@khoinguyen2002/doc-mcp 1.0.2 → 1.0.4

package/dist/mcp-server.js

@@ -2,8 +2,8 @@
 import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
 import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
 import { z } from "zod";
-import { listDriveFiles, readDriveDocument, } from "./tools/driveTools.js";
-import { saveAgentNote, searchKnowledge } from "./tools/knowledgeTools.js";
+import { listDriveFiles, readDriveDocument } from "./tools/driveTools.js";
+import { searchKnowledge } from "./tools/knowledgeTools.js";
 import { config } from "./config.js";
 const DRIVE_FOLDER_ID = config.DOC_MCP_DRIVE_FOLDER_ID;
 if (!DRIVE_FOLDER_ID) {
@@ -12,7 +12,7 @@ if (!DRIVE_FOLDER_ID) {
 }
 const server = new McpServer({
     name: "doc-agent",
-    version: "1.0.0",
+    version: "1.0.4",
 });
 // Register tools
 server.registerTool("list_drive_files", {
@@ -40,29 +40,20 @@ server.registerTool("list_drive_files", {
     };
 });
 server.registerTool("read_drive_document", {
-    description: "Read the content of a specific Google Drive document. The document will also be automatically ingested into vector memory for future semantic search.",
+    description: "Read the content of a specific Google Drive document. You can use the 'offset' parameter (obtained from search_knowledge) to read a specific chunk of text.",
     inputSchema: {
         fileId: z.string().describe("The Google Drive file ID to read"),
+        offset: z
+            .number()
+            .optional()
+            .describe("Starting character index (default: 0)"),
+        limit: z
+            .number()
+            .optional()
+            .describe("Maximum number of characters to return (default: 10000)"),
     },
-}, async ({ fileId }) => {
-    const res = await readDriveDocument(fileId);
-    if (!res.success) {
-        return {
-            content: [{ type: "text", text: `Error: ${res.error}` }],
-            isError: true,
-        };
-    }
-    return {
-        content: [{ type: "text", text: res.content || "No content found." }],
-    };
-});
-server.registerTool("save_agent_note", {
-    description: "Save an agent note, thought, or summary directly into the vector memory.",
-    inputSchema: {
-        content: z.string().describe("The note or knowledge content to store"),
-    },
-}, async ({ content }) => {
-    const res = await saveAgentNote(content);
+}, async ({ fileId, offset, limit }) => {
+    const res = await readDriveDocument(fileId, offset, limit);
     if (!res.success) {
         return {
             content: [{ type: "text", text: `Error: ${res.error}` }],
@@ -70,11 +61,11 @@ server.registerTool("save_agent_note", {
         };
     }
     return {
-        content: [{ type: "text", text: res.message || "Saved successfully" }],
+        content: [{ type: "text", text: JSON.stringify(res.data, null, 2) }],
     };
 });
 server.registerTool("search_knowledge", {
-    description: "Search the folder's vector memory for relevant context or knowledge.",
+    description: "Search the folder's vector memory for relevant context or knowledge. Returns structured JSON array of matching chunks.",
     inputSchema: {
         query: z.string().describe("The search query"),
         topK: z

package/dist/tools/driveTools.d.ts

@@ -16,14 +16,23 @@ export declare function syncSingleDocument(fileId: string, folderId: string): Pr
     driveModifiedTime: string;
     content?: undefined;
 }>;
-export declare function readDriveDocument(fileId: string): Promise<{
+export declare function readDriveDocument(fileId: string, offset?: number, limit?: number): Promise<{
     success: boolean;
-    content: string;
+    data: {
+        content: string;
+        metadata: {
+            totalSize: number;
+            offset: number;
+            limit: number;
+            isTruncated: boolean;
+            warning: string | undefined;
+        };
+    };
     error?: undefined;
 } | {
     success: boolean;
     error: any;
-    content?: undefined;
+    data?: undefined;
 }>;
 export declare function syncFolderState(folderId: string): Promise<{
     success: boolean;

