Distribute and run LLMs with a single file.
C++ cross-platformggufllama-cpplocal-ailocal-inferencelocal-llmopen-source-aisingle-file-executablespeech-to-text
Put new server code in its own namespace
Justine Tunney8a0ec61bd41d
diff --git a/llamafile/chatbot_eval.cpp b/llamafile/chatbot_eval.cppindex 397f8cb4..03aadc78 100644--- a/llamafile/chatbot_eval.cpp+++ b/llamafile/chatbot_eval.cpp@@ -89,6 +89,7 @@ bool eval_image_embed(const llava_image_embed *image_embed) {g_history.push_back(IMAGE_PLACEHOLDER_TOKEN);}clear_ephemeral();+ llama_synchronize(g_ctx);return true;}diff --git a/llamafile/server/atob.cpp b/llamafile/server/atob.cppindex 8d125127..e1705b8e 100644--- a/llamafile/server/atob.cpp+++ b/llamafile/server/atob.cpp@@ -18,6 +18,9 @@#include "utils.h"#include <string_view>+namespace lf {+namespace server {+boolatob(std::string_view val, bool dflt){@@ -31,3 +34,6 @@ atob(std::string_view val, bool dflt)return true;return dflt;}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/atom.cpp b/llamafile/server/atom.cppindex da9f9a9e..e6273457 100644--- a/llamafile/server/atom.cpp+++ b/llamafile/server/atom.cpp@@ -19,6 +19,9 @@#include "image.h"#include <cassert>+namespace lf {+namespace server {+/*** @fileoverview Context window content container.*@@ -139,3 +142,6 @@ operator==(const Atom& lhs, const Atom& rhs)return lhs.image() == rhs.image();}}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/atom.h b/llamafile/server/atom.hindex 444020d4..a09f40ed 100644--- a/llamafile/server/atom.h+++ b/llamafile/server/atom.h@@ -16,7 +16,9 @@// limitations under the License.#pragma once-#include <__fwd/string.h>++namespace lf {+namespace server {class Image;@@ -45,3 +47,6 @@ operator<(const Atom&, const Atom&);booloperator==(const Atom&, const Atom&);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/atom_test.cpp b/llamafile/server/atom_test.cppindex d9e61c00..59f8a3b3 100644--- a/llamafile/server/atom_test.cpp+++ b/llamafile/server/atom_test.cpp@@ -19,7 +19,11 @@#include "image.h"#include <cstdlib>-static const Atom atoms[] = {+namespace lf {+namespace server {+namespace {++const Atom atoms[] = {Atom(),Atom(-1),Atom(0),@@ -101,9 +105,19 @@ test_atom_operator_eq()exit(8);}-int-main()+void+atom_test(){test_atom_operator_lt();test_atom_operator_eq();}++} // namespace+} // namespace server+} // namespace lf++int+main()+{+ lf::server::atom_test();+}diff --git a/llamafile/server/atomize.cpp b/llamafile/server/atomize.cppindex a4c4fe35..8c2e37cb 100644--- a/llamafile/server/atomize.cpp+++ b/llamafile/server/atomize.cpp@@ -25,6 +25,9 @@#include <string>#include <vector>+namespace lf {+namespace server {+static voidappend_tokens(const llama_model* model,std::vector<Atom>* result,@@ -63,7 +66,7 @@ atomize(const llama_model* model,} catch (const base64_error& e) {continue;}- if (!lf::get_image_type(image))+ if (!get_image_type(image))continue;append_tokens(model, result, s.substr(0, pos), parse_special);result->emplace_back(new Image(image, -1));@@ -71,3 +74,6 @@ atomize(const llama_model* model,i = 0;}}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/buffer.cpp b/llamafile/server/buffer.cppindex ecd41226..30d0e1fd 100644--- a/llamafile/server/buffer.cpp+++ b/llamafile/server/buffer.cpp@@ -19,6 +19,9 @@#include <sys/mman.h>#include <unistd.h>+namespace lf {+namespace server {+static int pagesz = getpagesize();Buffer::Buffer(size_t capacity) noexcept@@ -45,3 +48,6 @@ Buffer::~Buffer() noexceptif (munmap(p, c + pagesz))__builtin_trap();}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/buffer.h b/llamafile/server/buffer.hindex 5a53dd91..340ee4dd 100644--- a/llamafile/server/buffer.h+++ b/llamafile/server/buffer.h@@ -18,6 +18,9 @@#pragma once#include <cstddef>+namespace lf {+namespace server {+struct Buffer{size_t i;@@ -28,3 +31,6 @@ struct Bufferexplicit Buffer(size_t) noexcept;~Buffer() noexcept;};++} // namespace server+} // namespace lfdiff --git a/llamafile/server/cleanup.cpp b/llamafile/server/cleanup.cppindex 6d2d7b02..f3d30071 100644--- a/llamafile/server/cleanup.cpp+++ b/llamafile/server/cleanup.cpp@@ -16,11 +16,12 @@// limitations under the License.