withcatai · giladgd · Jul 19, 2026 · Jul 19, 2026 · Jul 19, 2026 · Jul 19, 2026
diff --git a/.gitignore b/.gitignore
@@ -5,6 +5,7 @@ node_modules
 *.cpuprofile
 
 /dist
+/tsconfig.tsbuildinfo
 /docs-site
 /docs/api
 /templates/packed

diff --git a/.vitepress/config.ts b/.vitepress/config.ts
@@ -414,7 +414,8 @@ export default defineConfig({
                         ],
                         module: ts.ModuleKind.ES2022,
                         target: ts.ScriptTarget.ES2022,
-                        moduleDetection: ts.ModuleDetectionKind.Force
+                        moduleDetection: ts.ModuleDetectionKind.Force,
+                        rootDir: undefined
                     },
                     tsModule: ts
                 }
@@ -468,6 +469,27 @@ export default defineConfig({
                 detailedView: true,
                 miniSearch: {
                     searchOptions: {
+                        fuzzy: 0.05,
+                        tokenize(query: string) {
+                            const maxRemovedSpaces = 3;
+                            const words = query.match(/[\p{L}\p{N}]+/gu) ?? [];
+                            const terms = new Set(words);
+
+                            for (let start = 0; start < words.length; start++) {
+                                let joined = words[start]!;
+
+                                for (let removedSpaces = 1; removedSpaces <= maxRemovedSpaces; removedSpaces++) {
+                                    const nextWord = words[start + removedSpaces];
+                                    if (nextWord == null)
+                                        break;
+
+                                    joined += nextWord;
+                                    terms.add(joined);
+                                }
+                            }
+
+                            return [...terms];
+                        },
                         boostDocument(term, documentId, storedFields) {
                             const firstTitle = (storedFields?.titles as string[])?.[0];
                             if (firstTitle?.startsWith("Type Alias: "))

diff --git a/.vitepress/tsconfig.json b/.vitepress/tsconfig.json
@@ -15,11 +15,12 @@
         "noUncheckedIndexedAccess": true,
         "moduleDetection": "force",
         "skipLibCheck": true,
-        "moduleResolution": "bundler",
+        "moduleResolution": "node16",
         "resolveJsonModule": true,
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "..",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": true,

diff --git a/eslint.config.js b/eslint.config.js
@@ -188,6 +188,7 @@ export default defineConfig({
             allow: []
         }],
         "@typescript-eslint/explicit-member-accessibility": ["warn"],
+        "@typescript-eslint/consistent-type-definitions": ["warn", "type"],
         "@stylistic/member-delimiter-style": ["warn", {
             multiline: {
                 delimiter: "comma",

diff --git a/llama/addon/AddonModel.cpp b/llama/addon/AddonModel.cpp
@@ -530,7 +530,7 @@ Napi::Value AddonModel::Dispose(const Napi::CallbackInfo& info) {
         return worker->GetPromise();
     } else {
         disposeMT();
-    
+
         Napi::Promise::Deferred deferred = Napi::Promise::Deferred::New(info.Env());
         deferred.Resolve(info.Env().Undefined());
         return deferred.Promise();

diff --git a/llama/addon/addonGlobals.cpp b/llama/addon/addonGlobals.cpp
@@ -1,5 +1,7 @@
 #include <sstream>
 #include <vector>
+#include <algorithm>
+#include <limits>
 #include "addonGlobals.h"
 #include "napi.h"
 

