sync : vendor

common/CMakeLists.txt

@@ -58,25 +58,25 @@ add_library(${TARGET} STATIC
     arg.cpp
     arg.h
     base64.hpp
-    chat.cpp
-    chat.h
     chat-parser.cpp
     chat-parser.h
+    chat.cpp
+    chat.h
     common.cpp
     common.h
     console.cpp
     console.h
-    json-schema-to-grammar.cpp
-    json.hpp
-    json-partial.h
     json-partial.cpp
+    json-partial.h
+    json-schema-to-grammar.cpp
     llguidance.cpp
     log.cpp
     log.h
     minja/chat-template.hpp
     minja/minja.hpp
     ngram-cache.cpp
     ngram-cache.h
+    nlohmann/json.hpp
     regex-partial.cpp
     regex-partial.h
     sampling.cpp

common/chat-parser.h

@@ -2,9 +2,10 @@
 
 #include "chat.h"
 #include "json-partial.h"
-#include "json.hpp"
 #include "regex-partial.h"
 
+#include "nlohmann/json.hpp"
+
 #include <optional>
 #include <string>
 #include <vector>

common/json-partial.cpp

@@ -3,7 +3,7 @@
 #include "log.h"
 #include <string>
 
-#include <json.hpp>
+#include <nlohmann/json.hpp>
 
 using json = nlohmann::ordered_json;
 

common/json-partial.h

@@ -1,5 +1,5 @@
 #pragma once
-#include <json.hpp>
+#include <nlohmann/json.hpp>
 
 // Healing marker (empty if the JSON was fully parsed / wasn't healed).
 struct common_healing_marker {

common/json-schema-to-grammar.h

@@ -3,7 +3,7 @@
 #include "ggml.h"
 // Change JSON_ASSERT from assert() to GGML_ASSERT:
 #define JSON_ASSERT GGML_ASSERT
-#include "json.hpp"
+#include "nlohmann/json.hpp"
 
 std::string json_schema_to_grammar(const nlohmann::ordered_json & schema,
                                    bool force_gbnf = false);

common/minja/chat-template.hpp

@@ -22,7 +22,7 @@
 #include <string>
 #include <vector>
 
-#include <json.hpp>
+#include <nlohmann/json.hpp>
 
 using json = nlohmann::ordered_json;
 

common/minja/minja.hpp

@@ -29,7 +29,7 @@
 #include <utility>
 #include <vector>
 
-#include <json.hpp>
+#include <nlohmann/json.hpp>
 
 using json = nlohmann::ordered_json;
 

scripts/sync_vendor.py

@@ -0,0 +1,21 @@
+#!/usr/bin/env python3
+
+import urllib.request
+
+vendor = {
+    "https://github.com/nlohmann/json/releases/latest/download/json.hpp": "common/nlohmann/json.hpp",
+
+    # sync manually
+    # "https://raw.githubusercontent.com/ochafik/minja/refs/heads/main/include/minja/minja.hpp":         "common/minja/minja.hpp",
+    # "https://raw.githubusercontent.com/ochafik/minja/refs/heads/main/include/minja/chat-template.hpp": "common/minja/chat-template.hpp",
+
+    "https://raw.githubusercontent.com/nothings/stb/refs/heads/master/stb_image.h": "tools/mtmd/vendor/stb_image.h",
+
+    "https://github.com/mackron/miniaudio/raw/refs/tags/0.11.22/miniaudio.h": "tools/mtmd/vendor/miniaudio.h",
+
+    "https://raw.githubusercontent.com/yhirose/cpp-httplib/refs/tags/v0.20.1/httplib.h": "tools/server/httplib.h",
+}
+
+for url, filename in vendor.items():
+    print(f"downloading {url} to {filename}") # noqa: NP100
+    urllib.request.urlretrieve(url, filename)

tests/test-chat.cpp

@@ -7,9 +7,10 @@
 //
 #include <fstream>
 #include <iostream>
-#include <json.hpp>
 #include <string>
 
+#include <nlohmann/json.hpp>
+
 #include "chat.h"
 
