sync : ggml

* xcframework : fix build

* mtmd : remove unused include path

* vendor : use vendor::hash alias target in cmake

CMake reserves "::" in target names for imported/alias targets, so the real
target keeps the name vendor-hash and a vendor::hash ALIAS target is added.
Consumers (mtmd, llama-gguf-hash) now link against the namespaced alias.

Assisted-by: pi:llama.cpp/Qwen3.8-27B

* vendor : add cmake targets for all vendored libs with vendor:: aliases

Add INTERFACE targets for the header-only vendor libs (miniaudio, nlohmann,
sheredom, stb) and ALIAS targets named vendor::<lib> for all of them,
including cpp-httplib and hash. Each exposes the vendor/ root so includes
are namespaced, e.g. <nlohmann/json.hpp>.

Consolidate the per-lib add_subdirectory calls into a single
add_subdirectory(vendor), keeping the cpp-httplib gate on LLAMA_BUILD_COMMON.
Consumers (llama-common, mtmd) now link the aliases instead of relying on
raw vendor/ include paths.

hash: consumers now include via "hash/hash.h"; the vendor/hash dir is kept
as a PRIVATE include so the synced upstream sources compile unmodified.

Assisted-by: pi:llama.cpp/Qwen3.8-27B

* readme : use foo/bar names in acknowledgements

Assisted-by: pi:llama.cpp/Qwen3.8-27B

* ocd : fix valign

.github/actions/linux-setup-vulkan/action.yml

@@ -1,20 +0,0 @@
-name: "Linux - Setup Vulkan SDK"
-description: "Setup Vulkan SDK for Linux"
-inputs:
-  path:
-    description: "Installation path"
-    required: true
-  version:
-    description: "Vulkan SDK version"
-    required: true
-
-runs:
-  using: "composite"
-  steps:
-    - name: Setup Vulkan SDK
-      id: setup
-      uses: ./.github/actions/unarchive-tar
-      with:
-        url: https://sdk.lunarg.com/sdk/download/${{ inputs.version }}/linux/vulkan_sdk.tar.xz
-        path: ${{ inputs.path }}
-        strip: 1

.github/workflows/build-cache.yml

@@ -10,33 +10,6 @@ concurrency:
   cancel-in-progress: true
 
 jobs:
-  ubuntu-24-vulkan-cache:
-    runs-on: ubuntu-24.04
-
-    steps:
-      - name: Clone
-        id: checkout
-        uses: actions/checkout@v6
-
-      - name: Get latest Vulkan SDK version
-        id: vulkan_sdk_version
-        run: |
-          echo "VULKAN_SDK_VERSION=$(curl https://vulkan.lunarg.com/sdk/latest/linux.txt)" >> "$GITHUB_ENV"
-
-      - name: Setup Cache
-        uses: actions/cache@v5
-        id: cache-sdk
-        with:
-          path: ./vulkan_sdk
-          key: cache-gha-vulkan-sdk-${{ env.VULKAN_SDK_VERSION }}-${{ runner.os }}
-
-      - name: Setup Vulkan SDK
-        if: steps.cache-sdk.outputs.cache-hit != 'true'
-        uses: ./.github/actions/linux-setup-vulkan
-        with:
-          path: ./vulkan_sdk
-          version: ${{ env.VULKAN_SDK_VERSION }}
-
   #ubuntu-24-spacemit-cache:
   #  runs-on: ubuntu-24.04
 

.github/workflows/build-cpu.yml

@@ -21,6 +21,7 @@ on:
     paths: [
       '.github/workflows/build-cpu.yml',
       '.github/workflows/build-cmake-pkg.yml',
+      'ggml/src/ggml-rpc/**',
       '**/CMakeLists.txt',
       '**/.cmake',
       '**/*.h',
@@ -123,7 +124,6 @@ jobs:
     env:
       OPENBLAS_VERSION: 0.3.23
       SDE_VERSION: 9.33.0-2024-01-07
-      VULKAN_VERSION: 1.4.357.0
 
     strategy:
       matrix:
@@ -134,9 +134,6 @@ jobs:
           - build: 'x64-openblas'
             arch: 'x64'
             defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/x64-windows-llvm.cmake -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DGGML_OPENMP=OFF -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS -DBLAS_INCLUDE_DIRS="$env:RUNNER_TEMP/openblas/include" -DBLAS_LIBRARIES="$env:RUNNER_TEMP/openblas/lib/openblas.lib"'
-          - build: 'x64-vulkan'
-            arch: 'x64'
-            defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/x64-windows-llvm.cmake -DCMAKE_BUILD_TYPE=Release -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DGGML_RPC=ON -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DGGML_VULKAN=ON'
           - build: 'arm64'
             arch: 'arm64'
             defines: '-G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/arm64-windows-llvm.cmake -DGGML_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON'
@@ -167,15 +164,6 @@ jobs:
           $lib =  $(join-path $msvc 'bin\Hostx64\x64\lib.exe')
           & $lib /machine:x64 "/def:${env:RUNNER_TEMP}/openblas/lib/libopenblas.def" "/out:${env:RUNNER_TEMP}/openblas/lib/openblas.lib" /name:openblas.dll
 
-      - name: Install Vulkan SDK
-        id: get_vulkan
-        if: ${{ matrix.build == 'x64-vulkan' }}
-        run: |
-          curl.exe -o $env:RUNNER_TEMP/VulkanSDK-Installer.exe -L "https://sdk.lunarg.com/sdk/download/${env:VULKAN_VERSION}/windows/vulkansdk-windows-X64-${env:VULKAN_VERSION}.exe"
-          & "$env:RUNNER_TEMP\VulkanSDK-Installer.exe" --accept-licenses --default-answer --confirm-command install
-          Add-Content $env:GITHUB_ENV "VULKAN_SDK=C:\VulkanSDK\${env:VULKAN_VERSION}"
-          Add-Content $env:GITHUB_PATH "C:\VulkanSDK\${env:VULKAN_VERSION}\bin"
-
       - name: Install Ninja
         id: install_ninja
         run: |

.github/workflows/build-rpc.yml

@@ -1,66 +0,0 @@
-name: CI (rpc)
-
-on:
-  workflow_dispatch: # allows manual triggering
-  push:
-    branches:
-      - master
-    paths: [
-      '.github/workflows/build-rpc.yml',
-      '**/CMakeLists.txt',
-      '**/.cmake',
-      '**/*.h',
-      '**/*.hpp',
-      '**/*.c',
-      '**/*.cpp'
-    ]
-
-  pull_request:
-    types: [opened, synchronize, reopened]
-    paths: [
-      '.github/workflows/build-rpc.yml',
-      'ggml/src/ggml-rpc/**'
-    ]
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
-  cancel-in-progress: true
-
-env:
-  GGML_NLOOP: 3
-  GGML_N_THREADS: 1
-  LLAMA_ARG_LOG_COLORS: 1
-  LLAMA_ARG_LOG_PREFIX: 1
-  LLAMA_ARG_LOG_TIMESTAMPS: 1
-
-jobs:
-  ubuntu-24-rpc:
-    runs-on: ${{ 'ubuntu-24.04-arm' || 'ubuntu-24.04' }}
-
-    continue-on-error: true
-
-    steps:
-      - name: Clone
-        id: checkout
-        uses: actions/checkout@v6
-
-      - name: Dependencies
-        id: depends
-        run: |
-          sudo apt-get update
-          sudo apt-get install build-essential libssl-dev ninja-build
-
-      - name: Build
-        id: cmake_build
-        run: |
-          cmake -B build \
-            -G "Ninja" \
-            -DCMAKE_BUILD_TYPE=Release \
-            -DGGML_RPC=ON
-          time cmake --build build --config Release -j $(nproc)
-
-      - name: Test
-        id: cmake_test
-        run: |
-          cd build
-          ctest -L main --verbose

.github/workflows/build-vulkan.yml

@@ -93,19 +93,13 @@ jobs:
         run: |
           echo "VULKAN_SDK_VERSION=$(curl https://vulkan.lunarg.com/sdk/latest/linux.txt)" >> "$GITHUB_ENV"
 
-      - name: Use Vulkan SDK Cache
-        uses: actions/cache@v5
-        id: cache-sdk
-        with:
-          path: ./vulkan_sdk
-          key: cache-gha-vulkan-sdk-${{ env.VULKAN_SDK_VERSION }}-${{ runner.os }}
-
       - name: Setup Vulkan SDK
-        if: steps.cache-sdk.outputs.cache-hit != 'true'
-        uses: ./.github/actions/linux-setup-vulkan
+        id: setup
+        uses: ./.github/actions/unarchive-tar
         with:
+          url: https://sdk.lunarg.com/sdk/download/${{ env.VULKAN_SDK_VERSION }}/linux/vulkan_sdk.tar.xz
           path: ./vulkan_sdk
-          version: ${{ env.VULKAN_SDK_VERSION }}
+          strip: 1
 
       - name: ccache
         uses: ggml-org/ccache-action@v1.2.21
@@ -133,3 +127,56 @@ jobs:
           # This is using llvmpipe and runs slower than other backends
           # test-backend-ops is too slow on llvmpipe, skip it
           ctest -L main -E test-backend-ops --verbose --timeout 900
+
+  windows:
+    runs-on: windows-2025
+
+    env:
+      VULKAN_VERSION: 1.4.357.0
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v6
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.21
+        with:
+          key: cpu-windows-2025-x64-vulkan
+          variant: ccache
+          evict-old-files: 1d
+          save: ${{ github.event_name == 'push' && github.ref == 'refs/heads/master' }}
+
+      - name: Install Vulkan SDK
+        id: get_vulkan
+        run: |
+          curl.exe -o $env:RUNNER_TEMP/VulkanSDK-Installer.exe -L "https://sdk.lunarg.com/sdk/download/${env:VULKAN_VERSION}/windows/vulkansdk-windows-X64-${env:VULKAN_VERSION}.exe"
+          & "$env:RUNNER_TEMP\VulkanSDK-Installer.exe" --accept-licenses --default-answer --confirm-command install
+          Add-Content $env:GITHUB_ENV "VULKAN_SDK=C:\VulkanSDK\${env:VULKAN_VERSION}"
+          Add-Content $env:GITHUB_PATH "C:\VulkanSDK\${env:VULKAN_VERSION}\bin"
+
+      - name: Install Ninja
+        id: install_ninja
+        run: |
+          choco install ninja
+
+      - name: Build
+        id: cmake_build
+        run: |
+          cmake -S . -B build -G "Ninja Multi-Config" `
+            -D CMAKE_TOOLCHAIN_FILE=cmake/x64-windows-llvm.cmake `
+            -DCMAKE_BUILD_TYPE=Release `
+            -DGGML_NATIVE=OFF `
+            -DLLAMA_BUILD_SERVER=ON `
+            -DGGML_RPC=ON `
+            -DGGML_BACKEND_DL=ON `
+            -DGGML_CPU_ALL_VARIANTS=ON `
+            -DGGML_VULKAN=ON `
+            -DLLAMA_BUILD_BORINGSSL=ON
+          cmake --build build --config Release -j ${env:NUMBER_OF_PROCESSORS}
+
+      - name: Test
+        id: cmake_test
+        run: |
+          cd build
+          ctest -L main -C Release --verbose --timeout 900

.github/workflows/make-release.yml

@@ -49,6 +49,33 @@ jobs:
           git push origin "${VERSION}"
           echo "Created and pushed tag ${VERSION}"
 
+      - name: Generate release description
+        id: desc
+        run: bash scripts/make-release-desc.sh "${{ steps.checks.outputs.version }}"
+        env:
+          GITHUB_REPOSITORY: ${{ github.repository }}
+
+      - name: Create release
+        if: ${{ github.event.inputs.dry_run == 'false' }}
+        uses: ggml-org/action-create-release@v1
+        env:
+          GITHUB_TOKEN: ${{ github.token }}
+        with:
+          tag_name: ${{ steps.checks.outputs.version }}
+          # TODO: remove the prerelease flag once the semantic versioning workflow is ready
+          # ref: https://github.com/ggml-org/ggml/discussions/1579
+          prerelease: true
+          body: |
+            > [!NOTE]
+            > Semantic versioning is still work in progress.
+            > More info can be found in https://github.com/ggml-org/ggml/discussions/1579
+
+            ${{ steps.desc.outputs.nightly }}
+
+            ## ${{ steps.desc.outputs.changelog_title }}
+
+            ${{ steps.desc.outputs.changelog }}
+
       - name: Dry run summary
         if: ${{ github.event.inputs.dry_run == 'true' }}
         run: |

CMakeLists.txt

@@ -5,7 +5,7 @@ include(CheckIncludeFileCXX)
 ### llama.cpp version
 set(LLAMA_VERSION_MAJOR 0)
 set(LLAMA_VERSION_MINOR 1)
-set(LLAMA_VERSION_PATCH 1)
+set(LLAMA_VERSION_PATCH 2)
 set(LLAMA_VERSION_BASE "${LLAMA_VERSION_MAJOR}.${LLAMA_VERSION_MINOR}.${LLAMA_VERSION_PATCH}")
 
 # whether this is a development/nightly build
@@ -224,12 +224,10 @@ add_subdirectory(src)
 # utils, programs, examples and tests
 #
 
-# mtmd needs this even when common is not built
-add_subdirectory(vendor/hash)
+add_subdirectory(vendor)
 
 if (LLAMA_BUILD_COMMON)
     add_subdirectory(common)
-    add_subdirectory(vendor/cpp-httplib)
 endif()
 
 if (LLAMA_BUILD_COMMON AND LLAMA_BUILD_TESTS AND NOT CMAKE_JS_VERSION)

README.md

@@ -119,7 +119,7 @@ The `llama.cpp` project is build on top of the [ggml](https://github.com/ggml-or
 ## Acknowledgements
 
 - [yhirose/cpp-httplib](https://github.com/yhirose/cpp-httplib) - Single-header HTTP server, used by `llama-server` - MIT license
-- [stb-image](https://github.com/nothings/stb) - Single-header image format decoder, used by multimodal subsystem - Public domain
+- [nothings/stb](https://github.com/nothings/stb) - Single-header image format decoder, used by multimodal subsystem - Public domain
 - [nlohmann/json](https://github.com/nlohmann/json) - Single-header JSON library, used by various tools/examples - MIT License
-- [miniaudio.h](https://github.com/mackron/miniaudio) - Single-header audio format decoder, used by multimodal subsystem - Public domain
-- [subprocess.h](https://github.com/sheredom/subprocess.h) - Single-header process launching solution for C and C++ - Public domain
+- [mackron/miniaudio](https://github.com/mackron/miniaudio) - Single-header audio format decoder, used by multimodal subsystem - Public domain
+- [sheredom/subprocess.h](https://github.com/sheredom/subprocess.h) - Single-header process launching solution for C and C++ - Public domain

build-xcframework.sh

@@ -290,6 +290,7 @@ combine_static_libraries() {
         "${base_dir}/${build_dir}/ggml/src/ggml-metal/${release_dir}/libggml-metal.a"
         "${base_dir}/${build_dir}/ggml/src/ggml-blas/${release_dir}/libggml-blas.a"
         "${base_dir}/${build_dir}/tools/mtmd/${release_dir}/libmtmd.a"
+        "${base_dir}/${build_dir}/vendor/hash/${release_dir}/libvendor-hash.a"
     )
 
     # Create temporary directory for processing

common/CMakeLists.txt

@@ -126,7 +126,8 @@ set_target_properties(${TARGET} PROPERTIES
     MACHO_CURRENT_VERSION 0 # keep macOS linker from seeing oversized version number
 )
 
-target_include_directories(${TARGET} PUBLIC . ../vendor)
+target_include_directories(${TARGET} PUBLIC .)
+target_link_libraries     (${TARGET} PUBLIC vendor::nlohmann vendor::sheredom)
 target_compile_features   (${TARGET} PUBLIC cxx_std_17)
 
 if (LLAMA_SUBPROCESS)

common/arg.cpp

@@ -1710,38 +1710,6 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.cache_ram_mib = value;
         }
     ).set_env("LLAMA_ARG_CACHE_RAM").set_examples({LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_CLI}));
-    add_opt(common_arg(
-        {"-cdisk", "--cache-disk"}, "PATH",
-        "directory for the disk prompt cache; prompts evicted from the RAM cache are saved here and restored on later requests, including across restarts (default: disabled, requires cache-ram)",
-        [](common_params & params, const std::string & value) {
-            params.cache_disk_path = value;
-            if (!fs_is_directory(params.cache_disk_path)) {
-                throw std::invalid_argument("not a directory: " + value);
-            }
-            // if doesn't end with DIRECTORY_SEPARATOR, add it
-            if (params.cache_disk_path[params.cache_disk_path.size() - 1] != DIRECTORY_SEPARATOR) {
-                params.cache_disk_path += DIRECTORY_SEPARATOR;
-            }
-        }
-    ).set_env("LLAMA_ARG_CACHE_DISK").set_examples({LLAMA_EXAMPLE_SERVER}));
-    add_opt(common_arg(
-        {"--cache-disk-limit"}, "N",
-        string_format("total size budget of the disk prompt cache directory in MiB; oldest entries are deleted when exceeded (default: %d, -1 - no limit)", params.cache_disk_limit_mib),
-        [](common_params & params, int value) {
-            if (value == 0 || value < -1) {
-                throw std::invalid_argument("cache-disk-limit must be positive or -1 (no limit)");
-            }
-            params.cache_disk_limit_mib = value;
-        }
-    ).set_env("LLAMA_ARG_CACHE_DISK_LIMIT").set_examples({LLAMA_EXAMPLE_SERVER}));
-    add_opt(common_arg(
-        {"--cache-disk-write-through"},
-        {"--no-cache-disk-write-through"},
-        "write prompts to the disk cache every time they are saved to the RAM cache, instead of only when evicted from it (default: disabled)",
-        [](common_params & params, bool value) {
-            params.cache_disk_write_through = value;
-        }
-    ).set_env("LLAMA_ARG_CACHE_DISK_WRITE_THROUGH").set_examples({LLAMA_EXAMPLE_SERVER}));
     add_opt(common_arg(
         {"-kvu", "--kv-unified"},
         {"-no-kvu", "--no-kv-unified"},

common/common.h

@@ -614,10 +614,6 @@ struct common_params {
     int32_t checkpoint_min_step = 8192;  // minimum spacing between context checkpoints
     int32_t cache_ram_mib       = 8192;  // -1 = no limit, 0 - disable, 1 = 1 MiB, etc.
 
-    std::string cache_disk_path;                 // disk prompt cache directory, empty = disabled
-    int32_t     cache_disk_limit_mib     = -1;   // total size budget for the disk prompt cache dir, -1 = no limit
-    bool        cache_disk_write_through = false; // also write to disk whenever a prompt is saved to the RAM cache
-
     std::string hostname      = "127.0.0.1";
     std::string public_path   = "";                                                                         // NOLINT
     std::string api_prefix    = "";                                                                         // NOLINT

examples/gguf-hash/CMakeLists.txt

@@ -2,5 +2,5 @@ set(TARGET llama-gguf-hash)
 add_executable(${TARGET} gguf-hash.cpp)
 install(TARGETS ${TARGET} RUNTIME)
 
-target_link_libraries(${TARGET} PRIVATE vendor-hash ggml ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PRIVATE vendor::hash ggml ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_17)

examples/gguf-hash/gguf-hash.cpp

@@ -17,15 +17,15 @@
 extern "C" {
 #endif
 
-#include "xxhash/xxhash.h"
-#include "sha256/sha256.h"
+#include "hash/xxhash/xxhash.h"
+#include "hash/sha256/sha256.h"
 
 #ifdef __cplusplus
 }
 #endif
 
 // sha1 is compiled as C++ and lives in a namespace, see scripts/sync_vendor.py
-#include "sha1/sha1.h"
+#include "hash/sha1/sha1.h"
 using namespace vendor_hash;
 
 

ggml/CMakeLists.txt

@@ -5,7 +5,7 @@ project("ggml" C CXX ASM)
 ### GGML Version
 set(GGML_VERSION_MAJOR 0)
 set(GGML_VERSION_MINOR 20)
-set(GGML_VERSION_PATCH 1)
+set(GGML_VERSION_PATCH 2)
 set(GGML_VERSION_BASE "${GGML_VERSION_MAJOR}.${GGML_VERSION_MINOR}.${GGML_VERSION_PATCH}")
 
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/cmake/")

ggml/src/ggml-cuda/mmvq.cu

@@ -4,6 +4,7 @@
 #include "vecdotq.cuh"
 
 #include <cstdint>
+#include <type_traits>
 
 typedef float (*vec_dot_q_cuda_t)(const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs);
 
@@ -69,7 +70,8 @@ enum mmvq_parameter_table_id {
     MMVQ_PARAMETERS_GCN,
     MMVQ_PARAMETERS_RDNA2,
     MMVQ_PARAMETERS_RDNA3_0,
-    MMVQ_PARAMETERS_RDNA4
+    MMVQ_PARAMETERS_RDNA4,
+    MMVQ_PARAMETERS_GB10
 };
 
 static constexpr __device__ mmvq_parameter_table_id get_device_table_id() {
@@ -83,6 +85,8 @@ static constexpr __device__ mmvq_parameter_table_id get_device_table_id() {
     return MMVQ_PARAMETERS_GCN;
 #elif defined(__CUDA_ARCH__) && __CUDA_ARCH__ >= GGML_CUDA_CC_TURING && __CUDA_ARCH__ < GGML_CUDA_CC_AMPERE
     return MMVQ_PARAMETERS_TURING;
+#elif defined(__CUDA_ARCH__) && __CUDA_ARCH__ == GGML_CUDA_CC_DGX_SPARK
+    return MMVQ_PARAMETERS_GB10;
 #else
     return MMVQ_PARAMETERS_GENERIC;
 #endif
@@ -104,6 +108,9 @@ static __host__ mmvq_parameter_table_id get_device_table_id(int cc) {
     if (GGML_CUDA_CC_IS_NVIDIA(cc) && ggml_cuda_highest_compiled_arch(cc) >= GGML_CUDA_CC_TURING && ggml_cuda_highest_compiled_arch(cc) < GGML_CUDA_CC_AMPERE) {
         return MMVQ_PARAMETERS_TURING;
     }
+    if (GGML_CUDA_CC_IS_NVIDIA(cc) && ggml_cuda_highest_compiled_arch(cc) == GGML_CUDA_CC_DGX_SPARK) {
+        return MMVQ_PARAMETERS_GB10;
+    }
     return MMVQ_PARAMETERS_GENERIC;
 }
 
@@ -351,7 +358,7 @@ static constexpr __device__ int get_mmvq_mmid_max_batch_for_device() {
 #endif
 }
 
-static constexpr __host__ __device__ int calc_nwarps(ggml_type type, int ncols_dst, mmvq_parameter_table_id table_id) {
+static constexpr __host__ __device__ int calc_nwarps(ggml_type type, int ncols_dst, mmvq_parameter_table_id table_id, bool small_k = false, bool halve_iters = false) {
     if (table_id == MMVQ_PARAMETERS_GENERIC) {
         switch (ncols_dst) {
             case 1:
@@ -454,11 +461,32 @@ static constexpr __host__ __device__ int calc_nwarps(ggml_type type, int ncols_d
                 return 1;
         }
     }
+    if (table_id == MMVQ_PARAMETERS_GB10) {
+        const int generic = calc_nwarps(type, ncols_dst, MMVQ_PARAMETERS_GENERIC);
+        // Only worth the wider block when it actually retires the K loop in half the trips (Observation)
+        if (ncols_dst == 1 && !small_k && halve_iters) {
+            switch (type) {
+                case GGML_TYPE_Q4_0:
+                case GGML_TYPE_Q4_1:
+                case GGML_TYPE_Q5_0:
+                case GGML_TYPE_Q5_1:
+                case GGML_TYPE_Q8_0:
+                case GGML_TYPE_Q4_K:
+                case GGML_TYPE_Q5_K:
+                case GGML_TYPE_Q6_K:
+                case GGML_TYPE_IQ4_NL:
+                    return 2 * generic;
+                default:
+                    break;
+            }
+        }
+        return generic;
+    }
     return 1;
 }
 
 static constexpr __host__ __device__ int calc_rows_per_block(int ncols_dst, int table_id, bool small_k = false, int nwarps = 1) {
-    if (table_id == MMVQ_PARAMETERS_GENERIC || table_id == MMVQ_PARAMETERS_GCN || table_id == MMVQ_PARAMETERS_TURING) {
+    if (table_id == MMVQ_PARAMETERS_GENERIC || table_id == MMVQ_PARAMETERS_GCN || table_id == MMVQ_PARAMETERS_TURING || table_id == MMVQ_PARAMETERS_GB10) {
         switch (ncols_dst) {
             case 1:
                 return small_k ? nwarps : 1;
@@ -477,8 +505,8 @@ static constexpr __host__ __device__ int calc_rows_per_block(int ncols_dst, int
     return 1;
 }
 
