app : introduce the llama unified executable (#23296)

* app : introduce the llama unified executable

Signed-off-by: Adrien Gallouët <angt@huggingface.co>

* Use serve for server

Signed-off-by: Adrien Gallouët <angt@huggingface.co>

* Hide completion and bench, add help command

Signed-off-by: Adrien Gallouët <angt@huggingface.co>

* Remove STATIC

Signed-off-by: Adrien Gallouët <angt@huggingface.co>

* Use -impl targets instead of -lib

Signed-off-by: Adrien Gallouët <angt@huggingface.co>

* Revert "Remove STATIC"

This reverts commit cc44caccb9902b34a3531633edac911e5b3d65cd.

---------

Signed-off-by: Adrien Gallouët <angt@huggingface.co>

CMakeLists.txt

@@ -108,6 +108,7 @@ option(LLAMA_BUILD_TESTS            "llama: build tests"
 option(LLAMA_BUILD_TOOLS     "llama: build tools"                                                                ${LLAMA_STANDALONE})
 option(LLAMA_BUILD_EXAMPLES  "llama: build examples"                                                             ${LLAMA_STANDALONE})
 option(LLAMA_BUILD_SERVER    "llama: build server example"                                                       ${LLAMA_STANDALONE})
+option(LLAMA_BUILD_APP       "llama: build the unified binary"                                                   OFF)
 option(LLAMA_BUILD_UI        "llama: build the embedded Web UI for server"                                       ON)
 option(LLAMA_USE_PREBUILT_UI "llama: use prebuilt UI from HF Bucket when available (requires LLAMA_BUILD_UI=ON)" ON)
 
@@ -120,6 +121,7 @@ if(DEFINED LLAMA_USE_PREBUILT_WEBUI)
     set(LLAMA_USE_PREBUILT_UI ${LLAMA_USE_PREBUILT_WEBUI})
     message(DEPRECATION "LLAMA_USE_PREBUILT_WEBUI is deprecated, use LLAMA_USE_PREBUILT_UI instead")
 endif()
+
 option(LLAMA_TOOLS_INSTALL "llama: install tools" ${LLAMA_TOOLS_INSTALL_DEFAULT})
 option(LLAMA_TESTS_INSTALL "llama: install tests" ON)
 
@@ -226,6 +228,10 @@ if (LLAMA_BUILD_COMMON AND LLAMA_BUILD_TOOLS)
     add_subdirectory(tools)
 endif()
 
+if (LLAMA_BUILD_APP)
+    add_subdirectory(app)
+endif()
+
 # Automatically add all files from the 'licenses' directory
 file(GLOB EXTRA_LICENSES "${CMAKE_SOURCE_DIR}/licenses/LICENSE-*")
 

app/CMakeLists.txt

@@ -0,0 +1,11 @@
+set(TARGET llama-app)
+
+add_executable(${TARGET} llama.cpp)
+set_target_properties(${TARGET} PROPERTIES OUTPUT_NAME llama)
+
+target_link_libraries(${TARGET} PRIVATE llama-server-impl llama-cli-impl llama-completion-impl llama-bench-impl)
+target_compile_features(${TARGET} PRIVATE cxx_std_17)
+
+if(LLAMA_TOOLS_INSTALL)
+    install(TARGETS ${TARGET} RUNTIME)
+endif()

