@fugood/llama.node 1.1.0 → 1.1.1

package/lib/binding.js

@@ -15,13 +15,23 @@ var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (
 }) : function(o, v) {
     o["default"] = v;
 });
-var __importStar = (this && this.__importStar) || function (mod) {
-    if (mod && mod.__esModule) return mod;
-    var result = {};
-    if (mod != null) for (var k in mod) if (k !== "default" && Object.prototype.hasOwnProperty.call(mod, k)) __createBinding(result, mod, k);
-    __setModuleDefault(result, mod);
-    return result;
-};
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
 var __awaiter = (this && this.__awaiter) || function (thisArg, _arguments, P, generator) {
     function adopt(value) { return value instanceof P ? value : new P(function (resolve) { resolve(value); }); }
     return new (P || (P = Promise))(function (resolve, reject) {

package/lib/binding.ts

@@ -129,6 +129,7 @@ export type LlamaCompletionResult = {
   text: string
   reasoning_content?: string
   content?: string
+  chat_format: number
   tokens_predicted: number
   tokens_evaluated: number
   truncated: boolean

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@fugood/llama.node",
   "access": "public",
-  "version": "1.1.0",
+  "version": "1.1.1",
   "description": "An another Node binding of llama.cpp",
   "main": "lib/index.js",
   "scripts": {
@@ -71,19 +71,19 @@
     "CMakeLists.txt"
   ],
   "optionalDependencies": {
-    "@fugood/node-llama-linux-x64": "1.1.0",
-    "@fugood/node-llama-linux-x64-vulkan": "1.1.0",
-    "@fugood/node-llama-linux-x64-cuda": "1.1.0",
-    "@fugood/node-llama-linux-arm64": "1.1.0",
-    "@fugood/node-llama-linux-arm64-vulkan": "1.1.0",
-    "@fugood/node-llama-linux-arm64-cuda": "1.1.0",
-    "@fugood/node-llama-win32-x64": "1.1.0",
-    "@fugood/node-llama-win32-x64-vulkan": "1.1.0",
-    "@fugood/node-llama-win32-x64-cuda": "1.1.0",
-    "@fugood/node-llama-win32-arm64": "1.1.0",
-    "@fugood/node-llama-win32-arm64-vulkan": "1.1.0",
-    "@fugood/node-llama-darwin-x64": "1.1.0",
-    "@fugood/node-llama-darwin-arm64": "1.1.0"
+    "@fugood/node-llama-linux-x64": "1.1.1",
+    "@fugood/node-llama-linux-x64-vulkan": "1.1.1",
+    "@fugood/node-llama-linux-x64-cuda": "1.1.1",
+    "@fugood/node-llama-linux-arm64": "1.1.1",
+    "@fugood/node-llama-linux-arm64-vulkan": "1.1.1",
+    "@fugood/node-llama-linux-arm64-cuda": "1.1.1",
+    "@fugood/node-llama-win32-x64": "1.1.1",
+    "@fugood/node-llama-win32-x64-vulkan": "1.1.1",
+    "@fugood/node-llama-win32-x64-cuda": "1.1.1",
+    "@fugood/node-llama-win32-arm64": "1.1.1",
+    "@fugood/node-llama-win32-arm64-vulkan": "1.1.1",
+    "@fugood/node-llama-darwin-x64": "1.1.1",
+    "@fugood/node-llama-darwin-arm64": "1.1.1"
   },
   "devDependencies": {
     "@babel/preset-env": "^7.24.4",

package/src/LlamaCompletionWorker.cpp

@@ -157,10 +157,26 @@ void LlamaCompletionWorker::Execute() {
     // For multimodal input, n_past might already be set
     // Only decode text tokens if we have any input left
     if (n_input > 0) {
-      int ret = llama_decode(ctx, llama_batch_get_one(embd->data() + n_cur, n_input));
-      if (ret < 0) {
-        SetError("Failed to decode token, code: " + std::to_string(ret));
-        break;
+      // Decode tokens in batches using n_batch as chunk size
+      int n_past_batch = n_cur;
+      int n_remaining = n_input;
+      
+      while (n_remaining > 0) {
+        int n_eval = n_remaining;
+        if (n_eval > _params.n_batch) {
+          n_eval = _params.n_batch;
+        }
+        
+        int ret = llama_decode(ctx, llama_batch_get_one(embd->data() + n_past_batch, n_eval));
+        if (ret < 0) {
+          SetError("Failed to decode token batch, code: " + std::to_string(ret) + 
+                   ", n_eval: " + std::to_string(n_eval) + 
+                   ", n_past_batch: " + std::to_string(n_past_batch));
+          break;
+        }
+        
+        n_past_batch += n_eval;
+        n_remaining -= n_eval;
       }
     }
 
@@ -255,6 +271,8 @@ void LlamaCompletionWorker::OnOK() {
     try {
       common_chat_syntax chat_syntax;
       chat_syntax.format = static_cast<common_chat_format>(_chat_format);
+      result.Set("chat_format", Napi::Number::New(env, _chat_format));
+
       chat_syntax.thinking_forced_open = _thinking_forced_open;
 
       if (_reasoning_format == "deepseek") {

package/src/LlamaContext.cpp

@@ -247,7 +247,7 @@ LlamaContext::LlamaContext(const Napi::CallbackInfo &info)
   params.cache_type_v = kv_cache_type_from_str(
       get_option<std::string>(options, "cache_type_v", "f16").c_str());
   params.ctx_shift = get_option<bool>(options, "ctx_shift", true);
-  params.kv_unified = get_option<bool>(options, "kv_unified", true);
+  params.kv_unified = get_option<bool>(options, "kv_unified", false);
 
   params.use_mlock = get_option<bool>(options, "use_mlock", false);
   params.use_mmap = get_option<bool>(options, "use_mmap", true);