@softerist/heuristic-mcp 3.2.3 → 3.2.5

package/lib/embedding-process.js

@@ -1,451 +1,434 @@
-import { pipeline, env } from '@huggingface/transformers';
-import { configureNativeOnnxBackend } from './onnx-backend.js';
-import {
-  EMBEDDING_PROCESS_DEFAULT_GC_MAX_REQUESTS_WITHOUT_COLLECTION,
-  EMBEDDING_PROCESS_DEFAULT_GC_MIN_INTERVAL_MS,
-  EMBEDDING_PROCESS_DEFAULT_GC_RSS_THRESHOLD_MB,
-  EMBEDDING_PROCESS_GC_STATE_INITIAL,
-} from './constants.js';
-import readline from 'readline';
-import path from 'path';
-import os from 'os';
-import { pathToFileURL } from 'url';
-
-
-let currentRequestId = -1;
-const log = (...args) => {
-  if (currentRequestId > 0 && !process.env.EMBEDDING_PROCESS_VERBOSE) {
-    return;
-  }
-  console.error(...args);
-};
-
-function formatMemory() {
-  const usage = process.memoryUsage();
-  return `rss=${(usage.rss / 1024 / 1024).toFixed(1)}MB heap=${(usage.heapUsed / 1024 / 1024).toFixed(1)}MB`;
-}
-
-function readStdin() {
-  return new Promise((resolve, reject) => {
-    let data = '';
-    process.stdin.setEncoding('utf8');
-    process.stdin.on('data', (chunk) => {
-      data += chunk;
-    });
-    process.stdin.on('end', () => resolve(data));
-    process.stdin.on('error', reject);
-  });
-}
-
-const persistent = process.env.EMBEDDING_PROCESS_PERSISTENT === 'true';
-let embedderPromise = null;
-let configuredThreads = null;
-let configuredModel = null;
-let requestCounter = 0;
-let gcSupported = typeof global.gc === 'function';
-let nativeBackendConfigured = false;
-const gcState = { ...EMBEDDING_PROCESS_GC_STATE_INITIAL };
-
-function getGlobalCacheDir() {
-  if (process.platform === 'win32') {
-    return process.env.LOCALAPPDATA || path.join(os.homedir(), 'AppData', 'Local');
-  }
-  if (process.platform === 'darwin') {
-    return path.join(os.homedir(), 'Library', 'Caches');
-  }
-  return process.env.XDG_CACHE_HOME || path.join(os.homedir(), '.cache');
-}
-
-function toPositiveNumber(value, fallback) {
-  const parsed = Number(value);
-  return Number.isFinite(parsed) && parsed > 0 ? parsed : fallback;
-}
-
-function toNonNegativeInteger(value, fallback) {
-  const parsed = Number.parseInt(value, 10);
-  return Number.isInteger(parsed) && parsed >= 0 ? parsed : fallback;
-}
-
-function toPositiveInteger(value, fallback) {
-  const parsed = Number.parseInt(value, 10);
-  return Number.isInteger(parsed) && parsed > 0 ? parsed : fallback;
-}
-
-function resolveGcPolicy(payload) {
-  return {
-    rssThresholdMb: toPositiveNumber(
-      payload?.gcRssThresholdMb,
-      EMBEDDING_PROCESS_DEFAULT_GC_RSS_THRESHOLD_MB
-    ),
-    minIntervalMs: toNonNegativeInteger(
-      payload?.gcMinIntervalMs,
-      EMBEDDING_PROCESS_DEFAULT_GC_MIN_INTERVAL_MS
-    ),
-    maxRequestsWithoutCollection: toPositiveInteger(
-      payload?.gcMaxRequestsWithoutCollection,
-      EMBEDDING_PROCESS_DEFAULT_GC_MAX_REQUESTS_WITHOUT_COLLECTION
-    ),
-  };
-}
-
-function maybeRunGc(policy, { reason = 'unknown', force = false } = {}) {
-  if (!gcSupported) return false;
-
-  const before = process.memoryUsage();
-  const rssBeforeMb = before.rss / 1024 / 1024;
-  const rssTrigger = rssBeforeMb >= policy.rssThresholdMb;
-  const requestTrigger = gcState.requestsSinceLastRun >= policy.maxRequestsWithoutCollection;
-
-  if (!force && !rssTrigger && !requestTrigger) {
-    return false;
-  }
-
-  const now = Date.now();
