node-llama-cpp 3.9.0 → 3.11.0

package/llama/addon/AddonContext.cpp

@@ -587,7 +587,7 @@ Napi::Value AddonContext::DisposeSequence(const Napi::CallbackInfo& info) {
 
     int32_t sequenceId = info[0].As<Napi::Number>().Int32Value();
 
-    bool result = llama_kv_self_seq_rm(ctx, sequenceId, -1, -1);
+    bool result = llama_memory_seq_rm(llama_get_memory(ctx), sequenceId, -1, -1);
 
     if (!result) {
         Napi::Error::New(info.Env(), "Failed to dispose sequence").ThrowAsJavaScriptException();
@@ -606,7 +606,7 @@ Napi::Value AddonContext::RemoveTokenCellsFromSequence(const Napi::CallbackInfo&
     int32_t startPos = info[1].As<Napi::Number>().Int32Value();
     int32_t endPos = info[2].As<Napi::Number>().Int32Value();
 
-    bool result = llama_kv_self_seq_rm(ctx, sequenceId, startPos, endPos);
+    bool result = llama_memory_seq_rm(llama_get_memory(ctx), sequenceId, startPos, endPos);
 
     return Napi::Boolean::New(info.Env(), result);
 }
@@ -621,7 +621,7 @@ Napi::Value AddonContext::ShiftSequenceTokenCells(const Napi::CallbackInfo& info
     int32_t endPos = info[2].As<Napi::Number>().Int32Value();
     int32_t shiftDelta = info[3].As<Napi::Number>().Int32Value();
 
-    llama_kv_self_seq_add(ctx, sequenceId, startPos, endPos, shiftDelta);
+    llama_memory_seq_add(llama_get_memory(ctx), sequenceId, startPos, endPos, shiftDelta);
 
     return info.Env().Undefined();
 }
@@ -634,7 +634,7 @@ Napi::Value AddonContext::GetSequenceKvCacheMinPosition(const Napi::CallbackInfo
     int32_t sequenceId = info[0].As<Napi::Number>().Int32Value();
 
 
-    const auto minPosition = llama_kv_self_seq_pos_min(ctx, sequenceId);
+    const auto minPosition = llama_memory_seq_pos_min(llama_get_memory(ctx), sequenceId);
 
     return Napi::Number::New(info.Env(), minPosition);
 }
@@ -647,7 +647,7 @@ Napi::Value AddonContext::GetSequenceKvCacheMaxPosition(const Napi::CallbackInfo
     int32_t sequenceId = info[0].As<Napi::Number>().Int32Value();
 
 
-    const auto maxPosition = llama_kv_self_seq_pos_max(ctx, sequenceId);
+    const auto maxPosition = llama_memory_seq_pos_max(llama_get_memory(ctx), sequenceId);
 
     return Napi::Number::New(info.Env(), maxPosition);
 }

package/llama/addon/addon.cpp

@@ -196,6 +196,36 @@ Napi::Value addonLoadBackends(const Napi::CallbackInfo& info) {
     return info.Env().Undefined();
 }
 
+Napi::Value addonSetNuma(const Napi::CallbackInfo& info) {
+    const bool numaDisabled = info.Length() == 0
+        ? true
+        : info[0].IsBoolean()
+            ? !info[0].As<Napi::Boolean>().Value()
+            : false;
+
+    if (numaDisabled)
+        return info.Env().Undefined();
+
+    const auto numaType = info[0].IsString()
+        ? info[0].As<Napi::String>().Utf8Value()
+        : "";
+
+    if (numaType == "distribute") {
+        llama_numa_init(GGML_NUMA_STRATEGY_DISTRIBUTE);
+    } else if (numaType == "isolate") {
+        llama_numa_init(GGML_NUMA_STRATEGY_ISOLATE);
+    } else if (numaType == "numactl") {
+        llama_numa_init(GGML_NUMA_STRATEGY_NUMACTL);
+    } else if (numaType == "mirror") {
+        llama_numa_init(GGML_NUMA_STRATEGY_MIRROR);
+    } else {
+        Napi::Error::New(info.Env(), std::string("Invalid NUMA strategy \"") + numaType + "\"").ThrowAsJavaScriptException();
+        return info.Env().Undefined();
+    }
+
+    return info.Env().Undefined();
+}
+
 Napi::Value addonInit(const Napi::CallbackInfo& info) {
     if (backendInitialized) {
         Napi::Promise::Deferred deferred = Napi::Promise::Deferred::New(info.Env());
@@ -255,6 +285,7 @@ Napi::Object registerCallback(Napi::Env env, Napi::Object exports) {
         Napi::PropertyDescriptor::Function("getSwapInfo", getSwapInfo),
         Napi::PropertyDescriptor::Function("getMemoryInfo", getMemoryInfo),
         Napi::PropertyDescriptor::Function("loadBackends", addonLoadBackends),
+        Napi::PropertyDescriptor::Function("setNuma", addonSetNuma),
         Napi::PropertyDescriptor::Function("init", addonInit),
         Napi::PropertyDescriptor::Function("dispose", addonDispose),
     });