 #include "../src/unicode.h"

tools/run/run.cpp

@@ -26,11 +26,12 @@
 
 #include "chat.h"
 #include "common.h"
-#include "json.hpp"
-#include "linenoise.cpp/linenoise.h"
 #include "llama-cpp.h"
 #include "log.h"
 
+#include "nlohmann/json.hpp"
+#include "linenoise.cpp/linenoise.h"
+
 #if defined(__unix__) || (defined(__APPLE__) && defined(__MACH__)) || defined(_WIN32)
 [[noreturn]] static void sigint_handler(int) {
     printf("\n" LOG_COL_DEFAULT);

tools/server/httplib.h

@@ -8,7 +8,7 @@
 #ifndef CPPHTTPLIB_HTTPLIB_H
 #define CPPHTTPLIB_HTTPLIB_H
 
-#define CPPHTTPLIB_VERSION "0.20.0"
+#define CPPHTTPLIB_VERSION "0.20.1"
 
 /*
  * Configuration
@@ -145,6 +145,10 @@
 #define CPPHTTPLIB_LISTEN_BACKLOG 5
 #endif
 
+#ifndef CPPHTTPLIB_MAX_LINE_LENGTH
+#define CPPHTTPLIB_MAX_LINE_LENGTH 32768
+#endif
+
 /*
  * Headers
  */
@@ -3067,6 +3071,11 @@ inline bool stream_line_reader::getline() {
 #endif
 
   for (size_t i = 0;; i++) {
+    if (size() >= CPPHTTPLIB_MAX_LINE_LENGTH) {
+      // Treat exceptionally long lines as an error to
+      // prevent infinite loops/memory exhaustion
+      return false;
+    }
     char byte;
     auto n = strm_.read(&byte, 1);
 
@@ -6055,6 +6064,8 @@ inline void calc_actual_timeout(time_t max_timeout_msec, time_t duration_msec,
   auto actual_timeout_msec =
       (std::min)(max_timeout_msec - duration_msec, timeout_msec);
 
+  if (actual_timeout_msec < 0) { actual_timeout_msec = 0; }
+
   actual_timeout_sec = actual_timeout_msec / 1000;
   actual_timeout_usec = (actual_timeout_msec % 1000) * 1000;
 }
@@ -7327,8 +7338,9 @@ Server::process_request(Stream &strm, const std::string &remote_addr,
   }
 
   // Setup `is_connection_closed` method
-  req.is_connection_closed = [&]() {
-    return !detail::is_socket_alive(strm.socket());
+  auto sock = strm.socket();
+  req.is_connection_closed = [sock]() {
+    return !detail::is_socket_alive(sock);
   };
 
   // Routing

tools/server/server.cpp

@@ -13,7 +13,7 @@
 
 // Change JSON_ASSERT from assert() to GGML_ASSERT:
 #define JSON_ASSERT GGML_ASSERT
-#include "json.hpp"
+#include "nlohmann/json.hpp"
 // mime type for sending response
 #define MIMETYPE_JSON "application/json; charset=utf-8"
 

tools/server/utils.hpp

@@ -7,6 +7,7 @@
 #include "base64.hpp"
 #include "mtmd.h"
 #include "mtmd-helper.h"
+#include "chat.h"
 
 // increase max payload length to allow use of larger context size
 #define CPPHTTPLIB_FORM_URL_ENCODED_PAYLOAD_MAX_LENGTH 1048576
@@ -16,8 +17,7 @@
 
 // Change JSON_ASSERT from assert() to GGML_ASSERT:
 #define JSON_ASSERT GGML_ASSERT
-#include "json.hpp"
-#include "chat.h"
+#include "nlohmann/json.hpp"
 
 #include <random>
 #include <sstream>

tools/tts/tts.cpp

@@ -5,7 +5,8 @@
 #include "sampling.h"
 #include "log.h"
 #include "llama.h"
-#include "json.hpp"
+
+#include "nlohmann/json.hpp"
 
 #include <algorithm>
 #include <cmath>