@fugood/llama.node 1.5.0-rc.0 → 1.6.0-rc.0

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@fugood/llama.node",
   "access": "public",
-  "version": "1.5.0-rc.0",
+  "version": "1.6.0-rc.0",
   "description": "An another Node binding of llama.cpp",
   "main": "lib/index.js",
   "scripts": {
@@ -72,20 +72,20 @@
     "CMakeLists.txt"
   ],
   "optionalDependencies": {
-    "@fugood/node-llama-darwin-arm64": "1.5.0-rc.0",
-    "@fugood/node-llama-darwin-x64": "1.5.0-rc.0",
-    "@fugood/node-llama-linux-arm64": "1.5.0-rc.0",
-    "@fugood/node-llama-linux-arm64-cuda": "1.5.0-rc.0",
-    "@fugood/node-llama-linux-arm64-snapdragon": "1.5.0-rc.0",
-    "@fugood/node-llama-linux-arm64-vulkan": "1.5.0-rc.0",
-    "@fugood/node-llama-linux-x64": "1.5.0-rc.0",
-    "@fugood/node-llama-linux-x64-cuda": "1.5.0-rc.0",
-    "@fugood/node-llama-linux-x64-vulkan": "1.5.0-rc.0",
-    "@fugood/node-llama-win32-arm64": "1.5.0-rc.0",
-    "@fugood/node-llama-win32-arm64-vulkan": "1.5.0-rc.0",
-    "@fugood/node-llama-win32-x64": "1.5.0-rc.0",
-    "@fugood/node-llama-win32-x64-cuda": "1.5.0-rc.0",
-    "@fugood/node-llama-win32-x64-vulkan": "1.5.0-rc.0"
+    "@fugood/node-llama-darwin-arm64": "1.6.0-rc.0",
+    "@fugood/node-llama-darwin-x64": "1.6.0-rc.0",
+    "@fugood/node-llama-linux-arm64": "1.6.0-rc.0",
+    "@fugood/node-llama-linux-arm64-cuda": "1.6.0-rc.0",
+    "@fugood/node-llama-linux-arm64-snapdragon": "1.6.0-rc.0",
+    "@fugood/node-llama-linux-arm64-vulkan": "1.6.0-rc.0",
+    "@fugood/node-llama-linux-x64": "1.6.0-rc.0",
+    "@fugood/node-llama-linux-x64-cuda": "1.6.0-rc.0",
+    "@fugood/node-llama-linux-x64-vulkan": "1.6.0-rc.0",
+    "@fugood/node-llama-win32-arm64": "1.6.0-rc.0",
+    "@fugood/node-llama-win32-arm64-vulkan": "1.6.0-rc.0",
+    "@fugood/node-llama-win32-x64": "1.6.0-rc.0",
+    "@fugood/node-llama-win32-x64-cuda": "1.6.0-rc.0",
+    "@fugood/node-llama-win32-x64-vulkan": "1.6.0-rc.0"
   },
   "devDependencies": {
     "@babel/preset-env": "^7.24.4",

package/scripts/llama.cpp.patch

@@ -83,10 +83,10 @@ index b29544dac..5fa2c6c17 100644
  
      jinja::global_from_json(ctx, inp, inputs.mark_input);
 diff --git a/src/llama.cpp/common/chat.h b/src/llama.cpp/common/chat.h
-index 148801738..0317f1ab1 100644
+index ac19348ec..f6f9f612f 100644
 --- a/src/llama.cpp/common/chat.h
 +++ b/src/llama.cpp/common/chat.h
-@@ -222,6 +222,20 @@ common_chat_tool_choice common_chat_tool_choice_parse_oaicompat(const std::strin
+@@ -231,6 +231,20 @@ common_chat_tool_choice common_chat_tool_choice_parse_oaicompat(const std::strin
  
  bool common_chat_templates_support_enable_thinking(const common_chat_templates * chat_templates);
  
@@ -120,10 +120,10 @@ index 26250abb6..72ceddcc7 100644
      mparams.main_gpu        = params.main_gpu;
      mparams.split_mode      = params.split_mode;
 diff --git a/src/llama.cpp/common/common.h b/src/llama.cpp/common/common.h
-index b9566df62..c9425ad2f 100644
+index 96c990c05..c0b0b3093 100644
 --- a/src/llama.cpp/common/common.h
 +++ b/src/llama.cpp/common/common.h
-@@ -314,6 +314,7 @@ struct lr_opt {
+@@ -317,6 +317,7 @@ struct lr_opt {
  struct ggml_opt_optimizer_params common_opt_lr_pars(void * userdata);
  
  struct common_params {

package/src/llama.cpp/common/chat-parser.cpp

@@ -129,7 +129,7 @@ static void parse_json_tool_calls(
     }
 }
 
-common_chat_msg_parser::common_chat_msg_parser(const std::string & input, bool is_partial, const common_chat_syntax & syntax)
+common_chat_msg_parser::common_chat_msg_parser(const std::string & input, bool is_partial, const common_chat_parser_params & syntax)
     : input_(input), is_partial_(is_partial), syntax_(syntax)
 {
     result_.role = "assistant";
@@ -1611,7 +1611,7 @@ static void common_chat_parse(common_chat_msg_parser & builder) {
     builder.finish();
 }
 
