@dpesch/mantisbt-mcp-server 1.5.0 → 1.5.1

package/CHANGELOG.md

@@ -7,6 +7,13 @@ This project adheres to [Semantic Versioning](https://semver.org/).
 
 ---
 
+## [1.5.1] – 2026-03-17
+
+### Fixed
+- Semantic search: ONNX thread pool now defaults to 1 thread (`intra_op_num_threads=1`) instead of auto-detecting all available CPU cores. On WSL and multi-core machines the unrestricted default caused CPU saturation (700%+ CPU, 12 GB VM) during the initial index build. The number of threads is configurable via the new `MANTIS_SEARCH_THREADS` environment variable (default: `1`). `inter_op_num_threads` is always kept at 1 because Transformer model graphs are sequential and inter-op parallelism provides no benefit.
+
+---
+
 ## [1.5.0] – 2026-03-17
 
 ### Added

package/README.de.md

@@ -73,6 +73,7 @@ npm run build
 | `MANTIS_SEARCH_BACKEND` | – | `vectra` | Vektorspeicher: `vectra` (reines JS) oder `sqlite-vec` (manuelle Installation erforderlich) |
 | `MANTIS_SEARCH_DIR` | – | `{MANTIS_CACHE_DIR}/search` | Verzeichnis für den Suchindex |
 | `MANTIS_SEARCH_MODEL` | – | `Xenova/paraphrase-multilingual-MiniLM-L12-v2` | Embedding-Modell (wird beim ersten Start einmalig heruntergeladen, ~80 MB) |
+| `MANTIS_SEARCH_THREADS` | – | `1` | Anzahl der ONNX-Intra-Op-Threads für das Embedding-Modell. Standard ist 1, um CPU-Sättigung auf Mehrkernsystemen und in WSL zu verhindern. Nur erhöhen, wenn die Indexierungsgeschwindigkeit kritisch ist und der Host ausschließlich für diese Last vorgesehen ist. |
 
 ### Config-Datei (Fallback)
 

package/README.md

@@ -73,6 +73,7 @@ npm run build
 | `MANTIS_SEARCH_BACKEND` | – | `vectra` | Vector store backend: `vectra` (pure JS) or `sqlite-vec` (requires manual install) |
 | `MANTIS_SEARCH_DIR` | – | `{MANTIS_CACHE_DIR}/search` | Directory for the search index |
 | `MANTIS_SEARCH_MODEL` | – | `Xenova/paraphrase-multilingual-MiniLM-L12-v2` | Embedding model name (downloaded once on first use, ~80 MB) |
+| `MANTIS_SEARCH_THREADS` | – | `1` | Number of ONNX intra-op threads for the embedding model. Default is 1 to prevent CPU saturation on multi-core machines and WSL. Increase only if index rebuild speed matters and the host is dedicated to this workload. |
 
 ### Config file (fallback)
 

package/dist/config.js

@@ -76,6 +76,7 @@ export async function getConfig() {
     const searchDir = process.env.MANTIS_SEARCH_DIR ?? join(cacheDir, 'search');
     const searchModelName = process.env.MANTIS_SEARCH_MODEL ??
         'Xenova/paraphrase-multilingual-MiniLM-L12-v2';
+    const searchNumThreads = Math.max(1, parseInt(process.env.MANTIS_SEARCH_THREADS ?? '', 10) || 1);
     cachedConfig = {
         baseUrl: baseUrl.replace(/\/$/, ''), // strip trailing slash
         apiKey,
@@ -86,6 +87,7 @@ export async function getConfig() {
             backend: searchBackend,
             dir: searchDir,
             modelName: searchModelName,
+            numThreads: searchNumThreads,
         },
     };
     return cachedConfig;