diff --git a/packages/@node-llama-cpp/linux-arm64/.gitignore b/packages/@node-llama-cpp/linux-arm64/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/linux-arm64/tsconfig.json b/packages/@node-llama-cpp/linux-arm64/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/linux-armv7l/.gitignore b/packages/@node-llama-cpp/linux-armv7l/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/linux-armv7l/tsconfig.json b/packages/@node-llama-cpp/linux-armv7l/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/linux-riscv64/.gitignore b/packages/@node-llama-cpp/linux-riscv64/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/linux-riscv64/tsconfig.json b/packages/@node-llama-cpp/linux-riscv64/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/linux-x64-cuda-ext/.gitignore b/packages/@node-llama-cpp/linux-x64-cuda-ext/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/linux-x64-cuda-ext/tsconfig.json b/packages/@node-llama-cpp/linux-x64-cuda-ext/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/linux-x64-cuda/.gitignore b/packages/@node-llama-cpp/linux-x64-cuda/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/linux-x64-cuda/tsconfig.json b/packages/@node-llama-cpp/linux-x64-cuda/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/linux-x64-vulkan/.gitignore b/packages/@node-llama-cpp/linux-x64-vulkan/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/linux-x64-vulkan/tsconfig.json b/packages/@node-llama-cpp/linux-x64-vulkan/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/linux-x64/.gitignore b/packages/@node-llama-cpp/linux-x64/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/linux-x64/tsconfig.json b/packages/@node-llama-cpp/linux-x64/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/mac-arm64-metal/.gitignore b/packages/@node-llama-cpp/mac-arm64-metal/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/mac-arm64-metal/tsconfig.json b/packages/@node-llama-cpp/mac-arm64-metal/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/mac-x64/.gitignore b/packages/@node-llama-cpp/mac-x64/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/mac-x64/tsconfig.json b/packages/@node-llama-cpp/mac-x64/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/win-arm64/.gitignore b/packages/@node-llama-cpp/win-arm64/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/win-arm64/tsconfig.json b/packages/@node-llama-cpp/win-arm64/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/win-x64-cuda-ext/.gitignore b/packages/@node-llama-cpp/win-x64-cuda-ext/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/win-x64-cuda-ext/tsconfig.json b/packages/@node-llama-cpp/win-x64-cuda-ext/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/win-x64-cuda/.gitignore b/packages/@node-llama-cpp/win-x64-cuda/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/win-x64-cuda/tsconfig.json b/packages/@node-llama-cpp/win-x64-cuda/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/win-x64-vulkan/.gitignore b/packages/@node-llama-cpp/win-x64-vulkan/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/win-x64-vulkan/tsconfig.json b/packages/@node-llama-cpp/win-x64-vulkan/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/@node-llama-cpp/win-x64/.gitignore b/packages/@node-llama-cpp/win-x64/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/@node-llama-cpp/win-x64/tsconfig.json b/packages/@node-llama-cpp/win-x64/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": false,

diff --git a/packages/create-node-llama-cpp/.gitignore b/packages/create-node-llama-cpp/.gitignore
@@ -1 +1,2 @@
 /dist
+/tsconfig.tsbuildinfo
diff --git a/packages/create-node-llama-cpp/tsconfig.json b/packages/create-node-llama-cpp/tsconfig.json
@@ -18,6 +18,7 @@
         "strictNullChecks": true,
         "isolatedModules": true,
         "noEmit": false,
+        "rootDir": "./src",
         "outDir": "./dist",
         "strict": true,
         "sourceMap": true,

diff --git a/src/bindings/Llama.ts b/src/bindings/Llama.ts
@@ -402,15 +402,15 @@ export class Llama {
     /**
      * Cap the amount of VRAM that this Llama instance is allowed to use in bytes.
      * This is useful for constraining the resource usage of models and contexts created with the Llama instance.
-     * 
+     *
      * Capping to a value that's too low may cause model loads and context creations to either fail or not fully offload to VRAM,
      * causing inference to be significantly slower.
-     * 
+     *
      * Setting a cap will only affect future model loads and context creations.
-     * 
+     *
      * Use with caution.
      * Setting to `null` disables the cap.
-     * 
+     *
      * Defaults to `null`.
      */
     public async setVramCap(bytes: number | null) {
@@ -425,7 +425,7 @@ export class Llama {
 