package/dist/tools/driveTools.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"driveTools.d.ts","sourceRoot":"","sources":["../../src/tools/driveTools.ts"],"names":[],"mappings":"AAiCA,wBAAsB,cAAc,CAAC,OAAO,CAAC,EAAE,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM;;;;;;;;GAoC7E;AAED,wBAAsB,kBAAkB,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM;;;;;;;;GA6CxE;AAED,wBAAsB,iBAAiB,CAAC,MAAM,EAAE,MAAM;;;;;;;;GAsCrD;AAED,wBAAsB,eAAe,CAAC,QAAQ,EAAE,MAAM;;;;;;GAqDrD"}
+{"version":3,"file":"driveTools.d.ts","sourceRoot":"","sources":["../../src/tools/driveTools.ts"],"names":[],"mappings":"AAiCA,wBAAsB,cAAc,CAAC,OAAO,CAAC,EAAE,MAAM,EAAE,cAAc,CAAC,EAAE,MAAM;;;;;;;;GAoC7E;AAED,wBAAsB,kBAAkB,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM;;;;;;;;GAoDxE;AAED,wBAAsB,iBAAiB,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,GAAE,MAAU,EAAE,KAAK,GAAE,MAAc;;;;;;;;;;;;;;;;;GAsDhG;AAED,wBAAsB,eAAe,CAAC,QAAQ,EAAE,MAAM;;;;;;GAqDrD"}

package/dist/tools/driveTools.js

@@ -79,19 +79,25 @@ export async function syncSingleDocument(fileId, folderId) {
             chunkOverlap: config.CHUNK_OVERLAP,
         });
         const chunks = await splitter.splitText(content);
+        let currentOffset = 0;
         for (const chunk of chunks) {
+            const offset = content.indexOf(chunk, currentOffset);
+            if (offset !== -1) {
+                currentOffset = offset;
+            }
             await upsertProjectDocument(folderId, chunk, {
                 title: fileInfo.data.name || "Untitled Google Doc",
                 source: "google_drive",
                 file_id: fileId,
                 modified_time: driveModifiedTime,
+                offset: offset !== -1 ? offset : 0,
             });
         }
         return { synced: true, content, driveModifiedTime };
     }
     return { synced: false, driveModifiedTime };
 }
-export async function readDriveDocument(fileId) {
+export async function readDriveDocument(fileId, offset = 0, limit = 10000) {
     const folderId = config.DOC_MCP_DRIVE_FOLDER_ID;
     if (!folderId) {
         return {
@@ -112,15 +118,28 @@ export async function readDriveDocument(fileId) {
             content = typeof res.data === "string" ? res.data : "";
         }
         let finalContent = content;
-        const MAX_CHARS = 10000;
-        if (finalContent && finalContent.length > MAX_CHARS) {
-            finalContent =
-                finalContent.substring(0, MAX_CHARS) +
-                    "\n\n... [Content truncated due to length. The full document has been automatically ingested into Vector Memory. Use search_knowledge to query specific details.]";
+        const totalSize = finalContent ? finalContent.length : 0;
+        if (finalContent) {
+            finalContent = finalContent.substring(offset, offset + limit);
+        }
+        const isTruncated = offset + (finalContent?.length || 0) < totalSize;
+        let warning = undefined;
+        if (isTruncated) {
+            warning = `[WARNING]: This is not the entire document. Content has been truncated from character ${offset} to ${offset + finalContent.length} out of ${totalSize} total characters. Please use 'offset' and 'limit' parameters to read the rest of the document, or use search_knowledge to query specific details.`;
+            finalContent += `\n\n${warning}`;
         }
         return {
             success: true,
-            content: finalContent || "Empty file",
+            data: {
+                content: finalContent || "Empty file",
+                metadata: {
+                    totalSize,
+                    offset,
+                    limit,
+                    isTruncated,
+                    warning,
+                },
+            },
         };
     }
     catch (err) {

package/dist/tools/knowledgeTools.d.ts

@@ -15,5 +15,14 @@ export declare function searchKnowledge(query: string, topK?: number): Promise<{
     success: boolean;
     results: string;
     error?: undefined;
+} | {
+    success: boolean;
+    results: {
+        title: string;
+        fileId: any;
+        offset: any;
+        text: any;
+    }[];
+    error?: undefined;
 }>;
 //# sourceMappingURL=knowledgeTools.d.ts.map

package/dist/tools/knowledgeTools.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"knowledgeTools.d.ts","sourceRoot":"","sources":["../../src/tools/knowledgeTools.ts"],"names":[],"mappings":"AAIA,wBAAsB,aAAa,CAAC,OAAO,EAAE,MAAM;;;;;;;;GAmBlD;AAED,wBAAsB,eAAe,CAAC,KAAK,EAAE,MAAM,EAAE,IAAI,GAAE,MAAU;;;;;;;;GAoCpE"}
+{"version":3,"file":"knowledgeTools.d.ts","sourceRoot":"","sources":["../../src/tools/knowledgeTools.ts"],"names":[],"mappings":"AAIA,wBAAsB,aAAa,CAAC,OAAO,EAAE,MAAM;;;;;;;;GAmBlD;AAED,wBAAsB,eAAe,CAAC,KAAK,EAAE,MAAM,EAAE,IAAI,GAAE,MAAU;;;;;;;;;;;;;;;;;GA0CpE"}