package/dist/search/embedder.js

@@ -8,14 +8,16 @@
 // ---------------------------------------------------------------------------
 export class Embedder {
     modelName;
+    numThreads;
     pipe = null;
-    constructor(modelName) {
+    constructor(modelName, numThreads = 1) {
         this.modelName = modelName;
+        this.numThreads = numThreads;
     }
     async load() {
         if (this.pipe)
             return this.pipe;
-        process.stderr.write(`[mantisbt-search] Loading embedding model ${this.modelName}...\n`);
+        process.stderr.write(`[mantisbt-search] Loading embedding model ${this.modelName} (threads: ${this.numThreads})...\n`);
         let transformers;
         try {
             // eslint-disable-next-line @typescript-eslint/no-unsafe-assignment
@@ -25,7 +27,12 @@ export class Embedder {
             const msg = err instanceof Error ? err.message : String(err);
             throw new Error(`Failed to load @huggingface/transformers: ${msg}`);
         }
-        this.pipe = await transformers.pipeline('feature-extraction', this.modelName);
+        this.pipe = await transformers.pipeline('feature-extraction', this.modelName, {
+            session_options: {
+                intra_op_num_threads: this.numThreads,
+                inter_op_num_threads: 1, // Transformer graphs are sequential — no benefit from inter-op parallelism
+            },
+        });
         return this.pipe;
     }
     async embed(text) {

package/dist/search/index.js

@@ -9,7 +9,7 @@ export async function initializeSearchModule(server, client, config) {
     if (!config.enabled)
         return;
     const store = createVectorStore(config.backend, config.dir);
-    const embedder = new Embedder(config.modelName);
+    const embedder = new Embedder(config.modelName, config.numThreads);
     registerSearchTools(server, client, store, embedder);
     // Pre-initialize lastKnownTotal so get_search_index_status shows a value
     // immediately on startup, even while the background sync is still running.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@dpesch/mantisbt-mcp-server",
-  "version": "1.5.0",
+  "version": "1.5.1",
   "description": "MCP server for MantisBT REST API – read and manage bug tracker issues",
   "author": "Dominik Pesch",
   "license": "MIT",

package/tests/search/embedder.test.ts

@@ -0,0 +1,81 @@
+import { describe, it, expect, vi, beforeEach } from 'vitest';
+import { Embedder } from '../../src/search/embedder.js';
+
+// ---------------------------------------------------------------------------
+// Mock @huggingface/transformers (dynamic import)
+// ---------------------------------------------------------------------------
+
+const mockPipelineFn = vi.fn(async (texts: string | string[]) => {
+  if (Array.isArray(texts)) {
+    return texts.map(() => ({ data: new Float32Array(4).fill(0.1), dims: [1, 4] }));
+  }
+  return { data: new Float32Array(4).fill(0.1), dims: [4] };
+});
+
+const mockPipelineFactory = vi.fn(async (_task: string, _model: string, _opts?: unknown) => mockPipelineFn);
+
+vi.mock('@huggingface/transformers', () => ({
+  pipeline: mockPipelineFactory,
+}));
+
+beforeEach(() => {
+  vi.clearAllMocks();
+});
+
+// ---------------------------------------------------------------------------
+// Thread configuration
+// ---------------------------------------------------------------------------
+
+describe('Embedder – thread configuration', () => {
+  it('passes intra_op_num_threads=1 by default', async () => {
+    const embedder = new Embedder('test-model');
+    await embedder.embed('hello');
+
+    expect(mockPipelineFactory).toHaveBeenCalledWith(
+      'feature-extraction',
+      'test-model',
+      expect.objectContaining({
+        session_options: { intra_op_num_threads: 1, inter_op_num_threads: 1 },
+      }),
+    );
+  });
+
+  it('passes configured numThreads to intra_op_num_threads; inter stays 1', async () => {
+    const embedder = new Embedder('test-model', 4);
+    await embedder.embed('hello');
+
+    expect(mockPipelineFactory).toHaveBeenCalledWith(
+      'feature-extraction',
+      'test-model',
+      expect.objectContaining({
+        session_options: { intra_op_num_threads: 4, inter_op_num_threads: 1 },
+      }),
+    );
+  });
+
+  it('loads the pipeline only once (lazy singleton)', async () => {
+    const embedder = new Embedder('test-model', 1);
+    await embedder.embed('first');
+    await embedder.embed('second');
+
+    expect(mockPipelineFactory).toHaveBeenCalledTimes(1);
+  });
+});
+
+// ---------------------------------------------------------------------------
+// Embedder default — numThreads omitted
+// ---------------------------------------------------------------------------
+
+describe('Embedder – numThreads default', () => {
+  it('uses intra_op_num_threads=1 when numThreads is not passed', async () => {
+    const embedder = new Embedder('m');
+    await embedder.embed('x');
+    expect(mockPipelineFactory).toHaveBeenCalledWith(
+      'feature-extraction',
+      'm',
+      expect.objectContaining({
+        session_options: expect.objectContaining({ intra_op_num_threads: 1 }),
+      }),
+    );
+  });
+});