-template <ggml_type type, int ncols_dst, bool has_fusion, bool small_k = false>
-__launch_bounds__(calc_nwarps(type, ncols_dst, get_device_table_id())*ggml_cuda_get_physical_warp_size(), 1)
+template <ggml_type type, int ncols_dst, bool has_fusion, bool small_k = false, bool halve_iters = false>
+__launch_bounds__(calc_nwarps(type, ncols_dst, get_device_table_id(), small_k, halve_iters)*ggml_cuda_get_physical_warp_size(), 1)
 static __global__ void mul_mat_vec_q(
         const void * vx_ptr, const void * vy_ptr, const int32_t * ids_ptr, const ggml_cuda_mm_fusion_args_device fusion, float * dst_ptr,
         const uint32_t ncols_x, const uint3 nchannels_y, const uint32_t stride_row_x, const uint32_t stride_col_y,
@@ -495,7 +523,7 @@ static __global__ void mul_mat_vec_q(
     constexpr int qi  = ggml_cuda_type_traits<type>::qi;
     constexpr int vdr = get_vdr_mmvq(type);
     constexpr mmvq_parameter_table_id table_id = get_device_table_id();
-    constexpr int nwarps = calc_nwarps(type, ncols_dst, table_id);
+    constexpr int nwarps = calc_nwarps(type, ncols_dst, table_id, small_k, halve_iters);
     constexpr int rows_per_cuda_block = calc_rows_per_block(ncols_dst, table_id, small_k, nwarps);
     constexpr int warp_size = ggml_cuda_get_physical_warp_size();
 
@@ -773,8 +801,8 @@ static __global__ void mul_mat_vec_q_moe(
 template<ggml_type type>
 static std::pair<dim3, dim3> calc_launch_params(
         const int ncols_dst, const int nrows_x, const int nchannels_dst, const int nsamples_or_ntokens,
-        const int warp_size, const mmvq_parameter_table_id table_id, const bool small_k = false) {
-    const int nwarps = calc_nwarps(type, ncols_dst, table_id);
+        const int warp_size, const mmvq_parameter_table_id table_id, const bool small_k = false, const bool halve_iters = false) {
+    const int nwarps = calc_nwarps(type, ncols_dst, table_id, small_k, halve_iters);
     const int rpb = calc_rows_per_block(ncols_dst, table_id, small_k, nwarps);
     const int64_t nblocks = (nrows_x + rpb - 1) / rpb;
     const dim3 block_nums(nblocks, nchannels_dst, nsamples_or_ntokens);
@@ -782,7 +810,7 @@ static std::pair<dim3, dim3> calc_launch_params(
     return {block_nums, block_dims};
 }
 
-template<ggml_type type, int c_ncols_dst, bool small_k = false>
+template<ggml_type type, int c_ncols_dst, bool small_k = false, bool halve_iters = false>
 static void mul_mat_vec_q_switch_fusion(
         const void * vx, const void * vy, const int32_t * ids, const ggml_cuda_mm_fusion_args_device fusion, float * dst,
         const uint32_t ncols_x, const uint3 nchannels_y, const uint32_t stride_row_x, const uint32_t stride_col_y,
@@ -797,7 +825,7 @@ static void mul_mat_vec_q_switch_fusion(
     if constexpr (c_ncols_dst == 1) {
         if (has_fusion) {
             const ggml_cuda_kernel_launch_params launch_params = ggml_cuda_kernel_launch_params(block_nums, block_dims, nbytes_shared, stream);
-            ggml_cuda_kernel_launch(mul_mat_vec_q<type, c_ncols_dst, true, small_k>, launch_params,
+            ggml_cuda_kernel_launch(mul_mat_vec_q<type, c_ncols_dst, true, small_k, halve_iters>, launch_params,
                  vx, vy, ids, fusion, dst, ncols_x, nchannels_y, stride_row_x, stride_col_y, stride_col_dst,
                  channel_ratio, stride_channel_x, stride_channel_y, stride_channel_dst,
                  sample_ratio, stride_sample_x, stride_sample_y, stride_sample_dst, ids_stride);
@@ -808,7 +836,7 @@ static void mul_mat_vec_q_switch_fusion(
     GGML_ASSERT(!has_fusion && "fusion only supported for ncols_dst=1");
 
     const ggml_cuda_kernel_launch_params launch_params = ggml_cuda_kernel_launch_params(block_nums, block_dims, nbytes_shared, stream);
-    ggml_cuda_kernel_launch(mul_mat_vec_q<type, c_ncols_dst, false, small_k>, launch_params,
+    ggml_cuda_kernel_launch(mul_mat_vec_q<type, c_ncols_dst, false, small_k, halve_iters>, launch_params,
         vx, vy, ids, fusion, dst, ncols_x, nchannels_y, stride_row_x, stride_col_y, stride_col_dst,
         channel_ratio, stride_channel_x, stride_channel_y, stride_channel_dst,
         sample_ratio, stride_sample_x, stride_sample_y, stride_sample_dst, ids_stride);
@@ -860,16 +888,18 @@ static void mul_mat_vec_q_switch_ncols_dst(
 
     const bool has_ids = ids != nullptr;
 
+    // How the K loop divides up at the baseline block width, both decisions below use these.
+    constexpr int qk                    = ggml_cuda_type_traits<type>::qk;
+    constexpr int qi                    = ggml_cuda_type_traits<type>::qi;
+    constexpr int vdr                   = get_vdr_mmvq(type);
+    const int     blocks_per_row_x      = ncols_x / qk;
+    const int     blocks_per_iter_1warp = vdr * warp_size / qi;
+
     const auto should_use_small_k = [&](int c_ncols_dst) {
         // When K is small, increase rows_per_block to match nwarps so each warp has more work to do
         // Trigger when the full thread block covers all K blocks in a single loop iteration and few threads remain idle.
-        constexpr int qk                    = ggml_cuda_type_traits<type>::qk;
-        constexpr int qi                    = ggml_cuda_type_traits<type>::qi;
-        constexpr int vdr                   = get_vdr_mmvq(type);
-        const int     blocks_per_row_x      = ncols_x / qk;
-        const int     blocks_per_iter_1warp = vdr * warp_size / qi;
-        const int     nwarps                = calc_nwarps(type, c_ncols_dst, table_id);
-        bool          use                   = nwarps > 1 && blocks_per_row_x < nwarps * blocks_per_iter_1warp;
+        const int  nwarps = calc_nwarps(type, c_ncols_dst, table_id);
+        bool       use    = nwarps > 1 && blocks_per_row_x < nwarps * blocks_per_iter_1warp;
 
         constexpr std::array<ggml_type, 2> iq_slow_turing = {
             GGML_TYPE_IQ3_XXS,
@@ -902,6 +932,28 @@ static void mul_mat_vec_q_switch_ncols_dst(
         return use;
     };
 
+    // Whether doubling nwarps pays off on the ncols_dst == 1 path, where K sets the K loop trip count.
+    const auto should_halve_iters = [&] {
+        if (table_id != MMVQ_PARAMETERS_GB10) {
+            return false;
+        }
+
+        // Expert rows are gathered per token, so a wider block adds reduction work without reuse.
+        if (has_ids) {
+            return false;
+        }
+
+        const int blocks_per_iter = calc_nwarps(type, 1, table_id) * blocks_per_iter_1warp;
+        const int iters           = (blocks_per_row_x + blocks_per_iter - 1) /  blocks_per_iter;
+        const int iters_wide      = (blocks_per_row_x + blocks_per_iter * 2 - 1) / (blocks_per_iter * 2);
+
+        // An odd trip count leaves half the wider block idle for its last iteration, that tail is
+        // only affordable once the loop is long enough to dilute it to an eighth of the work (observation).
+        const int idle = iters_wide * 2 - iters;
+
+        return idle * 8 <= iters_wide * 2;
+    };
+
     if (has_ids && ncols_dst > 1) {
         // Multi-token MUL_MAT_ID path - dedicated MoE kernel
         mul_mat_vec_q_moe_launch<type>(
@@ -914,26 +966,34 @@ static void mul_mat_vec_q_switch_ncols_dst(
 
     switch (ncols_dst) {
         case 1: {
-            constexpr int c_ncols_dst = 1;
+            // static, else MSVC lambda capture breaks the constexpr uses below
+            static constexpr int c_ncols_dst = 1;
 
-            bool use_small_k = should_use_small_k(c_ncols_dst);
+            // Tag types keep the flags compile-time, so __launch_bounds__ matches what is launched.
+            const auto launch = [&](auto small_k_tag, auto halve_iters_tag) {
+                constexpr bool c_small_k = decltype(small_k_tag)::value;
+                // Types the table does not promote would compile a second, identical kernel.
+                constexpr bool c_promoted =
+                    calc_nwarps(type, c_ncols_dst, MMVQ_PARAMETERS_GB10, false, true) !=
+                    calc_nwarps(type, c_ncols_dst, MMVQ_PARAMETERS_GB10, false, false);
 
-            if (use_small_k) {
-                std::pair<dim3, dim3> dims = calc_launch_params<type>(c_ncols_dst, nrows_x, nchannels_dst,
-                                                                        nsamples_dst, warp_size, table_id, true);
-                mul_mat_vec_q_switch_fusion<type, c_ncols_dst, true>(
+                constexpr bool c_halve_iters = decltype(halve_iters_tag)::value && c_promoted;
+
+                const std::pair<dim3, dim3> dims = calc_launch_params<type>(c_ncols_dst, nrows_x, nchannels_dst,
+                                                                              nsamples_dst, warp_size, table_id, c_small_k, c_halve_iters);
+                mul_mat_vec_q_switch_fusion<type, c_ncols_dst, c_small_k, c_halve_iters>(
                     vx, vy, ids, fusion, dst, ncols_x, nchannels_y_fd, stride_row_x, stride_col_y, stride_col_dst,
                     channel_ratio_fd, stride_channel_x, stride_channel_y, stride_channel_dst, sample_ratio_fd,
                     stride_sample_x, stride_sample_y, stride_sample_dst, dims.first, dims.second, 0, ids_stride,
                     stream);
+            };
+
+            if (should_use_small_k(c_ncols_dst)) {
+                launch(std::true_type{},  std::false_type{});
+            } else if (should_halve_iters()) {
+                launch(std::false_type{}, std::true_type{});
             } else {
-                std::pair<dim3, dim3> dims = calc_launch_params<type>(c_ncols_dst, nrows_x, nchannels_dst,
-                                                                        nsamples_dst, warp_size, table_id);
-                mul_mat_vec_q_switch_fusion<type, c_ncols_dst>(
-                    vx, vy, ids, fusion, dst, ncols_x, nchannels_y_fd, stride_row_x, stride_col_y, stride_col_dst,
-                    channel_ratio_fd, stride_channel_x, stride_channel_y, stride_channel_dst, sample_ratio_fd,
-                    stride_sample_x, stride_sample_y, stride_sample_dst, dims.first, dims.second, 0, ids_stride,
-                    stream);
+                launch(std::false_type{}, std::false_type{});
             }
         } break;
         case 2: {

scripts/make-release-desc.sh

@@ -0,0 +1,87 @@
+#!/bin/bash
+# Generate the description of a release: the previous release version, the
+# change log and the link to the nightly release corresponding to the commit being released.
+#
+# Usage: make-release-desc.sh <version>
+#   <version>: current release version (v<maj>.<min>.<pat>, the leading v is optional)
+#
+# The previous version is the highest plain semver tag (v<maj>.<min>.<pat>)
+# strictly below <version>. The change log lists all commits between the
+# previous version tag and the release commit, one line per commit.
+#
+# The release commit is the commit <version> points at when the tag exists,
+# HEAD otherwise. The nightly release is the b* tag pointing at that commit
+# (release.yml tags the same commit); the link is only generated when that
+# tag exists.
+#
+# Env (when running in GitHub Actions):
+#   GITHUB_OUTPUT: previous_tag, changelog_title, changelog and nightly are written here
+#   GITHUB_REPOSITORY: owner/repo, used to build the nightly release URL (skipped when unset)
+set -euo pipefail
+
+if [[ $# -ne 1 ]]; then
+    echo "Usage: $(basename "$0") <version>"
+    exit 1
+fi
+VERSION="$1"
+
+# Accept the version with or without the leading v, reject anything else
+if [[ "${VERSION}" =~ ^[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
+    VERSION="v${VERSION}"
+elif [[ ! "${VERSION}" =~ ^v[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
+    echo "Error: invalid version '${VERSION}' (expected v<maj>.<min>.<pat>)"
+    exit 1
+fi
+
+# Make sure all remote tags are available locally (skipped on local runs without origin)
+if ! git fetch --tags origin 2>/dev/null; then
+    echo "Warning: could not fetch tags from origin (local run?)"
+fi
+
+# Release commit: the commit <version> points at when the tag exists, HEAD otherwise.
+if ! RELEASE_COMMIT="$(git rev-parse -q --verify "refs/tags/${VERSION}^{commit}" 2>/dev/null)"; then
+    RELEASE_COMMIT="$(git rev-parse HEAD)"
+fi
+
+echo "Release commit: $(git rev-parse --short "${RELEASE_COMMIT}")"
+
+PREV="$( { git tag --list; echo "${VERSION}"; } \
+    | grep -E '^v[0-9]+\.[0-9]+\.[0-9]+$' \
+    | sort -V \
+    | awk -v cur="${VERSION}" '$0 == cur { exit } { prev = $0 } END { print prev }')"
+
+if [[ -n "${PREV}" ]]; then
+    CHANGELOG="$(git log --oneline "${PREV}..${RELEASE_COMMIT}")"
+    CHANGELOG_TITLE="Change log since ${PREV}"
+else
+    CHANGELOG="(no previous release tag found)"
+    CHANGELOG_TITLE="Change log"
+fi
+
+# Nightly release: the b* tag pointing at the release commit (|| true: no match is not an error)
+NIGHTLY_TAG="$(git tag --points-at "${RELEASE_COMMIT}" | grep -E '(^|-)b[0-9]+(-[0-9a-f]{7})?$' | head -n 1 || true)"
+
+NIGHTLY=""
+if [[ -n "${NIGHTLY_TAG}" ]]; then
+    if [[ -n "${GITHUB_REPOSITORY:-}" ]]; then
+        NIGHTLY_URL="https://github.com/${GITHUB_REPOSITORY}/releases/tag/${NIGHTLY_TAG}"
+        NIGHTLY="**Nightly build:** [${NIGHTLY_TAG}](${NIGHTLY_URL})"
+        echo "Nightly release: ${NIGHTLY_URL}"
+    fi
+else
+    echo "No nightly release found for commit $(git rev-parse --short "${RELEASE_COMMIT}")"
+fi
+
+echo "Previous version: ${PREV:-none}"
+echo "${CHANGELOG}"
+
+if [[ -n "${GITHUB_OUTPUT:-}" ]]; then
+    {
+        echo "previous_tag=${PREV}"
+        echo "changelog_title=${CHANGELOG_TITLE}"
+        echo "nightly=${NIGHTLY}"
+        echo "changelog<<CHANGELOG_EOF"
+        echo "${CHANGELOG}"
+        echo "CHANGELOG_EOF"
+    } >> "${GITHUB_OUTPUT}"
+fi

scripts/sync-ggml.last

@@ -1 +1 @@
-3834fd814e74e8af277939dabd69ecc780affd21
+8c63e70982c95ceb862e3a1073a2c1beef75d60a

tools/mtmd/CMakeLists.txt

@@ -78,10 +78,8 @@ set_target_properties(mtmd PROPERTIES
 )
 
 target_link_libraries     (mtmd PUBLIC ggml llama)
-target_link_libraries     (mtmd PRIVATE Threads::Threads vendor-hash)
+target_link_libraries     (mtmd PRIVATE Threads::Threads vendor::hash vendor::miniaudio vendor::stb vendor::sheredom)
 target_include_directories(mtmd PUBLIC  .)
-target_include_directories(mtmd PRIVATE ../..)
-target_include_directories(mtmd PRIVATE ../../vendor)
 target_compile_features   (mtmd PRIVATE cxx_std_17)
 
 if (MTMD_VIDEO)

tools/mtmd/mtmd-helper.cpp

@@ -12,7 +12,7 @@
 #include "mtmd-helper-common.h"
 #include "llama.h"
 
-#include "hash.h"
+#include "hash/hash.h"
 
 #include <algorithm>
 #include <cinttypes>

tools/server/CMakeLists.txt

@@ -5,8 +5,6 @@ include_directories(${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR})
 set(TARGET server-context)
 
 add_library(${TARGET} STATIC
-    server-cache-disk.cpp
-    server-cache-disk.h
     server-chat.cpp
     server-chat.h
     server-task.cpp
@@ -33,7 +31,7 @@ endif()
 
 target_include_directories(${TARGET} PRIVATE ../mtmd)
 target_include_directories(${TARGET} PRIVATE ${CMAKE_SOURCE_DIR})
-target_link_libraries(${TARGET} PUBLIC llama-common mtmd vendor-hash ${CMAKE_THREAD_LIBS_INIT})
+target_link_libraries(${TARGET} PUBLIC llama-common mtmd ${CMAKE_THREAD_LIBS_INIT})
 
 # llama-server-impl: server logic, reusable by app
 

tools/server/README-dev.md

@@ -189,7 +189,7 @@ This endpoint is intended to be used internally by the Web UI and subject to cha
 Get a list of tools, each tool has these fields:
 - `tool` (string): the ID name of the tool, to be used in POST call. Example: `read_file`
 - `display_name` (string): the name to be displayed on UI. Example: `Read file`
-- `type` (string): `"builtin"` for a built-in tool, or `"mcp"` for a tool exposed by an MCP server
+- `type` (string): `"server"` for a server tool, or `"mcp"` for a tool exposed by an MCP server
 - `permissions` (object): a mapping string --> boolean that indicates the permission required by this tool. This is useful for the UI to ask the user before calling the tool. For now, the only permission supported is `"write"`
 - `definition` (object): the OAI-compat definition of this tool
 

tools/server/README.md

@@ -164,9 +164,6 @@ For the full list of features, please refer to [server's changelog](https://gith
 | `-ctxcp, --ctx-checkpoints, --swa-checkpoints N` | max number of context checkpoints to create per slot (default: 32)[(more info)](https://github.com/ggml-org/llama.cpp/pull/15293)<br/>(env: LLAMA_ARG_CTX_CHECKPOINTS) |
 | `-cms, --checkpoint-min-step N` | minimum spacing between context checkpoints in tokens (default: 8192, 0 = no minimum)<br/>(env: LLAMA_ARG_CHECKPOINT_MIN_SPACING_NT) |
 | `-cram, --cache-ram N` | set the maximum cache size in MiB (default: 8192, -1 - no limit, 0 - disable)[(more info)](https://github.com/ggml-org/llama.cpp/pull/16391)<br/>(env: LLAMA_ARG_CACHE_RAM) |
-| `-cdisk, --cache-disk PATH` | directory for the disk prompt cache; prompts evicted from the RAM cache are saved here and restored on later requests, including across restarts (default: disabled, requires cache-ram)<br/>(env: LLAMA_ARG_CACHE_DISK) |
-| `--cache-disk-limit N` | total size budget of the disk prompt cache directory in MiB; oldest entries are deleted when exceeded (default: -1, -1 - no limit)<br/>(env: LLAMA_ARG_CACHE_DISK_LIMIT) |
-| `--cache-disk-write-through, --no-cache-disk-write-through` | write prompts to the disk cache every time they are saved to the RAM cache, instead of only when evicted from it (default: disabled)<br/>(env: LLAMA_ARG_CACHE_DISK_WRITE_THROUGH) |
 | `-kvu, --kv-unified, -no-kvu, --no-kv-unified` | use single unified KV buffer shared across all sequences (default: enabled if number of slots is auto)<br/>(env: LLAMA_ARG_KV_UNIFIED) |
 | `--cache-idle-slots, --no-cache-idle-slots` | save idle slots to the prompt cache on new task, and clear them when using unified KV (default: enabled, requires cache-ram)<br/>(env: LLAMA_ARG_CACHE_IDLE_SLOTS) |
 | `--context-shift, --no-context-shift` | whether to use context shift on infinite text generation (default: disabled)<br/>(env: LLAMA_ARG_CONTEXT_SHIFT) |
@@ -199,11 +196,11 @@ For the full list of features, please refer to [server's changelog](https://gith
 | `--ui-config, --webui-config JSON` | JSON that provides default UI settings (overrides UI defaults)<br/>(env: LLAMA_ARG_UI_CONFIG) |
 | `--ui-config-file, --webui-config-file PATH` | JSON file that provides default UI settings (overrides UI defaults)<br/>(env: LLAMA_ARG_UI_CONFIG_FILE) |
 | `--ui-mcp-proxy, --webui-mcp-proxy, --no-ui-mcp-proxy, --no-webui-mcp-proxy` | experimental: whether to enable MCP CORS proxy - do not enable in untrusted environments (default: disabled)<br/>(env: LLAMA_ARG_UI_MCP_PROXY) |
-| `--tools TOOL1,TOOL2,...` | experimental: whether to enable built-in tools for AI agents - do not enable in untrusted environments (default: no tools)<br/>specify "all" to enable all tools<br/>available tools: read_file, file_glob_search, grep_search, exec_shell_command, write_file, edit_file, get_info<br/>note: for security reasons, this will limit --cors-origins to localhost by default<br/>(env: LLAMA_ARG_TOOLS) |
+| `--tools TOOL1,TOOL2,...` | experimental: whether to enable server tools for AI agents - do not enable in untrusted environments (default: no tools)<br/>specify "all" to enable all tools<br/>available tools: read_file, file_glob_search, grep_search, exec_shell_command, write_file, edit_file, get_info<br/>note: for security reasons, this will limit --cors-origins to localhost by default<br/>(env: LLAMA_ARG_TOOLS) |
 | `--tools-runtime OPTION` | experimental: run tools in a separate runtime environment (default: none, use host environment)<br/>available options:<br/>  'docker:<image>', 'podman:<image>': spin up a new container and reuse it for all invocations, clean up on server exit<br/>  'docker-container:<id>', 'podman-container:<id>': use an existing container by ID, won't stop on server exit<br/>  'ssh:<target>': run tools on a remote POSIX host over SSH, key-based auth and a trusted host key are required<br/><br/>(env: LLAMA_ARG_TOOLS_RUNTIME) |
 | `--mcp-servers-config PATH` | experimental: path to JSON file with MCP server definitions (Cursor-compatible format) - do not enable in untrusted environments (default: none)<br/>note: for security reasons, this will limit --cors-origins to localhost by default<br/>(env: LLAMA_ARG_MCP_SERVERS_CONFIG) |
 | `--mcp-servers-json JSON` | experimental: inline JSON with MCP server definitions (Cursor-compatible format) - do not enable in untrusted environments (default: none)<br/>note: for security reasons, this will limit --cors-origins to localhost by default<br/>(env: LLAMA_ARG_MCP_SERVERS_JSON) |
-| `-ag, --agent, -no-ag, --no-agent` | whether to enable CORS proxy and all built-in tools - do not enable in untrusted environments (default: disabled)<br/>note: for security reasons, this will limit --cors-origins to localhost by default<br/>(env: LLAMA_ARG_AGENT) |
+| `-ag, --agent, -no-ag, --no-agent` | whether to enable CORS proxy and all server tools - do not enable in untrusted environments (default: disabled)<br/>note: for security reasons, this will limit --cors-origins to localhost by default<br/>(env: LLAMA_ARG_AGENT) |
 | `--ui, --webui, --no-ui, --no-webui` | whether to enable the Web UI (default: enabled)<br/>(env: LLAMA_ARG_UI) |
 | `--embedding, --embeddings` | restrict to only support embedding use case; use only with dedicated embedding models (default: disabled)<br/>(env: LLAMA_ARG_EMBEDDINGS) |
 | `--rerank, --reranking` | enable reranking endpoint on server (default: disabled)<br/>(env: LLAMA_ARG_RERANKING) |
@@ -330,22 +327,6 @@ services:
       LLAMA_ARG_PORT: 8080
 ```
 
-### Prompt disk cache
-
-The server keeps recently used prompts (their processed KV cache state) in RAM, controlled by `--cache-ram`. With `--cache-disk PATH`, a disk tier is added below the RAM cache: entries evicted from RAM are written to the given directory, and all RAM entries are flushed there on graceful shutdown. On later requests - including after a server restart - the longest cached prefix of the incoming prompt is restored from disk instead of being re-processed.
-
-```sh
-llama-server -m model.gguf --cache-disk /path/to/cache --cache-disk-limit 32768
-```
-
-Details:
-
-- Files are named `{compat_hash}-{n_tokens}-{chain_hash}.kvc`, where the hashes identify the server configuration and the exact token prefix the file contains. Lookup is a single directory scan at startup plus one hash pass per prompt - no database is used.
-- The cache is invalidated automatically when the model file, mmproj, LoRA adapters, KV cache types, or rope parameters change (stale files are ignored, and deleted once the size budget is exceeded).
-- `--cache-disk-limit` bounds the total size of the directory in MiB; the oldest files (by modification time) are deleted first, including files left over from other models or configurations. The same directory can be shared by multiple servers.
-- By default, files are only written when an entry is evicted from the RAM cache (or on shutdown). With `--cache-disk-write-through`, every prompt saved to the RAM cache is also written to disk immediately, which is more crash-resilient at the cost of extra I/O.
-- Note that KV cache states can be large (potentially multiple GiB per prompt, depending on the model and prompt length), so make sure the disk budget is sized accordingly.
-
 ### Multimodal support
 
 Multimodal support was added in [#12898](https://github.com/ggml-org/llama.cpp/pull/12898) and is currently an experimental feature.
@@ -356,9 +337,9 @@ It is currently available in the following endpoints:
 
 For more details, please refer to [multimodal documentation](../../docs/multimodal.md)
 
-### Built-in tools support
+### Server tools support
 
-The server includes a set of built-in tools that enable the LLM to access the local file system directly from the Web UI.
+The server includes a set of server tools that enable the LLM to access the local file system directly from the Web UI.
 
 To use this feature, start the server with `--tools all`. You can also enable only specific tools by passing a comma-separated list: `--tools name1,name2,...`. Run `--help` for the full list of available tool names.
 
@@ -1650,9 +1631,9 @@ curl http://localhost:8080/v1/messages/count_tokens \
 {"input_tokens": 10}
 ```
 
-## Server built-in tools
+## Server tools
 
-The server exposes a REST API under `/tools` that allows the Web UI to call built-in tools. This endpoint is intended to be used internally by the Web UI and subject to change or to be removed in the future.
+The server exposes a REST API under `/tools` that allows the Web UI to call server tools. This endpoint is intended to be used internally by the Web UI and subject to change or to be removed in the future.
 