package/dist/tools/knowledgeTools.js

@@ -41,20 +41,27 @@ export async function searchKnowledge(query, topK = 3) {
             success: true,
             results: results.map((r) => {
                 let title = "Unknown Source";
+                let offset = undefined;
                 if (r.metadata) {
                     try {
                         const metaObj = JSON.parse(r.metadata);
                         if (metaObj.title)
                             title = metaObj.title;
+                        if (metaObj.offset !== undefined)
+                            offset = metaObj.offset;
                     }
                     catch (e) { }
                 }
-                return `[File: ${title} | File ID: ${r.file_id || 'N/A'}]\n${r.text}`;
-            }).join("\n\n---\n\n"),
+                return {
+                    title,
+                    fileId: r.file_id || "N/A",
+                    offset,
+                    text: r.text,
+                };
+            }),
         };
     }
     catch (err) {
-        console.error("searchKnowledge outer catch:", err.message);
         return { success: false, error: `Failed to search: ${err.message}` };
     }
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@khoinguyen2002/doc-mcp",
-  "version": "1.0.2",
+  "version": "1.0.4",
   "type": "module",
   "main": "dist/index.js",
   "bin": {

package/src/mcp-server.ts

@@ -2,10 +2,7 @@
 import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
 import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
 import { z } from "zod";
-import {
-  listDriveFiles,
-  readDriveDocument,
-} from "./tools/driveTools.js";
+import { listDriveFiles, readDriveDocument } from "./tools/driveTools.js";
 import { saveAgentNote, searchKnowledge } from "./tools/knowledgeTools.js";
 import { config } from "./config.js";
 
@@ -20,14 +17,15 @@ if (!DRIVE_FOLDER_ID) {
 
 const server = new McpServer({
   name: "doc-agent",
-  version: "1.0.0",
+  version: "1.0.4",
 });
 
 // Register tools
 server.registerTool(
   "list_drive_files",
   {
-    description: "List and search for Google Drive documents and subfolders in a specific folder.",
+    description:
+      "List and search for Google Drive documents and subfolders in a specific folder.",
     inputSchema: {
       keyword: z
         .string()
@@ -36,7 +34,9 @@ server.registerTool(
       targetFolderId: z
         .string()
         .optional()
-        .describe("Optional Google Drive folder ID to list contents from. Defaults to the root knowledge folder."),
+        .describe(
+          "Optional Google Drive folder ID to list contents from. Defaults to the root knowledge folder.",
+        ),
     },
   },
   async ({ keyword, targetFolderId }) => {
@@ -57,13 +57,21 @@ server.registerTool(
   "read_drive_document",
   {
     description:
-      "Read the content of a specific Google Drive document. The document will also be automatically ingested into vector memory for future semantic search.",
+      "Read the content of a specific Google Drive document. You can use the 'offset' parameter (obtained from search_knowledge) to read a specific chunk of text.",
     inputSchema: {
       fileId: z.string().describe("The Google Drive file ID to read"),
+      offset: z
+        .number()
+        .optional()
+        .describe("Starting character index (default: 0)"),
+      limit: z
+        .number()
+        .optional()
+        .describe("Maximum number of characters to return (default: 10000)"),
     },
   },
-  async ({ fileId }) => {
-    const res = await readDriveDocument(fileId);
+  async ({ fileId, offset, limit }) => {
+    const res = await readDriveDocument(fileId, offset, limit);
     if (!res.success) {
       return {
         content: [{ type: "text", text: `Error: ${res.error}` }],
@@ -71,38 +79,18 @@ server.registerTool(
       };
     }
     return {
-      content: [{ type: "text", text: res.content || "No content found." }],
+      content: [{ type: "text", text: JSON.stringify(res.data, null, 2) }],
     };
   },
 );
 