-  if (
-    !force &&
-    policy.minIntervalMs > 0 &&
-    gcState.lastRunAtMs > 0 &&
-    now - gcState.lastRunAtMs < policy.minIntervalMs
-  ) {
-    return false;
-  }
-
-  global.gc();
-  const after = process.memoryUsage();
-  gcState.lastRunAtMs = now;
-  gcState.requestsSinceLastRun = 0;
-
-  let trigger = 'forced';
-  if (!force) {
-    if (rssTrigger && requestTrigger) trigger = 'rss+requests';
-    else if (rssTrigger) trigger = 'rss';
-    else trigger = 'requests';
-  }
-
-  log(
-    `[Child:${process.pid}] GC ${reason}: trigger=${trigger} rss ${(before.rss / 1024 / 1024).toFixed(1)}MB -> ${(after.rss / 1024 / 1024).toFixed(1)}MB`
-  );
-  return true;
-}
-
-function ensureNativeBackend(threads) {
-  if (nativeBackendConfigured && !threads) return;
-  configureNativeOnnxBackend({
-    log,
-    label: '[Child]',
-    threads,
-  });
-  nativeBackendConfigured = true;
-}
-
-function setThreads(numThreads) {
-  ensureNativeBackend({
-    intraOpNumThreads: numThreads,
-    interOpNumThreads: 1,
-  });
-  configuredThreads = numThreads;
-}
-
-async function getEmbedder(embeddingModel, numThreads) {
-  if (!embedderPromise) {
-    configuredModel = embeddingModel;
-    setThreads(numThreads);
-    env.cacheDir = path.join(getGlobalCacheDir(), 'xenova');
-    log(`Loading model ${embeddingModel}...`);
-    const loadStart = Date.now();
-    embedderPromise = pipeline('feature-extraction', embeddingModel, {
-      quantized: true,
-      dtype: 'fp32',
-      session_options: {
-        numThreads,
-        intraOpNumThreads: numThreads,
-        interOpNumThreads: 1,
-      },
-    }).then((model) => {
-      const loadSec = ((Date.now() - loadStart) / 1000).toFixed(1);
-      log(`Model ready in ${loadSec}s, ${formatMemory()}`);
-      return model;
-    });
-  } else if (configuredModel && embeddingModel !== configuredModel) {
-    log(`Model changed (${configuredModel} -> ${embeddingModel}); reloading embedder`);
-    embedderPromise = null;
-    configuredModel = null;
-    return getEmbedder(embeddingModel, numThreads);
-  } else if (configuredThreads !== null && numThreads !== configuredThreads) {
-    log(`Warning: numThreads changed (${configuredThreads} -> ${numThreads})`);
-  }
-
-  return embedderPromise;
-}
-
-function resetEmbeddingProcessState() {
-  embedderPromise = null;
-  configuredThreads = null;
-  configuredModel = null;
-  requestCounter = 0;
-  currentRequestId = -1;
-  nativeBackendConfigured = false;
-  gcState.lastRunAtMs = 0;
-  gcState.requestsSinceLastRun = 0;
-}
-
-
-async function unloadModel() {
-  if (!embedderPromise) {
-    log('[Child] No model loaded, nothing to unload');
-    return { success: true, wasLoaded: false };
-  }
-
-  try {
-    const embedder = await embedderPromise;
-    
-    
-    if (embedder && typeof embedder.dispose === 'function') {
-      try {
-        await embedder.dispose();
-        log('[Child] Model disposed successfully');
-      } catch (disposeErr) {
-        log(`[Child] Model dispose warning: ${disposeErr.message}`);
-      }
-    }
-  } catch (err) {
-    log(`[Child] Error during model unload: ${err.message}`);
-  }
-
-  
-  embedderPromise = null;
-  configuredModel = null;
-  configuredThreads = null;
-
-  
-  if (gcSupported) {
-    maybeRunGc(resolveGcPolicy(), { reason: 'post-unload', force: true });
-  }
-