 **Please do NOT use this endpoint in a downstream application**
 

tools/server/server-cache-disk.cpp

@@ -1,580 +0,0 @@
-#include "server-cache-disk.h"
-
-#include "common.h"
-#include "llama.h"
-
-#include "xxhash/xxhash.h"
-
-#include <algorithm>
-#include <cinttypes>
-#include <cstdio>
-#include <cstring>
-#include <filesystem>
-#include <fstream>
-#include <functional>
-
-namespace {
-
-constexpr uint32_t SERVER_CACHE_DISK_MAGIC   = 0x3143564B; // "KVC1"
-constexpr uint32_t SERVER_CACHE_DISK_VERSION = 1;
-
-// seed for the chained prefix hash - changing it invalidates all filenames
-constexpr uint64_t SERVER_CACHE_DISK_CHAIN_SEED = 0x6b7663636861696eULL;
-
-struct server_cache_disk_file_header {
-    uint32_t magic       = SERVER_CACHE_DISK_MAGIC;
-    uint32_t version     = SERVER_CACHE_DISK_VERSION;
-    uint64_t compat_hash = 0; // full 64-bit value (the filename only carries the low 32 bits)
-    uint64_t chain_hash  = 0;
-    uint32_t n_tokens    = 0;
-    uint32_t pad         = 0;
-    uint64_t tokens_size = 0; // bytes of the server_tokens::serialize() section
-    uint64_t state_size  = 0; // bytes of the llama_state_seq_get_data section
-};
-
-static_assert(sizeof(server_cache_disk_file_header) == 48, "unexpected header size");
-
-std::string make_filename(uint64_t compat_hash, uint32_t n_tokens, uint64_t chain_hash) {
-    char buf[64];
-    snprintf(buf, sizeof(buf), "%08x-%u-%016" PRIx64 ".kvc", (uint32_t) compat_hash, n_tokens, chain_hash);
-    return buf;
-}
-
-bool parse_filename(const std::string & name, uint32_t & compat32, uint32_t & n_tokens, uint64_t & chain_hash) {
-    if (sscanf(name.c_str(), "%8x-%u-%16" SCNx64 ".kvc", &compat32, &n_tokens, &chain_hash) != 3) {
-        return false;
-    }
-
-    // reject padding/case/suffix variations by requiring the canonical spelling
-    return name == make_filename(compat32, n_tokens, chain_hash);
-}
-
-int64_t file_mtime(const std::filesystem::path & path) {
-    std::error_code ec;
-    const auto t = std::filesystem::last_write_time(path, ec);
-    return ec ? 0 : (int64_t) t.time_since_epoch().count();
-}
-
-uint64_t covered_key(uint32_t n_tokens, uint64_t chain_hash) {
-    const uint64_t buf[2] = { n_tokens, chain_hash };
-    return XXH64(buf, sizeof(buf), 0);
-}
-
-// walk the chained hash over the token list, invoking cb(n, h) at every valid prefix boundary:
-// after each text token and after each complete media chunk (never mid-chunk)
-// returns true if the walk reached n_max
-bool tokens_chain_hash_walk(const server_tokens & tokens, size_t n_max, const std::function<bool(size_t, uint64_t)> & cb) {
-    uint64_t h = SERVER_CACHE_DISK_CHAIN_SEED;
-
-    size_t i = 0;
-
-    try {
-        while (i < n_max) {
-            const llama_token tok = tokens[i];
-
-            if (tok == LLAMA_TOKEN_NULL) {
-                // media chunk - fold in its content id instead of the placeholder token ids,
-                // otherwise different images would hash identically
-                const auto & chunk = tokens.find_chunk(i);
-
-                const char * id    = mtmd_input_chunk_get_id(chunk.get());
-                const size_t n_tok = mtmd_input_chunk_get_n_tokens(chunk.get());
-
-                if (id == nullptr || id[0] == '\0' || n_tok == 0 || i + n_tok > n_max) {
-                    return false;
-                }
-
-                std::vector<uint8_t> buf;
-                buf.reserve(5 + strlen(id));
-                buf.push_back(0x01);
-                for (int b = 0; b < 4; ++b) {
-                    buf.push_back((uint8_t) (n_tok >> (8*b)));
-                }
-                buf.insert(buf.end(), id, id + strlen(id));
-
-                h = XXH64(buf.data(), buf.size(), h);
-
-                i += n_tok;
-            } else {
-                uint8_t buf[5] = { 0x00 };
-                memcpy(buf + 1, &tok, sizeof(tok));
-
-                h = XXH64(buf, sizeof(buf), h);
-
-                i += 1;
-            }
-
-            if (!cb(i, h)) {
-                return false;
-            }
-        }
-    } catch (const std::exception & e) {
-        SRV_WRN("failed to hash token list: %s\n", e.what());
-        return false;
-    }
-
-    return true;
-}
-
-} // namespace
-
-server_prompt_cache_disk::server_prompt_cache_disk(const std::string & dir_, uint64_t compat_hash, bool has_mtmd, int32_t limit_mib, bool write_through) :
-        write_through(write_through),
-        dir(dir_.empty() || dir_.back() == DIRECTORY_SEPARATOR ? dir_ : dir_ + DIRECTORY_SEPARATOR),
-        compat_hash(compat_hash),
-        has_mtmd(has_mtmd),
-        limit_bytes(limit_mib < 0 ? 0 : 1024ull*1024ull*limit_mib) {
-    scan_dir();
-}
-
-void server_prompt_cache_disk::scan_dir() {
-    namespace fs = std::filesystem;
-
-    std::error_code ec;
-
-    for (const auto & ent : fs::directory_iterator(dir, ec)) {
-        if (!ent.is_regular_file(ec)) {
-            continue;
-        }
-
-        const std::string name = ent.path().filename().string();
-
-        // leftover temporary files from a previous crash
-        if (name.size() > 4 && name.compare(name.size() - 4, 4, ".tmp") == 0 && name[0] == '.') {
-            fs::remove(ent.path(), ec);
-            continue;
-        }
-
-        uint32_t compat32  = 0;
-        uint32_t n_tokens  = 0;
-        uint64_t chain     = 0;
-
-        if (!parse_filename(name, compat32, n_tokens, chain)) {
-            continue;
-        }
-
-        server_cache_disk_file file;
-        file.name       = name;
-        file.chain_hash = chain;
-        file.n_tokens   = n_tokens;
-        file.n_bytes    = ent.file_size(ec);
-        file.mtime      = file_mtime(ent.path());
-
-        total_bytes += file.n_bytes;
-
-        if (compat32 == (uint32_t) compat_hash) {
-            index[n_tokens][chain] = std::move(file);
-        } else {
-            foreign.push_back(std::move(file));
-        }
-    }
-
-    SRV_INF("disk prompt cache '%s': %zu usable entries, %zu from other configurations, %.3f MiB total (budget: %.3f MiB)\n",
-            dir.c_str(), n_files(), foreign.size(), total_bytes / (1024.0 * 1024.0), limit_bytes / (1024.0 * 1024.0));
-}
-
-size_t server_prompt_cache_disk::n_files() const {
-    size_t res = 0;
-
-    for (const auto & [n, files] : index) {
-        res += files.size();
-    }
-
-    return res;
-}
-
-server_cache_disk_file * server_prompt_cache_disk::find_file(uint32_t n_tokens, uint64_t chain_hash) {
-    const auto it = index.find(n_tokens);
-    if (it == index.end()) {
-        return nullptr;
-    }
-
-    const auto it_file = it->second.find(chain_hash);
-
-    return it_file == it->second.end() ? nullptr : &it_file->second;
-}
-
-const server_cache_disk_file * server_prompt_cache_disk::lookup(const server_tokens & tokens, size_t n_max) const {
-    if (index.empty()) {
-        return nullptr;
-    }
-
-    // no file can be longer than the largest indexed length - cap the walk
-    n_max = std::min<size_t>(n_max, index.rbegin()->first);
-
-    const server_cache_disk_file * best = nullptr;
-
-    tokens_chain_hash_walk(tokens, n_max, [&](size_t n, uint64_t h) {
-        const auto it = index.find((uint32_t) n);
-        if (it != index.end()) {
-            const auto it_file = it->second.find(h);
-            if (it_file != it->second.end()) {
-                best = &it_file->second;
-            }
-        }
-
-        return true;
-    });
-
-    return best;
-}
-
-void server_prompt_cache_disk::touch(const server_cache_disk_file & file) {
-    std::error_code ec;
-    std::filesystem::last_write_time(dir + file.name, std::filesystem::file_time_type::clock::now(), ec);
-
-    if (auto * f = find_file(file.n_tokens, file.chain_hash)) {
-        f->mtime = file_mtime(dir + file.name);
-    }
-}
-
-void server_prompt_cache_disk::forget(const server_cache_disk_file & file) {
-    // copy the fields first - the reference may point into the index entry being erased
-    const uint32_t n_tokens = file.n_tokens;
-    const uint64_t chain    = file.chain_hash;
-    const uint64_t n_bytes  = file.n_bytes;
-
-    const auto it = index.find(n_tokens);
-    if (it == index.end()) {
-        return;
-    }
-
-    if (it->second.erase(chain) > 0) {
-        total_bytes -= std::min<size_t>(total_bytes, n_bytes);
-    }
-
-    if (it->second.empty()) {
-        index.erase(it);
-    }
-}
-
-void server_prompt_cache_disk::remove_file(const server_cache_disk_file & file) {
-    SRV_WRN("disk prompt cache: removing '%s'\n", file.name.c_str());
-
-    std::error_code ec;
-    std::filesystem::remove(dir + file.name, ec);
-
-    forget(file);
-}
-
-void server_prompt_cache_disk::enforce_budget(const std::string & name_protected) {
-    if (limit_bytes == 0) {
-        return;
-    }
-
-    while (total_bytes > limit_bytes) {
-        // find the oldest file, ours and foreign alike
-        const server_cache_disk_file * oldest = nullptr;
-        bool oldest_foreign = false;
-
-        for (const auto & [n, files] : index) {
-            for (const auto & [h, file] : files) {
-                if (file.name != name_protected && (!oldest || file.mtime < oldest->mtime)) {
-                    oldest = &file;
-                    oldest_foreign = false;
-                }
-            }
-        }
-
-        for (const auto & file : foreign) {
-            if (file.name != name_protected && (!oldest || file.mtime < oldest->mtime)) {
-                oldest = &file;
-                oldest_foreign = true;
-            }
-        }
-
-        if (!oldest) {
-            break;
-        }
-
-        SRV_INF("disk prompt cache: size %.3f MiB over budget %.3f MiB, evicting oldest entry '%s'\n",
-                total_bytes / (1024.0 * 1024.0), limit_bytes / (1024.0 * 1024.0), oldest->name.c_str());
-
-        if (oldest_foreign) {
-            std::error_code ec;
-            std::filesystem::remove(dir + oldest->name, ec);
-
-            total_bytes -= std::min<size_t>(total_bytes, oldest->n_bytes);
-
-            foreign.erase(foreign.begin() + (oldest - foreign.data()));
-        } else {
-            remove_file(*oldest);
-        }
-    }
-}
-
-bool server_prompt_cache_disk::store(const server_tokens & tokens, const std::vector<uint8_t> & state_main) {
-    if (tokens.empty() || state_main.empty()) {
-        return false;
-    }
-
-    std::vector<std::pair<size_t, uint64_t>> bounds;
-
-    if (!tokens_chain_hash_walk(tokens, tokens.size(), [&](size_t n, uint64_t h) { bounds.emplace_back(n, h); return true; }) ||
-        bounds.empty() || bounds.back().first != tokens.size()) {
-        SRV_WRN("%s", "disk prompt cache: token list cannot be hashed, skipping\n");
-        return false;
-    }
-
-    const uint32_t n_tokens = (uint32_t) tokens.size();
-    const uint64_t chain    = bounds.back().second;
-
-    if (auto * existing = find_file(n_tokens, chain)) {
-        SRV_TRC("disk prompt cache: '%s' already exists, refreshing\n", existing->name.c_str());
-        touch(*existing);
-        return true;
-    }
-
-    if (covered.count(covered_key(n_tokens, chain)) > 0) {
-        SRV_TRC(" - prompt with %u tokens is a prefix of an already persisted entry, skipping\n", n_tokens);
-        return true;
-    }
-
-    std::vector<char> tok_data;
-    try {
-        tok_data = tokens.serialize();
-    } catch (const std::exception & e) {
-        SRV_WRN("disk prompt cache: failed to serialize tokens: %s\n", e.what());
-        return false;
-    }
-
-    server_cache_disk_file_header header;
-    header.compat_hash = compat_hash;
-    header.chain_hash  = chain;
-    header.n_tokens    = n_tokens;
-    header.tokens_size = tok_data.size();
-    header.state_size  = state_main.size();
-
-    const std::string name = make_filename(compat_hash, n_tokens, chain);
-
-    char tmp_buf[64];
-    snprintf(tmp_buf, sizeof(tmp_buf), ".%08x-%u.tmp", (uint32_t) (uintptr_t) this, tmp_counter++);
-
-    const std::string path_tmp = dir + tmp_buf;
-    const std::string path     = dir + name;
-
-    {
-        std::ofstream out(path_tmp, std::ios::binary | std::ios::trunc);
-
-        out.write((const char *) &header, sizeof(header));
-        out.write(tok_data.data(), tok_data.size());
-        out.write((const char *) state_main.data(), state_main.size());
-
-        if (!out.good()) {
-            SRV_ERR("disk prompt cache: failed to write '%s'\n", path_tmp.c_str());
-
-            out.close();
-
-            std::error_code ec;
-            std::filesystem::remove(path_tmp, ec);
-
-            return false;
-        }
-    }
-
-    std::error_code ec;
-    std::filesystem::rename(path_tmp, path, ec);
-    if (ec) {
-        SRV_ERR("disk prompt cache: failed to rename '%s' to '%s': %s\n", path_tmp.c_str(), path.c_str(), ec.message().c_str());
-
-        std::filesystem::remove(path_tmp, ec);
-
-        return false;
-    }
-
-    server_cache_disk_file file;
-    file.name       = name;
-    file.chain_hash = chain;
-    file.n_tokens   = n_tokens;
-    file.n_bytes    = sizeof(header) + tok_data.size() + state_main.size();
-    file.mtime      = file_mtime(path);
-
-    total_bytes += file.n_bytes;
-
-    index[n_tokens][chain] = std::move(file);
-
-    for (const auto & [n, h] : bounds) {
-        covered.insert(covered_key((uint32_t) n, h));
-    }
-
-    SRV_INF("disk prompt cache: saved prompt with %u tokens, %.3f MiB to '%s'\n",
-            n_tokens, (sizeof(header) + tok_data.size() + state_main.size()) / (1024.0 * 1024.0), name.c_str());
-    SRV_DBG("%s", "__TEST_TAG_CACHE_DISK_STORE__\n");
-
-    enforce_budget(name);
-
-    return true;
-}
-
-server_prompt_cache_disk::load_status server_prompt_cache_disk::load(
-        server_cache_disk_file file, const server_tokens & tokens_new, llama_context * ctx, int32_t id_slot, server_tokens & tokens_out) {
-    const std::string path = dir + file.name;
-
-    std::error_code ec;
-    const uint64_t n_bytes = std::filesystem::file_size(path, ec);
-
-    if (ec) {
-        // deleted by another process - not an error, just a miss
-        forget(file);
-        return LOAD_MISS;
-    }
-
-    std::ifstream in(path, std::ios::binary);
-    if (!in.good()) {
-        forget(file);
-        return LOAD_MISS;
-    }
-
-    server_cache_disk_file_header header;
-    in.read((char *) &header, sizeof(header));
-
-    if (!in.good() ||
-        header.magic    != SERVER_CACHE_DISK_MAGIC   ||
-        header.version  != SERVER_CACHE_DISK_VERSION ||
-        header.chain_hash  != file.chain_hash ||
-        header.n_tokens    != file.n_tokens   ||
-        header.tokens_size % sizeof(llama_token) != 0 ||
-        sizeof(header) + header.tokens_size + header.state_size != n_bytes) {
-        SRV_WRN("disk prompt cache: '%s' is corrupt\n", file.name.c_str());
-        remove_file(file);
-        return LOAD_MISS;
-    }
-
-    if (header.compat_hash != compat_hash) {
-        // same low 32 bits, different configuration - leave the file for its owner
-        SRV_WRN("disk prompt cache: '%s' belongs to a different configuration, ignoring\n", file.name.c_str());
-        forget(file);
-        return LOAD_MISS;
-    }
-
-    llama_tokens packed(header.tokens_size / sizeof(llama_token));
-    in.read((char *) packed.data(), header.tokens_size);
-
-    if (!in.good()) {
-        SRV_WRN("disk prompt cache: '%s' is truncated\n", file.name.c_str());
-        remove_file(file);
-        return LOAD_MISS;
-    }
-
-    server_tokens loaded;
-    try {
-        loaded = server_tokens::deserialize(packed, has_mtmd);
-    } catch (const std::exception & e) {
-        SRV_WRN("disk prompt cache: failed to deserialize tokens from '%s': %s\n", file.name.c_str(), e.what());
-        remove_file(file);
-        return LOAD_MISS;
-    }
-
-    // the filename hash only proves an exact prefix probabilistically - verify against the actual tokens
-    if (loaded.size() != file.n_tokens ||
-        loaded.get_common_prefix(tokens_new) != file.n_tokens ||
-        !loaded.validate(ctx)) {
-        SRV_WRN("disk prompt cache: token mismatch in '%s' (hash collision?)\n", file.name.c_str());
-        remove_file(file);
-        return LOAD_MISS;
-    }
-
-    std::vector<uint8_t> state;
-    try {
-        state.resize(header.state_size);
-    } catch (const std::bad_alloc &) {
-        SRV_ERR("disk prompt cache: failed to allocate %" PRIu64 " bytes for '%s'\n", header.state_size, file.name.c_str());
-        return LOAD_MISS;
-    }
-
-    in.read((char *) state.data(), state.size());
-
-    if (!in.good()) {
-        SRV_WRN("disk prompt cache: '%s' is truncated\n", file.name.c_str());
-        remove_file(file);
-        return LOAD_MISS;
-    }
-
-    const size_t n = llama_state_seq_set_data_ext(ctx, state.data(), state.size(), id_slot, 0);
-    if (n != state.size()) {
-        SRV_WRN("disk prompt cache: failed to restore state from '%s' (%zu / %zu bytes)\n", file.name.c_str(), n, state.size());
-
-        // the sequence may hold a partial state now - clear it and let the caller recover
-        llama_memory_seq_rm(llama_get_memory(ctx), id_slot, -1, -1);
-
-        return LOAD_FAIL_SEQ_DIRTY;
-    }
-
-    tokens_out = std::move(loaded);
-
-    covered.insert(covered_key(file.n_tokens, file.chain_hash));
-
-    touch(file);
-
-    SRV_INF("disk prompt cache: restored prompt with %u tokens, %.3f MiB from '%s'\n",
-            file.n_tokens, state.size() / (1024.0 * 1024.0), file.name.c_str());
-    SRV_DBG("%s", "__TEST_TAG_CACHE_DISK_HIT__\n");
-
-    return LOAD_OK;
-}
-
-//
-// compat hash
-//
-
-namespace {
-
-template <typename T>
-void hash_pod(std::string & blob, const T & value) {
-    static_assert(std::is_trivially_copyable<T>::value, "hash_pod requires a POD type");
-    blob.append((const char *) &value, sizeof(value));
-}
-
-void hash_str(std::string & blob, const std::string & value) {
-    blob += value;
-    blob += '\0';
-}
-
-// path + size + mtime: conservative, but never misses a changed file
-void hash_file_meta(std::string & blob, const std::string & path) {
-    hash_str(blob, path);
-
-    std::error_code ec;
-
-    const uint64_t size = path.empty() ? 0 : (uint64_t) std::filesystem::file_size(path, ec);
-    hash_pod(blob, ec ? (uint64_t) 0 : size);
-
-    hash_pod(blob, path.empty() ? (int64_t) 0 : file_mtime(path));
-}
-
-} // namespace
-
-uint64_t server_cache_disk_compat_hash(const common_params & params) {
-    std::string blob;
-
-    // format versions
-    hash_pod(blob, (uint32_t) SERVER_CACHE_DISK_VERSION);
-    hash_pod(blob, (uint32_t) LLAMA_STATE_SEQ_VERSION);
-    hash_pod(blob, (uint32_t) server_tokens::SERVER_TOKENS_STATE_VERSION);
-
-    // model identity
-    hash_file_meta(blob, params.model.path);
-    hash_file_meta(blob, params.mmproj.path);
-
-    for (const auto & la : params.lora_adapters) {
-        hash_file_meta(blob, la.path);
-        hash_pod(blob, la.scale);
-    }
-
-    // KV cache layout
-    hash_pod(blob, (int32_t) params.cache_type_k);
-    hash_pod(blob, (int32_t) params.cache_type_v);
-    hash_pod(blob, (uint8_t) params.swa_full);
-
-    // rope params change the KV content for the same tokens
-    hash_pod(blob, params.rope_freq_base);
-    hash_pod(blob, params.rope_freq_scale);
-    hash_pod(blob, (int32_t) params.rope_scaling_type);
-    hash_pod(blob, params.yarn_ext_factor);
-    hash_pod(blob, params.yarn_attn_factor);
-    hash_pod(blob, params.yarn_beta_fast);
-    hash_pod(blob, params.yarn_beta_slow);
-    hash_pod(blob, params.yarn_orig_ctx);
-
-    return XXH64(blob.data(), blob.size(), 0);
-}

tools/server/server-cache-disk.h

@@ -1,96 +0,0 @@
-#pragma once
-
-#include "server-common.h"
-
-#include <cstdint>
-#include <map>
-#include <string>
-#include <unordered_map>
-#include <unordered_set>
-#include <vector>
-
-struct common_params;
-struct llama_context;
-
-// disk-backed prompt cache: a cold tier below the in-RAM server_prompt_cache
-//
-// each entry is one file in a flat directory, named after the exact token prefix it contains:
-//
-//     {compat_hash8}-{n_tokens}-{chain_hash16}.kvc
-//
-//   - compat_hash: hash of everything that invalidates a KV state (model file, mmproj, loras,
-//     cache types, rope params, ...) - see server_cache_disk_compat_hash()
-//   - chain_hash: chained hash over the first n_tokens tokens, so a filename identifies an exact
-//     prefix and lookup is a single rolling-hash pass over the incoming prompt plus an index probe
-//
-// file contents mirror what the RAM cache holds for the target context:
-//
-//     header | server_tokens::serialize() bytes | llama_state_seq_get_data (FLAGS_NONE) bytes
-
-struct server_cache_disk_file {
-    std::string name; // filename inside the cache directory
-
-    uint64_t chain_hash = 0;
-    uint32_t n_tokens   = 0;
-    uint64_t n_bytes    = 0;
-    int64_t  mtime      = 0; // only used for relative ordering during eviction
-};
-
-struct server_prompt_cache_disk {
-    server_prompt_cache_disk(const std::string & dir, uint64_t compat_hash, bool has_mtmd, int32_t limit_mib, bool write_through);
-
-    enum load_status {
-        LOAD_OK,             // state restored into the sequence
-        LOAD_MISS,           // file unusable (corrupt, collision, ...) - sequence untouched
-        LOAD_FAIL_SEQ_DIRTY, // restore failed mid-way - the sequence was cleared and must be re-filled
-    };
-
-    // largest exact-prefix hit for the first n_max tokens, or nullptr on miss
-    const server_cache_disk_file * lookup(const server_tokens & tokens, size_t n_max) const;
-
-    // restore the state from a file into sequence id_slot of ctx
-    // on LOAD_OK, tokens_out receives the cached token list (an exact prefix of tokens_new)
-    load_status load(server_cache_disk_file file, const server_tokens & tokens_new, llama_context * ctx, int32_t id_slot, server_tokens & tokens_out);
-
-    // write one entry; deduplicates against existing files and enforces the size budget
-    bool store(const server_tokens & tokens, const std::vector<uint8_t> & state_main);
-
-    size_t n_files() const;
-    size_t n_bytes_total() const { return total_bytes; }
-
-    const bool write_through;
-
-private:
-    void scan_dir();
-
-    server_cache_disk_file * find_file(uint32_t n_tokens, uint64_t chain_hash);
-
-    void touch (const server_cache_disk_file & file);        // bump mtime so eviction treats it as fresh
-    void forget(const server_cache_disk_file & file);        // drop from the index without touching the filesystem
-    void remove_file(const server_cache_disk_file & file);   // delete from disk and drop from the index
-
-    // delete oldest-mtime files (ours and foreign alike) while over the size budget
-    void enforce_budget(const std::string & name_protected);
-
-    const std::string dir;
-    const uint64_t    compat_hash;
-    const bool        has_mtmd;
-    const size_t      limit_bytes; // 0 = no limit
-
-    // n_tokens -> chain_hash -> file, for our compat hash only
-    std::map<uint32_t, std::unordered_map<uint64_t, server_cache_disk_file>> index;
-
-    // .kvc files with a different compat hash prefix - never opened, but counted toward the budget
-    std::vector<server_cache_disk_file> foreign;
-
-    size_t total_bytes = 0; // ours + foreign
-
-    // (n_tokens, chain_hash) prefixes known to be covered by a file written or loaded this
-    // session - lets store() skip prefixes of already-persisted prompts
-    std::unordered_set<uint64_t> covered;
-
-    uint32_t tmp_counter = 0;
-};
-
-// hash of everything that invalidates a saved KV state for the current server configuration
-uint64_t server_cache_disk_compat_hash(const common_params & params);

tools/server/server-common.cpp

@@ -266,6 +266,8 @@ static inline raw_buffer base64_decode(const std::string & encoded_string) {
 
 namespace {
 
+constexpr uint32_t SERVER_TOKENS_STATE_VERSION = 1;
+
 uint32_t server_tokens_state_u32(size_t value) {
     if (value > std::numeric_limits<uint32_t>::max()) {
         throw std::runtime_error("Server tokens state is too large");

tools/server/server-common.h

@@ -156,9 +156,6 @@ private: // disallow accessing these members directly, risking out-of-sync
     // map_idx_to_media will contain: {5, img0}, {8, img1}
 
 public:
-    // version of the serialize()/deserialize() format below
-    static constexpr uint32_t SERVER_TOKENS_STATE_VERSION = 1;
-
     server_tokens() = default;
     ~server_tokens() = default;
 

tools/server/server-context.cpp

@@ -275,13 +275,11 @@ struct server_slot {
             llama_state_seq_get_data_ext(ctx_dft, cur->data.drft.data(), cur_size_dft, id, LLAMA_STATE_SEQ_FLAGS_NONE);
         }
 