-common_chat_msg common_chat_parse(const std::string & input, bool is_partial, const common_chat_syntax & syntax) {
+common_chat_msg common_chat_parse(const std::string & input, bool is_partial, const common_chat_parser_params & syntax) {
     if (syntax.format == COMMON_CHAT_FORMAT_PEG_SIMPLE ||
         syntax.format == COMMON_CHAT_FORMAT_PEG_NATIVE ||
         syntax.format == COMMON_CHAT_FORMAT_PEG_CONSTRUCTED) {
@@ -1635,7 +1635,7 @@ common_chat_msg common_chat_parse(const std::string & input, bool is_partial, co
     return msg;
 }
 
-common_chat_msg common_chat_peg_parse(const common_peg_arena & parser, const std::string & input, bool is_partial, const common_chat_syntax & syntax) {
+common_chat_msg common_chat_peg_parse(const common_peg_arena & parser, const std::string & input, bool is_partial, const common_chat_parser_params & syntax) {
     if (parser.empty()) {
         throw std::runtime_error("Failed to parse due to missing parser definition.");
     }

package/src/llama.cpp/common/chat-parser.h

@@ -5,7 +5,7 @@
 #include "json-partial.h"
 #include "regex-partial.h"
 
-#include <nlohmann/json.hpp>
+#include <nlohmann/json_fwd.hpp>
 
 #include <optional>
 #include <string>
@@ -19,20 +19,20 @@ class common_chat_msg_partial_exception : public std::runtime_error {
 class common_chat_msg_parser {
     std::string input_;
     bool is_partial_;
-    common_chat_syntax syntax_;
+    common_chat_parser_params syntax_; // TODO: rename to params
     std::string healing_marker_;
 
     size_t pos_ = 0;
     common_chat_msg result_;
 
   public:
-    common_chat_msg_parser(const std::string & input, bool is_partial, const common_chat_syntax & syntax);
+    common_chat_msg_parser(const std::string & input, bool is_partial, const common_chat_parser_params & syntax);
     const std::string & input() const { return input_; }
     size_t pos() const { return pos_; }
     const std::string & healing_marker() const { return healing_marker_; }
     const bool & is_partial() const { return is_partial_; }
     const common_chat_msg & result() const { return result_; }
-    const common_chat_syntax & syntax() const { return syntax_; }
+    const common_chat_parser_params & syntax() const { return syntax_; }
 
     void move_to(size_t pos) {
         if (pos > input_.size()) {

package/src/llama.cpp/common/chat.h

@@ -145,7 +145,7 @@ struct common_chat_templates_inputs {
     std::vector<common_chat_tool> tools;
     common_chat_tool_choice tool_choice = COMMON_CHAT_TOOL_CHOICE_AUTO;
     bool parallel_tool_calls = false;
-    common_reasoning_format reasoning_format = COMMON_REASONING_FORMAT_NONE;
+    common_reasoning_format reasoning_format = COMMON_REASONING_FORMAT_NONE; // TODO: refactor this to "bool enable_thinking"
     bool enable_thinking = true;
     std::chrono::system_clock::time_point now = std::chrono::system_clock::now();
     std::map<std::string, std::string> chat_template_kwargs;
@@ -165,14 +165,21 @@ struct common_chat_params {
     std::string                         parser;
 };
 
-struct common_chat_syntax {
+// per-message parsing syntax
+// should be derived from common_chat_params
+struct common_chat_parser_params {
     common_chat_format       format                = COMMON_CHAT_FORMAT_CONTENT_ONLY;
-    common_reasoning_format  reasoning_format      = COMMON_REASONING_FORMAT_NONE;
+    common_reasoning_format  reasoning_format      = COMMON_REASONING_FORMAT_NONE; // TODO: refactor this to "bool parse_reasoning"
     // Whether reasoning_content should be inlined in the content (e.g. for reasoning_format=deepseek in stream mode)
     bool                     reasoning_in_content  = false;
     bool                     thinking_forced_open  = false;
     bool                     parse_tool_calls      = true;
     common_peg_arena         parser                = {};
+    common_chat_parser_params() = default;
+    common_chat_parser_params(const common_chat_params & chat_params) {
+        format               = chat_params.format;
+        thinking_forced_open = chat_params.thinking_forced_open;
+    }
 };
 