-  log(`[Child] Model unloaded, ${formatMemory()}`);
-  return { success: true, wasLoaded: true };
-}
-
-async function runEmbedding(payload) {
-  const {
-    embeddingModel,
-    chunks = [],
-    numThreads = 1,
-    batchSize = null,
-    enableExplicitGc = true,
-    gcRssThresholdMb = EMBEDDING_PROCESS_DEFAULT_GC_RSS_THRESHOLD_MB,
-    gcMinIntervalMs = EMBEDDING_PROCESS_DEFAULT_GC_MIN_INTERVAL_MS,
-    gcMaxRequestsWithoutCollection = EMBEDDING_PROCESS_DEFAULT_GC_MAX_REQUESTS_WITHOUT_COLLECTION,
-    requestId = null,
-  } = payload || {};
-  const shouldRunGc = enableExplicitGc !== false && gcSupported;
-  const gcPolicy = resolveGcPolicy({
-    gcRssThresholdMb,
-    gcMinIntervalMs,
-    gcMaxRequestsWithoutCollection,
-  });
-
-  if (!embeddingModel) {
-    throw new Error('Missing embeddingModel');
-  }
-
-  const reqId = requestId ?? requestCounter++;
-  currentRequestId = reqId;
-  const embedder = await getEmbedder(embeddingModel, numThreads);
-  log(`Request ${reqId}: embedding ${chunks.length} chunks, ${formatMemory()}`);
-
-  const results = [];
-  let disposeCount = 0;
-  const start = Date.now();
-  if (shouldRunGc) {
-    gcState.requestsSinceLastRun += 1;
-  }
-
-  
-  
-  const BATCH_SIZE =
-    Number.isInteger(batchSize) && batchSize > 0 ? Math.min(batchSize, 256) : 1;
-
-  for (let batchStart = 0; batchStart < chunks.length; batchStart += BATCH_SIZE) {
-    const batchEnd = Math.min(batchStart + BATCH_SIZE, chunks.length);
-    const batchChunks = chunks.slice(batchStart, batchEnd);
-    const batchTexts = batchChunks.map((c) => c.text);
-
-    try {
-      
-      const output = await embedder(batchTexts, { pooling: 'mean', normalize: true });
-
-      
-      const hiddenSize = output.dims[output.dims.length - 1];
-
-      for (let j = 0; j < batchChunks.length; j++) {
-        const chunk = batchChunks[j];
-        const vecStart = j * hiddenSize;
-        const vecEnd = vecStart + hiddenSize;
-        
-        const vector = new Float32Array(output.data.subarray(vecStart, vecEnd));
-
-        results.push({
-          file: chunk.file,
-          startLine: chunk.startLine,
-          endLine: chunk.endLine,
-          content: chunk.text,
-          vector: Array.from(vector),
-          success: true,
-        });
-      }
-
-      
-      if (typeof output.dispose === 'function') {
-        try {
-          output.dispose();
-        } catch {
-          
-        }
-      }
-      disposeCount++;
-    } catch (error) {
-      
-      log(`Batch failed, falling back to single: ${error.message}`);
-      for (const chunk of batchChunks) {
-        try {
-          const output = await embedder(chunk.text, { pooling: 'mean', normalize: true });
-          const vector = new Float32Array(output.data);
-          if (typeof output.dispose === 'function') {
-            try {
-              output.dispose();
-            } catch {
-              
-            }
-          }
-          disposeCount++;
-          results.push({
-            file: chunk.file,
-            startLine: chunk.startLine,
-            endLine: chunk.endLine,
-            content: chunk.text,
-            vector: Array.from(vector),
-            success: true,
-          });
-        } catch (innerErr) {
-          results.push({
-            file: chunk.file,
-            startLine: chunk.startLine,
-            endLine: chunk.endLine,