#include "cleanup.h"-+#include "llama.cpp/llama.h"#include <unistd.h>#include <vector>-#include "llama.cpp/llama.h"+namespace lf {+namespace server {voidcleanup_fildes(void* arg)@@ -53,3 +54,6 @@ cleanup_llama_context(void* arg){llama_free((llama_context*)arg);}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/cleanup.h b/llamafile/server/cleanup.hindex 5fc0221a..ca19a9fb 100644--- a/llamafile/server/cleanup.h+++ b/llamafile/server/cleanup.h@@ -17,6 +17,16 @@#pragma once+namespace lf {+namespace server {++struct Cleanup+{+ Cleanup* next;+ void (*func)(void*);+ void* arg;+};+voidcleanup_fildes(void*);@@ -31,3 +41,6 @@ cleanup_llama_batch(void*);voidcleanup_llama_context(void*);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/client.cpp b/llamafile/server/client.cppindex 30b42280..aee6ad52 100644--- a/llamafile/server/client.cpp+++ b/llamafile/server/client.cpp@@ -16,7 +16,19 @@// limitations under the License.#include "client.h"-+#include "llama.cpp/llama.h"+#include "llamafile/flags.h"+#include "llamafile/llamafile.h"+#include "llamafile/server/cleanup.h"+#include "llamafile/server/log.h"+#include "llamafile/server/server.h"+#include "llamafile/server/time.h"+#include "llamafile/server/tokenbucket.h"+#include "llamafile/server/worker.h"+#include "llamafile/string.h"+#include "llamafile/threadlocal.h"+#include "llamafile/trust.h"+#include "llamafile/version.h"#include <errno.h>#include <fcntl.h>#include <limits.h>@@ -28,26 +40,14 @@#include <unistd.h>#include <vector>-#include "llama.cpp/llama.h"-#include "llamafile/flags.h"-#include "llamafile/llamafile.h"-#include "llamafile/string.h"-#include "llamafile/threadlocal.h"-#include "llamafile/trust.h"-#include "llamafile/version.h"--#include "cleanup.h"-#include "log.h"-#include "server.h"-#include "time.h"-#include "tokenbucket.h"-#include "worker.h"-#define STANDARD_RESPONSE_HEADERS \"Server: llamafile/" LLAMAFILE_VERSION_STRING "\r\n" \"Referrer-Policy: origin\r\n" \"Cache-Control: private; max-age=0\r\n"+namespace lf {+namespace server {+static voidon_http_cancel(Client* client){@@ -644,7 +644,7 @@ Client::dispatcher()// serve static endpointsint infd;size_t size;- resolved_ = lf::resolve(FLAG_www_root, p1);+ resolved_ = resolve(FLAG_www_root, p1);for (;;) {infd = open(resolved_.c_str(), O_RDONLY);if (infd == -1) {@@ -670,7 +670,7 @@ Client::dispatcher()break;} else if (S_ISDIR(st.st_mode)) {::close(infd);- resolved_ = lf::resolve(resolved_, "index.html");+ resolved_ = resolve(resolved_, "index.html");} else {::close(infd);SLOG("won't serve special file: %s", resolved_.c_str());@@ -734,3 +734,6 @@ Client::param(std::string_view key)url_.params.p[i].val.n));return {};}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/client.h b/llamafile/server/client.hindex a7fd9fcf..1df06dbc 100644--- a/llamafile/server/client.h+++ b/llamafile/server/client.h@@ -16,9 +16,8 @@// limitations under the License.