     /**
      * Get the current VRAM cap in bytes. See {@link setVramCap `setVramCap`} for more information.
-     * 
+     *
      * Defaults to `null`, which means no cap is set.
      */
     public getVramCap() {
@@ -435,21 +435,21 @@ export class Llama {
     /**
      * Cap the amount of RAM that this Llama instance is allowed to use in bytes.
      * This is useful for constraining the resource usage of models and contexts created with the Llama instance.
-     * 
+     *
      * Capping to a value that's too low may cause model loads and context creations to fail.
      * Capping to any value will exclude swap from the resource calculations,
      * so extremely large models may not load at all even if you have enough swap available.
-     * 
+     *
      * Setting a cap will only affect future model loads and context creations.
-     * 
+     *
      * On unified memory systems, capping the RAM may also effectively cap the VRAM, as they are shared.
      * On such systems, it's recommended to either cap the VRAM or the RAM (but not both),
      * and if you need to cap both then make sure to set the RAM cap to a value greater than the VRAM cap.
      * > **Note:** You can detect a unified memory system by checking whether `getVramState().unifiedSize` is greater than 0.
-     * 
+     *
      * Use with caution.
      * Setting to `null` disables the cap.
-     * 
+     *
      * Defaults to `null`.
      */
     public async setRamCap(bytes: number | null) {
@@ -459,14 +459,14 @@ export class Llama {
             throw new RangeError("RAM cap must be a non-negative number or null");
         else if (bytes != null)
             bytes = Math.floor(bytes);
-        
+
         this._ramOrchestrator.memoryCap = bytes;
         this._swapOrchestrator.memoryCap = bytes == null ? null : 0; // if RAM is capped, we can't count on swap for calculation
     }
 
     /**
      * Get the current RAM cap in bytes. See {@link setRamCap `setRamCap`} for more information.
-     * 
+     *
      * Defaults to `null`, which means no cap is set.
      */
     public getRamCap() {
@@ -641,15 +641,23 @@ export class Llama {
         // llama.cpp uses dots to indicate progress, so we don't want to print them as different lines,
         // and instead, append to the same log line
         if (logMessageIsOnlyDots(message) && this._logger === Llama.defaultConsoleLogger) {
+            const stdout = LlamaLogLevelGreaterThanOrEqual(level, LlamaLogLevel.warn)
+                ? process.stderr
+                : process.stdout;
+
             if (logMessageIsOnlyDots(this._previousLog) && level === this._previousLogLevel) {
-                process.stdout.write(message);
+                stdout.write(message);
             } else {
                 this._nextLogNeedNewLine = true;
-                process.stdout.write(prefixAndColorMessage(message, getColorForLogLevel(level)));
+                stdout.write(prefixAndColorMessage(message, getColorForLogLevel(level)));
             }
         } else {
             if (this._nextLogNeedNewLine) {
-                process.stdout.write("\n");
+                const stdout = LlamaLogLevelGreaterThanOrEqual(this._previousLogLevel ?? level, LlamaLogLevel.warn)
+                    ? process.stderr
+                    : process.stdout;
+
+                stdout.write("\n");
                 this._nextLogNeedNewLine = false;
             }
 
@@ -707,7 +715,7 @@ export class Llama {
         const vramOrchestrator = new MemoryOrchestrator(getBalancedVramState.bind(undefined, bindings, true));
         const ramOrchestrator = new MemoryOrchestrator(async () => {
             const {total, wired} = await getBalancedRamState(bindings);
-            
+
             return {
                 total,
                 free: total - wired,
@@ -810,7 +818,7 @@ export class Llama {
             default:
                 void (level satisfies never);
                 console.warn(getConsoleLogPrefix() + getColorForLogLevel(LlamaLogLevel.warn)(`Unknown log level: ${level}`));
-                console.log(prefixAndColorMessage(message, getColorForLogLevel(level)));
+                console.warn(prefixAndColorMessage(message, getColorForLogLevel(level)));
         }
     }
 }
@@ -899,7 +907,7 @@ async function getBalancedVramState(bindings: BindingModule, balanceUnifiedMemor
 
             const nonUnifiedMemoryRam = systemMemoryInfo.total - unifiedSize;
             const lockedUnifiedVram = Math.max(0, Math.min(systemMemoryInfo.wired ?? 0, systemMemoryInfo.total) - nonUnifiedMemoryRam);
-    
+
             currentUsed = Math.max(currentUsed, Math.min(lockedUnifiedVram, unifiedSize));
         } catch (err) {
             // do nothing