-            error: innerErr.message,
-            success: false,
-          });
-        }
-      }
-    }
-
-    
-    if (batchEnd % 20 === 0 || batchEnd === chunks.length) {
-      const elapsed = ((Date.now() - start) / 1000).toFixed(1);
-      log(
-        `[Child:${process.pid}] Request ${reqId}: processed ${batchEnd}/${chunks.length} chunks in ${elapsed}s, ${formatMemory()}`
-      );
-    }
-
-    if (shouldRunGc && (batchEnd % 20 === 0 || batchEnd === chunks.length)) {
-      maybeRunGc(gcPolicy, {
-        reason: `request ${reqId} progress ${batchEnd}/${chunks.length}`,
-      });
-    }
-  }
-
-  const totalSec = ((Date.now() - start) / 1000).toFixed(1);
-  log(
-    `[Child:${process.pid}] Request ${reqId}: done ${results.length} chunks in ${totalSec}s, ${disposeCount} tensors disposed, ${formatMemory()}`
-  );
-  if (shouldRunGc) {
-    maybeRunGc(gcPolicy, { reason: `request ${reqId} end` });
-  }
-  const usage = process.memoryUsage();
-  return {
-    results,
-    meta: {
-      rssMb: usage.rss / 1024 / 1024,
-      heapMb: usage.heapUsed / 1024 / 1024,
-      heapTotalMb: usage.heapTotal / 1024 / 1024,
-    },
-  };
-}
-
-async function main() {
-  log(`[Child:${process.pid}] Starting, ${formatMemory()}`);
-
-  if (persistent) {
-    const rl = readline.createInterface({ input: process.stdin, crlfDelay: Infinity });
-    let chain = Promise.resolve();
-
-    rl.on('line', (line) => {
-      const trimmed = line.trim();
-      if (!trimmed) return;
-
-      let payload;
-      try {
-        payload = JSON.parse(trimmed);
-      } catch (err) {
-        log(`[Child:${process.pid}] Failed to parse payload: ${err.message}`);
-        process.stdout.write(`${JSON.stringify({ results: [] })}\n`);
-        return;
-      }
-
-      if (payload?.type === 'shutdown') {
-        rl.close();
-        process.exit(0);
-        return;
-      }
-
-      if (payload?.type === 'unload') {
-        chain = chain
-          .then(() => unloadModel())
-          .then((result) => {
-            process.stdout.write(`${JSON.stringify(result)}\n`);
-          })
-          .catch((err) => {
-            log(`[Child:${process.pid}] Error unloading model: ${err.message}`);
-            process.stdout.write(`${JSON.stringify({ success: false, error: err.message })}\n`);
-          });
-        return;
-      }
-
-      chain = chain
-        .then(() => runEmbedding(payload))
-        .then((output) => {
-          process.stdout.write(`${JSON.stringify(output)}\n`);
-        })
-        .catch((err) => {
-          log(`[Child:${process.pid}] Error processing payload: ${err.message}`);
-          process.stdout.write(`${JSON.stringify({ results: [] })}\n`);
-        });
-    });
-    return;
-  }
-
-  const raw = await readStdin();
-  if (!raw) return;
-
-  const payload = JSON.parse(raw);
-  const output = await runEmbedding(payload);
-  process.stdout.write(JSON.stringify(output));
-}
-
-function shouldRunMain() {
-  if (process.env.EMBEDDING_PROCESS_RUN_MAIN === 'true') return true;
-  if (process.env.VITEST) return false;
-  if (!process.argv[1]) return false;
-  const entryUrl = pathToFileURL(process.argv[1]).href;
-  return import.meta.url === entryUrl;
-}
-
-if (shouldRunMain()) {
-  main().catch((err) => {
-    log(`[Child:${process.pid}] Error: ${err?.message || err}`);
-    process.stderr.write(String(err?.message || err));