#pragma once-#include "buffer.h"-+#include <ctime>#include <libc/fmt/itoa.h>#include <libc/str/slice.h>#include <net/http/http.h>@@ -26,7 +25,6 @@#include <optional>#include <string>#include <sys/resource.h>-#include <time.h>#define HasHeader(H) (!!msg_.headers[H].a)#define HeaderData(H) (ibuf_.p + msg_.headers[H].a)@@ -36,20 +34,18 @@#define HeaderEqualCase(H, S) \SlicesEqualCase(S, strlen(S), HeaderData(H), HeaderLength(H))+struct llama_model;++namespace lf {+namespace server {++struct Cleanup;struct Slot;struct Worker;-struct llama_model;struct TokenizeParams;struct EmbeddingParams;struct V1ChatCompletionParams;-struct Cleanup-{- Cleanup* next;- void (*func)(void*);- void* arg;-};-struct Client{int fd_ = -1;@@ -117,3 +113,6 @@ struct Clientbool slotz() __wur;};++} // namespace server+} // namespace lfdiff --git a/llamafile/server/embedding.cpp b/llamafile/server/embedding.cppindex 94d927c4..fa986694 100644--- a/llamafile/server/embedding.cpp+++ b/llamafile/server/embedding.cpp@@ -16,18 +16,19 @@// limitations under the License.#include "client.h"--#include <math.h>-#include <string.h>-#include <sys/resource.h>-#include <vector>-#include "llama.cpp/llama.h"#include "llamafile/server/cleanup.h"#include "llamafile/server/fastjson.h"#include "llamafile/server/json.h"#include "llamafile/server/log.h"#include "llamafile/server/utils.h"+#include <cmath>+#include <cstring>+#include <sys/resource.h>+#include <vector>++namespace lf {+namespace server {struct EmbeddingParams{@@ -332,3 +333,6 @@ Client::embedding()p = stpcpy(p, "\r\n");return send_response(headers, p, content);}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/fastjson.cpp b/llamafile/server/fastjson.cppindex 2c2c21ff..62d46ce8 100644--- a/llamafile/server/fastjson.cpp+++ b/llamafile/server/fastjson.cpp@@ -16,13 +16,14 @@// limitations under the License.#include "fastjson.h"-+#include "double-conversion/double-to-string.h"+#include "double-conversion/utils.h"#include <cosmo.h>#include <net/http/escape.h>#include <string>-#include "double-conversion/double-to-string.h"-#include "double-conversion/utils.h"+namespace lf {+namespace server {static const char kEscapeLiteral[128] = {9, 9, 9, 9, 9, 9, 9, 9, 9, 1, 2, 9, 4, 3, 9, 9, // 0x00@@ -179,3 +180,6 @@ encode_js_string_literal(char* p, const std::string_view s) noexcept*p = 0;return p;}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/fastjson.h b/llamafile/server/fastjson.hindex c0b02ccc..c906a277 100644--- a/llamafile/server/fastjson.h+++ b/llamafile/server/fastjson.h@@ -18,6 +18,9 @@#pragma once#include <__fwd/string_view.h>+namespace lf {+namespace server {+char*encode_bool(char*, bool) noexcept;@@ -44,3 +47,6 @@ encode_json(char*, const std::string_view) noexcept;char*encode_js_string_literal(char*, const std::string_view) noexcept;++} // namespace server+} // namespace lfdiff --git a/llamafile/server/fastjson_test.cpp b/llamafile/server/fastjson_test.cppindex d63d3fb1..fb115108 100644--- a/llamafile/server/fastjson_test.cpp+++ b/llamafile/server/fastjson_test.cpp@@ -17,11 +17,14 @@#include "fastjson.h"-#include <limits.h>-#include <math.h>-#include <stdio.h>+#include <climits>+#include <cmath>+#include <cstdio>#include <string>+namespace lf {+namespace server {+std::stringencode_json(int x) noexcept{@@ -72,9 +75,8 @@ encode_js_string_literal(const std::string_view x) noexcept}int-main(int argc, char* argv[])+fastjson_test(){-if (encode_json(0) != "0")return 1;if (encode_json(INT_MAX) != "2147483647")@@ -124,3 +126,12 @@ main(int argc, char* argv[])return 0;}++} // namespace server+} // namespace lf++int+main()+{+ return lf::server::fastjson_test();+}diff --git a/llamafile/server/hextoint.cpp b/llamafile/server/hextoint.cppindex f6bec4f7..dc7eafb3 