@khoinguyen2002/doc-mcp 1.0.3 → 1.0.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -2,8 +2,8 @@
2
2
  import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
3
3
  import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
4
4
  import { z } from "zod";
5
- import { listDriveFiles, readDriveDocument, } from "./tools/driveTools.js";
6
- import { saveAgentNote, searchKnowledge } from "./tools/knowledgeTools.js";
5
+ import { listDriveFiles, readDriveDocument } from "./tools/driveTools.js";
6
+ import { searchKnowledge } from "./tools/knowledgeTools.js";
7
7
  import { config } from "./config.js";
8
8
  const DRIVE_FOLDER_ID = config.DOC_MCP_DRIVE_FOLDER_ID;
9
9
  if (!DRIVE_FOLDER_ID) {
@@ -12,7 +12,7 @@ if (!DRIVE_FOLDER_ID) {
12
12
  }
13
13
  const server = new McpServer({
14
14
  name: "doc-agent",
15
- version: "1.0.0",
15
+ version: "1.0.4",
16
16
  });
17
17
  // Register tools
18
18
  server.registerTool("list_drive_files", {
@@ -40,11 +40,17 @@ server.registerTool("list_drive_files", {
40
40
  };
41
41
  });
42
42
  server.registerTool("read_drive_document", {
43
- description: "Read the content of a specific Google Drive document. The document will also be automatically ingested into vector memory for future semantic search.",
43
+ 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.",
44
44
  inputSchema: {
45
45
  fileId: z.string().describe("The Google Drive file ID to read"),
46
- offset: z.number().optional().describe("Starting character index (default: 0)"),
47
- limit: z.number().optional().describe("Maximum number of characters to return (default: 10000)"),
46
+ offset: z
47
+ .number()
48
+ .optional()
49
+ .describe("Starting character index (default: 0)"),
50
+ limit: z
51
+ .number()
52
+ .optional()
53
+ .describe("Maximum number of characters to return (default: 10000)"),
48
54
  },
49
55
  }, async ({ fileId, offset, limit }) => {
50
56
  const res = await readDriveDocument(fileId, offset, limit);
@@ -58,25 +64,8 @@ server.registerTool("read_drive_document", {
58
64
  content: [{ type: "text", text: JSON.stringify(res.data, null, 2) }],
59
65
  };
60
66
  });
61
- server.registerTool("save_agent_note", {
62
- description: "Save an agent note, thought, or summary directly into the vector memory.",
63
- inputSchema: {
64
- content: z.string().describe("The note or knowledge content to store"),
65
- },
66
- }, async ({ content }) => {
67
- const res = await saveAgentNote(content);
68
- if (!res.success) {
69
- return {
70
- content: [{ type: "text", text: `Error: ${res.error}` }],
71
- isError: true,
72
- };
73
- }
74
- return {
75
- content: [{ type: "text", text: res.message || "Saved successfully" }],
76
- };
77
- });
78
67
  server.registerTool("search_knowledge", {
79
- description: "Search the folder's vector memory for relevant context or knowledge.",
68
+ description: "Search the folder's vector memory for relevant context or knowledge. Returns structured JSON array of matching chunks.",
80
69
  inputSchema: {
81
70
  query: z.string().describe("The search query"),
82
71
  topK: z
@@ -1 +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,EAAE,MAAM,GAAE,MAAU,EAAE,KAAK,GAAE,MAAc;;;;;;;;;;;;;;;;;GAsDhG;AAED,wBAAsB,eAAe,CAAC,QAAQ,EAAE,MAAM;;;;;;GAqDrD"}
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;;;;;;;;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"}
@@ -79,12 +79,18 @@ export async function syncSingleDocument(fileId, folderId) {
79
79
  chunkOverlap: config.CHUNK_OVERLAP,
80
80
  });
81
81
  const chunks = await splitter.splitText(content);
82
+ let currentOffset = 0;
82
83
  for (const chunk of chunks) {
84
+ const offset = content.indexOf(chunk, currentOffset);
85
+ if (offset !== -1) {
86
+ currentOffset = offset;
87
+ }
83
88
  await upsertProjectDocument(folderId, chunk, {
84
89
  title: fileInfo.data.name || "Untitled Google Doc",
85
90
  source: "google_drive",
86
91
  file_id: fileId,
87
92
  modified_time: driveModifiedTime,
93
+ offset: offset !== -1 ? offset : 0,
88
94
  });
89
95
  }
90
96
  return { synced: true, content, driveModifiedTime };
@@ -15,5 +15,14 @@ export declare function searchKnowledge(query: string, topK?: number): Promise<{
15
15
  success: boolean;
16
16
  results: string;
17
17
  error?: undefined;
18
+ } | {
19
+ success: boolean;
20
+ results: {
21
+ title: string;
22
+ fileId: any;
23
+ offset: any;
24
+ text: any;
25
+ }[];
26
+ error?: undefined;
18
27
  }>;
19
28
  //# sourceMappingURL=knowledgeTools.d.ts.map
@@ -1 +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;;;;;;;;GAqCpE"}
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;;;;;;;;;;;;;;;;;GA0CpE"}
@@ -39,20 +39,26 @@ export async function searchKnowledge(query, topK = 3) {
39
39
  }
40
40
  return {
41
41
  success: true,
42
- results: results
43
- .map((r) => {
42
+ results: results.map((r) => {
44
43
  let title = "Unknown Source";
44
+ let offset = undefined;
45
45
  if (r.metadata) {
46
46
  try {
47
47
  const metaObj = JSON.parse(r.metadata);
48
48
  if (metaObj.title)
49
49
  title = metaObj.title;
50
+ if (metaObj.offset !== undefined)
51
+ offset = metaObj.offset;
50
52
  }
51
53
  catch (e) { }
52
54
  }
53
- return `[File: ${title} | File ID: ${r.file_id || "N/A"}]\n${r.text}`;
54
- })
55
- .join("\n\n---\n\n"),
55
+ return {
56
+ title,
57
+ fileId: r.file_id || "N/A",
58
+ offset,
59
+ text: r.text,
60
+ };
61
+ }),
56
62
  };
57
63
  }
58
64
  catch (err) {
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@khoinguyen2002/doc-mcp",
3
- "version": "1.0.3",
3
+ "version": "1.0.4",
4
4
  "type": "module",
5
5
  "main": "dist/index.js",
6
6
  "bin": {
package/src/mcp-server.ts CHANGED
@@ -2,10 +2,7 @@
2
2
  import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
3
3
  import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
4
4
  import { z } from "zod";
5
- import {
6
- listDriveFiles,
7
- readDriveDocument,
8
- } from "./tools/driveTools.js";
5
+ import { listDriveFiles, readDriveDocument } from "./tools/driveTools.js";
9
6
  import { saveAgentNote, searchKnowledge } from "./tools/knowledgeTools.js";
10
7
  import { config } from "./config.js";
11
8
 