-    process.exit(1);
-  });
-}
-
-export { getEmbedder, resetEmbeddingProcessState, unloadModel };
+import { pipeline, env } from '@huggingface/transformers';
+import { configureNativeOnnxBackend } from './onnx-backend.js';
+import {
+  EMBEDDING_PROCESS_DEFAULT_GC_MAX_REQUESTS_WITHOUT_COLLECTION,
+  EMBEDDING_PROCESS_DEFAULT_GC_MIN_INTERVAL_MS,
+  EMBEDDING_PROCESS_DEFAULT_GC_RSS_THRESHOLD_MB,
+  EMBEDDING_PROCESS_GC_STATE_INITIAL,
+} from './constants.js';
+import readline from 'readline';
+import path from 'path';
+import os from 'os';
+import { pathToFileURL } from 'url';
+
+let currentRequestId = -1;
+const log = (...args) => {
+  if (currentRequestId > 0 && !process.env.EMBEDDING_PROCESS_VERBOSE) {
+    return;
+  }
+  console.error(...args);
+};
+
+function formatMemory() {
+  const usage = process.memoryUsage();
+  return `rss=${(usage.rss / 1024 / 1024).toFixed(1)}MB heap=${(usage.heapUsed / 1024 / 1024).toFixed(1)}MB`;
+}
+
+function readStdin() {
+  return new Promise((resolve, reject) => {
+    let data = '';
+    process.stdin.setEncoding('utf8');
+    process.stdin.on('data', (chunk) => {
+      data += chunk;
+    });
+    process.stdin.on('end', () => resolve(data));
+    process.stdin.on('error', reject);
+  });
+}
+
+const persistent = process.env.EMBEDDING_PROCESS_PERSISTENT === 'true';
+let embedderPromise = null;
+let configuredThreads = null;
+let configuredModel = null;
+let requestCounter = 0;
+let gcSupported = typeof global.gc === 'function';
+let nativeBackendConfigured = false;
+const gcState = { ...EMBEDDING_PROCESS_GC_STATE_INITIAL };
+
+function getGlobalCacheDir() {
+  if (process.platform === 'win32') {
+    return process.env.LOCALAPPDATA || path.join(os.homedir(), 'AppData', 'Local');
+  }
+  if (process.platform === 'darwin') {
+    return path.join(os.homedir(), 'Library', 'Caches');
+  }
+  return process.env.XDG_CACHE_HOME || path.join(os.homedir(), '.cache');
+}
+
+function toPositiveNumber(value, fallback) {
+  const parsed = Number(value);
+  return Number.isFinite(parsed) && parsed > 0 ? parsed : fallback;
+}
+
+function toNonNegativeInteger(value, fallback) {
+  const parsed = Number.parseInt(value, 10);
+  return Number.isInteger(parsed) && parsed >= 0 ? parsed : fallback;
+}
+
+function toPositiveInteger(value, fallback) {
+  const parsed = Number.parseInt(value, 10);
+  return Number.isInteger(parsed) && parsed > 0 ? parsed : fallback;
+}
+
+function resolveGcPolicy(payload) {
+  return {
+    rssThresholdMb: toPositiveNumber(
+      payload?.gcRssThresholdMb,
+      EMBEDDING_PROCESS_DEFAULT_GC_RSS_THRESHOLD_MB
+    ),
+    minIntervalMs: toNonNegativeInteger(
+      payload?.gcMinIntervalMs,
+      EMBEDDING_PROCESS_DEFAULT_GC_MIN_INTERVAL_MS
+    ),
+    maxRequestsWithoutCollection: toPositiveInteger(
+      payload?.gcMaxRequestsWithoutCollection,
+      EMBEDDING_PROCESS_DEFAULT_GC_MAX_REQUESTS_WITHOUT_COLLECTION
+    ),
+  };
+}
+
+function maybeRunGc(policy, { reason = 'unknown', force = false } = {}) {
+  if (!gcSupported) return false;
+
+  const before = process.memoryUsage();
+  const rssBeforeMb = before.rss / 1024 / 1024;
+  const rssTrigger = rssBeforeMb >= policy.rssThresholdMb;
+  const requestTrigger = gcState.requestsSinceLastRun >= policy.maxRequestsWithoutCollection;
+
+  if (!force && !rssTrigger && !requestTrigger) {
+    return false;
+  }
