cui-llama.rn 1.4.3 → 1.4.4

package/android/src/main/build-arm64/cmake_install.cmake

@@ -0,0 +1,66 @@
+# Install script for directory: D:/dev/react-native/cui-llama.rn/android/src/main
+
+# Set the install prefix
+if(NOT DEFINED CMAKE_INSTALL_PREFIX)
+  set(CMAKE_INSTALL_PREFIX "C:/Program Files (x86)/llama.rn")
+endif()
+string(REGEX REPLACE "/$" "" CMAKE_INSTALL_PREFIX "${CMAKE_INSTALL_PREFIX}")
+
+# Set the install configuration name.
+if(NOT DEFINED CMAKE_INSTALL_CONFIG_NAME)
+  if(BUILD_TYPE)
+    string(REGEX REPLACE "^[^A-Za-z0-9_]+" ""
+           CMAKE_INSTALL_CONFIG_NAME "${BUILD_TYPE}")
+  else()
+    set(CMAKE_INSTALL_CONFIG_NAME "Release")
+  endif()
+  message(STATUS "Install configuration: \"${CMAKE_INSTALL_CONFIG_NAME}\"")
+endif()
+
+# Set the component getting installed.
+if(NOT CMAKE_INSTALL_COMPONENT)
+  if(COMPONENT)
+    message(STATUS "Install component: \"${COMPONENT}\"")
+    set(CMAKE_INSTALL_COMPONENT "${COMPONENT}")
+  else()
+    set(CMAKE_INSTALL_COMPONENT)
+  endif()
+endif()
+
+# Install shared libraries without execute permission?
+if(NOT DEFINED CMAKE_INSTALL_SO_NO_EXE)
+  set(CMAKE_INSTALL_SO_NO_EXE "0")
+endif()
+
+# Is this installation the result of a crosscompile?
+if(NOT DEFINED CMAKE_CROSSCOMPILING)
+  set(CMAKE_CROSSCOMPILING "TRUE")
+endif()
+
+# Set path to fallback-tool for dependency-resolution.
+if(NOT DEFINED CMAKE_OBJDUMP)
+  set(CMAKE_OBJDUMP "D:/Android/Sdk/ndk/26.3.11579264/toolchains/llvm/prebuilt/windows-x86_64/bin/llvm-objdump.exe")
+endif()
+
+string(REPLACE ";" "\n" CMAKE_INSTALL_MANIFEST_CONTENT
+       "${CMAKE_INSTALL_MANIFEST_FILES}")
+if(CMAKE_INSTALL_LOCAL_ONLY)
+  file(WRITE "D:/dev/react-native/cui-llama.rn/android/src/main/build-arm64/install_local_manifest.txt"
+     "${CMAKE_INSTALL_MANIFEST_CONTENT}")
+endif()
+if(CMAKE_INSTALL_COMPONENT)
+  if(CMAKE_INSTALL_COMPONENT MATCHES "^[a-zA-Z0-9_.+-]+$")
+    set(CMAKE_INSTALL_MANIFEST "install_manifest_${CMAKE_INSTALL_COMPONENT}.txt")
+  else()
+    string(MD5 CMAKE_INST_COMP_HASH "${CMAKE_INSTALL_COMPONENT}")
+    set(CMAKE_INSTALL_MANIFEST "install_manifest_${CMAKE_INST_COMP_HASH}.txt")
+    unset(CMAKE_INST_COMP_HASH)
+  endif()
+else()
+  set(CMAKE_INSTALL_MANIFEST "install_manifest.txt")
+endif()
+
+if(NOT CMAKE_INSTALL_LOCAL_ONLY)
+  file(WRITE "D:/dev/react-native/cui-llama.rn/android/src/main/build-arm64/${CMAKE_INSTALL_MANIFEST}"
+     "${CMAKE_INSTALL_MANIFEST_CONTENT}")
+endif()

package/android/src/main/java/com/rnllama/LlamaContext.java

@@ -28,6 +28,32 @@ public class LlamaContext {
 
   private static String loadedLibrary = "";
 