100644--- a/llamafile/server/hextoint.cpp+++ b/llamafile/server/hextoint.cpp@@ -17,6 +17,9 @@#include "utils.h"+namespace lf {+namespace server {+alignas(signed char) const signed char kHexToInt[256] = {-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, // 0x00-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, // 0x10@@ -35,3 +38,6 @@ alignas(signed char) const signed char kHexToInt[256] = {-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, // 0xe0-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, // 0xf0};++} // namespace server+} // namespace lfdiff --git a/llamafile/server/image.cpp b/llamafile/server/image.cppindex be684429..abb84005 100644--- a/llamafile/server/image.cpp+++ b/llamafile/server/image.cpp@@ -20,6 +20,9 @@#include <cassert>#include <utility>+namespace lf {+namespace server {+Image::~Image() = default;Image::Image(const Image& old) : Image(old.bytes_, old.ctx_used_)@@ -68,3 +71,6 @@ operator==(const Image& lhs, const Image& rhs){return lhs.bytes() == rhs.bytes();}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/image.h b/llamafile/server/image.hindex 547444c6..2251b8fb 100644--- a/llamafile/server/image.h+++ b/llamafile/server/image.h@@ -18,6 +18,9 @@#pragma once#include <string>+namespace lf {+namespace server {+class Image{public:@@ -37,3 +40,6 @@ operator<(const Image&, const Image&);booloperator==(const Image&, const Image&);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/image_test.cpp b/llamafile/server/image_test.cppindex 6f0982fc..812c5b7d 100644--- a/llamafile/server/image_test.cpp+++ b/llamafile/server/image_test.cpp@@ -19,7 +19,11 @@#include "image.h"#include <cstdlib>-static const Image images[] = {+namespace lf {+namespace server {+namespace {++const Image images[] = {Image("", 0),Image("hello", 1),Image("there", 1),@@ -99,9 +103,19 @@ test_image_operator_eq()exit(8);}-int-main()+void+image_test(){test_image_operator_lt();test_image_operator_eq();}++} // namespace+} // namespace server+} // namespace lf++int+main()+{+ lf::server::image_test();+}diff --git a/llamafile/server/listen.cpp b/llamafile/server/listen.cppindex 1c50ec08..d39d4986 100644--- a/llamafile/server/listen.cpp+++ b/llamafile/server/listen.cpp@@ -15,19 +15,19 @@// See the License for the specific language governing permissions and// limitations under the License.-#include "server.h"-+#include "llamafile/llamafile.h"+#include "llamafile/server/log.h"+#include "llamafile/server/server.h"+#include <cstdio>+#include <cstdlib>+#include <cstring>#include <netdb.h>#include <netinet/in.h>#include <netinet/tcp.h>-#include <stdio.h>-#include <stdlib.h>-#include <string.h>#include <sys/socket.h>-#include "llamafile/llamafile.h"-#include "llamafile/server/log.h"-#include "llamafile/server/server.h"+namespace lf {+namespace server {voidprint_listening_url(unsigned ip, int port)@@ -116,3 +116,6 @@ create_listening_socket(const char* hostport)freeaddrinfo(ai);return fd;}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/log.cpp b/llamafile/server/log.cppindex a6f07530..f39ff3d6 100644--- a/llamafile/server/log.cpp+++ b/llamafile/server/log.cpp@@ -17,8 +17,11 @@#include "log.h"#include "time.h"+#include <cstring>#include <pthread.h>-#include <string.h>++namespace lf {+namespace server {static thread_local char g_thread_name[128];@@ -86,3 +89,6 @@ get_log_timestamp(void)s[25] = '0' + x / 1000 % 10;return s;}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/log.h b/llamafile/server/log.hindex 16423414..775c334c 100644--- a/llamafile/server/log.h+++ b/llamafile/server/log.h@@ -26,6 +26,9 @@get_thread_name(), \##__VA_ARGS__)+namespace lf {+namespace server {+const char*get_thread_name(void);@@ -34,3 +37,6 @@ get_log_timestamp(void);voidset_thread_name(const char*);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/main.cpp b/llamafile/server/main.cppindex cbe87c10..4659c994 100644--- a/llamafile/server/main.cpp+++ b/llamafile/server/main.cpp@@ -15,9 +15,6 @@// See the License for the specific language governing permissions and// limitations under the License.