+
+  const now = Date.now();
+  if (
+    !force &&
+    policy.minIntervalMs > 0 &&
+    gcState.lastRunAtMs > 0 &&
+    now - gcState.lastRunAtMs < policy.minIntervalMs
+  ) {
+    return false;
+  }
+
+  global.gc();
+  const after = process.memoryUsage();
+  gcState.lastRunAtMs = now;
+  gcState.requestsSinceLastRun = 0;
+
+  let trigger = 'forced';
+  if (!force) {
+    if (rssTrigger && requestTrigger) trigger = 'rss+requests';
+    else if (rssTrigger) trigger = 'rss';
+    else trigger = 'requests';
+  }
+
+  log(
+    `[Child:${process.pid}] GC ${reason}: trigger=${trigger} rss ${(before.rss / 1024 / 1024).toFixed(1)}MB -> ${(after.rss / 1024 / 1024).toFixed(1)}MB`
+  );
+  return true;
+}
+
+function ensureNativeBackend(threads) {
+  if (nativeBackendConfigured && !threads) return;
+  configureNativeOnnxBackend({
+    log,
+    label: '[Child]',
+    threads,
+  });
+  nativeBackendConfigured = true;
+}
+
+function setThreads(numThreads) {
+  ensureNativeBackend({
+    intraOpNumThreads: numThreads,
+    interOpNumThreads: 1,
+  });
+  configuredThreads = numThreads;
+}
+
+async function getEmbedder(embeddingModel, numThreads) {
+  if (!embedderPromise) {
+    configuredModel = embeddingModel;
+    setThreads(numThreads);
+    env.cacheDir = path.join(getGlobalCacheDir(), 'xenova');
+    log(`Loading model ${embeddingModel}...`);
+    const loadStart = Date.now();
+    embedderPromise = pipeline('feature-extraction', embeddingModel, {
+      quantized: true,
+      dtype: 'fp32',
+      session_options: {
+        numThreads,
+        intraOpNumThreads: numThreads,
+        interOpNumThreads: 1,
+      },
+    }).then((model) => {
+      const loadSec = ((Date.now() - loadStart) / 1000).toFixed(1);
+      log(`Model ready in ${loadSec}s, ${formatMemory()}`);
+      return model;
+    });
+  } else if (configuredModel && embeddingModel !== configuredModel) {
+    log(`Model changed (${configuredModel} -> ${embeddingModel}); reloading embedder`);
+    embedderPromise = null;
+    configuredModel = null;
+    return getEmbedder(embeddingModel, numThreads);
+  } else if (configuredThreads !== null && numThreads !== configuredThreads) {
+    log(`Warning: numThreads changed (${configuredThreads} -> ${numThreads})`);
+  }
+
+  return embedderPromise;
+}
+
+function resetEmbeddingProcessState() {
+  embedderPromise = null;
+  configuredThreads = null;
+  configuredModel = null;
+  requestCounter = 0;
+  currentRequestId = -1;
+  nativeBackendConfigured = false;
+  gcState.lastRunAtMs = 0;
+  gcState.requestsSinceLastRun = 0;
+}
+
+async function unloadModel() {
+  if (!embedderPromise) {
+    log('[Child] No model loaded, nothing to unload');
+    return { success: true, wasLoaded: false };
+  }
+
+  try {
+    const embedder = await embedderPromise;
+
+    if (embedder && typeof embedder.dispose === 'function') {
+      try {
+        await embedder.dispose();
+        log('[Child] Model disposed successfully');
+      } catch (disposeErr) {
+        log(`[Child] Model dispose warning: ${disposeErr.message}`);
+      }
+    }
+  } catch (err) {
+    log(`[Child] Error during model unload: ${err.message}`);
+  }
+
+  embedderPromise = null;
+  configuredModel = null;
+  configuredThreads = null;
+
+  if (gcSupported) {
+    maybeRunGc(resolveGcPolicy(), { reason: 'post-unload', force: true });
+  }
+