app/llama.cpp

@@ -0,0 +1,67 @@
+#include <cstdio>
+#include <string>
+#include <vector>
+
+int llama_server(int argc, char ** argv);
+int llama_cli(int argc, char ** argv);
+
+// hidden
+int llama_completion(int argc, char ** argv);
+int llama_bench(int argc, char ** argv);
+static int help(int argc, char ** argv);
+
+struct command {
+    const char * name;
+    const char * desc;
+    std::vector<std::string> aliases;
+    bool hidden;
+    int (*func)(int, char **);
+};
+
+static const command cmds[] = {
+    {"serve",      "HTTP API server",                    {"server"},   false, llama_server     },
+    {"cli",        "Command-line interactive interface", {"client"},   false, llama_cli        },
+    {"completion", "Text completion",                    {"complete"}, true,  llama_completion },
+    {"bench",      "Benchmarking tool",                  {},           true,  llama_bench      },
+    {"help",       "Show available commands",            {},           true,  help             },
+};
+
+static int help(int argc, char ** argv) {
+    const bool show_all = argc >= 2 && std::string(argv[1]) == "all";
+
+    printf("Usage: llama <command> [options]\n\nAvailable commands:\n");
+
+    for (const auto & cmd : cmds) {
+        if (show_all || !cmd.hidden) {
+            printf("  %-15s %s\n", cmd.name, cmd.desc);
+        }
+    }
+    printf("\nRun 'llama <command> --help' for command-specific usage.\n");
+
+    return 0;
+}
+
+static bool matches(const std::string & arg, const command & cmd) {
+    if (arg == cmd.name) {
+        return true;
+    }
+    for (const auto & alias : cmd.aliases) {
+        if (arg == alias) {
+            return true;
+        }
+    }
+    return false;
+}
+
+int main(int argc, char ** argv) {
+    const std::string arg = argc >= 2 ? argv[1] : "help";
+
+    for (const auto & cmd : cmds) {
+        if (matches(arg, cmd)) {
+            return cmd.func(argc - 1, argv + 1);
+        }
+    }
+
+    fprintf(stderr, "error: unknown command '%s'\n", arg.c_str());
+    return 1;
+}

tools/cli/CMakeLists.txt

@@ -1,9 +1,19 @@
-set(TARGET llama-cli)
+# llama-cli-impl: CLI logic, reusable by app
-add_executable(${TARGET} cli.cpp)
-target_link_libraries(${TARGET} PRIVATE server-context PUBLIC llama-common ${CMAKE_THREAD_LIBS_INIT})
-target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
-include_directories(../server)
+set(TARGET llama-cli-impl)
+
+add_library(${TARGET} STATIC cli.cpp)
+
+target_include_directories(${TARGET} PUBLIC ${CMAKE_CURRENT_SOURCE_DIR} ../server)
+target_link_libraries(${TARGET} PUBLIC server-context llama-common ${CMAKE_THREAD_LIBS_INIT})
+
+# llama-cli executable
+
+set(TARGET llama-cli)
+
+add_executable(${TARGET} main.cpp)
+target_link_libraries(${TARGET} PRIVATE llama-cli-impl)
+target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 if(LLAMA_TOOLS_INSTALL)
     install(TARGETS ${TARGET} RUNTIME)

tools/cli/cli.cpp

@@ -342,7 +342,10 @@ static std::vector<std::pair<std::string, size_t>> auto_completion_callback(std:
 
 static constexpr size_t FILE_GLOB_MAX_RESULTS = 100;
 
-int main(int argc, char ** argv) {
+// satisfies -Wmissing-declarations
+int llama_cli(int argc, char ** argv);
+
+int llama_cli(int argc, char ** argv) {
     common_params params;
 
     params.verbosity = LOG_LEVEL_ERROR; // by default, less verbose logs

tools/cli/main.cpp

@@ -0,0 +1,5 @@
+int llama_cli(int argc, char ** argv);
+
+int main(int argc, char ** argv) {
+    return llama_cli(argc, argv);
+}

tools/completion/CMakeLists.txt

@@ -1,6 +1,18 @@
+# llama-completion-impl: completion logic, reusable by app
+
+set(TARGET llama-completion-impl)
+
+add_library(${TARGET} STATIC completion.cpp)
+
+target_include_directories(${TARGET} PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
+target_link_libraries(${TARGET} PUBLIC llama-common llama ${CMAKE_THREAD_LIBS_INIT})
+
+# llama-completion executable
+
 set(TARGET llama-completion)
-add_executable(${TARGET} completion.cpp)
+
-target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
+add_executable(${TARGET} main.cpp)
+target_link_libraries(${TARGET} PRIVATE llama-completion-impl)
 target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 if(LLAMA_TOOLS_INSTALL)

tools/completion/completion.cpp

@@ -84,7 +84,10 @@ static void sigint_handler(int signo) {
 }
 #endif
 
-int main(int argc, char ** argv) {
+// satisfies -Wmissing-declarations
+int llama_completion(int argc, char ** argv);
+
+int llama_completion(int argc, char ** argv) {
     std::setlocale(LC_NUMERIC, "C");
 