package/llama/binariesGithubRelease.json

@@ -1,3 +1,3 @@
 {
-    "release": "b5590"
+    "release": "b6018"
 }

package/llama/llama.cpp.info.json

@@ -1,4 +1,4 @@
 {
-    "tag": "b5590",
+    "tag": "b6018",
     "llamaCppGithubRepo": "ggml-org/llama.cpp"
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "node-llama-cpp",
-  "version": "3.9.0",
+  "version": "3.11.0",
   "description": "Run AI models locally on your machine with node.js bindings for llama.cpp. Enforce a JSON schema on the model output on the generation level",
   "main": "./dist/index.js",
   "type": "module",
@@ -157,7 +157,7 @@
     "@types/yargs": "^17.0.33",
     "@vitest/coverage-v8": "^3.1.3",
     "@vitest/ui": "^3.1.3",
-    "electron": "^36.2.0",
+    "electron": "^37.2.4",
     "eslint": "^9.26.0",
     "eslint-import-resolver-typescript": "^4.3.4",
     "eslint-plugin-import": "^2.31.0",
@@ -178,12 +178,12 @@
     "typescript-eslint": "^8.32.0",
     "vite-node": "^3.1.3",
     "vitepress": "^1.6.3",
-    "vitepress-plugin-llms": "https://pkg.pr.new/vitepress-plugin-llms@51",
+    "vitepress-plugin-llms": "^1.7.2",
     "vitest": "^3.1.3",
     "zx": "^8.5.4"
   },
   "dependencies": {
-    "@huggingface/jinja": "^0.5.0",
+    "@huggingface/jinja": "^0.5.1",
     "async-retry": "^1.3.3",
     "bytes": "^3.1.2",
     "chalk": "^5.4.1",
@@ -197,7 +197,7 @@
     "ignore": "^7.0.4",
     "ipull": "^3.9.2",
     "is-unicode-supported": "^2.1.0",
-    "lifecycle-utils": "^2.0.0",
+    "lifecycle-utils": "^3.0.1",
     "log-symbols": "^7.0.0",
     "nanoid": "^5.1.5",
     "node-addon-api": "^8.3.1",
@@ -223,16 +223,16 @@
     }
   },
   "optionalDependencies": {
-    "@node-llama-cpp/linux-arm64": "3.9.0",
-    "@node-llama-cpp/linux-armv7l": "3.9.0",
-    "@node-llama-cpp/linux-x64": "3.9.0",
-    "@node-llama-cpp/linux-x64-cuda": "3.9.0",
-    "@node-llama-cpp/linux-x64-vulkan": "3.9.0",
-    "@node-llama-cpp/mac-arm64-metal": "3.9.0",
-    "@node-llama-cpp/mac-x64": "3.9.0",
-    "@node-llama-cpp/win-arm64": "3.9.0",
-    "@node-llama-cpp/win-x64": "3.9.0",
-    "@node-llama-cpp/win-x64-cuda": "3.9.0",
-    "@node-llama-cpp/win-x64-vulkan": "3.9.0"
+    "@node-llama-cpp/linux-arm64": "3.11.0",
+    "@node-llama-cpp/linux-armv7l": "3.11.0",
+    "@node-llama-cpp/linux-x64": "3.11.0",
+    "@node-llama-cpp/linux-x64-cuda": "3.11.0",
+    "@node-llama-cpp/linux-x64-vulkan": "3.11.0",
+    "@node-llama-cpp/mac-arm64-metal": "3.11.0",
+    "@node-llama-cpp/mac-x64": "3.11.0",
+    "@node-llama-cpp/win-arm64": "3.11.0",
+    "@node-llama-cpp/win-x64": "3.11.0",
+    "@node-llama-cpp/win-x64-cuda": "3.11.0",
+    "@node-llama-cpp/win-x64-vulkan": "3.11.0"
   }
 }