@@ -20,14 +17,15 @@ if (!DRIVE_FOLDER_ID) {
20
17
 
21
18
  const server = new McpServer({
22
19
  name: "doc-agent",
23
- version: "1.0.0",
20
+ version: "1.0.4",
24
21
  });
25
22
 
26
23
  // Register tools
27
24
  server.registerTool(
28
25
  "list_drive_files",
29
26
  {
30
- description: "List and search for Google Drive documents and subfolders in a specific folder.",
27
+ description:
28
+ "List and search for Google Drive documents and subfolders in a specific folder.",
31
29
  inputSchema: {
32
30
  keyword: z
33
31
  .string()
@@ -36,7 +34,9 @@ server.registerTool(
36
34
  targetFolderId: z
37
35
  .string()
38
36
  .optional()
39
- .describe("Optional Google Drive folder ID to list contents from. Defaults to the root knowledge folder."),
37
+ .describe(
38
+ "Optional Google Drive folder ID to list contents from. Defaults to the root knowledge folder.",
39
+ ),
40
40
  },
41
41
  },
42
42
  async ({ keyword, targetFolderId }) => {
@@ -57,11 +57,17 @@ server.registerTool(
57
57
  "read_drive_document",
58
58
  {
59
59
  description:
60
- "Read the content of a specific Google Drive document. The document will also be automatically ingested into vector memory for future semantic search.",
60
+ "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.",
61
61
  inputSchema: {
62
62
  fileId: z.string().describe("The Google Drive file ID to read"),
63
- offset: z.number().optional().describe("Starting character index (default: 0)"),
64
- limit: z.number().optional().describe("Maximum number of characters to return (default: 10000)"),
63
+ offset: z
64
+ .number()
65
+ .optional()
66
+ .describe("Starting character index (default: 0)"),
67
+ limit: z
68
+ .number()
69
+ .optional()
70
+ .describe("Maximum number of characters to return (default: 10000)"),
65
71
  },
66
72
  },
67
73
  async ({ fileId, offset, limit }) => {
@@ -78,33 +84,13 @@ server.registerTool(
78
84
  },
79
85
  );
80
86
 