     common_params params;

tools/completion/main.cpp

@@ -0,0 +1,5 @@
+int llama_completion(int argc, char ** argv);
+
+int main(int argc, char ** argv) {
+    return llama_completion(argc, argv);
+}

tools/llama-bench/CMakeLists.txt

@@ -1,6 +1,18 @@
+# llama-bench-impl: benchmark logic, reusable by app
+
+set(TARGET llama-bench-impl)
+
+add_library(${TARGET} STATIC llama-bench.cpp)
+
+target_include_directories(${TARGET} PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
+target_link_libraries(${TARGET} PUBLIC llama-common llama ${CMAKE_THREAD_LIBS_INIT})
+
+# llama-bench executable
+
 set(TARGET llama-bench)
-add_executable(${TARGET} llama-bench.cpp)
+
-target_link_libraries(${TARGET} PRIVATE llama-common llama ${CMAKE_THREAD_LIBS_INIT})
+add_executable(${TARGET} main.cpp)
+target_link_libraries(${TARGET} PRIVATE llama-bench-impl)
 target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 if(LLAMA_TOOLS_INSTALL)

tools/llama-bench/llama-bench.cpp

@@ -2136,7 +2136,10 @@ static std::unique_ptr<printer> create_printer(output_formats format) {
     GGML_ABORT("fatal error");
 }
 
-int main(int argc, char ** argv) {
+// satisfies -Wmissing-declarations
+int llama_bench(int argc, char ** argv);
+
+int llama_bench(int argc, char ** argv) {
     std::setlocale(LC_NUMERIC, "C");
     // try to set locale for unicode characters in markdown
     std::setlocale(LC_CTYPE, ".UTF-8");

tools/llama-bench/main.cpp

@@ -0,0 +1,5 @@
+int llama_bench(int argc, char ** argv);
+
+int main(int argc, char ** argv) {
+    return llama_bench(argc, argv);
+}

tools/server/CMakeLists.txt

@@ -27,12 +27,11 @@ target_include_directories(${TARGET} PRIVATE ../mtmd)
 target_include_directories(${TARGET} PRIVATE ${CMAKE_SOURCE_DIR})
 target_link_libraries(${TARGET} PUBLIC llama-common mtmd ${CMAKE_THREAD_LIBS_INIT})
 
+# llama-server-impl: server logic, reusable by app
 
-# llama-server executable
+set(TARGET llama-server-impl)
 
-set(TARGET llama-server)
+add_library(${TARGET} STATIC
-
-set(TARGET_SRCS
     server.cpp
     server-http.cpp
     server-http.h
@@ -40,11 +39,16 @@ set(TARGET_SRCS
     server-models.h
 )
 
-add_executable(${TARGET} ${TARGET_SRCS})
+target_include_directories(${TARGET} PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
+target_include_directories(${TARGET} PRIVATE ../mtmd ${CMAKE_SOURCE_DIR})
+target_link_libraries(${TARGET} PUBLIC server-context llama-ui cpp-httplib ${CMAKE_THREAD_LIBS_INIT})
+
+# llama-server executable
+
+set(TARGET llama-server)
+
+add_executable(${TARGET} main.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 
-target_include_directories(${TARGET} PRIVATE ../mtmd)
+target_link_libraries(${TARGET} PRIVATE llama-server-impl)
-target_include_directories(${TARGET} PRIVATE ${CMAKE_SOURCE_DIR})
-target_link_libraries(${TARGET} PRIVATE server-context llama-ui PUBLIC llama-common cpp-httplib ${CMAKE_THREAD_LIBS_INIT})
-
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

tools/server/main.cpp

@@ -0,0 +1,5 @@
+int llama_server(int argc, char ** argv);
+
+int main(int argc, char ** argv) {
+    return llama_server(argc, argv);
+}

tools/server/server.cpp

@@ -71,7 +71,10 @@ static server_http_context::handler_t ex_wrapper(server_http_context::handler_t
     };
 }
 
-int main(int argc, char ** argv) {
+// satisfies -Wmissing-declarations
+int llama_server(int argc, char ** argv);
+
+int llama_server(int argc, char ** argv) {
     std::setlocale(LC_NUMERIC, "C");
 
     // own arguments required by this example