NetBSD build support (#19589)

* Refactoring to use new llama_put_adapter_loras

* cont : alternative lora API

---------

Co-authored-by: Jake Chavis <jakechavis6@gmail.com>
Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

.github/ISSUE_TEMPLATE/010-bug-compilation.yml

@@ -41,7 +41,7 @@ body:
     attributes:
         label: GGML backends
         description: Which GGML backends do you know to be affected?
-        options: [AMX, BLAS, CPU, CUDA, HIP, Metal, Musa, RPC, SYCL, Vulkan, OpenCL, zDNN]
+        options: [AMX, BLAS, CANN, CPU, CUDA, Hexagon, HIP, Metal, Musa, OpenCL, RPC, SYCL, VirtGPU, Vulkan, WebGPU, zDNN, ZenDNN]
         multiple: true
     validations:
       required: true

.github/ISSUE_TEMPLATE/011-bug-results.yml

@@ -42,7 +42,7 @@ body:
     attributes:
         label: GGML backends
         description: Which GGML backends do you know to be affected?
-        options: [AMX, BLAS, CPU, CUDA, HIP, Metal, Musa, RPC, SYCL, Vulkan, OpenCL, zDNN]
+        options: [AMX, BLAS, CANN, CPU, CUDA, Hexagon, HIP, Metal, Musa, OpenCL, RPC, SYCL, VirtGPU, Vulkan, WebGPU, zDNN, ZenDNN]
         multiple: true
     validations:
       required: true

build-xcframework.sh

@@ -43,11 +43,6 @@ COMMON_CMAKE_ARGS=(
     -DGGML_OPENMP=${GGML_OPENMP}
 )
 
-XCODE_VERSION=$(xcodebuild -version 2>/dev/null | head -n1 | awk '{ print $2 }')
-MAJOR_VERSION=$(echo $XCODE_VERSION | cut -d. -f1)
-MINOR_VERSION=$(echo $XCODE_VERSION | cut -d. -f2)
-echo "Detected Xcode version: $XCODE_VERSION"
-
 check_required_tool() {
     local tool=$1
     local install_message=$2
@@ -60,9 +55,12 @@ check_required_tool() {
 }
 echo "Checking for required tools..."
 check_required_tool "cmake" "Please install CMake 3.28.0 or later (brew install cmake)"
-check_required_tool "xcodebuild" "Please install Xcode and Xcode Command Line Tools (xcode-select --install)"
-check_required_tool "libtool" "Please install libtool which should be available with Xcode Command Line Tools (CLT). Make sure Xcode CLT is installed (xcode-select --install)"
-check_required_tool "dsymutil" "Please install Xcode and Xcode Command Line Tools (xcode-select --install)"
+check_required_tool "xcrun" "Please install Xcode and Xcode Command Line Tools (xcode-select --install)"
+
+XCODE_VERSION=$(xcrun xcodebuild -version 2>/dev/null | head -n1 | awk '{ print $2 }')
+MAJOR_VERSION=$(echo $XCODE_VERSION | cut -d. -f1)
+MINOR_VERSION=$(echo $XCODE_VERSION | cut -d. -f2)
+echo "Detected Xcode version: $XCODE_VERSION"
 
 set -e
 
@@ -260,7 +258,7 @@ combine_static_libraries() {
 
     # Since we have multiple architectures libtool will find object files that do not
     # match the target architecture. We suppress these warnings.
-    libtool -static -o "${temp_dir}/combined.a" "${libs[@]}" 2> /dev/null
+    xcrun libtool -static -o "${temp_dir}/combined.a" "${libs[@]}" 2> /dev/null
 
     # Determine SDK, architectures, and install_name based on platform and simulator flag.
     local sdk=""
@@ -333,7 +331,7 @@ combine_static_libraries() {
 
     # Platform-specific post-processing for device builds
     if [[ "$is_simulator" == "false" ]]; then
-        if command -v xcrun vtool &>/dev/null; then
+        if xcrun -f vtool &>/dev/null; then
             case "$platform" in
                 "ios")
                     echo "Marking binary as a framework binary for iOS..."
@@ -528,7 +526,7 @@ combine_static_libraries "build-tvos-device" "Release-appletvos" "tvos" "false"
 
 # Create XCFramework with correct debug symbols paths
 echo "Creating XCFramework..."
-xcodebuild -create-xcframework \
+xcrun xcodebuild -create-xcframework \
     -framework $(pwd)/build-ios-sim/framework/llama.framework \
     -debug-symbols $(pwd)/build-ios-sim/dSYMs/llama.dSYM \
     -framework $(pwd)/build-ios-device/framework/llama.framework \

common/common.cpp

@@ -879,7 +879,8 @@ std::string fs_get_cache_directory() {
     if (getenv("LLAMA_CACHE")) {
         cache_directory = std::getenv("LLAMA_CACHE");
     } else {
-#if defined(__linux__) || defined(__FreeBSD__) || defined(_AIX) || defined(__OpenBSD__)
+#if defined(__linux__) || defined(__FreeBSD__) || defined(_AIX) || \
+        defined(__OpenBSD__) || defined(__NetBSD__)
         if (std::getenv("XDG_CACHE_HOME")) {
             cache_directory = std::getenv("XDG_CACHE_HOME");
         } else if (std::getenv("HOME")) {
@@ -1223,7 +1224,7 @@ common_init_result_ptr common_init_from_params(common_params & params) {
             return res;
         }
 
