  /**
+   * Concatenate all page texts for a DocPixie document, ordered by page number.
+   * Queries the docpixie_pages collection directly.
+   */
+  async getPageTexts(documentId) {
+    const pageRepo = this.db.getRepository("docpixie_pages");
+    const pages = await pageRepo.find({
+      filter: { documentId },
+      sort: ["pageNumber"]
+    });
+    return pages.map((p) => p.get("structuredText") || "").filter(Boolean).join("\n\n");
+  }
+  /**
+   * Fetch the summary + page texts for a set of DocPixie document IDs.
+   * Used in Stage 2 deep retrieval during AI chat work context.
+   *
+   * Returns a formatted string ready to be injected into the AI prompt.
+   * Caps at maxDocs documents to avoid blowing the context window.
+   */
+  async buildDeepContext(documentIds, maxDocs = 3) {
+    if (!documentIds.length) return "";
+    const docRepo = this.db.getRepository("docpixie_documents");
+    const ids = documentIds.slice(0, maxDocs);
+    const parts = [];
+    for (const id of ids) {
+      try {
+        const docRecord = await docRepo.findOne({ filter: { id } });
+        if (!docRecord) continue;
+        const status = docRecord.get("status");
+        const name = docRecord.get("name");
+        const summary = docRecord.get("summary");
+        if (status !== "ready") continue;
+        const pageText = await this.getPageTexts(id);
+        if (!pageText && !summary) continue;
+        const content = pageText || summary || "";
+        parts.push(
+          `<docpixie_document id="${id}" name="${name}">
+${content}
+</docpixie_document>`
+        );
+      } catch {
+      }
+    }
+    return parts.join("\n\n");
+  }
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  DocPixieExtractor
+});

package/package.json

@@ -6,7 +6,7 @@
   "description": "Provides Knowledge Base management, Vector Store, Vector Database (PGVector), and RAG retrieval capabilities for AI Employees.",
   "description.vi-VN": "Cung cấp quản lý Cơ sở tri thức, Vector Store, Vector Database (PGVector), và khả năng truy xuất RAG cho nhân viên AI.",
   "description.zh-CN": "为 AI 员工提供知识库管理、向量存储、向量数据库 (PGVector) 和 RAG 检索功能。",
-  "version": "1.1.2",
+  "version": "1.1.4",
   "license": "Apache-2.0",
   "main": "./dist/server/index.js",
   "files": [
@@ -18,18 +18,20 @@
     "README.md"
   ],
   "dependencies": {
-    "pg": "^8.13.0"
+    "pg": "^8.13.0",
+    "@langchain/community": "^1.1.0",
+    "@langchain/core": "^1.1.24",
+    "@langchain/textsplitters": "^0.1.0"
   },
   "peerDependencies": {
     "@nocobase/client": "2.x",
     "@nocobase/server": "2.x",
     "@nocobase/database": "2.x",
     "@nocobase/plugin-ai": "2.x",
-    "@nocobase/plugin-file-manager": "2.x",
-    "@nocobase/test": "2.x",
-    "@langchain/community": "^1.1.0",
-    "@langchain/core": "^1.1.24",
-    "@langchain/textsplitters": "^0.1.0"
+    "@nocobase/plugin-file-manager": "2.x"
+  },
+  "devDependencies": {
+    "@nocobase/test": "2.x"
   },
   "keywords": [
     "AI",

package/README.md

@@ -1,35 +0,0 @@
-# plugin-knowledge-base
-
-AI Knowledge Base plugin for NocoBase. Provides RAG (Retrieval Augmented Generation) capabilities for AI Employees.
-
-## Features
-
-- **Knowledge Base Management** — Create and manage knowledge bases with document upload
-- **Access Control (ACL)** — 4 access levels: Private, Basic, Shared (role-based), Public
-  - **Shared mode**: `allowedRoles` (read/search) and `uploadRoles` (upload documents) per KB
-  - Permission enforcement on all operations: list, create, destroy, reprocess, RAG search
-- **Vector Store** — Bind Embedding models with Vector Databases
-- **Vector Database** — PGVector connection management with connection testing
-- **Document Processing** — Automatic document parsing, chunking, and vectorization
-  - File upload (.pdf, .txt, .md, .doc, .docx, .ppt, .csv, .json)
-  - Paste text documents directly
-  - Auto-delete source file option after successful embedding
-- **RAG Retrieval** — Semantic search for AI Employee conversations via Work Context
-- **KB Management Popup in Chat** — Users can browse, upload, and select KBs directly from AI chat
-  - Left sidebar with KB list and access level badges
-  - Document table with status, chunk count, and actions
-  - File upload and drag-and-drop support
-  - "Select for Chat" toggle for RAG context
-
-## Requirements
-
-- NocoBase 2.x
-- Plugin AI (`@nocobase/plugin-ai`) enabled
-- PostgreSQL with `pgvector` extension (for PGVector provider)
-
-## Installation
-
-```bash
-yarn pm add plugin-knowledge-base
-yarn pm enable plugin-knowledge-base
-```