+  log(`[Child] Model unloaded, ${formatMemory()}`);
+  return { success: true, wasLoaded: true };
+}
+
+async function runEmbedding(payload) {
+  const {
+    embeddingModel,
+    chunks = [],
+    numThreads = 1,
+    batchSize = null,
+    enableExplicitGc = true,
+    gcRssThresholdMb = EMBEDDING_PROCESS_DEFAULT_GC_RSS_THRESHOLD_MB,
+    gcMinIntervalMs = EMBEDDING_PROCESS_DEFAULT_GC_MIN_INTERVAL_MS,
+    gcMaxRequestsWithoutCollection = EMBEDDING_PROCESS_DEFAULT_GC_MAX_REQUESTS_WITHOUT_COLLECTION,
+    requestId = null,
+  } = payload || {};
+  const shouldRunGc = enableExplicitGc !== false && gcSupported;
+  const gcPolicy = resolveGcPolicy({
+    gcRssThresholdMb,
+    gcMinIntervalMs,
+    gcMaxRequestsWithoutCollection,
+  });
+
+  if (!embeddingModel) {
+    throw new Error('Missing embeddingModel');
+  }
+
+  const reqId = requestId ?? requestCounter++;
+  currentRequestId = reqId;
+  const embedder = await getEmbedder(embeddingModel, numThreads);
+  log(`Request ${reqId}: embedding ${chunks.length} chunks, ${formatMemory()}`);
+
+  const results = [];
+  let disposeCount = 0;
+  const start = Date.now();
+  if (shouldRunGc) {
+    gcState.requestsSinceLastRun += 1;
+  }
+
+  const BATCH_SIZE = Number.isInteger(batchSize) && batchSize > 0 ? Math.min(batchSize, 256) : 1;
+
+  for (let batchStart = 0; batchStart < chunks.length; batchStart += BATCH_SIZE) {
+    const batchEnd = Math.min(batchStart + BATCH_SIZE, chunks.length);
+    const batchChunks = chunks.slice(batchStart, batchEnd);
+    const batchTexts = batchChunks.map((c) => c.text);
+
+    try {
+      const output = await embedder(batchTexts, { pooling: 'mean', normalize: true });
+
+      const hiddenSize = output.dims[output.dims.length - 1];
+
+      for (let j = 0; j < batchChunks.length; j++) {
+        const chunk = batchChunks[j];
+        const vecStart = j * hiddenSize;
+        const vecEnd = vecStart + hiddenSize;
+
+        const vector = new Float32Array(output.data.subarray(vecStart, vecEnd));
+
+        results.push({
+          file: chunk.file,
+          startLine: chunk.startLine,
+          endLine: chunk.endLine,
+          content: chunk.text,
+          vector: Array.from(vector),
+          success: true,
+        });
+      }
+
+      if (typeof output.dispose === 'function') {
+        try {
+          output.dispose();
+        } catch {}
+      }
+      disposeCount++;
+    } catch (error) {
+      log(`Batch failed, falling back to single: ${error.message}`);
+      for (const chunk of batchChunks) {
+        try {
+          const output = await embedder(chunk.text, { pooling: 'mean', normalize: true });
+          const vector = new Float32Array(output.data);
+          if (typeof output.dispose === 'function') {
+            try {
+              output.dispose();
+            } catch {}
+          }
+          disposeCount++;
+          results.push({
+            file: chunk.file,
+            startLine: chunk.startLine,
+            endLine: chunk.endLine,
+            content: chunk.text,
+            vector: Array.from(vector),
+            success: true,
+          });
+        } catch (innerErr) {
+          results.push({
+            file: chunk.file,
+            startLine: chunk.startLine,
+            endLine: chunk.endLine,
+            error: innerErr.message,
+            success: false,
+          });
+        }
+      }
+    }