-        int err = llama_apply_adapter_cvec(
+        int err = llama_set_adapter_cvec(
                 lctx,
                 cvec.data.data(),
                 cvec.data.size(),
@@ -1325,12 +1326,15 @@ std::string get_model_endpoint() {
 }
 
 void common_set_adapter_lora(struct llama_context * ctx, std::vector<common_adapter_lora_info> & lora) {
-    llama_clear_adapter_lora(ctx);
-    for (auto & la : lora) {
-        if (la.scale != 0.0f) {
-            llama_set_adapter_lora(ctx, la.ptr, la.scale);
-        }
+    std::vector<llama_adapter_lora *> loras;
+    std::vector<float> scales;
+
+    for (auto & la: lora) {
+        loras.push_back(la.ptr);
+        scales.push_back(la.scale);
     }
+
+    llama_set_adapters_lora(ctx, loras.data(), loras.size(), scales.data());
 }
 
 struct llama_model_params common_model_params_to_llama(common_params & params) {

common/download.cpp

@@ -114,44 +114,18 @@ static void write_etag(const std::string & path, const std::string & etag) {
 }
 
 static std::string read_etag(const std::string & path) {
-    std::string none;
     const std::string etag_path = path + ".etag";
-
-    if (std::filesystem::exists(etag_path)) {
-        std::ifstream etag_in(etag_path);
-        if (!etag_in) {
-            LOG_ERR("%s: could not open .etag file for reading: %s\n", __func__, etag_path.c_str());
-            return none;
-        }
-        std::string etag;
-        std::getline(etag_in, etag);
-        return etag;
+    if (!std::filesystem::exists(etag_path)) {
+        return {};
     }
-
-    // no etag file, but maybe there is an old .json
-    // remove this code later
-    const std::string metadata_path = path + ".json";
-
-    if (std::filesystem::exists(metadata_path)) {
-        std::ifstream metadata_in(metadata_path);
-        try {
-            nlohmann::json metadata_json;
-            metadata_in >> metadata_json;
-            LOG_DBG("%s: previous metadata file found %s: %s\n", __func__, metadata_path.c_str(),
-                    metadata_json.dump().c_str());
-            if (metadata_json.contains("etag") && metadata_json.at("etag").is_string()) {
-                std::string etag = metadata_json.at("etag");
-                write_etag(path, etag);
-                if (!std::filesystem::remove(metadata_path)) {
-                    LOG_WRN("%s: failed to delete old .json metadata file: %s\n", __func__, metadata_path.c_str());
-                }
-                return etag;
-            }
-        } catch (const nlohmann::json::exception & e) {
-            LOG_ERR("%s: error reading metadata file %s: %s\n", __func__, metadata_path.c_str(), e.what());
-        }
+    std::ifstream etag_in(etag_path);
+    if (!etag_in) {
+        LOG_ERR("%s: could not open .etag file for reading: %s\n", __func__, etag_path.c_str());
+        return {};
     }
-    return none;
+    std::string etag;
+    std::getline(etag_in, etag);
+    return etag;
 }
 
 static bool is_http_status_ok(int status) {
@@ -347,62 +321,64 @@ static int common_download_file_single_online(const std::string        & url,
         LOG_INF("%s: no previous model file found %s\n", __func__, path.c_str());
     }
 
