@2kw/ai-mcp-server 8.0.0-dev.2 → 8.0.0-dev.5

package/dist/tools/conversion.js

@@ -89,9 +89,9 @@ export function register(server, client) {
         options: z
             .object({
             outputFormats: z.array(z.string()).optional().describe("Output formats, e.g. ['md', 'text', 'html', 'json']. Only relevant for ocr/vlm pipelines."),
-            ocrEngine: z.string().optional().describe("OCR engine (e.g. 'easyocr', 'tesseract')"),
+            ocrEngine: z.string().optional().describe("OCR engine to read text from scanned pages and images. Omit for the default."),
             ocrLanguages: z.array(z.string()).optional().describe("OCR languages (e.g. ['en', 'de'])"),
-            pdfBackend: z.string().optional().describe("PDF backend (e.g. 'dlparser', 'pypdfium2')"),
+            pdfBackend: z.string().optional().describe("PDF parser to read the document layout. Omit for the default."),
             tableStructure: z.boolean().optional().describe("Enable table structure detection"),
             maxPages: z.number().optional().describe("Maximum number of pages to process"),
             imageExportMode: z.string().optional().describe("Image export mode (e.g. 'placeholder', 'embedded')"),

package/dist/tools/files.js

@@ -6,22 +6,23 @@ import { getMimeType } from "../mime.js";
 const FilePurpose = z.enum(["agent_input", "knowledge"]);
 export function register(server, client) {
     // ── list_files ────────────────────────────────────────────────────────────
-    server.tool("2kw_list_files", "List the organization's files as a page, newest first by default, optionally narrowed by purpose.", {
+    server.tool("2kw_list_files", "List the organization's files, newest first by default, optionally narrowed by purpose. When more files exist, pass the returned last id as after.", {
         purpose: FilePurpose.optional().describe("Narrow the list to one purpose"),
-        page: z.number().optional().default(0).describe("Page number (0-based)"),
-        size: z.number().optional().default(20).describe("Page size"),
-    }, async ({ purpose, page, size }) => {
+        limit: z.number().optional().default(20).describe("How many files to list (1-100)"),
+        after: z.string().optional().describe("List the files after this file id (the previous list's last id)"),
+        order: z.enum(["asc", "desc"]).optional().default("desc").describe("asc = oldest first, desc = newest first"),
+    }, async ({ purpose, limit, after, order }) => {
         try {
             const { data } = await client.GET("/v1/files", {
-                params: { query: { purpose, pageable: { page, size } } },
+                params: { query: { purpose, limit, after, order } },
             });
-            const result = data;
-            const lines = (result.content ?? []).map((f) => `- ${f.filename} (id: ${f.id}, purpose: ${f.purpose}, bytes: ${f.bytes})`);
+            const lines = (data?.data ?? []).map((f) => `- ${f.filename} (id: ${f.id}, purpose: ${f.purpose}, bytes: ${f.bytes})`);
+            const more = data?.has_more ? `\nMore files: pass after=${data.last_id}` : "";
             return {
                 content: [
                     {
                         type: "text",
-                        text: `Files (${result.totalElements} total, page ${(result.number ?? 0) + 1}/${result.totalPages}):\n${lines.join("\n") || "(none)"}`,
+                        text: `Files:\n${lines.join("\n") || "(none)"}${more}`,
                     },
                 ],
             };

package/dist/tools/knowledge.js

@@ -58,7 +58,7 @@ const knowledgeBaseFields = {
     parentChunkSize: z.number().int().min(100).max(8000).optional().describe("Parent chunk size for hierarchical chunking (100-8000), must exceed chunkSize"),
     hybridSearchEnabled: z.boolean().optional().describe("Enable hybrid dense + lexical search"),
     rerankerProviderId: z.string().optional().describe("Provider id used for reranking. builtin is refused: there is no built-in reranker."),
-    textSearchLanguage: z.enum(TEXT_SEARCH_LANGUAGES).optional().describe("Text-search language for the lexical leg (default simple on create). On update, omit to leave the active and any pending language untouched — resending the current value would cancel a reindex in progress; changing it starts one in the background."),
+    textSearchLanguage: z.enum(TEXT_SEARCH_LANGUAGES).optional().describe("Text-search language for the lexical leg (default simple on create). On update, a language other than the active and the pending one starts a reindex in the background; omitting it, or sending the active or pending language back, leaves both untouched."),
 };
 function buildKnowledgeBaseBody(params) {
     const body = {
@@ -77,6 +77,7 @@ function buildKnowledgeBaseBody(params) {
         "hybridSearchEnabled",
         "rerankerProviderId",
         "textSearchLanguage",
+        "cancelTextSearchReindex",
     ]) {
         if (params[key] !== undefined)
             body[key] = params[key];
@@ -161,7 +162,11 @@ export function register(server, client) {
         }
     });
     // ── update_knowledge_base ────────────────────────────────────────────────
-    server.tool("2kw_update_knowledge_base", "Apply the request's fields to a knowledge base and snapshot the result as the next configuration version. name, slug, embeddingProviderId, embeddingModel and embeddingDim are required on every update; embeddingDim cannot actually change — it must match the existing value or the update is rejected.", { knowledgeBaseId: z.string().describe("The knowledge base ID"), ...knowledgeBaseFields }, async ({ knowledgeBaseId, ...params }) => {
+    server.tool("2kw_update_knowledge_base", "Apply the request's fields to a knowledge base and snapshot the result as the next configuration version. name, slug, embeddingProviderId, embeddingModel and embeddingDim are required on every update; embeddingDim cannot actually change — it must match the existing value or the update is rejected.", {
+        knowledgeBaseId: z.string().describe("The knowledge base ID"),
+        ...knowledgeBaseFields,
+        cancelTextSearchReindex: z.boolean().optional().describe("true abandons the text-search reindex in progress and keeps the active language. Applied before textSearchLanguage."),
+    }, async ({ knowledgeBaseId, ...params }) => {
         try {
             const { data } = await client.PATCH("/v1/knowledge-bases/{id}", {
                 params: { path: { id: knowledgeBaseId } },
@@ -338,7 +343,7 @@ export function register(server, client) {
         }
     });
     // ── search_knowledge_base ─────────────────────────────────────────────────
-    server.tool("2kw_search_knowledge_base", "Hybrid dense + lexical retrieval with Reciprocal Rank Fusion over a knowledge base. topK is clamped to 1..100. Not paginated — raise topK instead of asking for a later page. Each hit carries a chunkId that can be passed to 2kw_resolve_citation to re-fetch the passage later.", {
+    server.tool("2kw_search_knowledge_base", "Hybrid semantic and keyword search over a knowledge base. topK is clamped to 1..100. Not paginated — raise topK instead of asking for a later page. Each hit carries a chunkId that can be passed to 2kw_resolve_citation to re-fetch the passage later.", {
         knowledgeBaseId: z.string().describe("The knowledge base ID"),
         query: z.string().min(1).describe("The search query"),
         topK: z.number().int().min(1).max(100).optional().describe("Number of results to return (default backend-defined, clamped 1-100)"),

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@2kw/ai-mcp-server",
-  "version": "8.0.0-dev.2",
+  "version": "8.0.0-dev.5",
   "description": "MCP server for 2kw.ai — EU-hosted AI platform: OpenAI-compatible LLM gateway, schema-driven document extraction, transcription, agents with a knowledge base, and cost observability. 199 tools for Claude Code, Cursor, and Windsurf.",
   "mcpName": "ai.2kw/mcp-server",
   "keywords": [