81
- server.registerTool(
82
- "save_agent_note",
83
- {
84
- description: "Save an agent note, thought, or summary directly into the vector memory.",
85
- inputSchema: {
86
- content: z.string().describe("The note or knowledge content to store"),
87
- },
88
- },
89
- async ({ content }) => {
90
- const res = await saveAgentNote(content);
91
- if (!res.success) {
92
- return {
93
- content: [{ type: "text", text: `Error: ${res.error}` }],
94
- isError: true,
95
- };
96
- }
97
- return {
98
- content: [{ type: "text", text: res.message || "Saved successfully" }],
99
- };
100
- },
101
- );
87
+
102
88
 
103
89
  server.registerTool(
104
90
  "search_knowledge",
105
91
  {
106
92
  description:
107
- "Search the folder's vector memory for relevant context or knowledge.",
93
+ "Search the folder's vector memory for relevant context or knowledge. Returns structured JSON array of matching chunks.",
108
94
  inputSchema: {
109
95
  query: z.string().describe("The search query"),
110
96
  topK: z
@@ -102,12 +102,19 @@ export async function syncSingleDocument(fileId: string, folderId: string) {
102
102
  });
103
103
  const chunks = await splitter.splitText(content);
104
104
 
105
+ let currentOffset = 0;
105
106
  for (const chunk of chunks) {
107
+ const offset = content.indexOf(chunk, currentOffset);
108
+ if (offset !== -1) {
109
+ currentOffset = offset;
110
+ }
111
+
106
112
  await upsertProjectDocument(folderId, chunk, {
107
113
  title: fileInfo.data.name || "Untitled Google Doc",
108
114
  source: "google_drive",
109
115
  file_id: fileId,
110
116
  modified_time: driveModifiedTime,
117
+ offset: offset !== -1 ? offset : 0,
111
118
  });
112
119
  }
113
120
  return { synced: true, content, driveModifiedTime };
@@ -44,18 +44,23 @@ export async function searchKnowledge(query: string, topK: number = 3) {
44
44
 
45
45
  return {
46
46
  success: true,
47
- results: results
48
- .map((r: any) => {
49
- let title = "Unknown Source";
50
- if (r.metadata) {
51
- try {
52
- const metaObj = JSON.parse(r.metadata);
53
- if (metaObj.title) title = metaObj.title;
54
- } catch (e) {}
55
- }
56
- return `[File: ${title} | File ID: ${r.file_id || "N/A"}]\n${r.text}`;
57
- })
58
- .join("\n\n---\n\n"),
47
+ results: results.map((r: any) => {
48
+ let title = "Unknown Source";
49
+ let offset = undefined;
50
+ if (r.metadata) {
51
+ try {
52
+ const metaObj = JSON.parse(r.metadata);
53
+ if (metaObj.title) title = metaObj.title;
54
+ if (metaObj.offset !== undefined) offset = metaObj.offset;
55
+ } catch (e) {}
56
+ }
57
+ return {
58
+ title,
59
+ fileId: r.file_id || "N/A",
60
+ offset,
61
+ text: r.text,
62
+ };
63
+ }),
59
64
  };
60
65
  } catch (err: any) {
61
66
  return { success: false, error: `Failed to search: ${err.message}` };