openclaw-amem 1.4.1 → 1.4.2

package/dist/index.js

@@ -56,6 +56,12 @@ var init_config = __esm({
 function getEmbeddingModel() {
   return process.env.AMEM_EMBED_MODEL?.trim() || DEFAULT_EMBEDDING_MODEL;
 }
+function getEmbeddingPooling() {
+  const explicit = process.env.AMEM_EMBED_POOLING?.trim().toLowerCase();
+  if (explicit === "mean" || explicit === "cls") return explicit;
+  const basename2 = getEmbeddingModel().split("/").pop()?.toLowerCase() ?? "";
+  return CLS_POOLED_MODELS.has(basename2) ? "cls" : "mean";
+}
 async function getExtractor() {
   const wanted = getEmbeddingModel();
   if (extractor && loadedModelName === wanted) return extractor;
@@ -76,21 +82,25 @@ async function getEmbeddingDim() {
   cachedDim = probe.length;
   return cachedDim;
 }
-function meanPoolingNormalize(output, attentionMask) {
+function poolNormalize(output, attentionMask, mode) {
   const seqLen = output.length;
   const dim = output[0].length;
   const pooled = new Array(dim).fill(0);
-  let maskSum = 0;
-  for (let i = 0; i < seqLen; i++) {
-    const m = attentionMask[i];
-    maskSum += m;
+  if (mode === "cls") {
+    for (let j = 0; j < dim; j++) pooled[j] = output[0][j];
+  } else {
+    let maskSum = 0;
+    for (let i = 0; i < seqLen; i++) {
+      const m = attentionMask[i];
+      maskSum += m;
+      for (let j = 0; j < dim; j++) {
+        pooled[j] += output[i][j] * m;
+      }
+    }
     for (let j = 0; j < dim; j++) {
-      pooled[j] += output[i][j] * m;
+      pooled[j] /= Math.max(maskSum, 1e-9);
     }
   }
-  for (let j = 0; j < dim; j++) {
-    pooled[j] /= Math.max(maskSum, 1e-9);
-  }
   let norm = 0;
   for (const v of pooled) norm += v * v;
   norm = Math.sqrt(norm);
@@ -98,7 +108,8 @@ function meanPoolingNormalize(output, attentionMask) {
 }
 async function encode(text) {
   const ext = await getExtractor();
-  const result = await ext(text, { pooling: "mean", normalize: true });
+  const pooling = getEmbeddingPooling();
+  const result = await ext(text, { pooling, normalize: true });
   if (result && result.data) {
     return Array.from(result.data);
   }
@@ -114,7 +125,7 @@ async function encode(text) {
       }
       raw.push(row);
     }
-    return meanPoolingNormalize(raw, new Array(seqLen).fill(1));
+    return poolNormalize(raw, new Array(seqLen).fill(1), pooling);
   }
   throw new Error("Unexpected embedding output shape");
 }
@@ -129,7 +140,7 @@ function cosineSimilarity(a, b) {
   for (let i = 0; i < a.length; i++) dot += a[i] * b[i];
   return dot;
 }
-var pipeline, extractor, loadedModelName, cachedDim, DEFAULT_EMBEDDING_MODEL;
+var pipeline, extractor, loadedModelName, cachedDim, DEFAULT_EMBEDDING_MODEL, CLS_POOLED_MODELS;
 var init_embedding = __esm({
   "../amem-core/src/embedding.ts"() {
     "use strict";
@@ -138,6 +149,19 @@ var init_embedding = __esm({
     loadedModelName = null;
     cachedDim = null;
     DEFAULT_EMBEDDING_MODEL = "Xenova/paraphrase-multilingual-MiniLM-L12-v2";
+    CLS_POOLED_MODELS = /* @__PURE__ */ new Set([
+      "bge-m3",
+      "bge-base-zh-v1.5",
+      "bge-small-zh-v1.5",
+      "bge-base-en-v1.5",
+      "bge-small-en-v1.5",
+      "bge-large-en-v1.5",
+      "gte-multilingual-base",
+      "gte-modernbert-base",
+      "gte-large-en-v1.5",
+      "snowflake-arctic-embed-m",
+      "snowflake-arctic-embed-l"
+    ]);
   }
 });
 
@@ -2395,6 +2419,7 @@ __export(src_exports, {
   generateReviewBatch: () => generateReviewBatch,
   getEmbeddingDim: () => getEmbeddingDim,
   getEmbeddingModel: () => getEmbeddingModel,
+  getEmbeddingPooling: () => getEmbeddingPooling,
   getNote: () => getNote,
   invalidateNote: () => invalidateNote,
   isModelLoaded: () => isModelLoaded,

package/openclaw.plugin.json

@@ -2,7 +2,7 @@
   "id": "openclaw-amem",
   "name": "amem",
   "description": "Catches memories that contradict each other. Notes rewrite themselves as new ones arrive, link into a graph, and stay separated per agent and per person. Runs on a small model. Requires a local Qdrant. 中文走 jieba 分词，提示词有中文版。",
-  "version": "1.4.1",
+  "version": "1.4.2",
   "kind": "memory",
   "openclaw": {
     "compat": {
@@ -39,6 +39,7 @@
           "AMEM_LLM_TIMEOUT",
           "AMEM_CRUD_UPDATE_MIN_SIM",
           "AMEM_EMBED_MODEL",
+          "AMEM_EMBED_POOLING",
           "AMEM_COLLECTION",
           "AMEM_DATA_DIR",
           "AMEM_EVO_COUNTER_PATH",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "openclaw-amem",
-  "version": "1.4.1",
+  "version": "1.4.2",
   "description": "Catches memories that contradict each other. Notes rewrite themselves as new ones arrive, link into a graph, and stay separated per agent and per person. Runs on a small model. Requires a local Qdrant. 中文走 jieba 分词，提示词有中文版。",
   "license": "MIT",
   "main": "dist/index.js",
@@ -16,7 +16,7 @@
     }
   },
   "dependencies": {
-    "@anthropic-ai/sdk": "^0.112.1",
+    "@anthropic-ai/sdk": "^0.112.5",
     "@huggingface/transformers": "^4.2.0",
     "@node-rs/jieba": "^2.0.1",
     "@qdrant/js-client-rest": "^1.18.0",
@@ -28,7 +28,7 @@
     "@types/node": "^26.1.1",
     "@types/uuid": "^11.0.0",
     "eslint": "^10.7.0",
-    "prettier": "^3.9.5",
+    "prettier": "^3.9.6",
     "tsup": "^8.4.0",
     "tsx": "^4.23.1",
     "typescript": "^6.0.3",