-server.registerTool(
-  "save_agent_note",
-  {
-    description: "Save an agent note, thought, or summary directly into the vector memory.",
-    inputSchema: {
-      content: z.string().describe("The note or knowledge content to store"),
-    },
-  },
-  async ({ content }) => {
-    const res = await saveAgentNote(content);
-    if (!res.success) {
-      return {
-        content: [{ type: "text", text: `Error: ${res.error}` }],
-        isError: true,
-      };
-    }
-    return {
-      content: [{ type: "text", text: res.message || "Saved successfully" }],
-    };
-  },
-);
+
 
 server.registerTool(
   "search_knowledge",
   {
     description:
-      "Search the folder's vector memory for relevant context or knowledge.",
+      "Search the folder's vector memory for relevant context or knowledge. Returns structured JSON array of matching chunks.",
     inputSchema: {
       query: z.string().describe("The search query"),
       topK: z

package/src/tools/driveTools.ts

@@ -102,12 +102,19 @@ export async function syncSingleDocument(fileId: string, folderId: string) {
     });
     const chunks = await splitter.splitText(content);
 
+    let currentOffset = 0;
     for (const chunk of chunks) {
+      const offset = content.indexOf(chunk, currentOffset);
+      if (offset !== -1) {
+        currentOffset = offset;
+      }
+
       await upsertProjectDocument(folderId, chunk, {
         title: fileInfo.data.name || "Untitled Google Doc",
         source: "google_drive",
         file_id: fileId,
         modified_time: driveModifiedTime,
+        offset: offset !== -1 ? offset : 0,
       });
     }
     return { synced: true, content, driveModifiedTime };
@@ -116,7 +123,7 @@ export async function syncSingleDocument(fileId: string, folderId: string) {
   return { synced: false, driveModifiedTime };
 }
 
-export async function readDriveDocument(fileId: string) {
+export async function readDriveDocument(fileId: string, offset: number = 0, limit: number = 10000) {
   const folderId = config.DOC_MCP_DRIVE_FOLDER_ID;
   if (!folderId) {
     return {
@@ -140,16 +147,32 @@ export async function readDriveDocument(fileId: string) {
     }
 
     let finalContent = content;
-    const MAX_CHARS = 10000;
-    if (finalContent && finalContent.length > MAX_CHARS) {
-      finalContent =
-        finalContent.substring(0, MAX_CHARS) +
-        "\n\n... [Content truncated due to length. The full document has been automatically ingested into Vector Memory. Use search_knowledge to query specific details.]";
+    const totalSize = finalContent ? finalContent.length : 0;
+
+    if (finalContent) {
+      finalContent = finalContent.substring(offset, offset + limit);
+    }
+
+    const isTruncated = offset + (finalContent?.length || 0) < totalSize;
+    let warning = undefined;
+
+    if (isTruncated) {
+      warning = `[WARNING]: This is not the entire document. Content has been truncated from character ${offset} to ${offset + finalContent!.length} out of ${totalSize} total characters. Please use 'offset' and 'limit' parameters to read the rest of the document, or use search_knowledge to query specific details.`;
+      finalContent += `\n\n${warning}`;
     }
 
     return {
       success: true,
-      content: finalContent || "Empty file",
+      data: {
+        content: finalContent || "Empty file",
+        metadata: {
+          totalSize,
+          offset,
+          limit,
+          isTruncated,
+          warning,
+        },
+      },
     };
   } catch (err: any) {
     return { success: false, error: err.message };

package/src/tools/knowledgeTools.ts

@@ -44,18 +44,23 @@ export async function searchKnowledge(query: string, topK: number = 3) {
 
     return {
       success: true,
-      results: results
-        .map((r: any) => {
-          let title = "Unknown Source";
-          if (r.metadata) {
-            try {
-              const metaObj = JSON.parse(r.metadata);
-              if (metaObj.title) title = metaObj.title;
-            } catch (e) {}
-          }
-          return `[File: ${title} | File ID: ${r.file_id || "N/A"}]\n${r.text}`;
-        })
-        .join("\n\n---\n\n"),
+      results: results.map((r: any) => {
+        let title = "Unknown Source";
+        let offset = undefined;
+        if (r.metadata) {
+          try {
+            const metaObj = JSON.parse(r.metadata);
+            if (metaObj.title) title = metaObj.title;
+            if (metaObj.offset !== undefined) offset = metaObj.offset;
+          } catch (e) {}
+        }
+        return {
+          title,
+          fileId: r.file_id || "N/A",
+          offset,
+          text: r.text,
+        };
+      }),
     };
   } catch (err: any) {
     return { success: false, error: `Failed to search: ${err.message}` };