 // Check if the template supplied via "--chat-template" is supported or not. Returns true if it's valid
@@ -213,10 +220,12 @@ std::string common_chat_format_example(
     const std::map<std::string, std::string> & chat_template_kwargs);
 
 const char*               common_chat_format_name(common_chat_format format);
-const char*               common_reasoning_format_name(common_reasoning_format format);
-common_reasoning_format   common_reasoning_format_from_name(const std::string & format);
-common_chat_msg           common_chat_parse(const std::string & input, bool is_partial, const common_chat_syntax & syntax);
-common_chat_msg           common_chat_peg_parse(const common_peg_arena & parser, const std::string & input, bool is_partial, const common_chat_syntax & syntax);
+common_chat_msg           common_chat_parse(const std::string & input, bool is_partial, const common_chat_parser_params & syntax);
+common_chat_msg           common_chat_peg_parse(const common_peg_arena & parser, const std::string & input, bool is_partial, const common_chat_parser_params & syntax);
+
+// used by arg and server
+const char *             common_reasoning_format_name(common_reasoning_format format);
+common_reasoning_format  common_reasoning_format_from_name(const std::string & format);
 
 common_chat_tool_choice common_chat_tool_choice_parse_oaicompat(const std::string & tool_choice);
 

package/src/llama.cpp/common/common.h

@@ -57,6 +57,8 @@ extern const char * LLAMA_COMMIT;
 extern const char * LLAMA_COMPILER;
 extern const char * LLAMA_BUILD_TARGET;
 
+const static std::string build_info("b" + std::to_string(LLAMA_BUILD_NUMBER) + "-" + LLAMA_COMMIT);
+
 struct common_control_vector_load_info;
 
 //
@@ -284,6 +286,7 @@ struct common_params_diffusion {
 };
 
 // reasoning API response format (not to be confused as chat template's reasoning format)
+// only used by server
 enum common_reasoning_format {
     COMMON_REASONING_FORMAT_NONE,
     COMMON_REASONING_FORMAT_AUTO,            // Same as deepseek, using `message.reasoning_content`

package/src/llama.cpp/common/download.cpp

@@ -314,23 +314,26 @@ static bool common_pull_file(httplib::Client & cli,
 
 // download one single file from remote URL to local path
 // returns status code or -1 on error
-static int common_download_file_single_online(const std::string & url,
-                                               const std::string & path,
-                                               const std::string & bearer_token,
-                                               const common_header_list & custom_headers) {
+static int common_download_file_single_online(const std::string        & url,
+                                              const std::string        & path,
+                                              const std::string        & bearer_token,
+                                              const common_header_list & custom_headers) {
     static const int max_attempts        = 3;
     static const int retry_delay_seconds = 2;
 
     auto [cli, parts] = common_http_client(url);
 
-    httplib::Headers default_headers = {{"User-Agent", "llama-cpp"}};
-    if (!bearer_token.empty()) {
-        default_headers.insert({"Authorization", "Bearer " + bearer_token});
-    }
+    httplib::Headers headers;
     for (const auto & h : custom_headers) {
-        default_headers.emplace(h.first, h.second);
+        headers.emplace(h.first, h.second);
     }
-    cli.set_default_headers(default_headers);
+    if (headers.find("User-Agent") == headers.end()) {
+        headers.emplace("User-Agent", "llama-cpp/" + build_info);
+    }
+    if (!bearer_token.empty()) {
+        headers.emplace("Authorization", "Bearer " + bearer_token);
+    }
+    cli.set_default_headers(headers);
 
     const bool file_exists = std::filesystem::exists(path);
 
@@ -437,10 +440,12 @@ std::pair<long, std::vector<char>> common_remote_get_content(const std::string
                                                              const common_remote_params & params) {
     auto [cli, parts] = common_http_client(url);
 
-    httplib::Headers headers = {{"User-Agent", "llama-cpp"}};
-
-    for (const auto & header : params.headers) {
-        headers.emplace(header.first, header.second);
+    httplib::Headers headers;
+    for (const auto & h : params.headers) {
+        headers.emplace(h.first, h.second);
+    }
+    if (headers.find("User-Agent") == headers.end()) {
+        headers.emplace("User-Agent", "llama-cpp/" + build_info);
     }
 
     if (params.timeout > 0) {

package/src/llama.cpp/common/json-partial.h

@@ -1,5 +1,6 @@
 #pragma once
 
+// TODO: use json_fwd.hpp when possible
 #include <nlohmann/json.hpp>
 
 // Healing marker (empty if the JSON was fully parsed / wasn't healed).