-#include <assert.h>-#include <cosmo.h>-#include "llama.cpp/llama.h"#include "llamafile/llamafile.h"#include "llamafile/pool.h"@@ -29,6 +26,11 @@#include "llamafile/server/tokenbucket.h"#include "llamafile/server/utils.h"#include "llamafile/version.h"+#include <cassert>+#include <cosmo.h>++namespace lf {+namespace server {Server* g_server;@@ -136,4 +138,14 @@ main(int argc, char* argv[])while (!pthread_orphan_np())pthread_decimate_np();CheckForMemoryLeaks();+ return 0;+}++} // namespace server+} // namespace lf++int+main(int argc, char* argv[])+{+ return lf::server::main(argc, argv);}diff --git a/llamafile/server/or_empty.cpp b/llamafile/server/or_empty.cppindex 93c76dfd..67d60a03 100644--- a/llamafile/server/or_empty.cpp+++ b/llamafile/server/or_empty.cpp@@ -16,9 +16,11 @@// limitations under the License.#include "utils.h"-#include <string_view>+namespace lf {+namespace server {+std::string_viewor_empty(std::optional<std::string_view> x){@@ -26,3 +28,6 @@ or_empty(std::optional<std::string_view> x)return x.value();return {};}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/server.cpp b/llamafile/server/server.cppindex 32d06bbf..67f02884 100644--- a/llamafile/server/server.cpp+++ b/llamafile/server/server.cpp@@ -16,21 +16,22 @@// limitations under the License.#include "server.h"-#include "slots.h"--#include <assert.h>-#include <netinet/in.h>-#include <netinet/tcp.h>-#include <stdio.h>-#include <sys/socket.h>-#include <time.h>-#include <unistd.h>-#include "llamafile/crash.h"#include "llamafile/llamafile.h"#include "llamafile/server/log.h"#include "llamafile/server/server.h"+#include "llamafile/server/slots.h"#include "llamafile/server/worker.h"+#include <cassert>+#include <cstdio>+#include <ctime>+#include <netinet/in.h>+#include <netinet/tcp.h>+#include <sys/socket.h>+#include <unistd.h>++namespace lf {+namespace server {Server::Server(int fd, Slots* slots, llama_model* model): fd(fd), slots_(slots), model_(model)@@ -194,3 +195,6 @@ Server::shutdown()unlock();}}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/server.h b/llamafile/server/server.hindex ebce3a41..05a515c5 100644--- a/llamafile/server/server.h+++ b/llamafile/server/server.h@@ -20,9 +20,13 @@#include <cosmo.h>#include <pthread.h>-struct Slots;struct llama_model;+namespace lf {+namespace server {++struct Slots;+struct Server{Server(int, Slots*, llama_model*);@@ -54,3 +58,6 @@ extern Server* g_server;intcreate_listening_socket(const char*);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/signals.cpp b/llamafile/server/signals.cppindex 09bf00d8..554f1170 100644--- a/llamafile/server/signals.cpp+++ b/llamafile/server/signals.cpp@@ -16,14 +16,15 @@// limitations under the License.