feat(server): add read doc tool (#12811)

close AI-186















#### PR Dependency Tree


* **PR #12811** 👈

This tree was auto-generated by
[Charcoal](https://github.com/danerwilliams/charcoal)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

- **New Features**
- Added a new tool enabling users to read document content and metadata
within a workspace, with enforced access control.
- **Improvements**
- Updated tool interfaces and outputs to support the document reading
functionality seamlessly.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

packages/backend/server/src/plugins/copilot/providers/provider.ts

@@ -11,14 +11,17 @@ import {
 } from '../../../base';
 import { DocReader } from '../../../core/doc';
 import { AccessController } from '../../../core/permission';
+import { Models } from '../../../models';
 import { IndexerService } from '../../indexer';
 import { CopilotContextService } from '../context';
 import {
   buildContentGetter,
+  buildDocContentGetter,
   buildDocKeywordSearchGetter,
   buildDocSearchGetter,
   createDocEditTool,
   createDocKeywordSearchTool,
+  createDocReadTool,
   createDocSemanticSearchTool,
   createExaCrawlTool,
   createExaSearchTool,
@@ -182,6 +185,14 @@ export abstract class CopilotProvider<C = any> {
             }
             break;
           }
+          case 'docRead': {
+            const ac = this.moduleRef.get(AccessController, { strict: false });
+            const models = this.moduleRef.get(Models, { strict: false });
+            const docReader = this.moduleRef.get(DocReader, { strict: false });
+            const getDoc = buildDocContentGetter(ac, docReader, models);
+            tools.doc_read = createDocReadTool(getDoc.bind(null, options));
+            break;
+          }
           case 'webSearch': {
             tools.web_search_exa = createExaSearchTool(this.AFFiNEConfig);
             tools.web_crawl_exa = createExaCrawlTool(this.AFFiNEConfig);

packages/backend/server/src/plugins/copilot/providers/utils.ts

@@ -13,6 +13,7 @@ import { ZodType } from 'zod';
 import {
   createDocEditTool,
   createDocKeywordSearchTool,
+  createDocReadTool,
   createDocSemanticSearchTool,
   createExaCrawlTool,
   createExaSearchTool,
@@ -386,6 +387,7 @@ export interface CustomAITools extends ToolSet {
   doc_edit: ReturnType<typeof createDocEditTool>;
   doc_semantic_search: ReturnType<typeof createDocSemanticSearchTool>;
   doc_keyword_search: ReturnType<typeof createDocKeywordSearchTool>;
+  doc_read: ReturnType<typeof createDocReadTool>;
   web_search_exa: ReturnType<typeof createExaSearchTool>;
   web_crawl_exa: ReturnType<typeof createExaCrawlTool>;
 }
@@ -451,6 +453,10 @@ export class TextStreamParser {
             result += `\nSearching the keyword "${chunk.args.query}"\n`;
             break;
           }
+          case 'doc_read': {
+            result += `\nReading the doc "${chunk.args.doc_id}"\n`;
+            break;
+          }
         }
         result = this.markAsCallout(result);
         break;

packages/backend/server/src/plugins/copilot/tools/doc-read.ts

@@ -0,0 +1,88 @@
+import { Logger } from '@nestjs/common';
+import { tool } from 'ai';
+import { z } from 'zod';
+
+import { DocReader } from '../../../core/doc';
+import { AccessController } from '../../../core/permission';
+import { Models, publicUserSelect } from '../../../models';
+import type { CopilotChatOptions } from '../providers';
+import { toolError } from './error';
+
+const logger = new Logger('DocReadTool');
+
+export const buildDocContentGetter = (
+  ac: AccessController,
+  docReader: DocReader,
+  models: Models
+) => {
+  const getDoc = async (options: CopilotChatOptions, docId?: string) => {
+    if (!options?.user || !options?.workspace || !docId) {
+      return;
+    }
+    const canAccess = await ac
+      .user(options.user)
+      .workspace(options.workspace)
+      .doc(docId)
+      .can('Doc.Read');
+    if (!canAccess) {
+      logger.warn(
+        `User ${options.user} does not have access to doc ${docId} in workspace ${options.workspace}`
+      );
+      return;
+    }
+
+    const docMeta = await models.doc.getSnapshot(options.workspace, docId, {
+      select: {
+        createdAt: true,
+        updatedAt: true,
+        createdByUser: {
+          select: publicUserSelect,
+        },
+        updatedByUser: {
+          select: publicUserSelect,
+        },
+      },
+    });
+    if (!docMeta) {
+      return;
+    }
+
+    const content = await docReader.getDocMarkdown(options.workspace, docId);
+    if (!content) {
+      return;
+    }
+
+    return {
+      title: content.title,
+      markdown: content.markdown,
+      createdAt: docMeta.createdAt,
+      updatedAt: docMeta.updatedAt,
+      createdByUser: docMeta.createdByUser,
+      updatedByUser: docMeta.updatedByUser,
+    };
+  };
+  return getDoc;
+};
+
+export const createDocReadTool = (
+  getDoc: (targetId?: string) => Promise<object | undefined>
+) => {
+  return tool({
+    description: 'Read the content of a doc in the current workspace',
+    parameters: z.object({
+      doc_id: z.string().describe('The target doc to read'),
+    }),
+    execute: async ({ doc_id }) => {
+      try {
+        const doc = await getDoc(doc_id);
+        if (!doc) {
+          return;
+        }
+        return { ...doc };
+      } catch (err: any) {
+        logger.error(`Failed to read the doc ${doc_id}`, err);
+        return toolError('Doc Read Failed', err.message);
+      }
+    },
+  });
+};

packages/backend/server/src/plugins/copilot/tools/index.ts

@@ -1,5 +1,6 @@
 export * from './doc-edit';
 export * from './doc-keyword-search';
+export * from './doc-read';
 export * from './doc-semantic-search';
 export * from './error';
 export * from './exa-crawl';