-        prompt_cache.disk_store_write_through(*cur);
-
         return true;
     }
 
     bool prompt_load(server_prompt_cache & prompt_cache, const server_tokens & tokens) {
-        bool res = prompt_cache.load(prompt, tokens, ctx_tgt, ctx_dft, id, n_ctx);
+        bool res = prompt_cache.load(prompt, tokens, ctx_tgt, ctx_dft, id);
         if (!res) {
             SLT_WRN(*this, "%s", "failed to load prompt from cache\n");
         }
@@ -1310,26 +1308,7 @@ private:
             SRV_TRC("%s", "use `--cache-ram 0` to disable the prompt cache\n");
 
             prompt_cache = std::make_unique<server_prompt_cache>(params_base.cache_ram_mib, n_ctx);
-
-            if (!params_base.cache_disk_path.empty()) {
-                const uint64_t compat_hash = server_cache_disk_compat_hash(params_base);
-
-                SRV_INF("disk prompt cache is enabled, dir: '%s', compat hash: %08x\n",
-                        params_base.cache_disk_path.c_str(), (uint32_t) compat_hash);
-
-                prompt_cache->disk = std::make_unique<server_prompt_cache_disk>(
-                        params_base.cache_disk_path,
-                        compat_hash,
-                        mctx != nullptr,
-                        params_base.cache_disk_limit_mib,
-                        params_base.cache_disk_write_through);
-            }
         } else {
-            if (!params_base.cache_disk_path.empty()) {
-                SRV_ERR("%s", "--cache-disk requires the RAM prompt cache - remove `--cache-ram 0`\n");
-                return false;
-            }
-
             SRV_TRC("%s", "prompt cache is disabled - use `--cache-ram N` to enable it\n");
         }
         SRV_TRC("%s", "for more info see https://github.com/ggml-org/llama.cpp/pull/16391\n");
@@ -4079,11 +4058,6 @@ bool server_context::load_model(common_params & params) {
 void server_context::start_loop() {
     auto & params = impl->params_base;
     impl->queue_tasks.start_loop(params.sleep_idle_seconds * 1000);
-
-    // on graceful shutdown, give the RAM prompt cache entries a chance to survive the restart
-    if (impl->prompt_cache) {
-        impl->prompt_cache->disk_flush();
-    }
 }
 
 void server_context::terminate() {

tools/server/server-task.cpp

@@ -1750,8 +1750,6 @@ server_prompt_cache_state * server_prompt_cache::alloc(const server_prompt & pro
             SRV_WRN(" - making room for prompt cache entry, removing oldest entry (size = %.3f MiB)\n",
                     states.front().size() / (1024.0 * 1024.0));
 
-            spill_front();
-
             states.pop_front();
         }
     }
@@ -1789,7 +1787,7 @@ server_prompt_cache_state * server_prompt_cache::alloc(const server_prompt & pro
     return &states.back();
 }
 
-bool server_prompt_cache::load(server_prompt & prompt, const server_tokens & tokens_new, llama_context * ctx_tgt, llama_context * ctx_dft, int32_t id_slot, int32_t n_ctx_slot) {
+bool server_prompt_cache::load(server_prompt & prompt, const server_tokens & tokens_new, llama_context * ctx_tgt, llama_context * ctx_dft, int32_t id_slot) {
     const int lcp_best = prompt.tokens.get_common_prefix(tokens_new);
 
     float f_keep_best = prompt.tokens.size() > 0 ? float(lcp_best) / prompt.tokens.size() : -1.0f; // empty slot: any cache entry wins
@@ -1799,8 +1797,6 @@ bool server_prompt_cache::load(server_prompt & prompt, const server_tokens & tok
 
     auto it_best = states.end();
 
-    int lcp_it_best = 0;
-
     // find the most similar cached prompt, that would also preserve the most context
     for (auto it = states.begin(); it != states.end(); ++it) {
         const int lcp_cur = it->prompt.tokens.get_common_prefix(tokens_new);
@@ -1819,41 +1815,7 @@ bool server_prompt_cache::load(server_prompt & prompt, const server_tokens & tok
             f_keep_best = f_keep_cur;
             f_sim_best  = f_sim_cur;
 
-            it_best     = it;
-            lcp_it_best = lcp_cur;
-        }
-    }
-
-    // check the disk tier for an exact-prefix match longer than what RAM (or the slot itself) offers
-    if (disk) {
-        const int lcp_sel = std::max(lcp_best, lcp_it_best);
-
-        const size_t n_max = std::min<size_t>(tokens_new.size(), std::max(0, n_ctx_slot));
-
-        const auto * file = disk->lookup(tokens_new, n_max);
-
-        if (file && (int64_t) file->n_tokens > (int64_t) lcp_sel) {
-            server_tokens tokens_disk;
-
-            const auto status = disk->load(*file, tokens_new, ctx_tgt, id_slot, tokens_disk);
-
-            if (status == server_prompt_cache_disk::LOAD_OK) {
-                // disk entries carry no draft state - clear the draft sequence so it re-prefills
-                if (ctx_dft) {
-                    llama_memory_seq_rm(llama_get_memory(ctx_dft), id_slot, -1, -1);
-                }
-
-                prompt.tokens = std::move(tokens_disk);
-                prompt.checkpoints.clear();
-
-                return true;
-            }
-
-            if (status == server_prompt_cache_disk::LOAD_FAIL_SEQ_DIRTY && it_best == states.end()) {
-                // the slot's sequence was cleared during the failed restore and there is no RAM
-                // candidate to restore over it - the caller has to clear the slot
-                return false;
-            }
+            it_best = it;
         }
     }
 
@@ -1907,8 +1869,6 @@ void server_prompt_cache::update() {
         while (!states.empty() && size() > limit_size) {
             SRV_WRN(" - cache size limit reached, removing oldest entry (size = %.3f MiB)\n", states.front().size() / (1024.0 * 1024.0));
 
-            spill_front();
-
             states.pop_front();
         }
     }
@@ -1924,8 +1884,6 @@ void server_prompt_cache::update() {
             SRV_WRN(" - cache token limit (%zu, est: %zu) reached, removing oldest entry (size = %.3f MiB)\n",
                     limit_tokens, limit_tokens_cur, states.front().size() / (1024.0 * 1024.0));
 
-            spill_front();
-
             states.pop_front();
         }
     }
@@ -1938,39 +1896,3 @@ void server_prompt_cache::update() {
                 (const void *)&state, state.prompt.n_tokens(), state.prompt.checkpoints.size(), state.size() / (1024.0 * 1024.0));
     }
 }
-
-void server_prompt_cache::disk_store(const server_prompt_cache_state & state) const {
-    if (!disk || state.data.main.empty()) {
-        return;
-    }
-
-    disk->store(state.prompt.tokens, state.data.main);
-}
-
-void server_prompt_cache::disk_store_write_through(const server_prompt_cache_state & state) const {
-    if (!disk || !disk->write_through) {
-        return;
-    }
-
-    disk_store(state);
-}
-
-void server_prompt_cache::disk_flush() const {
-    if (!disk) {
-        return;
-    }
-
-    SRV_INF("flushing %zu prompt cache entries to disk\n", states.size());
-
-    for (const auto & state : states) {
-        disk_store(state);
-    }
-}
-
-void server_prompt_cache::spill_front() const {
-    if (!disk || states.empty()) {
-        return;
-    }
-
-    disk_store(states.front());
-}

tools/server/server-task.h

@@ -7,10 +7,8 @@
 #include <unordered_set>
 #include <list>
 #include <map>
-#include <memory>
 
 // TODO: prevent including the whole server-common.h as we only use server_tokens
-#include "server-cache-disk.h"
 #include "server-common.h"
 
 using json = nlohmann::ordered_json;
@@ -614,10 +612,6 @@ struct server_prompt_cache {
 
     std::list<server_prompt_cache_state> states;
 
-    // optional cold tier - entries evicted from RAM are spilled here and can be restored later,
-    // including across server restarts
-    std::unique_ptr<server_prompt_cache_disk> disk;
-
     // in bytes, 0 = no limit
     size_t limit_size = 0;
 
@@ -630,22 +624,9 @@ struct server_prompt_cache {
 
     server_prompt_cache_state * alloc(const server_prompt & prompt, size_t state_size_main, size_t state_size_drft);
 
-    bool load(server_prompt & prompt, const server_tokens & tokens_new, llama_context * ctx_tgt, llama_context * ctx_dft, int32_t id_slot, int32_t n_ctx_slot);
+    bool load(server_prompt & prompt, const server_tokens & tokens_new, llama_context * ctx_tgt, llama_context * ctx_dft, int32_t id_slot);
 
     void update();
-
-    // write one RAM cache entry to the disk tier (no-op when the disk tier is disabled)
-    void disk_store(const server_prompt_cache_state & state) const;
-
-    // disk_store, but only when write-through mode is enabled
-    void disk_store_write_through(const server_prompt_cache_state & state) const;
-
-    // spill all RAM entries to the disk tier (e.g. on graceful shutdown)
-    void disk_flush() const;
-
-private:
-    // spill the entry that is about to be evicted
-    void spill_front() const;
 };
 
 // used exclusively by router mode

tools/server/server-tools.cpp

@@ -2035,7 +2035,7 @@ void server_tools::setup(const std::vector<std::string> & enabled_tools,
         }
     }
 
-    // append MCP tools, skipping any that collide with a built-in or another MCP tool of the same "<server>_<tool>" name
+    // append MCP tools, skipping any that collide with a server tool or another MCP tool of the same "<server>_<tool>" name
     if (!mcp_mgr.empty()) {
         std::unordered_set<std::string> seen_names;
         for (auto & t : tools) {

tools/server/server-tools.h

@@ -18,7 +18,7 @@ struct server_tool {
 
     virtual ~server_tool() = default;
     virtual json get_definition() const = 0;
-    virtual std::string type() const { return "builtin"; }
+    virtual std::string type() const { return "server"; }
 
     struct stream {
         server_response & qr;

tools/server/server.cpp

@@ -346,7 +346,7 @@ int llama_server(common_params & params, int argc, char ** argv) {
         ctx_http.get ("/tools",           ex_wrapper(tools.handle_get));
         ctx_http.post("/tools",           ex_wrapper(tools.handle_post));
         if (!params.server_tools.empty()) {
-            warn_names.push_back("built-in tools (experimental)");
+            warn_names.push_back("server tools (experimental)");
         }
         if (!params.server_tools_runtime.empty()) {
             warn_names.push_back("tools runtime (experimental)");

tools/server/tests/unit/test_cache_disk.py

@@ -1,268 +0,0 @@
-import base64
-import glob
-import os
-import shutil
-import tempfile
-import time
-
-import pytest
-import requests
-from utils import *
-
-server = ServerPreset.tinyllama2()
-
-cache_dir: str = ""
-
-
-class LogReader:
-    def __init__(self, path):
-        self.path = path
-        self.pos = 0
-    def drain(self):
-        with open(self.path) as f:
-            f.seek(self.pos)
-            content = f.read()
-            self.pos = f.tell()
-        return content
-    def wait_for(self, tag, timeout=10) -> bool:
-        # the server log is pumped to the file asynchronously - poll for the tag
-        deadline = time.time() + timeout
-        while time.time() < deadline:
-            if tag in self.drain():
-                return True
-            time.sleep(0.25)
-        return False
-
-
-def kvc_files() -> list[str]:
-    return sorted(glob.glob(os.path.join(cache_dir, "*.kvc")))
-
-
-@pytest.fixture(autouse=True)
-def create_server():
-    global server, cache_dir
-    cache_dir = tempfile.mkdtemp(prefix="llama_cache_disk_")
-    server = ServerPreset.tinyllama2()
-    server.n_slots = 1
-    server.temperature = 0.0
-    server.debug = True
-    server.cache_disk = cache_dir
-    fd, server.log_path = tempfile.mkstemp(suffix='.log')
-    os.close(fd)
-    yield
-    shutil.rmtree(cache_dir, ignore_errors=True)
-
-
-PROMPT_A = (
-    "Once upon a time in a land far away, there lived a brave knight "
-    "who traveled across mountains and rivers to find the legendary "
-    "golden sword hidden deep within the enchanted forest of whispers."
-)
-
-PROMPT_B = "The quick brown fox jumps over the lazy dog."
-
-
-def make_prompt_request(prompt, n_predict=0):
-    global server
-    res = server.make_request("POST", "/completion", data={
-        "prompt": prompt,
-        "n_predict": n_predict,  # 0 = evaluate the prompt into the KV cache only
-        "cache_prompt": True,
-    })
-    assert res.status_code == 200
-    return res
-
-
-def test_write_through_and_restart_hit():
-    global server
-    server.cache_disk_write_through = True
-    server.start()
-    log = LogReader(server.log_path)
-
-    res = make_prompt_request(PROMPT_A)
-    prompt_n_full = res.body["timings"]["prompt_n"]
-    assert prompt_n_full > 0
-
-    # nothing is written while the prompt is still live in the slot
-    assert len(kvc_files()) == 0
-
-    # a different prompt takes over the only slot - the previous one is saved
-    # to the RAM cache and, in write-through mode, to disk immediately
-    make_prompt_request(PROMPT_B)
-    assert log.wait_for("__TEST_TAG_CACHE_DISK_STORE__")
-    assert len(kvc_files()) == 1
-
-    # the state must survive a full server restart
-    server.stop()
-    server.start()
-    log = LogReader(server.log_path)
-
-    res = make_prompt_request(PROMPT_A)
-    assert log.wait_for("__TEST_TAG_CACHE_DISK_HIT__")
-    assert res.body["timings"]["prompt_n"] == 1  # only the last token is re-evaluated
-    assert res.body["timings"]["cache_n"] == prompt_n_full - 1
-
-
-def test_spill_on_shutdown_flush():
-    global server
-    server.start()
-    log = LogReader(server.log_path)
-
-    make_prompt_request(PROMPT_A)
-    make_prompt_request(PROMPT_B)  # forces PROMPT_A into the RAM cache
-
-    # without write-through, nothing reaches the disk while running
-    time.sleep(0.5)
-    assert "__TEST_TAG_CACHE_DISK_STORE__" not in log.drain()
-    assert len(kvc_files()) == 0
-
-    # a graceful shutdown flushes the RAM cache entries to disk
-    server.stop()
-    assert len(kvc_files()) == 1
-
-    server.start()
-    log = LogReader(server.log_path)
-
-    res = make_prompt_request(PROMPT_A)
-    assert log.wait_for("__TEST_TAG_CACHE_DISK_HIT__")
-    assert res.body["timings"]["prompt_n"] == 1
-
-
-def test_ram_cache_hit_takes_priority():
-    global server
-    server.cache_disk_write_through = True
-    server.start()
-    log = LogReader(server.log_path)
-
-    make_prompt_request(PROMPT_A)
-    make_prompt_request(PROMPT_B)
-    assert len(kvc_files()) == 1
-
-    # PROMPT_A is in both the RAM cache and on disk - the RAM copy must win
-    # (the disk entry is never longer than the RAM one here)
-    res = make_prompt_request(PROMPT_A)
-    time.sleep(0.5)
-    assert "__TEST_TAG_CACHE_DISK_HIT__" not in log.drain()
-    assert res.body["timings"]["cache_n"] > 0
-
-
-def test_budget_eviction():
-    global server
-    server.n_ctx = 2048
-    server.n_batch = 512
-    server.cache_disk_write_through = True
-    server.cache_disk_limit = 1  # MiB
-    server.start()
-
-    # three long, distinct token-array prompts; each state is close to 1 MiB
-    n_len = 1500
-    for i in range(3):
-        make_prompt_request([100 + i] * n_len)
-
-    # one final small prompt to force the last long prompt out of the slot
-    make_prompt_request(PROMPT_B)
-
-    files = kvc_files()
-    assert len(files) >= 1
-    assert len(files) < 3  # the oldest entries were evicted
-
-    # the budget is respected (a single over-budget file is allowed to remain)
-    if len(files) > 1:
-        assert sum(os.path.getsize(f) for f in files) <= 1024 * 1024
-
-
-def test_corrupt_file_is_removed():
-    global server
-    server.cache_disk_write_through = True
-    server.start()
-
-    make_prompt_request(PROMPT_A)
-    make_prompt_request(PROMPT_B)
-    files = kvc_files()
-    assert len(files) == 1
-
-    server.stop()
-
-    # corrupt the serialized token section (starts right after the 48-byte header)
-    with open(files[0], "r+b") as f:
-        f.seek(48 + 4)
-        f.write(b"\xff\xff\xff\xff")
-
-    server.start()
-    log = LogReader(server.log_path)
-
-    # the request must still succeed, with the prompt fully re-processed
-    res = make_prompt_request(PROMPT_A)
-    time.sleep(0.5)
-    assert "__TEST_TAG_CACHE_DISK_HIT__" not in log.drain()
-    assert res.body["timings"]["prompt_n"] > 1
-
-    # the corrupt file was deleted
-    assert len(kvc_files()) == 0
-
-
-IMG_URL_CAT = "https://huggingface.co/ggml-org/tinygemma3-GGUF/resolve/main/test/91_cat.png"
-
-
-def _get_img_base64(url: str) -> str:
-    response = requests.get(url)
-    response.raise_for_status()
-    return base64.b64encode(response.content).decode("utf-8")
-
-
-@pytest.fixture
-def mmproj_server():
-    global cache_dir
-    os.environ['LLAMA_MEDIA_MARKER'] = '<__media__>'
-    mm_server = ServerPreset.tinygemma3()
-    mm_server.n_slots = 1
-    mm_server.temperature = 0.0
-    mm_server.debug = True
-    # use the full SWA cache so the restored image prefix can be reused
-    mm_server.swa_full = True
-    mm_server.cache_disk = cache_dir
-    mm_server.cache_disk_write_through = True
-    fd, mm_server.log_path = tempfile.mkstemp(suffix='.log')
-    os.close(fd)
-    return mm_server
-
-
-def test_image_prompt_across_restart(mmproj_server):
-    server = mmproj_server
-    server.start()
-
-    prompt_cat = {
-        "prompt_string": "What is this: <__media__>\n",
-        "multimodal_data": [_get_img_base64(IMG_URL_CAT)],
-    }
-
-    res = server.make_request("POST", "/completions", data={
-        "n_predict": 0,
-        "cache_prompt": True,
-        "prompt": prompt_cat,
-    })
-    assert res.status_code == 200
-    prompt_n_full = res.body["timings"]["prompt_n"]
-
-    res = server.make_request("POST", "/completions", data={
-        "n_predict": 0,
-        "cache_prompt": True,
-        "prompt": "The quick brown fox",
-    })
-    assert res.status_code == 200
-    assert len(kvc_files()) == 1
-
-    server.stop()
-    server.start()
-    log = LogReader(server.log_path)
-
-    # the image KV must be restored from disk in the new process
-    res = server.make_request("POST", "/completions", data={
-        "n_predict": 0,
-        "cache_prompt": True,
-        "prompt": prompt_cat,
-    })
-    assert res.status_code == 200
-    assert log.wait_for("__TEST_TAG_CACHE_DISK_HIT__")
-    assert res.body["timings"]["prompt_n"] == 1
-    assert res.body["timings"]["cache_n"] == prompt_n_full - 1

tools/server/tests/utils.py

@@ -111,9 +111,6 @@ class ServerProcess:
     media_path: str | None = None
     sleep_idle_seconds: int | None = None
     cache_ram: int | None = None
-    cache_disk: str | None = None
-    cache_disk_limit: int | None = None
-    cache_disk_write_through: bool = False
     no_cache_idle_slots: bool = False
     log_path: str | None = None
     ui_mcp_proxy: bool = False
@@ -274,12 +271,6 @@ class ServerProcess:
             server_args.extend(["--sleep-idle-seconds", self.sleep_idle_seconds])
         if self.cache_ram is not None:
             server_args.extend(["--cache-ram", self.cache_ram])
-        if self.cache_disk is not None:
-            server_args.extend(["--cache-disk", self.cache_disk])
-        if self.cache_disk_limit is not None:
-            server_args.extend(["--cache-disk-limit", self.cache_disk_limit])
-        if self.cache_disk_write_through:
-            server_args.append("--cache-disk-write-through")
         if self.no_cache_idle_slots:
             server_args.append("--no-cache-idle-slots")
         if self.ui_mcp_proxy:

tools/ui/eslint.config.js

@@ -61,6 +61,9 @@ export default ts.config(
 				{ blankLine: 'always', next: ['return', 'throw', 'break', 'continue'], prev: '*' }
 			],
 
+			// Alphabetical order for enum members
+			'perfectionist/sort-enums': ['error', { type: 'natural' }],
+
 			'perfectionist/sort-objects': ['error', { type: 'natural' }],
 
 			// Alphabetical order for variable declarations and object keys

tools/ui/src/lib/components/app/chat/ChatForm/ChatFormActions/ChatFormActionAdd/ChatFormActionAddToolsSubmenu.svelte

@@ -35,7 +35,7 @@
 						<span>
 							Run llama-server with <code>{CLI_FLAGS.TOOLS}</code> flag to enable
 
-							<strong>Built-in Tools</strong>.
+							<strong>Server Tools</strong>.
 						</span>
 					</span>
 

tools/ui/src/lib/components/app/chat/ChatForm/ChatFormCurrentWorkingDirectory/ChatFormCurrentWorkingDirectory.svelte

@@ -62,7 +62,7 @@
 	// it, the picker still opens for manual entry but explains why search is
 	// unavailable instead of firing searches that would only fail. Browse is
 	// hidden too: it resolves the picked folder name through the same tool.
-	const fileSearchKey = $derived(toolsStore.getPermissionKey(BuiltInTool.FILE_GLOB_SEARCH));
+	const fileSearchKey = $derived(toolsStore.getPermissionKey(BuiltInTool.SERVER_FILE_GLOB_SEARCH));
 	const fileSearchEnabled = $derived(
 		fileSearchKey !== null && toolsStore.isToolEnabled(fileSearchKey)
 	);
@@ -212,7 +212,7 @@
 	// so the caller fails visibly instead of committing a bare leaf name.
 	async function resolveNativeName(name: string): Promise<string | null> {
 		try {
-			const res = await ToolsService.executeToolRaw(BuiltInTool.FILE_GLOB_SEARCH, {
+			const res = await ToolsService.executeToolRaw(BuiltInTool.SERVER_FILE_GLOB_SEARCH, {
 				include: buildCaseInsensitiveGlob(name),
 				limit: SEARCH.NATIVE_LIMIT,
 				max_depth: SEARCH.NATIVE_MAX_DEPTH,

tools/ui/src/lib/components/app/chat/ChatForm/ChatFormPickers/ChatFormPickerMention.svelte

@@ -51,7 +51,7 @@
 	// When the server does not expose file_glob_search (started without
 	// --tools) or the user disabled it, the picker still opens but explains
 	// why instead of firing searches that would only fail.
-	const fileSearchKey = $derived(toolsStore.getPermissionKey(BuiltInTool.FILE_GLOB_SEARCH));
+	const fileSearchKey = $derived(toolsStore.getPermissionKey(BuiltInTool.SERVER_FILE_GLOB_SEARCH));
 	const fileSearchEnabled = $derived(
 		fileSearchKey !== null && toolsStore.isToolEnabled(fileSearchKey)
 	);

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/ChatMessageToolCallBlock.svelte