+
+    if (batchEnd % 20 === 0 || batchEnd === chunks.length) {
+      const elapsed = ((Date.now() - start) / 1000).toFixed(1);
+      log(
+        `[Child:${process.pid}] Request ${reqId}: processed ${batchEnd}/${chunks.length} chunks in ${elapsed}s, ${formatMemory()}`
+      );
+    }
+
+    if (shouldRunGc && (batchEnd % 20 === 0 || batchEnd === chunks.length)) {
+      maybeRunGc(gcPolicy, {
+        reason: `request ${reqId} progress ${batchEnd}/${chunks.length}`,
+      });
+    }
+  }
+
+  const totalSec = ((Date.now() - start) / 1000).toFixed(1);
+  log(
+    `[Child:${process.pid}] Request ${reqId}: done ${results.length} chunks in ${totalSec}s, ${disposeCount} tensors disposed, ${formatMemory()}`
+  );
+  if (shouldRunGc) {
+    maybeRunGc(gcPolicy, { reason: `request ${reqId} end` });
+  }
+  const usage = process.memoryUsage();
+  return {
+    results,
+    meta: {
+      rssMb: usage.rss / 1024 / 1024,
+      heapMb: usage.heapUsed / 1024 / 1024,
+      heapTotalMb: usage.heapTotal / 1024 / 1024,
+    },
+  };
+}
+
+async function main() {
+  log(`[Child:${process.pid}] Starting, ${formatMemory()}`);
+
+  if (persistent) {
+    const rl = readline.createInterface({ input: process.stdin, crlfDelay: Infinity });
+    let chain = Promise.resolve();
+
+    rl.on('line', (line) => {
+      const trimmed = line.trim();
+      if (!trimmed) return;
+
+      let payload;
+      try {
+        payload = JSON.parse(trimmed);
+      } catch (err) {
+        log(`[Child:${process.pid}] Failed to parse payload: ${err.message}`);
+        process.stdout.write(`${JSON.stringify({ results: [] })}\n`);
+        return;
+      }
+
+      if (payload?.type === 'shutdown') {
+        rl.close();
+        process.exit(0);
+        return;
+      }
+
+      if (payload?.type === 'unload') {
+        chain = chain
+          .then(() => unloadModel())
+          .then((result) => {
+            process.stdout.write(`${JSON.stringify(result)}\n`);
+          })
+          .catch((err) => {
+            log(`[Child:${process.pid}] Error unloading model: ${err.message}`);
+            process.stdout.write(`${JSON.stringify({ success: false, error: err.message })}\n`);
+          });
+        return;
+      }
+
+      chain = chain
+        .then(() => runEmbedding(payload))
+        .then((output) => {
+          process.stdout.write(`${JSON.stringify(output)}\n`);
+        })
+        .catch((err) => {
+          log(`[Child:${process.pid}] Error processing payload: ${err.message}`);
+          process.stdout.write(`${JSON.stringify({ results: [] })}\n`);
+        });
+    });
+    return;
+  }
+
+  const raw = await readStdin();
+  if (!raw) return;
+
+  const payload = JSON.parse(raw);
+  const output = await runEmbedding(payload);
+  process.stdout.write(JSON.stringify(output));
+}
+
+function shouldRunMain() {
+  if (process.env.EMBEDDING_PROCESS_RUN_MAIN === 'true') return true;
+  if (process.env.VITEST) return false;
+  if (!process.argv[1]) return false;
+  const entryUrl = pathToFileURL(process.argv[1]).href;
+  return import.meta.url === entryUrl;
+}
+
+if (shouldRunMain()) {
+  main().catch((err) => {
+    log(`[Child:${process.pid}] Error: ${err?.message || err}`);
+    process.stderr.write(String(err?.message || err));
+    process.exit(1);
+  });
+}
+
+export { getEmbedder, resetEmbeddingProcessState, unloadModel };