+  private static class NativeLogCallback {
+    DeviceEventManagerModule.RCTDeviceEventEmitter eventEmitter;
+
+    public NativeLogCallback(ReactApplicationContext reactContext) {
+      this.eventEmitter = reactContext.getJSModule(DeviceEventManagerModule.RCTDeviceEventEmitter.class);
+    }
+
+    void emitNativeLog(String level, String text) {
+      WritableMap event = Arguments.createMap();
+      event.putString("level", level);
+      event.putString("text", text);
+      eventEmitter.emit("@RNLlama_onNativeLog", event);
+    }
+  }
+
+  static void toggleNativeLog(ReactApplicationContext reactContext, boolean enabled) {
+    if (LlamaContext.isArchNotSupported()) {
+      throw new IllegalStateException("Only 64-bit architectures are supported");
+    }
+    if (enabled) {
+      setupLog(new NativeLogCallback(reactContext));
+    } else {
+      unsetLog();
+    }
+  }
+
   private int id;
   private ReactApplicationContext reactContext;
   private long context;
@@ -73,7 +99,7 @@ public class LlamaContext {
   }
 
   public LlamaContext(int id, ReactApplicationContext reactContext, ReadableMap params) {
-    if (LlamaContext.isArm64V8a() == false && LlamaContext.isX86_64() == false) {
+    if (LlamaContext.isArchNotSupported()) {
       throw new IllegalStateException("Only 64-bit architectures are supported");
     }
     if (!params.hasKey("model")) {
@@ -95,13 +121,17 @@ public class LlamaContext {
         Log.e(NAME, "Failed to convert to FD!");
       }
     }
-    logToAndroid();
+    
     // Check if file has GGUF magic numbers
     this.id = id;
     eventEmitter = reactContext.getJSModule(DeviceEventManagerModule.RCTDeviceEventEmitter.class);
     this.context = initContext(
       // String model,
       modelName,
+      // String chat_template,
+      params.hasKey("chat_template") ? params.getString("chat_template") : "",
+      // String reasoning_format,
+      params.hasKey("reasoning_format") ? params.getString("reasoning_format") : "none",
       // boolean embedding,
       params.hasKey("embedding") ? params.getBoolean("embedding") : false,
       // int embd_normalize,
@@ -166,12 +196,24 @@ public class LlamaContext {
     return loadedLibrary;
   }
 
-  public String getFormattedChat(ReadableArray messages, String chatTemplate) {
-    ReadableMap[] msgs = new ReadableMap[messages.size()];
-    for (int i = 0; i < messages.size(); i++) {
-      msgs[i] = messages.getMap(i);
-    }
-    return getFormattedChat(this.context, msgs, chatTemplate == null ? "" : chatTemplate);
+  public WritableMap getFormattedChatWithJinja(String messages, String chatTemplate, ReadableMap params) {
+    String jsonSchema = params.hasKey("json_schema") ? params.getString("json_schema") : "";
+    String tools = params.hasKey("tools") ? params.getString("tools") : "";
+    Boolean parallelToolCalls = params.hasKey("parallel_tool_calls") ? params.getBoolean("parallel_tool_calls") : false;
+    String toolChoice = params.hasKey("tool_choice") ? params.getString("tool_choice") : "";
+    return getFormattedChatWithJinja(
+      this.context,
+      messages,
+      chatTemplate == null ? "" : chatTemplate,
+      jsonSchema,
+      tools,
+      parallelToolCalls,
+      toolChoice
+    );
+  }
+
+  public String getFormattedChat(String messages, String chatTemplate) {
+    return getFormattedChat(this.context, messages, chatTemplate == null ? "" : chatTemplate);
   }
 