@@ -35,19 +35,19 @@
 
 {#if isSearchCall}
 	<ChatMessageToolCallBlockSearchResults {section} {open} {isStreaming} {onToggle} />
-{:else if section.toolName === BuiltInTool.GET_DATETIME}
+{:else if section.toolName === BuiltInTool.BROWSER_GET_DATETIME}
 	<ChatMessageToolCallBlockGetDatetime {section} {isStreaming} />
-{:else if section.toolName === BuiltInTool.GET_INFO}
+{:else if section.toolName === BuiltInTool.SERVER_GET_INFO}
 	<ChatMessageToolCallBlockGetInfo {section} {isStreaming} />
-{:else if section.toolName === BuiltInTool.READ_FILE}
+{:else if section.toolName === BuiltInTool.SERVER_READ_FILE}
 	<ChatMessageToolCallBlockReadFile {section} {open} {isStreaming} {onToggle} />
-{:else if section.toolName === BuiltInTool.READ_MEDIA}
+{:else if section.toolName === BuiltInTool.BROWSER_READ_MEDIA}
 	<ChatMessageToolCallBlockReadMedia {section} {open} {isStreaming} {onToggle} />
-{:else if section.toolName === BuiltInTool.EDIT_FILE}
+{:else if section.toolName === BuiltInTool.SERVER_EDIT_FILE}
 	<ChatMessageToolCallBlockEditFile {section} {open} {isStreaming} {onToggle} />
-{:else if section.toolName === BuiltInTool.WRITE_FILE}
+{:else if section.toolName === BuiltInTool.SERVER_WRITE_FILE}
 	<ChatMessageToolCallBlockWriteFile {section} {open} {isStreaming} {onToggle} />
-{:else if section.toolName === BuiltInTool.EXEC_SHELL_COMMAND}
+{:else if section.toolName === BuiltInTool.SERVER_EXEC_SHELL_COMMAND}
 	<ChatMessageToolCallBlockExecShellCommand
 		{section}
 		{open}
@@ -56,11 +56,11 @@
 		{attachments}
 		{onToggle}
 	/>
-{:else if section.toolName === BuiltInTool.FILE_GLOB_SEARCH}
+{:else if section.toolName === BuiltInTool.SERVER_FILE_GLOB_SEARCH}
 	<ChatMessageToolCallBlockFileGlobSearch {section} {open} {isStreaming} {onToggle} />
-{:else if section.toolName === BuiltInTool.GREP_SEARCH}
+{:else if section.toolName === BuiltInTool.SERVER_GREP_SEARCH}
 	<ChatMessageToolCallBlockGrepSearch {section} {open} {isStreaming} {onToggle} />
-{:else if section.toolName === BuiltInTool.RUN_JAVASCRIPT}
+{:else if section.toolName === BuiltInTool.BROWSER_RUN_JAVASCRIPT}
 	<ChatMessageToolCallBlockRunJavascript {section} {open} {isStreaming} {onToggle} />
 {:else}
 	<ChatMessageToolCallBlockDefault {section} {open} {isStreaming} {attachments} {onToggle} />

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/ChatMessageToolCallBlockDefault.svelte

@@ -12,7 +12,7 @@
 	import {
 		classifyToolResult,
 		formatJsonPretty,
-		getBuiltinToolUi,
+		getToolUi,
 		parseToolResultWithMedia
 	} from '$lib/utils';
 	import { createBase64DataUrl } from '$lib/utils/data-url';
@@ -27,7 +27,7 @@
 
 	let { attachments, isStreaming, onToggle, open, section }: Props = $props();
 
-	const title = $derived(getBuiltinToolUi(section.toolName)?.label ?? section.toolName ?? '');
+	const title = $derived(getToolUi(section.toolName)?.label ?? section.toolName ?? '');
 	const outputKind = $derived(classifyToolResult(section.toolResult));
 	const parsedLines: ToolResultLine[] = $derived(
 		section.toolResult ? parseToolResultWithMedia(section.toolResult, attachments) : []

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/ChatMessageToolCallBlockRunJavascript.svelte

@@ -6,7 +6,7 @@
 	import { MAX_HEIGHT_CODE_BLOCK } from '$lib/constants';
 	import { FileTypeText } from '$lib/enums';
 	import type { AgenticSection } from '$lib/types';
-	import { getBuiltinToolUi } from '$lib/utils';
+	import { getToolUi } from '$lib/utils';
 
 	interface Props {
 		section: AgenticSection;
@@ -18,7 +18,7 @@
 	let { isStreaming, onToggle, open, section }: Props = $props();
 
 	const runJsMeta = $derived(parseRunJavascriptMeta(section));
-	const title = $derived(getBuiltinToolUi(section.toolName)?.label ?? section.toolName ?? '');
+	const title = $derived(getToolUi(section.toolName)?.label ?? section.toolName ?? '');
 </script>
 
 <ToolCallBlock {section} {open} {isStreaming} meta={runJsMeta} {title} {onToggle}>

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/ToolCallBlock.svelte

@@ -14,8 +14,8 @@
 	import { ICON_CLASS_DEFAULT, ICON_CLASS_SPIN } from '$lib/constants';
 	import { AgenticSectionType } from '$lib/enums';
 	import { mcpStore } from '$lib/stores';
-	import type { AgenticSection, BuiltinToolUiEntry } from '$lib/types';
-	import { getBuiltinToolUi } from '$lib/utils';
+	import type { AgenticSection, ToolUiEntry } from '$lib/types';
+	import { getToolUi } from '$lib/utils';
 	import type { Component, Snippet } from 'svelte';
 
 	type ToolCallBlockMetaWithError = TMeta & { errorMessage?: string };
@@ -82,7 +82,7 @@
 	const showSpinner = $derived(isPending || (isStreamingCall && isStreaming) || extraLiveStreaming);
 	const isCodeStreaming = $derived(isStreaming && (isPending || isStreamingCall));
 
-	const toolUi: BuiltinToolUiEntry | null = $derived(getBuiltinToolUi(section.toolName));
+	const toolUi: ToolUiEntry | null = $derived(getToolUi(section.toolName));
 	const toolIcon: Component = $derived(
 		spinIconWhenActive && showSpinner ? Loader2 : (toolUi?.icon ?? Wrench)
 	);

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/parsers/edit-file.ts

@@ -24,7 +24,7 @@ export type EditFileMeta = {
 };
 
 export function parseEditFileMeta(section: AgenticSection): EditFileMeta | null {
-	const args = parseToolArgs(BuiltInTool.EDIT_FILE, section, { partial: true });
+	const args = parseToolArgs(BuiltInTool.SERVER_EDIT_FILE, section, { partial: true });
 
 	if (!args) return null;
 

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/parsers/exec-shell-command.ts

@@ -14,7 +14,7 @@ export type ExecShellCommandMeta = {
 };
 
 export function parseExecShellCommandMeta(section: AgenticSection): ExecShellCommandMeta | null {
-	const args = parseToolArgs(BuiltInTool.EXEC_SHELL_COMMAND, section);
+	const args = parseToolArgs(BuiltInTool.SERVER_EXEC_SHELL_COMMAND, section);
 
 	if (!args) return null;
 

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/parsers/file-glob-search.ts

@@ -19,7 +19,7 @@ export type FileGlobSearchMeta = {
 };
 
 export function parseFileGlobSearchMeta(section: AgenticSection): FileGlobSearchMeta | null {
-	const args = parseToolArgs(BuiltInTool.FILE_GLOB_SEARCH, section);
+	const args = parseToolArgs(BuiltInTool.SERVER_FILE_GLOB_SEARCH, section);
 
 	if (!args) return null;
 

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/parsers/grep-search.ts

@@ -28,7 +28,7 @@ export type GrepSearchMeta = {
 };
 
 export function parseGrepSearchMeta(section: AgenticSection): GrepSearchMeta | null {
-	const args = parseToolArgs(BuiltInTool.GREP_SEARCH, section);
+	const args = parseToolArgs(BuiltInTool.SERVER_GREP_SEARCH, section);
 
 	if (!args) return null;
 

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/parsers/read-file.ts

@@ -16,7 +16,7 @@ export type ReadFileMeta = {
 };
 
 export function parseReadFileMeta(section: AgenticSection): ReadFileMeta | null {
-	const args = parseToolArgs(BuiltInTool.READ_FILE, section, { partial: true });
+	const args = parseToolArgs(BuiltInTool.SERVER_READ_FILE, section, { partial: true });
 
 	if (!args) return null;
 

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/parsers/run-javascript.ts

@@ -16,7 +16,7 @@ export type RunJavascriptMeta = {
 };
 
 export function parseRunJavascriptMeta(section: AgenticSection): RunJavascriptMeta | null {
-	const args = parseToolArgs(BuiltInTool.RUN_JAVASCRIPT, section);
+	const args = parseToolArgs(BuiltInTool.BROWSER_RUN_JAVASCRIPT, section);
 
 	if (!args) return null;
 

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/parsers/write-file.ts

@@ -20,7 +20,7 @@ export type WriteFileMeta = {
 };
 
 export function parseWriteFileMeta(section: AgenticSection): WriteFileMeta | null {
-	const args = parseToolArgs(BuiltInTool.WRITE_FILE, section, { partial: true });
+	const args = parseToolArgs(BuiltInTool.SERVER_WRITE_FILE, section, { partial: true });
 
 	if (!args) return null;
 

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessageActions/ChatMessageActionCard/ChatMessageActionCardPermissionRequest.svelte

@@ -61,8 +61,8 @@
 				{:else}
 					{@const source = toolsStore.getToolSource(toolName)}
 					{@const providerName =
-						source === ToolSource.BUILTIN
-							? TOOL_SERVER_LABELS[ToolSource.BUILTIN]
+						source === ToolSource.SERVER
+							? TOOL_SERVER_LABELS[ToolSource.SERVER]
 							: source === ToolSource.CUSTOM
 								? TOOL_SERVER_LABELS[ToolSource.CUSTOM]
 								: 'MCP Tools'}

tools/ui/src/lib/components/app/chat/index.ts

@@ -278,7 +278,7 @@ export { default as ChatFormInput } from './ChatForm/ChatFormInput/ChatFormInput
 /**
  * Working directory selector for agent mode. Renders a chip below the chat
  * form; clicking it opens a popover with a directory picker backed by the
- * server's `file_glob_search` built-in tool (POST /tools). The picked
+ * server's `file_glob_search` server tool (POST /tools). The picked
  * directory is exposed via `bind:directory`; changing it records a
  * synthetic "Set working directory to ..." user message into chat history
  * and is enforced on tool calls via the `x-tool-cwd` request header.
@@ -380,7 +380,7 @@ export { default as ChatFormPickerListItemSkeleton } from './ChatForm/ChatFormPi
 
 /**
  * `@`-triggered file/folder mention picker. Resolves `@<query>` in the chat
- * input to a filesystem match via the server's `file_glob_search` built-in
+ * input to a filesystem match via the server's `file_glob_search` server tool
  * tool, scoped to the conversation cwd (or server home when unset).
  * Selection splices a `[name](file:///<abs path>)` link into the input.
  */

tools/ui/src/lib/components/app/settings/SettingsChat/SettingsChatToolsTab.svelte

@@ -6,7 +6,7 @@
 	import { ICON_CLASS_DEFAULT } from '$lib/constants';
 	import { ToolSource } from '$lib/enums/tools.enums';
 	import { mcpStore, permissionsStore, toolsStore } from '$lib/stores';
-	import { getBuiltinToolUi } from '$lib/utils';
+	import { getToolUi } from '$lib/utils';
 	import { SvelteSet } from 'svelte/reactivity';
 
 	let expandedGroups = new SvelteSet<string>();
@@ -69,12 +69,12 @@
 
 						{#each group.tools as entry (entry.key)}
 							{@const toolName = entry.definition.function.name}
-							{@const builtinUi =
-								entry.source === ToolSource.BUILTIN || entry.source === ToolSource.FRONTEND
-									? getBuiltinToolUi(toolName)
+							{@const toolUi =
+								entry.source === ToolSource.SERVER || entry.source === ToolSource.BROWSER
+									? getToolUi(toolName)
 									: null}
-							{@const displayLabel = builtinUi?.label ?? toolName}
-							{@const IconComponent = builtinUi?.icon ?? null}
+							{@const displayLabel = toolUi?.label ?? toolName}
+							{@const IconComponent = toolUi?.icon ?? null}
 							{@const isEnabled = toolsStore.isToolEnabled(entry.key)}
 							{@const permissionKey = entry.key}
 							{@const isAlwaysAllowed = permissionsStore.hasTool(permissionKey)}

tools/ui/src/lib/components/app/settings/index.ts

@@ -69,7 +69,7 @@ export { default as SettingsChatFields } from './SettingsChat/SettingsChatFields
 /**
  * **SettingsChatToolsTab** - Tools configuration tab for chat settings
  *
- * Displays available tools grouped by source (built-in, MCP, custom) with
+ * Displays available tools grouped by source (server, browser, MCP, custom) with
  * toggles to enable/disable individual tools and tool groups. Shows MCP
  * server favicons and permission management controls.
  */

tools/ui/src/lib/constants/browser-info.ts

@@ -2,7 +2,9 @@ import { CLI_FLAGS } from './cli-flags.constants';
 import { BuiltInTool, JsonSchemaType, ToolCallType } from '$lib/enums';
 import type { OpenAIToolDefinition } from '$lib/types';
 
-export const BROWSER_INFO_TOOL_NAME = BuiltInTool.GET_INFO;
+// get_info is served by the server, but the browser falls back to this
+// implementation when the server does not provide it - same wire name.
+export const BROWSER_INFO_TOOL_NAME = BuiltInTool.SERVER_GET_INFO;
 
 /** UA token to OS name, first match wins - Android and iOS UAs also carry the Linux / Mac OS X tokens */
 export const BROWSER_INFO_OS_UA_PATTERNS: readonly [RegExp, string][] = [

tools/ui/src/lib/constants/built-in-tools.constants.ts

@@ -1,52 +0,0 @@
-// Registry of built-in and frontend (browser) tools whose renderer
-// shows a recognizable icon and friendly label inline in the chat UI.
-//
-// To add a new built-in tool, add an entry to BUILTIN_TOOL_UI. To give a
-// tool a custom title or body renderer, add a dedicated component under
-// ChatMessageToolCall/ and route it in ChatMessageToolCallBlock.svelte
-// (see ChatMessageToolCallBlockGetDatetime and
-// ChatMessageToolCallBlockSearchResults for prior art).
-
-import {
-	Braces,
-	Clock,
-	Eye,
-	FilePen,
-	FilePlus,
-	FileSearch,
-	FileText,
-	Info,
-	SearchCode,
-	Terminal
-} from '@lucide/svelte';
-import { BuiltInTool, ToolSource } from '$lib/enums';
-import type { BuiltinToolUiEntry } from '$lib/types';
-
-export const BUILTIN_TOOL_UI: Readonly<Record<BuiltInTool, BuiltinToolUiEntry>> = {
-	[BuiltInTool.EDIT_FILE]: { icon: FilePen, label: 'Edit file', source: ToolSource.BUILTIN },
-	[BuiltInTool.EXEC_SHELL_COMMAND]: {
-		icon: Terminal,
-		label: 'Run command',
-		source: ToolSource.BUILTIN
-	},
-	[BuiltInTool.FILE_GLOB_SEARCH]: {
-		icon: FileSearch,
-		label: 'Search files',
-		source: ToolSource.BUILTIN
-	},
-	[BuiltInTool.GET_DATETIME]: { icon: Clock, label: 'Current time', source: ToolSource.FRONTEND },
-	[BuiltInTool.GET_INFO]: { icon: Info, label: 'Runtime info', source: ToolSource.BUILTIN },
-	[BuiltInTool.GREP_SEARCH]: {
-		icon: SearchCode,
-		label: 'Search in files',
-		source: ToolSource.BUILTIN
-	},
-	[BuiltInTool.READ_FILE]: { icon: FileText, label: 'Read file', source: ToolSource.BUILTIN },
-	[BuiltInTool.READ_MEDIA]: { icon: Eye, label: 'Read media', source: ToolSource.FRONTEND },
-	[BuiltInTool.RUN_JAVASCRIPT]: {
-		icon: Braces,
-		label: 'Run JavaScript',
-		source: ToolSource.FRONTEND
-	},
-	[BuiltInTool.WRITE_FILE]: { icon: FilePlus, label: 'Write file', source: ToolSource.BUILTIN }
-} as const;

tools/ui/src/lib/constants/get-datetime.ts

@@ -1,7 +1,7 @@
 import { BuiltInTool, JsonSchemaType, ToolCallType } from '$lib/enums';
 import type { OpenAIToolDefinition } from '$lib/types';
 
-export const GET_DATETIME_TOOL_NAME = BuiltInTool.GET_DATETIME;
+export const GET_DATETIME_TOOL_NAME = BuiltInTool.BROWSER_GET_DATETIME;
 
 export function buildGetDatetimeToolDefinition(): OpenAIToolDefinition {
 	return {

tools/ui/src/lib/constants/index.ts

@@ -15,7 +15,7 @@ export * from './context-gauge-popup.constants';
 export * from './conversation-import.constants';
 export * from './binary-detection.constants';
 export * from './content-detection.constants';
-export * from './built-in-tools.constants';
+export * from './tool-ui.constants';
 export * from './cache.constants';
 export * from './chat-form.constants';
 export * from './cli-flags.constants';

tools/ui/src/lib/constants/read-media.ts

@@ -7,7 +7,7 @@ import {
 } from '$lib/enums';
 import type { OpenAIToolDefinition } from '$lib/types';
 
-export const READ_MEDIA_TOOL_NAME = BuiltInTool.READ_MEDIA;
+export const READ_MEDIA_TOOL_NAME = BuiltInTool.BROWSER_READ_MEDIA;
 
 // header lines of the tool result, parsed back by the read_media renderer
 export const PREFIX_FILE = 'File: ';

tools/ui/src/lib/constants/sandbox.constants.ts

@@ -1,6 +1,6 @@
 import { BuiltInTool } from '$lib/enums';
 
-export const SANDBOX_TOOL_NAME = BuiltInTool.RUN_JAVASCRIPT;
+export const SANDBOX_TOOL_NAME = BuiltInTool.BROWSER_RUN_JAVASCRIPT;
 
 export const SANDBOX_TIMEOUT_MS_DEFAULT = 10000;
 

tools/ui/src/lib/constants/tool-ui.constants.ts

@@ -0,0 +1,60 @@
+// Registry of server and browser tools whose renderer
+// shows a recognizable icon and friendly label inline in the chat UI.
+//
+// To add a new tool, add an entry to TOOL_UI. To give a
+// tool a custom title or body renderer, add a dedicated component under
+// ChatMessageToolCall/ and route it in ChatMessageToolCallBlock.svelte
+// (see ChatMessageToolCallBlockGetDatetime and
+// ChatMessageToolCallBlockSearchResults for prior art).
+
+import {
+	Braces,
+	Clock,
+	Eye,
+	FilePen,
+	FilePlus,
+	FileSearch,
+	FileText,
+	Info,
+	SearchCode,
+	Terminal
+} from '@lucide/svelte';
+import { BuiltInTool, ToolSource } from '$lib/enums';
+import type { ToolUiEntry } from '$lib/types';
+
+export const TOOL_UI: Readonly<Record<BuiltInTool, ToolUiEntry>> = {
+	[BuiltInTool.BROWSER_GET_DATETIME]: {
+		icon: Clock,
+		label: 'Current time',
+		source: ToolSource.BROWSER
+	},
+	[BuiltInTool.BROWSER_READ_MEDIA]: { icon: Eye, label: 'Read media', source: ToolSource.BROWSER },
+	[BuiltInTool.BROWSER_RUN_JAVASCRIPT]: {
+		icon: Braces,
+		label: 'Run JavaScript',
+		source: ToolSource.BROWSER
+	},
+	[BuiltInTool.SERVER_EDIT_FILE]: { icon: FilePen, label: 'Edit file', source: ToolSource.SERVER },
+	[BuiltInTool.SERVER_EXEC_SHELL_COMMAND]: {
+		icon: Terminal,
+		label: 'Run command',
+		source: ToolSource.SERVER
+	},
+	[BuiltInTool.SERVER_FILE_GLOB_SEARCH]: {
+		icon: FileSearch,
+		label: 'Search files',
+		source: ToolSource.SERVER
+	},
+	[BuiltInTool.SERVER_GET_INFO]: { icon: Info, label: 'Runtime info', source: ToolSource.SERVER },
+	[BuiltInTool.SERVER_GREP_SEARCH]: {
+		icon: SearchCode,
+		label: 'Search in files',
+		source: ToolSource.SERVER
+	},
+	[BuiltInTool.SERVER_READ_FILE]: { icon: FileText, label: 'Read file', source: ToolSource.SERVER },
+	[BuiltInTool.SERVER_WRITE_FILE]: {
+		icon: FilePlus,
+		label: 'Write file',
+		source: ToolSource.SERVER
+	}
+} as const;

tools/ui/src/lib/constants/ui.constants.ts

@@ -18,15 +18,15 @@ export const UI_DATA_ATTRS = {
 } as const;
 
 export const TOOL_GROUP_LABELS = {
-	[ToolSource.BUILTIN]: 'Built-in',
+	[ToolSource.BROWSER]: 'Browser',
 	[ToolSource.CUSTOM]: 'JSON Schema',
-	[ToolSource.FRONTEND]: 'Browser'
+	[ToolSource.SERVER]: 'Server'
 } as const;
 
 export const TOOL_SERVER_LABELS = {
-	[ToolSource.BUILTIN]: 'Built-in Tools',
+	[ToolSource.BROWSER]: 'Browser Tools',
 	[ToolSource.CUSTOM]: 'Custom Tools',
-	[ToolSource.FRONTEND]: 'Browser Tools'
+	[ToolSource.SERVER]: 'Server Tools'
 } as const;
 
 export const TOOLTIP_DELAY_DURATION = 500;

tools/ui/src/lib/enums/agentic.enums.ts

@@ -9,12 +9,12 @@ export enum ToolCallType {
  * Types of sections in agentic content display.
  */
 export enum AgenticSectionType {
+	REASONING = 'reasoning',
+	REASONING_PENDING = 'reasoning_pending',
 	TEXT = 'text',
 	TOOL_CALL = 'tool_call',
 	TOOL_CALL_PENDING = 'tool_call_pending',
-	TOOL_CALL_STREAMING = 'tool_call_streaming',
-	REASONING = 'reasoning',
-	REASONING_PENDING = 'reasoning_pending'
+	TOOL_CALL_STREAMING = 'tool_call_streaming'
 }
 
 /**
@@ -22,8 +22,8 @@ export enum AgenticSectionType {
  */
 export enum ContinueIntentKind {
 	APPEND_TEXT = 'append_text',
-	RERUN_TURN = 'rerun_turn',
-	NEXT_TURN = 'next_turn'
+	NEXT_TURN = 'next_turn',
+	RERUN_TURN = 'rerun_turn'
 }
 
 /**
@@ -39,7 +39,7 @@ export enum ToolResultKind {
  * Line classification for the unified-diff renderer of `edit_file` results.
  */
 export enum DiffLineKind {
-	CONTEXT = 'context',
 	ADD = 'add',
+	CONTEXT = 'context',
 	REMOVE = 'remove'
 }

tools/ui/src/lib/enums/attachment.enums.ts

@@ -4,12 +4,12 @@
 export enum AttachmentType {
 	AUDIO = 'AUDIO',
 	IMAGE = 'IMAGE',
-	VIDEO = 'VIDEO',
+	LEGACY_CONTEXT = 'context', // Legacy attachment type for backward compatibility
 	MCP_PROMPT = 'MCP_PROMPT',
 	MCP_RESOURCE = 'MCP_RESOURCE',
 	PDF = 'PDF',
 	TEXT = 'TEXT',
-	LEGACY_CONTEXT = 'context' // Legacy attachment type for backward compatibility
+	VIDEO = 'VIDEO'
 }
 
 /**
@@ -17,14 +17,14 @@ export enum AttachmentType {
  * Used to select which file upload or attachment action is triggered.
  */
 export enum AttachmentMenuItemId {
-	IMAGES = 'images',
 	AUDIO = 'audio',
-	VIDEO = 'video',
-	TEXT = 'text',
+	IMAGES = 'images',
+	MCP_PROMPT = 'mcp-prompt',
+	MCP_RESOURCES = 'mcp-resources',
 	PDF = 'pdf',
 	SYSTEM_MESSAGE = 'system-message',
-	MCP_PROMPT = 'mcp-prompt',
-	MCP_RESOURCES = 'mcp-resources'
+	TEXT = 'text',
+	VIDEO = 'video'
 }
 
 /**
@@ -32,9 +32,9 @@ export enum AttachmentMenuItemId {
  */
 export enum AttachmentItemEnabledWhen {
 	ALWAYS = 'always',
-	HAS_VISION_MODALITY = 'hasVisionModality',
 	HAS_AUDIO_MODALITY = 'hasAudioModality',
-	HAS_VIDEO_MODALITY = 'hasVideoModality'
+	HAS_VIDEO_MODALITY = 'hasVideoModality',
+	HAS_VISION_MODALITY = 'hasVisionModality'
 }
 
 /**
@@ -42,9 +42,9 @@ export enum AttachmentItemEnabledWhen {
  */
 export enum AttachmentAction {
 	FILE_UPLOAD = 'onFileUpload',
-	SYSTEM_PROMPT_CLICK = 'onSystemPromptClick',
 	MCP_PROMPT_CLICK = 'onMcpPromptClick',
-	MCP_RESOURCES_CLICK = 'onMcpResourcesClick'
+	MCP_RESOURCES_CLICK = 'onMcpResourcesClick',
+	SYSTEM_PROMPT_CLICK = 'onSystemPromptClick'
 }
 
 /**
@@ -52,9 +52,9 @@ export enum AttachmentAction {
  */
 export enum AttachmentLabel {
 	FILE = 'File',
-	PDF_FILE = 'PDF File',
 	MCP_PROMPT = 'MCP Prompt',
-	MCP_RESOURCE = 'MCP Resource'
+	MCP_RESOURCE = 'MCP Resource',
+	PDF_FILE = 'PDF File'
 }
 
