feat: sync llama.cpp

.github/actions/setup/action.yml

@@ -7,7 +7,7 @@ runs:
     - name: Setup Node.js
       uses: actions/setup-node@v3
       with:
-        node-version: 18.x
+        node-version: 20.x
 
     - name: Cache dependencies
       id: yarn-cache

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

@@ -70,6 +70,8 @@ public LlamaContext(int id, ReactApplicationContext reactContext, ReadableMap pa
       params.getString("model"),
       // 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,
@@ -301,6 +303,8 @@ public WritableMap completion(ReadableMap params) {
       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
@@ -468,6 +472,7 @@ protected static native WritableMap modelInfo(
   protected static native long initContext(
     String model,
     String chat_template,
+    String reasoning_format,
     boolean embedding,
     int embd_normalize,
     int n_ctx,
@@ -550,6 +555,7 @@ protected static native WritableMap doCompletion(
     float   dry_base,
     int dry_allowed_length,
     int dry_penalty_last_n,
+    float top_n_sigma,
     String[] dry_sequence_breakers,
     PartialCompletionCallback partial_completion_callback
   );

android/src/main/jni.cpp

@@ -223,6 +223,7 @@ Java_com_rnllama_LlamaContext_initContext(
     jobject thiz,
     jstring model_path_str,
     jstring chat_template,
+    jstring reasoning_format,
     jboolean embedding,
     jint embd_normalize,
     jint n_ctx,
@@ -259,6 +260,13 @@ Java_com_rnllama_LlamaContext_initContext(
     const char *chat_template_chars = env->GetStringUTFChars(chat_template, nullptr);
     defaultParams.chat_template = chat_template_chars;
 
+    const char *reasoning_format_chars = env->GetStringUTFChars(reasoning_format, nullptr);
+    if (strcmp(reasoning_format_chars, "deepseek") == 0) {
+        defaultParams.reasoning_format = COMMON_REASONING_FORMAT_DEEPSEEK;
+    } else {
+        defaultParams.reasoning_format = COMMON_REASONING_FORMAT_NONE;
+    }
+
     defaultParams.n_ctx = n_ctx;
     defaultParams.n_batch = n_batch;
     defaultParams.n_ubatch = n_ubatch;
@@ -326,6 +334,7 @@ Java_com_rnllama_LlamaContext_initContext(
 
     env->ReleaseStringUTFChars(model_path_str, model_path_chars);
     env->ReleaseStringUTFChars(chat_template, chat_template_chars);
+    env->ReleaseStringUTFChars(reasoning_format, reasoning_format_chars);
     env->ReleaseStringUTFChars(cache_type_k, cache_type_k_chars);
     env->ReleaseStringUTFChars(cache_type_v, cache_type_v_chars);
 
@@ -664,6 +673,7 @@ Java_com_rnllama_LlamaContext_doCompletion(
     jfloat   dry_base,
     jint dry_allowed_length,
     jint dry_penalty_last_n,
+    jfloat top_n_sigma,
     jobjectArray dry_sequence_breakers,
     jobject partial_completion_callback
 ) {
@@ -706,6 +716,7 @@ Java_com_rnllama_LlamaContext_doCompletion(
     sparams.dry_base = dry_base;
     sparams.dry_allowed_length = dry_allowed_length;
     sparams.dry_penalty_last_n = dry_penalty_last_n;
+    sparams.top_n_sigma = top_n_sigma;
 
     // grammar
     auto grammar_chars = env->GetStringUTFChars(grammar, nullptr);
@@ -882,10 +893,16 @@ Java_com_rnllama_LlamaContext_doCompletion(
     llama->is_predicting = false;
 
     auto toolCalls = createWritableArray(env);
+    std::string reasoningContent = "";
+    std::string *content = nullptr;
     auto toolCallsSize = 0;
     if (!llama->is_interrupted) {
         try {
             common_chat_msg message = common_chat_parse(llama->generated_text, static_cast<common_chat_format>(chat_format));
+            if (!message.reasoning_content.empty()) {
+                reasoningContent = message.reasoning_content;
+            }
+            content = &message.content;
             for (const auto &tc : message.tool_calls) {
                 auto toolCall = createWriteableMap(env);
                 putString(env, toolCall, "type", "function");
@@ -906,6 +923,12 @@ Java_com_rnllama_LlamaContext_doCompletion(
 
     auto result = createWriteableMap(env);
     putString(env, result, "text", llama->generated_text.c_str());
+    if (content) {
+        putString(env, result, "content", content->c_str());
+    }
+    if (!reasoningContent.empty()) {
+        putString(env, result, "reasoning_content", reasoningContent.c_str());
+    }
     if (toolCallsSize > 0) {
         putArray(env, result, "tool_calls", toolCalls);
     }

cpp/chat-template.hpp

@@ -249,16 +249,30 @@ class chat_template {
                     inputs.add_generation_prompt = false;
                     full = apply(inputs);
                 }
-
-                if (full.find(prefix) != 0) {
-                    if (prefix.rfind(eos_token_) == prefix.size() - eos_token_.size()) {
-                        prefix = prefix.substr(0, prefix.size() - eos_token_.size());
+                auto eos_pos_last = full.rfind(eos_token_);
+                if (eos_pos_last == prefix.size() - eos_token_.size() ||
+                      (full[full.size() - 1] == '\n' && (eos_pos_last == full.size() - eos_token_.size() - 1))) {
+                    full = full.substr(0, eos_pos_last);
+                }
+                size_t common_prefix_length = 0;
+                for (size_t i = 0; i < prefix.size() && i < full.size(); ++i) {
+                    if (prefix[i] != full[i]) {
+                        break;
                     }
+                    if (prefix[i] == '<') {
+                        // DeepSeek R1's template (as of 20250209) adds a trailing <think> if add_generation_prompt,
+                        // but it removes thinking tags for past messages.
+                        // The prefix and full strings diverge at <think> vs. <｜tool▁calls▁begin｜>, we avoid consuming the leading <.
+                        continue;
+                    }
+                    common_prefix_length = i + 1;
                 }
-                if (full.find(prefix) != 0) {
+                auto example = full.substr(common_prefix_length);
+                if (example.find("tool_name") == std::string::npos && example.find("some_value") == std::string::npos) {
                     fprintf(stderr, "Failed to infer a tool call example (possible template bug)\n");
+                } else {
+                    tool_call_example_ = example;
                 }
-                tool_call_example_ = full.substr(prefix.size());
             }
         } catch (const std::exception & e) {
             fprintf(stderr, "Failed to generate tool call example: %s\n", e.what());
@@ -363,7 +377,7 @@ class chat_template {
             if (polyfill_tools) {
                 adjusted_messages = add_system(inputs.messages,
                     "You can call any of the following tools to satisfy the user's requests: " + minja::Value(inputs.tools).dump(2, /* to_json= */ true) +
-                    (!polyfill_tool_call_example || tool_call_example_.empty() ? "" : "\n\nExample tool call syntax:\n\n" + tool_call_example_));
+                    (!polyfill_tool_call_example || tool_call_example_.empty() ? "" : "\n\nExample tool call syntax:\n\n" + tool_call_example_ + "\n\n"));
             } else {
                 adjusted_messages = inputs.messages;
             }