   private void emitLoadProgress(int progress) {
@@ -259,8 +301,18 @@ public class LlamaContext {
       this.context,
       // String prompt,
       params.getString("prompt"),
+      // int chat_format,
+      params.hasKey("chat_format") ? params.getInt("chat_format") : 0,
       // String grammar,
       params.hasKey("grammar") ? params.getString("grammar") : "",
+      // String json_schema,
+      params.hasKey("json_schema") ? params.getString("json_schema") : "",
+      // boolean grammar_lazy,
+      params.hasKey("grammar_lazy") ? params.getBoolean("grammar_lazy") : false,
+      // ReadableArray grammar_triggers,
+      params.hasKey("grammar_triggers") ? params.getArray("grammar_triggers") : null,
+      // ReadableArray preserved_tokens,
+      params.hasKey("preserved_tokens") ? params.getArray("preserved_tokens") : null,
       // float temperature,
       params.hasKey("temperature") ? (float) params.getDouble("temperature") : 0.7f,
       // int n_threads,
@@ -311,6 +363,8 @@ public class LlamaContext {
       params.hasKey("dry_allowed_length") ? params.getInt("dry_allowed_length") : 2,
       // int dry_penalty_last_n,
       params.hasKey("dry_penalty_last_n") ? params.getInt("dry_penalty_last_n") : -1,
+      // float top_n_sigma,
+      params.hasKey("top_n_sigma") ? (float) params.getDouble("top_n_sigma") : -1.0f,
       // String[] dry_sequence_breakers, when undef, we use the default definition from common.h
       params.hasKey("dry_sequence_breakers") ? params.getArray("dry_sequence_breakers").toArrayList().toArray(new String[0]) : new String[]{"\n", ":", "\"", "*"},
       // PartialCompletionCallback partial_completion_callback
@@ -431,15 +485,13 @@ public class LlamaContext {
       //  Log.d(NAME, "Loading librnllama_v8_7.so with runtime feature detection");
       //  System.loadLibrary("rnllama_v8_7");
     } else if (LlamaContext.isX86_64()) {
-        Log.d(NAME, "Loading librnllama_x86_64.so");
-        System.loadLibrary("rnllama_x86_64");
-        loadedLibrary = "rnllama_x86_64";
+      Log.d(NAME, "Loading librnllama_x86_64.so");
+      System.loadLibrary("rnllama_x86_64");
+      loadedLibrary = "rnllama_x86_64";
     } else {
-        Log.d(NAME, "Loading default librnllama.so");
-        System.loadLibrary("rnllama");
-        loadedLibrary = "rnllama";
+      Log.d(NAME, "ARM32 is not supported, skipping loading library");
     }
-}
+  }
 
   public static boolean isArm64V8a() {
     return Build.SUPPORTED_ABIS[0].equals("arm64-v8a");
@@ -449,6 +501,10 @@ public class LlamaContext {
     return Build.SUPPORTED_ABIS[0].equals("x86_64");
   }
 
+  private static boolean isArchNotSupported() {
+    return isArm64V8a() == false && isX86_64() == false;
+  }
+
   public static String getCpuFeatures() {
     File file = new File("/proc/cpuinfo");
     StringBuilder stringBuilder = new StringBuilder();
@@ -481,6 +537,8 @@ public class LlamaContext {
   );
   protected static native long initContext(
     String model,
+    String chat_template,
+    String reasoning_format,
     boolean embedding,
     int embd_normalize,
     int n_ctx,
@@ -506,9 +564,18 @@ public class LlamaContext {
   protected static native WritableMap loadModelDetails(
     long contextPtr
   );
+  protected static native WritableMap getFormattedChatWithJinja(
+    long contextPtr,
+    String messages,
+    String chatTemplate,
+    String jsonSchema,
+    String tools,
+    boolean parallelToolCalls,
+    String toolChoice
+  );
   protected static native String getFormattedChat(
     long contextPtr,
-    ReadableMap[] messages,
+    String messages,
     String chatTemplate
   );
   protected static native WritableMap loadSession(
@@ -523,7 +590,12 @@ public class LlamaContext {
   protected static native WritableMap doCompletion(
     long context_ptr,
     String prompt,
+    int chat_format,
     String grammar,
+    String json_schema,
+    boolean grammar_lazy,
+    ReadableArray grammar_triggers,
+    ReadableArray preserved_tokens,
     float temperature,
     int n_threads,
     int n_predict,
@@ -549,6 +621,7 @@ public class LlamaContext {
     float   dry_base,
     int dry_allowed_length,
     int dry_penalty_last_n,
+    float top_n_sigma,
     String[] dry_sequence_breakers,
     PartialCompletionCallback partial_completion_callback
   );
@@ -567,5 +640,6 @@ public class LlamaContext {
   protected static native void removeLoraAdapters(long contextPtr);
   protected static native WritableArray getLoadedLoraAdapters(long contextPtr);
   protected static native void freeContext(long contextPtr);
-  protected static native void logToAndroid();
+  protected static native void setupLog(NativeLogCallback logCallback);
+  protected static native void unsetLog();
 }