 /**

tools/ui/src/lib/enums/boolean-string.enums.ts

@@ -1,5 +1,5 @@
 /** String representation of a boolean used in data attributes and persisted values. */
 export enum BooleanString {
-	TRUE = 'true',
-	FALSE = 'false'
+	FALSE = 'false',
+	TRUE = 'true'
 }

tools/ui/src/lib/enums/chat.enums.ts

@@ -1,41 +1,41 @@
 export enum ChatMessageStatsView {
 	GENERATION = 'generation',
 	READING = 'reading',
-	TOOLS = 'tools',
-	SUMMARY = 'summary'
+	SUMMARY = 'summary',
+	TOOLS = 'tools'
 }
 
 export enum ChatMessageStatisticsMode {
-	SWITCHABLE = 'switchable',
+	GENERATION = 'generation',
 	READING = 'reading',
-	GENERATION = 'generation'
+	SWITCHABLE = 'switchable'
 }
 
 /**
  * Connection state of a streamed completion, drives the resume status indicator.
  */
 export enum StreamConnectionState {
-	STREAMING = 'streaming',
+	LOST = 'lost',
 	RESUMING = 'resuming',
-	LOST = 'lost'
+	STREAMING = 'streaming'
 }
 
 /**
  * Reasoning format options for API requests.
  */
 export enum ReasoningFormat {
-	NONE = 'none',
-	AUTO = 'auto'
+	AUTO = 'auto',
+	NONE = 'none'
 }
 
 /**
  * Message roles for chat messages.
  */
 export enum MessageRole {
-	USER = 'user',
 	ASSISTANT = 'assistant',
 	SYSTEM = 'system',
-	TOOL = 'tool'
+	TOOL = 'tool',
+	USER = 'user'
 }
 
 /**
@@ -43,27 +43,27 @@ export enum MessageRole {
  */
 export enum MessageType {
 	ROOT = 'root',
+	SYSTEM = 'system',
 	TEXT = 'text',
-	THINK = 'think',
-	SYSTEM = 'system'
+	THINK = 'think'
 }
 
 /**
  * Content part types for API chat message content.
  */
 export enum ContentPartType {
-	TEXT = 'text',
 	IMAGE_URL = 'image_url',
 	INPUT_AUDIO = 'input_audio',
-	INPUT_VIDEO = 'input_video'
+	INPUT_VIDEO = 'input_video',
+	TEXT = 'text'
 }
 
 /**
  * Error dialog types for displaying server/timeout errors.
  */
 export enum ErrorDialogType {
-	TIMEOUT = 'timeout',
-	SERVER = 'server'
+	SERVER = 'server',
+	TIMEOUT = 'timeout'
 }
 
 export enum ConversationSelectionMode {
@@ -75,27 +75,27 @@ export enum ConversationSelectionMode {
  * PDF view mode options for previewing PDF attachments.
  */
 export enum PdfViewMode {
-	TEXT = 'text',
-	PAGES = 'pages'
+	PAGES = 'pages',
+	TEXT = 'text'
 }
 
 export enum ChatFormCommandAction {
-	PROMPT = 'prompt',
 	CWD = 'cwd',
-	MODEL = 'model'
+	MODEL = 'model',
+	PROMPT = 'prompt'
 }
 
 export enum FileMentionEntryType {
-	FILE = 'file',
-	DIRECTORY = 'directory'
+	DIRECTORY = 'directory',
+	FILE = 'file'
 }
 
 /**
  * Kinds of tokens the chat-form-input-rich produces.
  */
 export enum ChatFormInputRichTokenKind {
-	TEXT = 'text',
 	BADGE = 'badge',
+	CODE_BLOCK = 'code_block',
 	CODE_INLINE = 'code_inline',
-	CODE_BLOCK = 'code_block'
+	TEXT = 'text'
 }

tools/ui/src/lib/enums/conversation-import.enums.ts

@@ -4,6 +4,6 @@
  * message record belongs to it.
  */
 export enum SessionRecordType {
-	SESSION = 'session',
-	MESSAGE = 'message'
+	MESSAGE = 'message',
+	SESSION = 'session'
 }

tools/ui/src/lib/enums/files.enums.ts

@@ -5,11 +5,11 @@
 
 // File type category enum
 export enum FileTypeCategory {
-	IMAGE = 'image',
 	AUDIO = 'audio',
-	VIDEO = 'video',
+	IMAGE = 'image',
 	PDF = 'pdf',
-	TEXT = 'text'
+	TEXT = 'text',
+	VIDEO = 'video'
 }
 
 /**
@@ -21,13 +21,13 @@ export enum SpecialFileType {
 
 // Specific file type enums for each category
 export enum FileTypeImage {
+	GIF = 'gif',
+	HEIC = 'heic',
+	HEIF = 'heif',
 	JPEG = 'jpeg',
 	PNG = 'png',
-	GIF = 'gif',
-	WEBP = 'webp',
 	SVG = 'svg',
-	HEIC = 'heic',
-	HEIF = 'heif'
+	WEBP = 'webp'
 }
 
 export enum FileTypeAudio {
@@ -46,55 +46,55 @@ export enum FileTypePdf {
 }
 
 export enum FileTypeText {
-	PLAIN_TEXT = 'plainText',
-	MARKDOWN = 'md',
 	ASCIIDOC = 'asciidoc',
-	JAVASCRIPT = 'js',
-	TYPESCRIPT = 'ts',
-	JSX = 'jsx',
-	TSX = 'tsx',
-	CSS = 'css',
-	HTML = 'html',
-	JSON = 'json',
-	XML = 'xml',
-	YAML = 'yaml',
-	CSV = 'csv',
-	LOG = 'log',
-	PYTHON = 'python',
-	JAVA = 'java',
+	BIBTEX = 'bibtex',
 	CPP = 'cpp',
-	PHP = 'php',
-	RUBY = 'ruby',
+	CSHARP = 'csharp',
+	CSS = 'css',
+	CSV = 'csv',
+	CUDA = 'cuda',
+	DART = 'dart',
 	GO = 'go',
+	HASKELL = 'haskell',
+	HTML = 'html',
+	JAVA = 'java',
+	JAVASCRIPT = 'js',
+	JSON = 'json',
+	JSX = 'jsx',
+	KOTLIN = 'kotlin',
+	LATEX = 'latex',
+	LOG = 'log',
+	MARKDOWN = 'md',
+	PHP = 'php',
+	PLAIN_TEXT = 'plainText',
+	PROPERTIES = 'properties',
+	PYTHON = 'python',
+	R = 'r',
+	RUBY = 'ruby',
 	RUST = 'rust',
+	SCALA = 'scala',
 	SHELL = 'shell',
 	SQL = 'sql',
-	R = 'r',
-	SCALA = 'scala',
-	KOTLIN = 'kotlin',
-	SWIFT = 'swift',
-	DART = 'dart',
-	VUE = 'vue',
 	SVELTE = 'svelte',
-	LATEX = 'latex',
-	BIBTEX = 'bibtex',
-	CUDA = 'cuda',
+	SWIFT = 'swift',
+	TSX = 'tsx',
+	TYPESCRIPT = 'ts',
+	VUE = 'vue',
 	VULKAN = 'vulkan',
-	HASKELL = 'haskell',
-	CSHARP = 'csharp',
-	PROPERTIES = 'properties'
+	XML = 'xml',
+	YAML = 'yaml'
 }
 
 // File extension enums
 export enum FileExtensionImage {
-	JPG = '.jpg',
-	JPEG = '.jpeg',
-	PNG = '.png',
 	GIF = '.gif',
-	WEBP = '.webp',
-	SVG = '.svg',
 	HEIC = '.heic',
-	HEIF = '.heif'
+	HEIF = '.heif',
+	JPEG = '.jpeg',
+	JPG = '.jpg',
+	PNG = '.png',
+	SVG = '.svg',
+	WEBP = '.webp'
 }
 
 export enum FileExtensionAudio {
@@ -112,64 +112,64 @@ export enum FileExtensionPdf {
 }
 
 export enum FileExtensionText {
-	TXT = '.txt',
-	MD = '.md',
 	ADOC = '.adoc',
-	JS = '.js',
-	TS = '.ts',
-	JSX = '.jsx',
-	TSX = '.tsx',
+	BAT = '.bat',
+	BIB = '.bib',
+	C = '.c',
+	COMP = '.comp',
+	CPP = '.cpp',
+	CS = '.cs',
 	CSS = '.css',
-	HTML = '.html',
+	CSV = '.csv',
+	CU = '.cu',
+	CUH = '.cuh',
+	DART = '.dart',
+	GO = '.go',
+	H = '.h',
+	HPP = '.hpp',
+	HS = '.hs',
 	HTM = '.htm',
+	HTML = '.html',
+	JAVA = '.java',
+	JS = '.js',
 	JSON = '.json',
 	JSONL = '.jsonl',
-	ZIP = '.zip',
+	JSX = '.jsx',
+	KT = '.kt',
+	LOG = '.log',
+	MD = '.md',
+	PHP = '.php',
+	PROPERTIES = '.properties',
+	PY = '.py',
+	R = '.r',
+	RB = '.rb',
+	RS = '.rs',
+	SCALA = '.scala',
+	SH = '.sh',
+	SQL = '.sql',
+	SVELTE = '.svelte',
+	SWIFT = '.swift',
+	TEX = '.tex',
+	TS = '.ts',
+	TSX = '.tsx',
+	TXT = '.txt',
+	VUE = '.vue',
 	XML = '.xml',
 	YAML = '.yaml',
 	YML = '.yml',
-	CSV = '.csv',
-	LOG = '.log',
-	PY = '.py',
-	JAVA = '.java',
-	CPP = '.cpp',
-	C = '.c',
-	H = '.h',
-	PHP = '.php',
-	RB = '.rb',
-	GO = '.go',
-	RS = '.rs',
-	SH = '.sh',
-	BAT = '.bat',
-	SQL = '.sql',
-	R = '.r',
-	SCALA = '.scala',
-	KT = '.kt',
-	SWIFT = '.swift',
-	DART = '.dart',
-	VUE = '.vue',
-	SVELTE = '.svelte',
-	TEX = '.tex',
-	BIB = '.bib',
-	CU = '.cu',
-	CUH = '.cuh',
-	COMP = '.comp',
-	HPP = '.hpp',
-	HS = '.hs',
-	PROPERTIES = '.properties',
-	CS = '.cs'
+	ZIP = '.zip'
 }
 
 // MIME type prefixes and includes for content detection
 export enum MimeTypePrefix {
-	IMAGE = 'image/',
 	AUDIO = 'audio/',
+	IMAGE = 'image/',
 	TEXT = 'text'
 }
 
 export enum MimeTypeIncludes {
-	JSON = 'json',
 	JAVASCRIPT = 'javascript',
+	JSON = 'json',
 	TYPESCRIPT = 'typescript'
 }
 
@@ -182,23 +182,23 @@ export enum UriPattern {
 // MIME type enums
 export enum MimeTypeApplication {
 	JSON = 'application/json',
-	PDF = 'application/pdf',
 	OCTET_STREAM = 'application/octet-stream',
+	PDF = 'application/pdf',
 	ZIP = 'application/zip'
 }
 
 export enum MimeTypeAudio {
-	MP3_MPEG = 'audio/mpeg',
 	MP3 = 'audio/mp3',
+	MP3_MPEG = 'audio/mpeg',
 	MP4 = 'audio/mp4',
+	VND_WAVE = 'audio/vnd.wave',
 	WAV = 'audio/wav',
 	WAVE = 'audio/wave',
-	X_WAV = 'audio/x-wav',
-	X_WAVE = 'audio/x-wave',
-	VND_WAVE = 'audio/vnd.wave',
-	X_PN_WAV = 'audio/x-pn-wav',
 	WEBM = 'audio/webm',
-	WEBM_OPUS = 'audio/webm;codecs=opus'
+	WEBM_OPUS = 'audio/webm;codecs=opus',
+	X_PN_WAV = 'audio/x-pn-wav',
+	X_WAV = 'audio/x-wav',
+	X_WAVE = 'audio/x-wave'
 }
 
 export enum MimeTypeVideo {
@@ -207,62 +207,62 @@ export enum MimeTypeVideo {
 }
 
 export enum MimeTypeImage {
+	GIF = 'image/gif',
+	HEIC = 'image/heic',
+	HEIF = 'image/heif',
+	ICO = 'image/x-icon',
+	ICO_MICROSOFT = 'image/vnd.microsoft.icon',
 	JPEG = 'image/jpeg',
 	JPG = 'image/jpg',
 	PNG = 'image/png',
-	GIF = 'image/gif',
-	WEBP = 'image/webp',
 	SVG = 'image/svg+xml',
-	ICO = 'image/x-icon',
-	ICO_MICROSOFT = 'image/vnd.microsoft.icon',
-	HEIC = 'image/heic',
-	HEIF = 'image/heif'
+	WEBP = 'image/webp'
 }
 
 export enum MimeTypeText {
-	PLAIN = 'text/plain',
-	MARKDOWN = 'text/markdown',
 	ASCIIDOC = 'text/asciidoc',
-	JAVASCRIPT = 'text/javascript',
-	JAVASCRIPT_APP = 'application/javascript',
-	TYPESCRIPT = 'text/typescript',
-	JSX = 'text/jsx',
-	TSX = 'text/tsx',
-	CSS = 'text/css',
-	HTML = 'text/html',
-	JSON = 'application/json',
-	JSONL = 'application/jsonl',
-	XML_TEXT = 'text/xml',
-	XML_APP = 'application/xml',
-	YAML_TEXT = 'text/yaml',
-	YAML_APP = 'application/yaml',
-	CSV = 'text/csv',
-	PYTHON = 'text/x-python',
-	JAVA = 'text/x-java-source',
+	BAT = 'application/x-bat',
+	BIBTEX = 'text/x-bibtex',
+	C_HDR = 'text/x-chdr',
+	C_SRC = 'text/x-csrc',
 	CPP_HDR = 'text/x-c++hdr',
 	CPP_SRC = 'text/x-c++src',
 	CSHARP = 'text/x-csharp',
-	HASKELL = 'text/x-haskell',
-	C_SRC = 'text/x-csrc',
-	C_HDR = 'text/x-chdr',
-	PHP = 'text/x-php',
-	RUBY = 'text/x-ruby',
-	GO = 'text/x-go',
-	RUST = 'text/x-rust',
-	SHELL = 'text/x-shellscript',
-	BAT = 'application/x-bat',
-	SQL = 'text/x-sql',
-	R = 'text/x-r',
-	SCALA = 'text/x-scala',
-	KOTLIN = 'text/x-kotlin',
-	SWIFT = 'text/x-swift',
+	CSS = 'text/css',
+	CSV = 'text/csv',
+	CUDA = 'text/x-cuda',
 	DART = 'text/x-dart',
-	VUE = 'text/x-vue',
+	GO = 'text/x-go',
+	HASKELL = 'text/x-haskell',
+	HTML = 'text/html',
+	JAVA = 'text/x-java-source',
+	JAVASCRIPT = 'text/javascript',
+	JAVASCRIPT_APP = 'application/javascript',
+	JSON = 'application/json',
+	JSONL = 'application/jsonl',
+	JSX = 'text/jsx',
+	KOTLIN = 'text/x-kotlin',
+	LATEX = 'application/x-latex',
+	MARKDOWN = 'text/markdown',
+	PHP = 'text/x-php',
+	PLAIN = 'text/plain',
+	PROPERTIES = 'text/properties',
+	PYTHON = 'text/x-python',
+	R = 'text/x-r',
+	RUBY = 'text/x-ruby',
+	RUST = 'text/x-rust',
+	SCALA = 'text/x-scala',
+	SHELL = 'text/x-shellscript',
+	SQL = 'text/x-sql',
 	SVELTE = 'text/x-svelte',
+	SWIFT = 'text/x-swift',
 	TEX = 'text/x-tex',
 	TEX_APP = 'application/x-tex',
-	LATEX = 'application/x-latex',
-	BIBTEX = 'text/x-bibtex',
-	CUDA = 'text/x-cuda',
-	PROPERTIES = 'text/properties'
+	TSX = 'text/tsx',
+	TYPESCRIPT = 'text/typescript',
+	VUE = 'text/x-vue',
+	XML_APP = 'application/xml',
+	XML_TEXT = 'text/xml',
+	YAML_APP = 'application/yaml',
+	YAML_TEXT = 'text/yaml'
 }

tools/ui/src/lib/enums/keyboard.enums.ts

@@ -2,19 +2,19 @@
  * Keyboard key names for event handling
  */
 export enum KeyboardKey {
-	ENTER = 'Enter',
-	ESCAPE = 'Escape',
-	ARROW_UP = 'ArrowUp',
 	ARROW_DOWN = 'ArrowDown',
 	ARROW_LEFT = 'ArrowLeft',
 	ARROW_RIGHT = 'ArrowRight',
-	TAB = 'Tab',
+	ARROW_UP = 'ArrowUp',
 	B_LOWER = 'b',
 	D_LOWER = 'd',
 	D_UPPER = 'D',
 	E_UPPER = 'E',
+	ENTER = 'Enter',
+	ESCAPE = 'Escape',
 	K_LOWER = 'k',
 	O_LOWER = 'o',
 	O_UPPER = 'O',
-	SPACE = ' '
+	SPACE = ' ',
+	TAB = 'Tab'
 }

tools/ui/src/lib/enums/mcp.enums.ts

@@ -2,61 +2,61 @@
  * Connection lifecycle phases for MCP protocol
  */
 export enum MCPConnectionPhase {
-	IDLE = 'idle',
-	TRANSPORT_CREATING = 'transport_creating',
-	TRANSPORT_READY = 'transport_ready',
-	INITIALIZING = 'initializing',
 	CAPABILITIES_EXCHANGED = 'capabilities_exchanged',
-	LISTING_TOOLS = 'listing_tools',
 	CONNECTED = 'connected',
+	DISCONNECTED = 'disconnected',
 	ERROR = 'error',
-	DISCONNECTED = 'disconnected'
+	IDLE = 'idle',
+	INITIALIZING = 'initializing',
+	LISTING_TOOLS = 'listing_tools',
+	TRANSPORT_CREATING = 'transport_creating',
+	TRANSPORT_READY = 'transport_ready'
 }
 
 /**
  * Log level for connection events
  */
 export enum MCPLogLevel {
+	ERROR = 'error',
 	INFO = 'info',
-	WARN = 'warn',
-	ERROR = 'error'
+	WARN = 'warn'
 }
 
 /**
  * Transport types for MCP connections
  */
 export enum MCPTransportType {
-	WEBSOCKET = 'websocket',
+	SSE = 'sse',
 	STREAMABLE_HTTP = 'streamable_http',
-	SSE = 'sse'
+	WEBSOCKET = 'websocket'
 }
 
 /**
  * Health check status for MCP servers
  */
 export enum HealthCheckStatus {
-	IDLE = 'idle',
 	CONNECTING = 'connecting',
-	SUCCESS = 'success',
-	ERROR = 'error'
+	ERROR = 'error',
+	IDLE = 'idle',
+	SUCCESS = 'success'
 }
 
 /**
  * Content types for MCP tool results
  */
 export enum MCPContentType {
-	TEXT = 'text',
 	IMAGE = 'image',
-	RESOURCE = 'resource'
+	RESOURCE = 'resource',
+	TEXT = 'text'
 }
 
 /**
  * JSON Schema types used in MCP tool definitions
  */
 export enum JsonSchemaType {
+	NUMBER = 'number',
 	OBJECT = 'object',
-	STRING = 'string',
-	NUMBER = 'number'
+	STRING = 'string'
 }
 
 /**

tools/ui/src/lib/enums/model.enums.ts

@@ -1,6 +1,6 @@
 export enum ModelModality {
-	TEXT = 'TEXT',
 	AUDIO = 'AUDIO',
-	VISION = 'VISION',
-	VIDEO = 'VIDEO'
+	TEXT = 'TEXT',
+	VIDEO = 'VIDEO',
+	VISION = 'VISION'
 }

tools/ui/src/lib/enums/reasoning-effort.enums.ts

@@ -4,9 +4,9 @@
  */
 export enum ReasoningEffort {
 	DEFAULT = 'default',
-	OFF = 'off',
-	LOW = 'low',
-	MEDIUM = 'medium',
 	HIGH = 'high',
-	MAX = 'max'
+	LOW = 'low',
+	MAX = 'max',
+	MEDIUM = 'medium',
+	OFF = 'off'
 }

tools/ui/src/lib/enums/server.enums.ts

@@ -13,11 +13,11 @@ export enum ServerRole {
  * Used as the `value` field in the status object from /models endpoint
  */
 export enum ServerModelStatus {
-	UNLOADED = 'unloaded',
-	LOADING = 'loading',
+	FAILED = 'failed',
 	LOADED = 'loaded',
+	LOADING = 'loading',
 	SLEEPING = 'sleeping',
-	FAILED = 'failed'
+	UNLOADED = 'unloaded'
 }
 
 /**
@@ -26,10 +26,10 @@ export enum ServerModelStatus {
  * tools/server/server-models.cpp from the C++ server.
  */
 export enum ServerModelsSseEventType {
-	STATUS_CHANGE = 'status_change',
-	MODEL_STATUS = 'model_status',
-	STATUS_UPDATE = 'status_update',
-	MODELS_RELOAD = 'models_reload',
+	DOWNLOAD_PROGRESS = 'download_progress',
 	MODEL_REMOVE = 'model_remove',
-	DOWNLOAD_PROGRESS = 'download_progress'
+	MODEL_STATUS = 'model_status',
+	MODELS_RELOAD = 'models_reload',
+	STATUS_CHANGE = 'status_change',
+	STATUS_UPDATE = 'status_update'
 }

tools/ui/src/lib/enums/settings.enums.ts

@@ -2,26 +2,26 @@
  * Parameter source - indicates whether a parameter uses default or custom value
  */
 export enum ParameterSource {
-	DEFAULT = 'default',
-	CUSTOM = 'custom'
+	CUSTOM = 'custom',
+	DEFAULT = 'default'
 }
 
 /**
  * Syncable parameter type - data types for parameters that can be synced with server
  */
 export enum SyncableParameterType {
+	BOOLEAN = 'boolean',
 	NUMBER = 'number',
-	STRING = 'string',
-	BOOLEAN = 'boolean'
+	STRING = 'string'
 }
 
 /**
  * Settings field type - defines the input type for settings fields
  */
 export enum SettingsFieldType {
-	INPUT = 'input',
-	TEXTAREA = 'textarea',
 	CHECKBOX = 'checkbox',
+	INPUT = 'input',
+	RADIO = 'radio',
 	SELECT = 'select',
-	RADIO = 'radio'
+	TEXTAREA = 'textarea'
 }

tools/ui/src/lib/enums/splash.enums.ts

@@ -2,6 +2,6 @@
  * Splash screen orientation for iOS apple-touch-startup-image
  */
 export enum SplashOrientation {
-	PORTRAIT = 'portrait',
-	LANDSCAPE = 'landscape'
+	LANDSCAPE = 'landscape',
+	PORTRAIT = 'portrait'
 }

tools/ui/src/lib/enums/tools.enums.ts

@@ -1,20 +1,20 @@
 export enum ToolSource {
-	BUILTIN = 'builtin',
-	MCP = 'mcp',
+	BROWSER = 'browser',
 	CUSTOM = 'custom',
-	FRONTEND = 'frontend'
+	MCP = 'mcp',
+	SERVER = 'server'
 }
 
 export enum ToolPermissionDecision {
 	ALWAYS = 'always',
 	ALWAYS_SERVER = 'always_server',
-	ONCE = 'once',
-	DENY = 'deny'
+	DENY = 'deny',
+	ONCE = 'once'
 }
 
 export enum ToolResponseField {
-	PLAIN_TEXT = 'plain_text_response',
-	ERROR = 'error'
+	ERROR = 'error',
+	PLAIN_TEXT = 'plain_text_response'
 }
 
 /**
@@ -22,28 +22,34 @@ export enum ToolResponseField {
  * Mirrors the server-side validation in server-tools.cpp.
  */
 export enum GlobSearchType {
-	FILE = 'file',
+	ALL = 'all',
 	DIR = 'dir',
-	ALL = 'all'
+	FILE = 'file'
 }
 