#include "signals.h"--#include <signal.h>-#include <ucontext.h>-#include "llamafile/crash.h"#include "llamafile/server/log.h"#include "llamafile/server/server.h"#include "llamafile/threadlocal.h"+#include <signal.h>+#include <ucontext.h>++namespace lf {+namespace server {static struct{@@ -106,3 +107,6 @@ signals_destroy(void)sigaction(SIGXCPU, &old.sigxcpu, 0);sigaction(SIGXFSZ, &old.sigxfsz, 0);}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/signals.h b/llamafile/server/signals.hindex 1cc1d288..4e584550 100644--- a/llamafile/server/signals.h+++ b/llamafile/server/signals.h@@ -18,10 +18,14 @@#pragma once#include <signal.h>-struct StackFrame;+namespace lf {+namespace server {voidsignals_init(void);voidsignals_destroy(void);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/slot.cpp b/llamafile/server/slot.cppindex a58eb9d9..82d57109 100644--- a/llamafile/server/slot.cpp+++ b/llamafile/server/slot.cpp@@ -31,6 +31,9 @@#include <cassert>#include <cosmo.h>+namespace lf {+namespace server {+static intchoose_ctx_size(llama_model* model){@@ -295,7 +298,10 @@ Slot::dump(std::string* result)*result +=llamafile_token_to_piece(ctx_, token, RENDER_SPECIAL_TOKENS);} else if (history_[i].is_image()) {- lf::convert_image_to_uri(result, history_[i].image().bytes());+ convert_image_to_uri(result, history_[i].image().bytes());}}}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/slot.h b/llamafile/server/slot.hindex 0e019070..593f7a3c 100644--- a/llamafile/server/slot.h+++ b/llamafile/server/slot.h@@ -16,16 +16,20 @@// limitations under the License.#pragma once+#include <ctime>#include <string>-#include <time.h>#include <vector>-struct Atom;-struct Image;struct llama_context;struct llama_model;struct clip_ctx;+namespace lf {+namespace server {++struct Atom;+struct Image;+struct Slot{enum@@ -60,3 +64,6 @@ struct Slotvoid tokenize(std::vector<Atom>*, std::string_view, bool);void dump(std::string*);};++} // namespace server+} // namespace lfdiff --git a/llamafile/server/slot_entry.cpp b/llamafile/server/slot_entry.cppindex a3dc2932..01451d4e 100644--- a/llamafile/server/slot_entry.cpp+++ b/llamafile/server/slot_entry.cpp@@ -16,9 +16,12 @@// limitations under the License.#include "slot_entry.h"-#include "atom.h"-#include "slot.h"-#include <assert.h>+#include "llamafile/server/atom.h"+#include "llamafile/server/slot.h"+#include <cassert>++namespace lf {+namespace server {SlotEntry::SlotEntry(Slot* slot) : slot_(slot), key_(nullptr){@@ -51,3 +54,6 @@ operator<(const SlotEntry& lhs, const SlotEntry& rhs){return *lhs.key() < *rhs.key();}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/slot_entry.h b/llamafile/server/slot_entry.hindex e30bed0d..9d7e8c08 100644--- a/llamafile/server/slot_entry.h+++ b/llamafile/server/slot_entry.h@@ -18,6 +18,9 @@#pragma once#include <vector>+namespace lf {+namespace server {+struct Slot;struct Atom;@@ -37,3 +40,6 @@ class SlotEntrybooloperator<(const SlotEntry&, const SlotEntry&);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/slots.cpp b/llamafile/server/slots.cppindex 56c4bd7d..c0db2279 100644--- a/llamafile/server/slots.cpp+++ b/llamafile/server/slots.cpp@@ -21,7 +21,10 @@#include "llamafile/server/slot.h"#include "llamafile/server/slot_entry.h"#include "llamafile/vector.h"-#include <assert.h>+#include <cassert>++namespace lf {+namespace server {Slots::Slots(llama_model* model) : model_(model){@@ -87,7 +90,7 @@ Slots::take(const std::vector<Atom>& prefix)--i;Slot* slot = i->slot();unassert(slot);- int cpl = lf::vector_common_prefix_length(slot->history_, prefix);+ int cpl = vector_common_prefix_length(slot->history_, prefix);if (cpl == slot->history_.size()) {slots_.erase(i);pthread_mutex_unlock(&lock_);@@ -121,3 +124,6 @@ Slots::give(Slot* slot)pthread_cond_signal(&cond_);pthread_mutex_unlock(&lock_);}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/slots.h b/llamafile/server/slots.hindex d7020761..de5a2bd2 100644--- a/llamafile/server/slots.h+++ b/llamafile/server/slots.h@@ -20,10 +20,14 @@#include <set>#include <vector>+struct llama_model;++namespace lf {+namespace server {+class Atom;struct Slot;class SlotEntry;-struct llama_model;struct Slots{@@ -40,3 +44,6 @@ struct SlotsSlot* take(const std::vector<Atom>&);void give(Slot*);};++} // namespace server+} // namespace lfdiff --git a/llamafile/server/slotz.cpp b/llamafile/server/slotz.cppindex f00b0a77..97ad0f90 100644--- a/llamafile/server/slotz.cpp+++ b/llamafile/server/slotz.cpp@@ -22,6 +22,9 @@#include "utils.h"#include "worker.h"+namespace lf {+namespace server {+boolClient::slotz(){@@ -38,3 +41,6 @@ Client::slotz()p = stpcpy(p, "Content-Type: text/plain\r\n");return send_response(obuf_.p, p, dump);}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/time.cpp b/llamafile/server/time.cppindex 1ef3a741..1159f27e 100644--- a/llamafile/server/time.cpp+++ b/llamafile/server/time.cpp@@ -16,19 +16,20 @@// limitations under the License.#include "time.h"-+#include "llamafile/crash.h"+#include "llamafile/server/log.h"#include <atomic>+#include <csignal>#include <pthread.h>-#include <signal.h>#include <unistd.h>-#include "llamafile/crash.h"-#include "llamafile/server/log.h"-//// lockless implementation of gmtime_r() and localtime_r()//+namespace lf {+namespace server {+struct Clock{std::atomic_uint roll;@@ -203,3 +204,6 @@ localtime_lockless(long now, tm* tm){time_lockless(&g_clck[1], now, tm);}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/time.h b/llamafile/server/time.hindex d23a5f51..ed0c5892 100644--- a/llamafile/server/time.h+++ b/llamafile/server/time.h@@ -18,6 +18,9 @@#pragma once#include <time.h>+namespace lf {+namespace server {+voidtime_init();@@ -29,3 +32,6 @@ gmtime_lockless(long, tm*);voidlocaltime_lockless(long, tm*);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/tokenbucket.cpp b/llamafile/server/tokenbucket.cppindex c04d5f44..a277e8d1 100644--- a/llamafile/server/tokenbucket.cpp+++ b/llamafile/server/tokenbucket.cpp@@ -23,6 +23,9 @@#include <signal.h>#include <stdatomic.h>+namespace lf {+namespace server {+union TokenBucket{atomic_uint_fast64_t* w;@@ -121,3 +124,6 @@ tokenbucket_acquire(unsigned ip){return acquire_token(g_tokens.b, ip, FLAG_token_cidr);}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/tokenbucket.h b/llamafile/server/tokenbucket.hindex 62fb290d..4baca060 100644--- a/llamafile/server/tokenbucket.h+++ b/llamafile/server/tokenbucket.h@@ -17,6 +17,9 @@#pragma once+namespace lf {+namespace server {+voidtokenbucket_init();@@ -28,3 +31,6 @@ tokenbucket_acquire(unsigned);voidtokenbucket_replenish();++} // namespace server+} // namespace lfdiff --git a/llamafile/server/tokenbucket_test.cpp b/llamafile/server/tokenbucket_test.cppindex 789e4b87..149f5f00 100644--- a/llamafile/server/tokenbucket_test.cpp+++ b/llamafile/server/tokenbucket_test.cpp@@ -15,50 +15,63 @@// See the License for the specific language governing permissions and// limitations under the License.-#include "tokenbucket.h"-+#include "llamafile/llamafile.h"+#include "llamafile/server/tokenbucket.h"+#include <climits>#include <cosmo.h>-#include <limits.h>+#include <cstdlib>-#include "llamafile/llamafile.h"+namespace lf {+namespace server {+namespace {-int-main(int argc, char* argv[])+void+tokenbucket_test(){FLAG_token_rate = .0001;FLAG_token_cidr = 24;tokenbucket_init();if (tokenbucket_acquire(0x7f000001) != 0)- return 1;+ exit(1);if (tokenbucket_acquire(0x7f000002) != 1)- return 1;+ exit(1);if (tokenbucket_acquire(0x7f000003) != 2)- return 1;+ exit(1);if (tokenbucket_acquire(0x7f000100) != 0)- return 1;+ exit(1);if (tokenbucket_acquire(0x7f000101) != 1)- return 1;+ exit(1);tokenbucket_replenish();tokenbucket_replenish();if (tokenbucket_acquire(0x7f000003) != 1)- return 1;+ exit(1);if (tokenbucket_acquire(0x7f000003) != 2)- return 1;+ exit(1);if (tokenbucket_acquire(0x7f000101) != 0)- return 1;+ exit(1);tokenbucket_replenish();tokenbucket_replenish();tokenbucket_replenish();if (tokenbucket_acquire(0x7f000003) != 0)- return 1;+ exit(1);if (tokenbucket_acquire(0x7f000101) != 0)- return 1;+ exit(1);tokenbucket_destroy();CheckForMemoryLeaks();}++} // namespace+} // namespace server+} // namespace lf++int+main()+{+ lf::server::tokenbucket_test();+}diff --git a/llamafile/server/tokenize.cpp b/llamafile/server/tokenize.cppindex d32a732d..cc20e829 100644--- a/llamafile/server/tokenize.cpp+++ b/llamafile/server/tokenize.cpp@@ -23,11 +23,14 @@#include "llamafile/server/log.h"#include "llamafile/server/signals.h"#include "llamafile/server/utils.h"-#include <string.h>+#include <cstring>#include <sys/resource.h>#include <utility>#include <vector>+namespace lf {+namespace server {+struct TokenizeParams{bool add_special;@@ -171,3 +174,6 @@ Client::tokenize()p = stpcpy(p, "\r\n");return send_response(headers, p, content);}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/utils.h b/llamafile/server/utils.hindex 30205316..bd00dad5 100644--- a/llamafile/server/utils.h+++ b/llamafile/server/utils.h@@ -21,9 +21,13 @@#include <__fwd/vector.h>#include <optional>-class Atom;struct llama_model;+namespace lf {+namespace server {++class Atom;+extern const signed char kHexToInt[256];bool@@ -37,3 +41,6 @@ atomize(const llama_model* model,std::vector<Atom>* result,std::string_view s,bool parse_special);++} // namespace server+} // namespace lfdiff --git a/llamafile/server/v1_chat_completions.cpp b/llamafile/server/v1_chat_completions.cppindex d880658b..c517ee4a 100644--- a/llamafile/server/v1_chat_completions.cpp+++ b/llamafile/server/v1_chat_completions.cpp@@ -32,11 +32,14 @@#include "llamafile/server/worker.h"#include "llamafile/string.h"#include "llamafile/vector.h"-#include <math.h>-#include <string.h>+#include <cmath>+#include <cstring>#include <sys/resource.h>#include <vector>+namespace lf {+namespace server {+struct V1ChatCompletionParams{bool stream = false;@@ -61,7 +64,7 @@ struct V1ChatCompletionParamsbool should_stop(const std::vector<Atom>& history){for (const auto& suffix : stop)- if (lf::vector_ends_with(history, suffix))+ if (vector_ends_with(history, suffix))return true;return false;}@@ -576,3 +579,6 @@ Client::v1_chat_completions()return send_response(obuf_.p, p, response->content);}}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/worker.cpp b/llamafile/server/worker.cppindex 1ae3240c..668a115c 100644--- a/llamafile/server/worker.cpp+++ b/llamafile/server/worker.cpp@@ -24,10 +24,13 @@#include "llamafile/server/tokenbucket.h"#include "llamafile/threadlocal.h"#include "llamafile/trust.h"-#include <assert.h>+#include <cassert>#include <exception>#include <pthread.h>+namespace lf {+namespace server {+Worker::Worker(Server* server, llama_model* model): server_(server), client_(model){@@ -156,3 +159,6 @@ Worker::run()cleanup.set(nullptr);retire();}++} // namespace server+} // namespace lfdiff --git a/llamafile/server/worker.h b/llamafile/server/worker.hindex a40b240b..e23b8630 100644--- a/llamafile/server/worker.h+++ b/llamafile/server/worker.h@@ -22,9 +22,13 @@#define WORKER(e) DLL_CONTAINER(Worker, elem_, e)-struct Server;struct llama_model;+namespace lf {+namespace server {++struct Server;+struct Worker{Server* server_;@@ -42,3 +46,6 @@ struct Workervoid retire();void kill();};++} // namespace server+} // namespace lf