package/android/src/main/java/com/rnllama/RNLlama.java

@@ -35,6 +35,32 @@ public class RNLlama implements LifecycleEventListener {
 
   private HashMap<Integer, LlamaContext> contexts = new HashMap<>();
 
+  public void toggleNativeLog(boolean enabled, Promise promise) {
+    new AsyncTask<Void, Void, Boolean>() {
+      private Exception exception;
+
+      @Override
+      protected Boolean doInBackground(Void... voids) {
+        try {
+          LlamaContext.toggleNativeLog(reactContext, enabled);
+          return true;
+        } catch (Exception e) {
+          exception = e;
+        }
+        return null;
+      }
+
+      @Override
+      protected void onPostExecute(Boolean result) {
+        if (exception != null) {
+          promise.reject(exception);
+          return;
+        }
+        promise.resolve(result);
+      }
+    }.executeOnExecutor(AsyncTask.THREAD_POOL_EXECUTOR);
+  }
+
   private int llamaContextLimit = -1;
 
   public void setContextLimit(double limit, Promise promise) {
@@ -116,18 +142,25 @@ public class RNLlama implements LifecycleEventListener {
     tasks.put(task, "initContext");
   }
 
-  public void getFormattedChat(double id, final ReadableArray messages, final String chatTemplate, Promise promise) {
+  public void getFormattedChat(double id, final String messages, final String chatTemplate, final ReadableMap params, Promise promise) {
     final int contextId = (int) id;
-    AsyncTask task = new AsyncTask<Void, Void, String>() {
+    AsyncTask task = new AsyncTask<Void, Void, Object>() {
       private Exception exception;
 
       @Override
-      protected String doInBackground(Void... voids) {
+      protected Object doInBackground(Void... voids) {
         try {
           LlamaContext context = contexts.get(contextId);
           if (context == null) {
             throw new Exception("Context not found");
           }
+          if (params.hasKey("jinja") && params.getBoolean("jinja")) {
+            ReadableMap result = context.getFormattedChatWithJinja(messages, chatTemplate, params);
+            if (result.hasKey("_error")) {
+              throw new Exception(result.getString("_error"));
+            }
+            return result;
+          }
           return context.getFormattedChat(messages, chatTemplate);
         } catch (Exception e) {
           exception = e;
@@ -136,7 +169,7 @@ public class RNLlama implements LifecycleEventListener {
       }
 
       @Override
-      protected void onPostExecute(String result) {
+      protected void onPostExecute(Object result) {
         if (exception != null) {
           promise.reject(exception);
           return;

package/android/src/main/jni-utils.h

@@ -16,6 +16,12 @@ jobject getMap(JNIEnv *env, jobject readableArray, int index) {
     return env->CallObjectMethod(readableArray, getMapMethod, index);
 }
 
+jstring getString(JNIEnv *env, jobject readableArray, int index) {
+    jclass arrayClass = env->GetObjectClass(readableArray);
+    jmethodID getStringMethod = env->GetMethodID(arrayClass, "getString", "(I)Ljava/lang/String;");
+    return (jstring) env->CallObjectMethod(readableArray, getStringMethod, index);
+}
+
 // Other methods not used yet
 
 }