 /**
- * Wire-format identifiers for built-in and frontend tools. The string
+ * Wire-format identifiers for server and browser tools. The string
  * value matches what the model emits in tool call names, so comparing
- * against `BuiltInTool.READ_FILE` is equivalent to comparing against the
- * raw `'read_file'` literal - the enum just keeps the two in lock-step
- * and gives TypeScript a single source of truth for autocomplete / rename
- * support.
+ * against `BuiltInTool.SERVER_READ_FILE` is equivalent to comparing
+ * against the raw `'read_file'` literal - the enum just keeps the two in
+ * lock-step and gives TypeScript a single source of truth for autocomplete
+ * / rename support.
+ *
+ * The `SERVER_` / `BROWSER_` prefixes mirror the tool's primary source
+ * (llama-server vs llama-ui). `get_info` is the exception: it is served by
+ * the server, but llama-ui falls back to a browser implementation when the
+ * server does not provide it, so it can surface under both categories in
+ * the UI while keeping a single wire name.
  */
 export enum BuiltInTool {
-	READ_FILE = 'read_file',
-	READ_MEDIA = 'read_media',
-	EDIT_FILE = 'edit_file',
-	WRITE_FILE = 'write_file',
-	GET_DATETIME = 'get_datetime',
-	GET_INFO = 'get_info',
-	FILE_GLOB_SEARCH = 'file_glob_search',
-	GREP_SEARCH = 'grep_search',
-	EXEC_SHELL_COMMAND = 'exec_shell_command',
-	RUN_JAVASCRIPT = 'run_javascript'
+	BROWSER_GET_DATETIME = 'get_datetime',
+	BROWSER_READ_MEDIA = 'read_media',
+	BROWSER_RUN_JAVASCRIPT = 'run_javascript',
+	SERVER_EDIT_FILE = 'edit_file',
+	SERVER_EXEC_SHELL_COMMAND = 'exec_shell_command',
+	SERVER_FILE_GLOB_SEARCH = 'file_glob_search',
+	SERVER_GET_INFO = 'get_info',
+	SERVER_GREP_SEARCH = 'grep_search',
+	SERVER_READ_FILE = 'read_file',
+	SERVER_WRITE_FILE = 'write_file'
 }

tools/ui/src/lib/enums/ui.enums.ts

@@ -1,22 +1,22 @@
 export enum ColorMode {
-	LIGHT = 'light',
 	DARK = 'dark',
+	LIGHT = 'light',
 	SYSTEM = 'system'
 }
 
 export enum TooltipSide {
-	TOP = 'top',
-	RIGHT = 'right',
 	BOTTOM = 'bottom',
-	LEFT = 'left'
+	LEFT = 'left',
+	RIGHT = 'right',
+	TOP = 'top'
 }
 
 /**
  * MCP prompt display variant
  */
 export enum McpPromptVariant {
-	MESSAGE = 'message',
-	ATTACHMENT = 'attachment'
+	ATTACHMENT = 'attachment',
+	MESSAGE = 'message'
 }
 
 /**
@@ -39,8 +39,8 @@ export enum HtmlInputType {
  * Alert level that drives the context gauge dial color.
  */
 export enum ColorLevel {
-	OK = 'ok',
-	WARNING = 'warning',
 	CRITICAL = 'critical',
-	NEUTRAL = 'neutral'
+	NEUTRAL = 'neutral',
+	OK = 'ok',
+	WARNING = 'warning'
 }

tools/ui/src/lib/hooks/use-tools-panel.svelte.ts

@@ -46,13 +46,13 @@ export function useToolsPanel(): UseToolsPanelReturn {
 
 		// Tools endpoint is unreachable (404) — server started without --tools
 		if (toolsStore.isToolsEndpointUnreachable) {
-			return `To enable Built-In Tools you need to run llama-server with ${CLI_FLAGS.TOOLS} all or ${CLI_FLAGS.TOOLS} <name> flag. To see MCP Tools you need to add / enable MCP Server(s).`;
+			return `To enable Server Tools you need to run llama-server with ${CLI_FLAGS.TOOLS} all or ${CLI_FLAGS.TOOLS} <name> flag. To see MCP Tools you need to add / enable MCP Server(s).`;
 		}
 
 		// Other errors — return null so UI shows "Failed to load tools"
 		if (toolsStore.error) return null;
 
-		return `To enable Built-In Tools you need to run llama-server with ${CLI_FLAGS.TOOLS} all or ${CLI_FLAGS.TOOLS} <name> flag. To see MCP Tools you need to add / enable MCP Server(s).`;
+		return `To enable Server Tools you need to run llama-server with ${CLI_FLAGS.TOOLS} all or ${CLI_FLAGS.TOOLS} <name> flag. To see MCP Tools you need to add / enable MCP Server(s).`;
 	});
 
 	function isGroupChecked(group: ToolGroup): boolean {
@@ -95,8 +95,8 @@ export function useToolsPanel(): UseToolsPanelReturn {
 	}
 
 	function handleOpen(): void {
-		if (toolsStore.builtinTools.length === 0 && !toolsStore.loading) {
-			toolsStore.fetchBuiltinTools();
+		if (toolsStore.serverTools.length === 0 && !toolsStore.loading) {
+			toolsStore.fetchServerTools();
 		}
 
 		mcpStore.runHealthChecksForServers(mcpStore.getServers().filter((s) => s.enabled));

tools/ui/src/lib/services/index.ts

@@ -262,9 +262,9 @@ export { ParameterSyncService } from './parameter-sync.service';
 export { MCPService } from './mcp.service';
 
 /**
- * **SandboxService** - Frontend JavaScript execution in a browser sandbox
+ * **SandboxService** - Browser JavaScript execution in a browser sandbox
  *
- * Stateless executor for the run_javascript frontend tool. Model generated
+ * Stateless executor for the run_javascript browser tool. Model generated
  * code runs in a Web Worker spawned inside a sandboxed iframe with an opaque
  * origin: no access to the app origin, its storage or its API, and outgoing
  * requests carry a null origin. The code never touches a main thread, so the
@@ -274,7 +274,7 @@ export { MCPService } from './mcp.service';
  * **Architecture & Relationships:**
  * - **SandboxService** (this class): Stateless sandbox execution
  * - **toolsStore**: Exposes the tool definition when the sandbox is enabled
- * - **agenticStore**: Dispatches ToolSource.FRONTEND calls here
+ * - **agenticStore**: Dispatches ToolSource.BROWSER calls here
  *
  * @see buildSandboxToolDefinition in utils/sandbox-tool - tool schema sent to the LLM
  * @see agenticStore in stores/agentic.svelte.ts - tool dispatch

tools/ui/src/lib/services/read-media.service.ts

@@ -28,15 +28,15 @@ function fileExtension(path: string): string {
 }
 
 /**
- * **ReadMediaService** - frontend executor for the `read_media` tool
+ * **ReadMediaService** - browser executor for the `read_media` tool
  *
  * The tool is synthetic: no such tool exists on the server. It reads the file
- * through the built-in `read_file` tool with the `base64` response type, then
+ * through the server `read_file` tool with the `base64` response type, then
  * turns the bytes into a data URI line. The agentic store lifts that line into
  * an image or audio attachment on the tool result message, which is what makes
  * the model perceive the file instead of reading a wall of base64.
  *
- * Living in the frontend is what lets it exist only for models that can
+ * Living in the browser is what lets it exist only for models that can
  * actually use the result - the server has no idea which model is selected.
  *
  * @see buildReadMediaToolDefinition in constants/read-media.ts - tool schema sent to the LLM
@@ -82,7 +82,7 @@ export class ReadMediaService {
 		}
 
 		const raw = await ToolsService.executeToolRaw(
-			BuiltInTool.READ_FILE,
+			BuiltInTool.SERVER_READ_FILE,
 			{ path },
 			signal,
 			cwd,

tools/ui/src/lib/services/sandbox.service.ts

@@ -68,7 +68,7 @@ function formatReply(reply: SandboxReply): ToolExecutionResult {
 
 export class SandboxService {
 	/**
-	 * Execute a frontend sandbox tool call and return its output.
+	 * Execute a browser sandbox tool call and return its output.
 	 * One disposable iframe per execution, removed on completion,
 	 * timeout or abort. Removing the iframe terminates the worker
 	 * at the browser level, so runaway code cannot outlive it.
@@ -79,7 +79,7 @@ export class SandboxService {
 		signal?: AbortSignal
 	): Promise<ToolExecutionResult> {
 		if (toolName !== SANDBOX_TOOL_NAME) {
-			return { content: `Unknown frontend tool: ${toolName}`, isError: true };
+			return { content: `Unknown browser tool: ${toolName}`, isError: true };
 		}
 
 		const code = typeof params.code === 'string' ? params.code : '';

tools/ui/src/lib/services/tools.service.ts

@@ -1,23 +1,23 @@
 import { base } from '$app/paths';
 import { API_TOOLS, HEADERS } from '$lib/constants';
 import { ToolResponseField } from '$lib/enums';
-import type { ServerBuiltinToolInfo, ToolExecutionResult } from '$lib/types';
+import type { ServerToolInfo, ToolExecutionResult } from '$lib/types';
 import { apiFetch } from '$lib/utils';
 import { getJsonHeaders } from '$lib/utils/api-headers';
 import { parseSseJsonStream, type SseJsonEvent } from '$lib/utils/sse';
 
 export class ToolsService {
 	/**
-	 * Fetch the list of built-in tools from the server.
+	 * Fetch the list of server tools from the server.
 	 *
 	 * @returns Array of tool definitions in OpenAI-compatible format
 	 */
-	static async list(): Promise<ServerBuiltinToolInfo[]> {
-		return apiFetch<ServerBuiltinToolInfo[]>(API_TOOLS.LIST);
+	static async list(): Promise<ServerToolInfo[]> {
+		return apiFetch<ServerToolInfo[]>(API_TOOLS.LIST);
 	}
 
 	/**
-	 * Execute a built-in tool on the server.
+	 * Execute a server tool on the server.
 	 *
 	 * @param cwd - Working directory for the tool call, sent as the
 	 * x-tool-cwd request header. The server resolves relative paths
@@ -48,7 +48,7 @@ export class ToolsService {
 	}
 
 	/**
-	 * Execute a built-in tool and return the raw JSON response. Unlike
+	 * Execute a server tool and return the raw JSON response. Unlike
 	 * executeTool, this preserves structured fields (e.g. file_glob_search's
 	 * `entries` and `base`) that the flattened ToolExecutionResult drops.
 	 *
@@ -77,7 +77,7 @@ export class ToolsService {
 	}
 
 	/**
-	 * Stream a built-in tool's output chunks from the server. The server
+	 * Stream a server tool's output chunks from the server. The server
 	 * `POST /tools` endpoint with `{stream: true}` emits `data: {"chunk": "..."}`
 	 * events followed by a terminal `data: {"done": true}` (optionally with
 	 * `error`). Yields the chunk string for each partial event.

tools/ui/src/lib/stores/agentic.svelte.ts

@@ -326,8 +326,8 @@ class AgenticStore {
 		const maxTurns = Number(settings.agenticMaxTurns) || DEFAULT_AGENTIC_CONFIG.maxTurns;
 		const hasTools =
 			mcpStore.hasEnabledServers(perChatOverrides) ||
-			toolsStore.builtinTools.length > 0 ||
-			toolsStore.frontendTools.length > 0 ||
+			toolsStore.serverTools.length > 0 ||
+			toolsStore.browserTools.length > 0 ||
 			toolsStore.customTools.length > 0;
 
 		return {
@@ -455,9 +455,9 @@ class AgenticStore {
 		this._continueResolvers.delete(conversationId);
 		this._steeringMessages.delete(conversationId);
 
-		// Ensure built-in tools are fetched before checking if agentic is enabled
-		if (toolsStore.builtinTools.length === 0 && !toolsStore.loading) {
-			await toolsStore.fetchBuiltinTools();
+		// Ensure server tools are fetched before checking if agentic is enabled
+		if (toolsStore.serverTools.length === 0 && !toolsStore.loading) {
+			await toolsStore.fetchServerTools();
 		}
 
 		const agenticConfig = this.getConfig(settingsStore.config, perChatOverrides);
@@ -906,8 +906,8 @@ class AgenticStore {
 				} else {
 					try {
 						if (
-							toolSource === ToolSource.BUILTIN &&
-							toolName === BuiltInTool.EXEC_SHELL_COMMAND &&
+							toolSource === ToolSource.SERVER &&
+							toolName === BuiltInTool.SERVER_EXEC_SHELL_COMMAND &&
 							createToolResultMessage &&
 							updateToolResultMessage
 						) {
@@ -938,7 +938,7 @@ class AgenticStore {
 								}
 							}
 							result = accumulated;
-						} else if (toolSource === ToolSource.BUILTIN) {
+						} else if (toolSource === ToolSource.SERVER) {
 							const args = this.parseToolArguments(toolCall.function.arguments);
 							const cwd = conversationsStore.activeConversation?.cwd;
 							const executionResult = await ToolsService.executeTool(toolName, args, signal, cwd);
@@ -946,16 +946,16 @@ class AgenticStore {
 							result = executionResult.content;
 
 							if (executionResult.isError) toolSuccess = false;
-						} else if (toolSource === ToolSource.FRONTEND) {
+						} else if (toolSource === ToolSource.BROWSER) {
 							const args = this.parseToolArguments(toolCall.function.arguments);
 
 							let executionResult: ToolExecutionResult;
 
-							if (toolName === BuiltInTool.GET_DATETIME) {
+							if (toolName === BuiltInTool.BROWSER_GET_DATETIME) {
 								executionResult = executeGetDatetimeTool();
-							} else if (toolName === BuiltInTool.GET_INFO) {
+							} else if (toolName === BuiltInTool.SERVER_GET_INFO) {
 								executionResult = executeBrowserInfoTool();
-							} else if (toolName === BuiltInTool.READ_MEDIA) {
+							} else if (toolName === BuiltInTool.BROWSER_READ_MEDIA) {
 								executionResult = await ReadMediaService.executeTool(
 									args,
 									{

tools/ui/src/lib/stores/tools.svelte.ts

@@ -28,11 +28,11 @@ import { SvelteMap, SvelteSet } from 'svelte/reactivity';
 /** Stable selection identity for a tool, shared by the disabled set and the permission store */
 
 class ToolsStore {
-	private _builtinTools = $state<OpenAIToolDefinition[]>([]);
+	private _serverTools = $state<OpenAIToolDefinition[]>([]);
 	private _loading = $state(false);
 	private _error = $state<string | null>(null);
 	private _disabledTools = $state(new SvelteSet<string>());
-	// builtin tools that resolve their paths against the working directory,
+	// server tools that resolve their paths against the working directory,
 	// as declared by the server in its `/tools` listing
 	private _cwdAwareTools = $state(new SvelteSet<string>());
 	private _toolsEndpointUnreachable = $state(false);
@@ -58,7 +58,7 @@ class ToolsStore {
 			console.error('[ToolsStore] Failed to load disabled tools from localStorage:', err);
 		}
 
-		this.fetchBuiltinTools();
+		this.fetchServerTools();
 	}
 
 	private persistDisabledTools(): void {
@@ -78,10 +78,10 @@ class ToolsStore {
 				return serverId ? `mcp-${serverId}:${name}` : `mcp:${name}`;
 			case ToolSource.CUSTOM:
 				return `custom:${name}`;
-			case ToolSource.FRONTEND:
-				return `frontend:${name}`;
+			case ToolSource.BROWSER:
+				return `browser:${name}`;
 			default:
-				return `builtin:${name}`;
+				return `server:${name}`;
 		}
 	}
 
@@ -164,8 +164,8 @@ class ToolsStore {
 		};
 	}
 
-	get builtinTools(): OpenAIToolDefinition[] {
-		return this._builtinTools;
+	get serverTools(): OpenAIToolDefinition[] {
+		return this._serverTools;
 	}
 
 	get serverHome(): string | null {
@@ -176,7 +176,7 @@ class ToolsStore {
 		return this.mcpEntries().map((e) => e.definition);
 	}
 
-	get frontendTools(): OpenAIToolDefinition[] {
+	get browserTools(): OpenAIToolDefinition[] {
 		const tools: OpenAIToolDefinition[] = [buildGetDatetimeToolDefinition()];
 
 		if (settingsStore.config.jsSandboxEnabled) {
@@ -188,25 +188,25 @@ class ToolsStore {
 		if (readMedia) tools.push(readMedia);
 
 		// provide browser's get_info tool if server doesn't provide one
-		if (!this.hasBuiltinTool(BuiltInTool.GET_INFO)) {
+		if (!this.hasServerTool(BuiltInTool.SERVER_GET_INFO)) {
 			tools.push(buildBrowserInfoToolDefinition());
 		}
 
 		return tools;
 	}
 
-	private hasBuiltinTool(name: BuiltInTool): boolean {
-		return this._builtinTools.some((def) => def.function.name === name);
+	private hasServerTool(name: BuiltInTool): boolean {
+		return this._serverTools.some((def) => def.function.name === name);
 	}
 
 	/**
-	 * `read_media` runs in the frontend on top of the server's `read_file`, so it
+	 * `read_media` runs in the browser on top of the server's `read_file`, so it
 	 * exists only when that tool is served and the active model can perceive the
 	 * bytes. The server cannot make this call - it does not know which model the
 	 * conversation uses.
 	 */
 	private readMediaTool(): OpenAIToolDefinition | null {
-		if (!this.hasBuiltinTool(BuiltInTool.READ_FILE)) return null;
+		if (!this.hasServerTool(BuiltInTool.SERVER_READ_FILE)) return null;
 
 		const model = modelsStore.selectedModelName ?? modelsStore.models[0]?.model ?? '';
 
@@ -304,23 +304,23 @@ class ToolsStore {
 			entries.push(entry);
 		};
 
-		for (const def of this._builtinTools) {
+		for (const def of this._serverTools) {
 			const name = def.function.name;
 
 			push({
 				definition: def,
-				key: this.toolKey(ToolSource.BUILTIN, name),
-				source: ToolSource.BUILTIN
+				key: this.toolKey(ToolSource.SERVER, name),
+				source: ToolSource.SERVER
 			});
 		}
 
-		for (const def of this.frontendTools) {
+		for (const def of this.browserTools) {
 			const name = def.function.name;
 
 			push({
 				definition: def,
-				key: this.toolKey(ToolSource.FRONTEND, name),
-				source: ToolSource.FRONTEND
+				key: this.toolKey(ToolSource.BROWSER, name),
+				source: ToolSource.BROWSER
 			});
 		}
 
@@ -384,17 +384,17 @@ class ToolsStore {
 				return entry.serverName ?? '';
 			case ToolSource.CUSTOM:
 				return TOOL_GROUP_LABELS[ToolSource.CUSTOM];
-			case ToolSource.FRONTEND:
-				return TOOL_GROUP_LABELS[ToolSource.FRONTEND];
+			case ToolSource.BROWSER:
+				return TOOL_GROUP_LABELS[ToolSource.BROWSER];
 			default:
-				return TOOL_GROUP_LABELS[ToolSource.BUILTIN];
+				return TOOL_GROUP_LABELS[ToolSource.SERVER];
 		}
 	}
 
 	/**
 	 * Enabled tool definitions for sending to the LLM.
 	 * MCP tool schemas are normalized here so the wire payload is consistent
-	 * across all four sources (built-in, frontend/sandbox, MCP, custom JSON).
+	 * across all four sources (server, browser/sandbox, MCP, custom JSON).
 	 * The API identifies tools by name, so a name is sent at most once.
 	 */
 	getEnabledToolsForLLM(): OpenAIToolDefinition[] {
@@ -417,8 +417,8 @@ class ToolsStore {
 			result.push(def);
 		};
 
-		for (const def of this._builtinTools) take(def);
-		for (const def of this.frontendTools) take(def);
+		for (const def of this._serverTools) take(def);
+		for (const def of this.browserTools) take(def);
 		// mcpEntries() over mcpStore directly so wire shape stays normalized and aligned with the tools UI.
 		for (const entry of this.mcpEntries()) take(entry.definition);
 		for (const def of this.customTools) take(def);
@@ -542,11 +542,11 @@ class ToolsStore {
 
 		if (entry.serverName) return mcpStore.getServerDisplayName(entry.serverName);
 
-		if (entry.source === ToolSource.BUILTIN) return TOOL_SERVER_LABELS[ToolSource.BUILTIN];
+		if (entry.source === ToolSource.SERVER) return TOOL_SERVER_LABELS[ToolSource.SERVER];
 
 		if (entry.source === ToolSource.CUSTOM) return TOOL_SERVER_LABELS[ToolSource.CUSTOM];
 
-		if (entry.source === ToolSource.FRONTEND) return TOOL_SERVER_LABELS[ToolSource.FRONTEND];
+		if (entry.source === ToolSource.BROWSER) return TOOL_SERVER_LABELS[ToolSource.BROWSER];
 
 		return '';
 	}
@@ -556,27 +556,27 @@ class ToolsStore {
 		return this.findEntryByName(toolName)?.key ?? null;
 	}
 
-	/** Check if there are any enabled tools available (builtin, MCP, or custom) */
+	/** Check if there are any enabled tools available (server, MCP, or custom) */
 	get hasEnabledTools(): boolean {
 		return this.getEnabledToolsForLLM().length > 0;
 	}
 
 	/**
-	 * Check if a working directory is worth setting: at least one builtin tool
+	 * Check if a working directory is worth setting: at least one server tool
 	 * that reads it is both served and left enabled by the user.
 	 */
 	get hasEnabledCwdTools(): boolean {
-		return this._builtinTools.some((def) => {
+		return this._serverTools.some((def) => {
 			const name = def.function.name;
 
 			return (
 				this._cwdAwareTools.has(name) &&
-				!this._disabledTools.has(this.toolKey(ToolSource.BUILTIN, name))
+				!this._disabledTools.has(this.toolKey(ToolSource.SERVER, name))
 			);
 		});
 	}
 
-	async fetchBuiltinTools(): Promise<void> {
+	async fetchServerTools(): Promise<void> {
 		if (this._loading) return;
 
 		this._loading = true;
@@ -586,7 +586,7 @@ class ToolsStore {
 		try {
 			const toolInfos = await ToolsService.list();
 
-			this._builtinTools = toolInfos.map((info) => info.definition);
+			this._serverTools = toolInfos.map((info) => info.definition);
 			this._cwdAwareTools = new SvelteSet(
 				toolInfos.filter((info) => info.uses_cwd).map((info) => info.tool)
 			);
@@ -599,9 +599,9 @@ class ToolsStore {
 			// TODO: check status code instead of relying on message
 			if (errorMessage.includes('this feature is disabled')) {
 				this._toolsEndpointUnreachable = true;
-				console.info('[ToolsStore] Built-in tools are disabled on the server');
+				console.info('[ToolsStore] Server tools are disabled on the server');
 			} else {
-				console.error('[ToolsStore] Failed to fetch built-in tools:', err);
+				console.error('[ToolsStore] Failed to fetch server tools:', err);
 			}
 		} finally {
 			this._loading = false;
@@ -618,7 +618,7 @@ class ToolsStore {
 		if (this._serverHome !== undefined) return this._serverHome;
 
 		try {
-			const res = await ToolsService.executeToolRaw(BuiltInTool.FILE_GLOB_SEARCH, {
+			const res = await ToolsService.executeToolRaw(BuiltInTool.SERVER_FILE_GLOB_SEARCH, {
 				limit: 1,
 				max_depth: 1,
 				path: HOME_TILDE,

tools/ui/src/lib/types/index.ts

@@ -147,7 +147,7 @@ export type {
 	ServerStatus,
 	ToolCallParams,
 	ToolExecutionResult,
-	ServerBuiltinToolInfo,
+	ServerToolInfo,
 	Tool,
 	Prompt,
 	GetPromptResult,
@@ -208,7 +208,7 @@ export type {
 export type { DesktopIconStripItem } from './navigation';
 
 // Tools types
-export type { ToolEntry, ToolGroup, BuiltinToolUiEntry } from './tools';
+export type { ToolEntry, ToolGroup, ToolUiEntry } from './tools';
 
 // Reasoning
 export type { ReasoningEffortLevel } from './reasoning';

tools/ui/src/lib/types/mcp.d.ts

@@ -285,10 +285,10 @@ export interface ToolExecutionResult {
 	isError: boolean;
 }
 
-export interface ServerBuiltinToolInfo {
+export interface ServerToolInfo {
 	display_name: string;
 	tool: string;
-	type: ToolSource.BUILTIN;
+	type: ToolSource.SERVER;
 	permissions: {
 		write: boolean;
 	};

tools/ui/src/lib/types/tools.d.ts

@@ -3,12 +3,12 @@ import type { ToolSource } from '$lib/enums';
 import type { Component } from 'svelte';
 
 /**
- * UI metadata for a built-in or frontend tool, keyed by its `BuiltInTool` id.
+ * UI metadata for a server or browser tool, keyed by its `BuiltInTool` id.
  */
-export interface BuiltinToolUiEntry {
+export interface ToolUiEntry {
 	icon: Component;
 	label: string;
-	source: ToolSource.BUILTIN | ToolSource.FRONTEND;
+	source: ToolSource.SERVER | ToolSource.BROWSER;
 }
 
 export interface ToolEntry {
@@ -17,7 +17,7 @@ export interface ToolEntry {
 	serverName?: string;
 	/** For MCP tools, the server ID (used for permission keys) */
 	serverId?: string;
-	/** Stable selection identity: builtin:name, mcp-<serverId>:name, mcp:name, custom:name */
+	/** Stable selection identity: server:name, mcp-<serverId>:name, mcp:name, custom:name */
 	key: string;
 	definition: OpenAIToolDefinition;
 }

tools/ui/src/lib/utils/built-in-tools.ts

@@ -1,13 +0,0 @@
-import { BUILTIN_TOOL_UI } from '$lib/constants';
-import type { BuiltinToolUiEntry } from '$lib/types';
-
-/**
- * Resolve the UI metadata (label + icon) for a built-in tool by its name.
- * Falls back to null for unknown or non-built-in tools so callers can render
- * a generic chrome instead.
- */
-export function getBuiltinToolUi(toolName: string | undefined): BuiltinToolUiEntry | null {
-	if (!toolName) return null;
-
-	return (BUILTIN_TOOL_UI as Record<string, BuiltinToolUiEntry>)[toolName] ?? null;
-}

tools/ui/src/lib/utils/get-datetime.ts

@@ -1,5 +1,5 @@
 /**
- * Frontend executor for the `get_datetime` tool. It runs in the browser, so it
+ * Browser executor for the `get_datetime` tool. It runs in the browser, so it
  * reports the user's own clock and time zone instead of the server's UTC time -
  * a chat about "tomorrow" means the user's tomorrow, not the host's.
  *

tools/ui/src/lib/utils/glob-search.ts

@@ -42,7 +42,7 @@ export async function runGlobSearch(
 	}
 
 	const res = await ToolsService.executeToolRaw(
-		BuiltInTool.FILE_GLOB_SEARCH,
+		BuiltInTool.SERVER_FILE_GLOB_SEARCH,
 		{ include: args.include, limit, max_depth: args.maxDepth, path: args.path, type },
 		signal
 	);

tools/ui/src/lib/utils/index.ts

@@ -127,7 +127,7 @@ export { sanitizeKeyValuePairKey, sanitizeKeyValuePairValue } from './sanitize';
 // Image error fallback utilities
 export { getImageErrorFallbackHtml } from './image-error-fallback';
 
-// SSE-with-JSON stream iterator (used by built-in tool streaming, decoupled
+// SSE-with-JSON stream iterator (used by server tool streaming, decoupled
 // from chat.service.ts which embeds its own SSE parser for resume support)
 export { parseSseJsonStream } from './sse';
 
@@ -310,7 +310,7 @@ export {
 	withAbortSignal
 } from './abort';
 
-// Tool-call meta utilities. Parsers for each built-in tool live next to
+// Tool-call meta utilities. Parsers for each server tool live next to
 // their renderer family under
 // `src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageToolCall/parsers/`.
 // This module only carries the helpers that genuinely cross tool
@@ -321,7 +321,7 @@ export { tryParseToolResultObject } from './tool-call-meta';
 // Per-tool UI metadata (label + icon) used by the tool-call chrome.
 // Re-exported through $lib/utils so renderer components can read the
 // label without depending on $lib/constants directly.
-export { getBuiltinToolUi } from './built-in-tools';
+export { getToolUi } from './tool-ui';
 
 // Chat command picker
 
@@ -331,7 +331,7 @@ export { getChatCommands } from './chat-commands';
 // SANDBOX_TOOL_DEFINITION is deprecated; kept for backward compatibility.
 export { buildSandboxToolDefinition, SANDBOX_TOOL_DEFINITION } from './sandbox-tool';
 
-// Frontend `get_datetime` executor (the browser clock, not the server's)
+// Browser `get_datetime` executor (the browser clock, not the server's)
 export { executeGetDatetimeTool } from './get-datetime';
 
 // Browser fallback for the server's get_info tool

tools/ui/src/lib/utils/search-results.ts

@@ -50,10 +50,10 @@ const FAVICON_PATH = '/favicon.ico';
 // (and that callers read off `SearchResult`), so `FieldKey.TITLE` is a
 // drop-in for the literal `'title'`.
 enum FieldKey {
-	TITLE = 'title',
-	URL = 'url',
+	AUTHOR = 'author',
 	PUBLISHED = 'published',
-	AUTHOR = 'author'
+	TITLE = 'title',
+	URL = 'url'
 }
 const FIELD_PREFIXES: ReadonlyArray<{ key: FieldKey; prefix: string }> = [
 	{ key: FieldKey.TITLE, prefix: 'Title:' },

tools/ui/src/lib/utils/tool-ui.ts

@@ -0,0 +1,13 @@
+import { TOOL_UI } from '$lib/constants';
+import type { ToolUiEntry } from '$lib/types';
+
+/**
+ * Resolve the UI metadata (label + icon) for a server or browser tool by its
+ * name. Falls back to null for unknown tools so callers can render a generic
+ * chrome instead.
+ */
+export function getToolUi(toolName: string | undefined): ToolUiEntry | null {
+	if (!toolName) return null;
+
+	return (TOOL_UI as Record<string, ToolUiEntry>)[toolName] ?? null;
+}

tools/ui/tests/client/chat-form-mention-picker-gate.svelte.test.ts

@@ -13,15 +13,15 @@ import { afterEach, describe, expect, it } from 'vitest';
 import { render } from 'vitest-browser-svelte';
 
 const FILE_SEARCH_DEF: OpenAIToolDefinition = {
-	function: { description: '', name: BuiltInTool.FILE_GLOB_SEARCH, parameters: {} },
+	function: { description: '', name: BuiltInTool.SERVER_FILE_GLOB_SEARCH, parameters: {} },
 	type: 'function'
 };
-const FILE_SEARCH_KEY = `builtin:${BuiltInTool.FILE_GLOB_SEARCH}`;
+const FILE_SEARCH_KEY = `server:${BuiltInTool.SERVER_FILE_GLOB_SEARCH}`;
 
-// The store keeps its builtin tool list private; tests inject it through
+// The store keeps its server tool list private; tests inject it through
 // the reactive field so the derived gates recompute.
-function setBuiltinTools(defs: OpenAIToolDefinition[]) {
-	(toolsStore as unknown as { _builtinTools: OpenAIToolDefinition[] })._builtinTools = defs;
+function setServerTools(defs: OpenAIToolDefinition[]) {
+	(toolsStore as unknown as { _serverTools: OpenAIToolDefinition[] })._serverTools = defs;
 }
 
 function renderPicker() {
@@ -34,14 +34,14 @@ function renderPicker() {
 }
 
 afterEach(() => {
-	setBuiltinTools([]);
+	setServerTools([]);
 	toolsStore.setToolEnabled(FILE_SEARCH_KEY, true);
 	localStorage.removeItem(DISABLED_TOOL_KEYS_LOCALSTORAGE_KEY);
 });
 
 describe('ChatFormPickerMention file_glob_search gate', () => {
 	it('explains that file search is unavailable when the server has no tools', async () => {
-		setBuiltinTools([]);
+		setServerTools([]);
 		renderPicker();
 		await tick();
 
@@ -51,7 +51,7 @@ describe('ChatFormPickerMention file_glob_search gate', () => {
 	});
 
 	it('explains that file search must be enabled when the user disabled it', async () => {
-		setBuiltinTools([FILE_SEARCH_DEF]);
+		setServerTools([FILE_SEARCH_DEF]);
 		toolsStore.setToolEnabled(FILE_SEARCH_KEY, false);
 		renderPicker();
 		await tick();

tools/ui/tests/unit/tool-calls.test.ts

@@ -16,7 +16,7 @@ import { describe, expect, it } from 'vitest';
 
 function makeSection(
 	overrides: Partial<AgenticSection> = {},
-	toolName = BuiltInTool.READ_FILE
+	toolName = BuiltInTool.SERVER_READ_FILE
 ): AgenticSection {
 	return {
 		content: '',
@@ -115,15 +115,18 @@ describe('formatCwdMessage / parseCwdMessage', () => {
 
 describe('parseToolArgs (shared)', () => {
 	it('returns null when the section has no toolArgs', () => {
-		const result = parseToolArgs(BuiltInTool.READ_FILE, makeSection({ toolArgs: undefined }));
+		const result = parseToolArgs(
+			BuiltInTool.SERVER_READ_FILE,
+			makeSection({ toolArgs: undefined })
+		);
 
 		expect(result).toBeNull();
 	});
 
 	it('returns null when the tool name does not match', () => {
 		const result = parseToolArgs(
-			BuiltInTool.READ_FILE,
-			makeSection({ toolArgs: '{"path":"/x"}' }, BuiltInTool.WRITE_FILE)
+			BuiltInTool.SERVER_READ_FILE,
+			makeSection({ toolArgs: '{"path":"/x"}' }, BuiltInTool.SERVER_WRITE_FILE)
 		);
 
 		expect(result).toBeNull();
@@ -131,7 +134,7 @@ describe('parseToolArgs (shared)', () => {
 
 	it('returns null when args are not valid final JSON (partial: false)', () => {
 		const result = parseToolArgs(
-			BuiltInTool.READ_FILE,
+			BuiltInTool.SERVER_READ_FILE,
 			makeSection({ toolArgs: '{"path": "/foo.tx' })
 		);
 
@@ -140,7 +143,7 @@ describe('parseToolArgs (shared)', () => {
 
 	it('returns parsed args when valid final JSON', () => {
 		const result = parseToolArgs(
-			BuiltInTool.READ_FILE,
+			BuiltInTool.SERVER_READ_FILE,
 			makeSection({ toolArgs: '{"path":"/foo.txt"}' })
 		);
 
@@ -149,7 +152,7 @@ describe('parseToolArgs (shared)', () => {
 
 	it('accepts partial JSON when partial: true', () => {
 		const result = parseToolArgs(
-			BuiltInTool.READ_FILE,
+			BuiltInTool.SERVER_READ_FILE,
 			makeSection({ toolArgs: '{"path": "/foo.tx' }),
 			{ partial: true }
 		);
@@ -162,7 +165,10 @@ describe('parseWriteFileMeta', () => {
 	it('returns null for sections with a different tool name', () => {
 		expect(
 			parseWriteFileMeta(
-				makeSection({ toolArgs: '{"path":"/x","content":"y"}', toolName: BuiltInTool.READ_FILE })
+				makeSection({
+					toolArgs: '{"path":"/x","content":"y"}',
+					toolName: BuiltInTool.SERVER_READ_FILE
+				})
 			)
 		).toBeNull();
 	});
@@ -170,14 +176,14 @@ describe('parseWriteFileMeta', () => {
 	it('returns null when args have no path-like field', () => {
 		expect(
 			parseWriteFileMeta(
-				makeSection({ toolArgs: '{"content":"x"}', toolName: BuiltInTool.WRITE_FILE })
+				makeSection({ toolArgs: '{"content":"x"}', toolName: BuiltInTool.SERVER_WRITE_FILE })
 			)
 		).toBeNull();
 	});
 
 	it('accepts partial args (renders incrementally as content streams in)', () => {
 		const meta = parseWriteFileMeta(
-			makeSection({ toolArgs: '{"path":"/foo.t', toolName: BuiltInTool.WRITE_FILE })
+			makeSection({ toolArgs: '{"path":"/foo.t', toolName: BuiltInTool.SERVER_WRITE_FILE })
 		);
 
 		expect(meta?.filePath).toBe('/foo.t');
@@ -188,10 +194,10 @@ describe('parseWriteFileMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"path":"/foo.ts","content":"x"}',
-					toolName: BuiltInTool.WRITE_FILE,
+					toolName: BuiltInTool.SERVER_WRITE_FILE,
 					toolResult: '{"result":"wrote","bytes":42}'
 				},
-				BuiltInTool.WRITE_FILE
+				BuiltInTool.SERVER_WRITE_FILE
 			)
 		);
 
@@ -208,7 +214,7 @@ describe('parseWriteFileMeta', () => {
 		const meta = parseWriteFileMeta(
 			makeSection({
 				toolArgs: '{"path":"/foo","content":"x"}',
-				toolName: BuiltInTool.WRITE_FILE,
+				toolName: BuiltInTool.SERVER_WRITE_FILE,
 				toolResult: '{"error":"permission denied"}'
 			})
 		);
@@ -223,10 +229,10 @@ describe('parseEditFileMeta', () => {
 			{
 				toolArgs:
 					'{"path":"/foo.ts","edits":[{"old_text":"a","new_text":"b"},{"old_text":"c","new_text":"d"}]}',
-				toolName: BuiltInTool.EDIT_FILE,
+				toolName: BuiltInTool.SERVER_EDIT_FILE,
 				toolResult: '{"result":"ok","edits_applied":2}'
 			},
-			BuiltInTool.EDIT_FILE
+			BuiltInTool.SERVER_EDIT_FILE
 		);
 		const meta = parseEditFileMeta(section);
 
@@ -242,9 +248,9 @@ describe('parseEditFileMeta', () => {
 		const section = makeSection(
 			{
 				toolArgs: '{"path":"/foo","edits":[{"old_text":""},{"old_text":"a","new_text":""}]}',
-				toolName: BuiltInTool.EDIT_FILE
+				toolName: BuiltInTool.SERVER_EDIT_FILE
 			},
-			BuiltInTool.EDIT_FILE
+			BuiltInTool.SERVER_EDIT_FILE
 		);
 		const meta = parseEditFileMeta(section);
 
@@ -257,10 +263,10 @@ describe('parseEditFileMeta', () => {
 		const section = makeSection(
 			{
 				toolArgs: '{"path":"/foo"}',
-				toolName: BuiltInTool.EDIT_FILE,
+				toolName: BuiltInTool.SERVER_EDIT_FILE,
 				toolResult: '{"error":"bad path","result":"ok"}'
 			},
-			BuiltInTool.EDIT_FILE
+			BuiltInTool.SERVER_EDIT_FILE
 		);
 		const meta = parseEditFileMeta(section);
 
@@ -272,7 +278,7 @@ describe('parseEditFileMeta', () => {
 describe('parseReadFileMeta', () => {
 	it('parses file name alone (no range)', () => {
 		const meta = parseReadFileMeta(
-			makeSection({ toolArgs: '{"path":"/foo.txt"}' }, BuiltInTool.READ_FILE)
+			makeSection({ toolArgs: '{"path":"/foo.txt"}' }, BuiltInTool.SERVER_READ_FILE)
 		);
 
 		expect(meta?.fileName).toBe('foo.txt');
@@ -283,7 +289,7 @@ describe('parseReadFileMeta', () => {
 		const meta = parseReadFileMeta(
 			makeSection(
 				{ toolArgs: '{"path":"/foo.ts","start_line":10,"end_line":20}' },
-				BuiltInTool.READ_FILE
+				BuiltInTool.SERVER_READ_FILE
 			)
 		);
 
@@ -294,7 +300,7 @@ describe('parseReadFileMeta', () => {
 		const meta = parseReadFileMeta(
 			makeSection(
 				{ toolArgs: '{"path":"/foo.ts","start_line":10,"line_count":5}' },
-				BuiltInTool.READ_FILE
+				BuiltInTool.SERVER_READ_FILE
 			)
 		);
 
@@ -302,7 +308,9 @@ describe('parseReadFileMeta', () => {
 	});
 
 	it('returns null when args cannot be parsed', () => {
-		expect(parseReadFileMeta(makeSection({ toolArgs: '{bad' }, BuiltInTool.READ_FILE))).toBeNull();
+		expect(
+			parseReadFileMeta(makeSection({ toolArgs: '{bad' }, BuiltInTool.SERVER_READ_FILE))
+		).toBeNull();
 	});
 });
 
@@ -310,12 +318,12 @@ describe('parseGrepSearchMeta', () => {
 	it('returns null when path or pattern is missing', () => {
 		expect(
 			parseGrepSearchMeta(
-				makeSection({ toolArgs: '{"pattern":"foo"}', toolName: BuiltInTool.GREP_SEARCH })
+				makeSection({ toolArgs: '{"pattern":"foo"}', toolName: BuiltInTool.SERVER_GREP_SEARCH })
 			)
 		).toBeNull();
 		expect(
 			parseGrepSearchMeta(
-				makeSection({ toolArgs: '{"path":"/x"}', toolName: BuiltInTool.GREP_SEARCH })
+				makeSection({ toolArgs: '{"path":"/x"}', toolName: BuiltInTool.SERVER_GREP_SEARCH })
 			)
 		).toBeNull();
 	});
@@ -325,10 +333,10 @@ describe('parseGrepSearchMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"path":"/x","pattern":"foo"}',
-					toolName: BuiltInTool.GREP_SEARCH,
+					toolName: BuiltInTool.SERVER_GREP_SEARCH,
 					toolResult: JSON.stringify({ plain_text_response: 'a.ts:hello\nb.ts:world' })
 				},
-				BuiltInTool.GREP_SEARCH
+				BuiltInTool.SERVER_GREP_SEARCH
 			)
 		);
 
@@ -341,10 +349,10 @@ describe('parseGrepSearchMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"path":"/x","pattern":"foo"}',
-					toolName: BuiltInTool.GREP_SEARCH,
+					toolName: BuiltInTool.SERVER_GREP_SEARCH,
 					toolResult: 'a.ts:hello\nb.ts:world'
 				},
-				BuiltInTool.GREP_SEARCH
+				BuiltInTool.SERVER_GREP_SEARCH
 			)
 		);
 
@@ -356,10 +364,10 @@ describe('parseGrepSearchMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"path":"/x","pattern":"foo","return_line_numbers":true}',
-					toolName: BuiltInTool.GREP_SEARCH,
+					toolName: BuiltInTool.SERVER_GREP_SEARCH,
 					toolResult: 'a.ts:12:hello'
 				},
-				BuiltInTool.GREP_SEARCH
+				BuiltInTool.SERVER_GREP_SEARCH
 			)
 		);
 
@@ -374,10 +382,10 @@ describe('parseFileGlobSearchMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"path":"/x"}',
-					toolName: BuiltInTool.FILE_GLOB_SEARCH,
+					toolName: BuiltInTool.SERVER_FILE_GLOB_SEARCH,
 					toolResult: 'a.ts\nb.ts'
 				},
-				BuiltInTool.FILE_GLOB_SEARCH
+				BuiltInTool.SERVER_FILE_GLOB_SEARCH
 			)
 		);
 
@@ -389,10 +397,10 @@ describe('parseFileGlobSearchMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"path":"/x"}',
-					toolName: BuiltInTool.FILE_GLOB_SEARCH,
+					toolName: BuiltInTool.SERVER_FILE_GLOB_SEARCH,
 					toolResult: JSON.stringify({ plain_text_response: 'a.ts\nb.ts' })
 				},
-				BuiltInTool.FILE_GLOB_SEARCH
+				BuiltInTool.SERVER_FILE_GLOB_SEARCH
 			)
 		);
 
@@ -404,10 +412,10 @@ describe('parseFileGlobSearchMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"path":"/x"}',
-					toolName: BuiltInTool.FILE_GLOB_SEARCH,
+					toolName: BuiltInTool.SERVER_FILE_GLOB_SEARCH,
 					toolResult: JSON.stringify({ error: 'permission denied' })
 				},
-				BuiltInTool.FILE_GLOB_SEARCH
+				BuiltInTool.SERVER_FILE_GLOB_SEARCH
 			)
 		);
 
@@ -418,15 +426,20 @@ describe('parseFileGlobSearchMeta', () => {
 describe('parseRunJavascriptMeta', () => {
 	it('returns null when code is missing', () => {
 		expect(
-			parseRunJavascriptMeta(makeSection({ toolArgs: '{}', toolName: BuiltInTool.RUN_JAVASCRIPT }))
+			parseRunJavascriptMeta(
+				makeSection({ toolArgs: '{}', toolName: BuiltInTool.BROWSER_RUN_JAVASCRIPT })
+			)
 		).toBeNull();
 	});
 
 	it('reads code and timeout', () => {
 		const meta = parseRunJavascriptMeta(
 			makeSection(
-				{ toolArgs: '{"code":"Math.PI","timeout_ms":5000}', toolName: BuiltInTool.RUN_JAVASCRIPT },
-				BuiltInTool.RUN_JAVASCRIPT
+				{
+					toolArgs: '{"code":"Math.PI","timeout_ms":5000}',
+					toolName: BuiltInTool.BROWSER_RUN_JAVASCRIPT
+				},
+				BuiltInTool.BROWSER_RUN_JAVASCRIPT
 			)
 		);
 
@@ -439,10 +452,10 @@ describe('parseRunJavascriptMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"code":"throw new Error()"}',
-					toolName: BuiltInTool.RUN_JAVASCRIPT,
+					toolName: BuiltInTool.BROWSER_RUN_JAVASCRIPT,
 					toolResult: JSON.stringify({ error: 'undefined is not a function' })
 				},
-				BuiltInTool.RUN_JAVASCRIPT
+				BuiltInTool.BROWSER_RUN_JAVASCRIPT
 			)
 		);
 
@@ -457,10 +470,10 @@ describe('parseRunJavascriptMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"code":"[1,2,3]"}',
-					toolName: BuiltInTool.RUN_JAVASCRIPT,
+					toolName: BuiltInTool.BROWSER_RUN_JAVASCRIPT,
 					toolResult: '[1,2,3]'
 				},
-				BuiltInTool.RUN_JAVASCRIPT
+				BuiltInTool.BROWSER_RUN_JAVASCRIPT
 			)
 		);
 
@@ -472,10 +485,10 @@ describe('parseRunJavascriptMeta', () => {
 			makeSection(
 				{
 					toolArgs: '{"code":"foo"}',
-					toolName: BuiltInTool.RUN_JAVASCRIPT,
+					toolName: BuiltInTool.BROWSER_RUN_JAVASCRIPT,
 					toolResult: 'Error: undefined is not a function\n  at <anonymous>:1:1'
 				},
-				BuiltInTool.RUN_JAVASCRIPT
+				BuiltInTool.BROWSER_RUN_JAVASCRIPT
 			)
 		);
 
@@ -487,8 +500,8 @@ describe('parseExecShellCommandMeta', () => {
 	it('reads command from the args', () => {
 		const meta = parseExecShellCommandMeta(
 			makeSection(
-				{ toolArgs: '{"command":"ls -la"}', toolName: BuiltInTool.EXEC_SHELL_COMMAND },
-				BuiltInTool.EXEC_SHELL_COMMAND
+				{ toolArgs: '{"command":"ls -la"}', toolName: BuiltInTool.SERVER_EXEC_SHELL_COMMAND },
+				BuiltInTool.SERVER_EXEC_SHELL_COMMAND
 			)
 		);
 
@@ -499,16 +512,16 @@ describe('parseExecShellCommandMeta', () => {
 		expect(
 			parseExecShellCommandMeta(
 				makeSection(
-					{ toolArgs: '{"cmd":"ls"}', toolName: BuiltInTool.EXEC_SHELL_COMMAND },
-					BuiltInTool.EXEC_SHELL_COMMAND
+					{ toolArgs: '{"cmd":"ls"}', toolName: BuiltInTool.SERVER_EXEC_SHELL_COMMAND },
+					BuiltInTool.SERVER_EXEC_SHELL_COMMAND
 				)
 			)?.command
 		).toBe('ls');
 		expect(
 			parseExecShellCommandMeta(
 				makeSection(
-					{ toolArgs: '{"shell_command":"ls"}', toolName: BuiltInTool.EXEC_SHELL_COMMAND },
-					BuiltInTool.EXEC_SHELL_COMMAND
+					{ toolArgs: '{"shell_command":"ls"}', toolName: BuiltInTool.SERVER_EXEC_SHELL_COMMAND },
+					BuiltInTool.SERVER_EXEC_SHELL_COMMAND
 				)
 			)?.command
 		).toBe('ls');
@@ -518,8 +531,8 @@ describe('parseExecShellCommandMeta', () => {
 		expect(
 			parseExecShellCommandMeta(
 				makeSection(
-					{ toolArgs: '{"cwd":"/x"}', toolName: BuiltInTool.EXEC_SHELL_COMMAND },
-					BuiltInTool.EXEC_SHELL_COMMAND
+					{ toolArgs: '{"cwd":"/x"}', toolName: BuiltInTool.SERVER_EXEC_SHELL_COMMAND },
+					BuiltInTool.SERVER_EXEC_SHELL_COMMAND
 				)
 			)
 		).toBeNull();

vendor/CMakeLists.txt

@@ -0,0 +1,11 @@
+# mtmd needs these even when common is not built
+add_subdirectory(hash)
+add_subdirectory(miniaudio)
+add_subdirectory(nlohmann)
+add_subdirectory(sheredom)
+add_subdirectory(stb)
+
+# only used by common
+if (LLAMA_BUILD_COMMON)
+    add_subdirectory(cpp-httplib)
+endif()

vendor/cpp-httplib/CMakeLists.txt

@@ -9,6 +9,8 @@ set(CMAKE_POSITION_INDEPENDENT_CODE ON)
 
 add_library(${TARGET} STATIC httplib.cpp httplib.h)
 
+add_library(vendor::cpp-httplib ALIAS ${TARGET})
+
 # disable warnings in 3rd party code
 if (CMAKE_CXX_COMPILER_ID STREQUAL "MSVC")
     target_compile_options(${TARGET} PRIVATE /w)

vendor/hash/CMakeLists.txt

@@ -16,6 +16,8 @@ add_library(${TARGET} STATIC
     ${VENDOR_SRCS}
 )
 
+add_library(vendor::hash ALIAS ${TARGET})
+
 target_compile_features(${TARGET} PRIVATE cxx_std_17)
 
 # disable warnings in 3rd party code, but keep them for hash.cpp
@@ -29,5 +31,8 @@ set_source_files_properties(${VENDOR_SRCS} PROPERTIES COMPILE_OPTIONS ${NO_WARN_
 # sha1 lives in a namespace to avoid a clash with boringssl, see scripts/sync_vendor.py
 set_source_files_properties(sha1/sha1.c PROPERTIES LANGUAGE CXX)
 
-# sha256.c includes "rotate-bits/rotate-bits.h", so consumers get this dir too
-target_include_directories(${TARGET} PUBLIC .)
+# expose the vendor/ root so consumers can include via "hash/hash.h"
+target_include_directories(${TARGET} PUBLIC ..)
+
+# internal includes of the vendored sources, e.g. sha256.c -> "rotate-bits/rotate-bits.h"
+target_include_directories(${TARGET} PRIVATE .)

vendor/miniaudio/CMakeLists.txt

@@ -0,0 +1,6 @@
+# header-only: interface target exposing the vendor/ root so consumers
+# can include via <miniaudio/miniaudio.h>
+add_library(miniaudio INTERFACE)
+add_library(vendor::miniaudio ALIAS miniaudio)
+
+target_include_directories(miniaudio INTERFACE ..)

vendor/nlohmann/CMakeLists.txt

@@ -0,0 +1,6 @@
+# header-only: interface target exposing the vendor/ root so consumers
+# can include via <nlohmann/json.hpp>
+add_library(nlohmann INTERFACE)
+add_library(vendor::nlohmann ALIAS nlohmann)
+
+target_include_directories(nlohmann INTERFACE ..)