-    for (int i = 0; i < max_attempts; ++i) {
-        auto head = cli.Head(parts.path);
-        bool head_ok = head && head->status >= 200 && head->status < 300;
-        if (!head_ok) {
-            LOG_WRN("%s: HEAD invalid http status code received: %d\n", __func__, head ? head->status : -1);
-            if (file_exists) {
-                LOG_INF("%s: Using cached file (HEAD failed): %s\n", __func__, path.c_str());
-                return 304; // 304 Not Modified - fake cached response
-            }
-            return head->status; // cannot use cached file, return raw status code
-            // TODO: maybe retry only on certain codes
-        }
-
-        std::string etag;
-        if (head_ok && head->has_header("ETag")) {
-            etag = head->get_header_value("ETag");
-        }
-
-        size_t total_size = 0;
-        if (head_ok && head->has_header("Content-Length")) {
-            try {
-                total_size = std::stoull(head->get_header_value("Content-Length"));
-            } catch (const std::exception& e) {
-                LOG_WRN("%s: Invalid Content-Length in HEAD response: %s\n", __func__, e.what());
-            }
-        }
-
-        bool supports_ranges = false;
-        if (head_ok && head->has_header("Accept-Ranges")) {
-            supports_ranges = head->get_header_value("Accept-Ranges") != "none";
-        }
-
-        bool should_download_from_scratch = false;
-        if (!last_etag.empty() && !etag.empty() && last_etag != etag) {
-            LOG_WRN("%s: ETag header is different (%s != %s): triggering a new download\n", __func__,
-                    last_etag.c_str(), etag.c_str());
-            should_download_from_scratch = true;
-        }
-
+    auto head = cli.Head(parts.path);
+    if (!head || head->status < 200 || head->status >= 300) {
+        LOG_WRN("%s: HEAD failed, status: %d\n", __func__, head ? head->status : -1);
         if (file_exists) {
-            if (!should_download_from_scratch) {
-                LOG_INF("%s: using cached file: %s\n", __func__, path.c_str());
-                return 304; // 304 Not Modified - fake cached response
-            }
-            LOG_WRN("%s: deleting previous downloaded file: %s\n", __func__, path.c_str());
-            if (remove(path.c_str()) != 0) {
-                LOG_ERR("%s: unable to delete file: %s\n", __func__, path.c_str());
-                return -1;
-            }
+            LOG_INF("%s: using cached file (HEAD failed): %s\n", __func__, path.c_str());
+            return 304; // 304 Not Modified - fake cached response
+        }
+        return head ? head->status : -1;
+    }
+
+    std::string etag;
+    if (head->has_header("ETag")) {
+        etag = head->get_header_value("ETag");
+    }
+
+    size_t total_size = 0;
+    if (head->has_header("Content-Length")) {
+        try {
+            total_size = std::stoull(head->get_header_value("Content-Length"));
+        } catch (const std::exception& e) {
+            LOG_WRN("%s: invalid Content-Length in HEAD response: %s\n", __func__, e.what());
+        }
+    }
+
+    bool supports_ranges = false;
+    if (head->has_header("Accept-Ranges")) {
+        supports_ranges = head->get_header_value("Accept-Ranges") != "none";
+    }
+
+    if (file_exists) {
+        if (etag.empty()) {
+            LOG_INF("%s: using cached file (no server etag): %s\n", __func__, path.c_str());
+            return 304; // 304 Not Modified - fake cached response
+        }
+        if (!last_etag.empty() && last_etag == etag) {
+            LOG_INF("%s: using cached file (same etag): %s\n", __func__, path.c_str());
+            return 304; // 304 Not Modified - fake cached response
+        }
+        if (remove(path.c_str()) != 0) {
+            LOG_ERR("%s: unable to delete file: %s\n", __func__, path.c_str());
+            return -1;
+        }
+    }
+
+    const std::string path_temporary = path + ".downloadInProgress";
+    int delay = retry_delay_seconds;
+
+    for (int i = 0; i < max_attempts; ++i) {
+        if (i) {
+            LOG_WRN("%s: retrying after %d seconds...\n", __func__, delay);
+            std::this_thread::sleep_for(std::chrono::seconds(delay));
+            delay *= retry_delay_seconds;
         }
 
-        const std::string path_temporary = path + ".downloadInProgress";
         size_t existing_size = 0;
 
         if (std::filesystem::exists(path_temporary)) {
-            if (supports_ranges && !should_download_from_scratch) {
+            if (supports_ranges) {
                 existing_size = std::filesystem::file_size(path_temporary);
             } else if (remove(path_temporary.c_str()) != 0) {
                 LOG_ERR("%s: unable to delete file: %s\n", __func__, path_temporary.c_str());
@@ -410,32 +386,23 @@ static int common_download_file_single_online(const std::string        & url,
             }
         }
 
-        // start the download
-        LOG_INF("%s: trying to download model from %s to %s (etag:%s)...\n",
-                __func__, common_http_show_masked_url(parts).c_str(), path_temporary.c_str(), etag.c_str());
-        const bool was_pull_successful = common_pull_file(cli, parts.path, path_temporary, supports_ranges, existing_size, total_size);
-        if (!was_pull_successful) {
-            if (i + 1 < max_attempts) {
-                const int exponential_backoff_delay = std::pow(retry_delay_seconds, i) * 1000;
-                LOG_WRN("%s: retrying after %d milliseconds...\n", __func__, exponential_backoff_delay);
-                std::this_thread::sleep_for(std::chrono::milliseconds(exponential_backoff_delay));
-            } else {
-                LOG_ERR("%s: download failed after %d attempts\n", __func__, max_attempts);
+        LOG_INF("%s: downloading from %s to %s (etag:%s)...\n",
+                __func__, common_http_show_masked_url(parts).c_str(),
+                path_temporary.c_str(), etag.c_str());
+
+        if (common_pull_file(cli, parts.path, path_temporary, supports_ranges, existing_size, total_size)) {
+            if (std::rename(path_temporary.c_str(), path.c_str()) != 0) {
+                LOG_ERR("%s: unable to rename file: %s to %s\n", __func__, path_temporary.c_str(), path.c_str());
+                return -1;
             }
-            continue;
+            if (!etag.empty()) {
+                write_etag(path, etag);
+            }
+            return head->status;
         }
-
-        if (std::rename(path_temporary.c_str(), path.c_str()) != 0) {
-            LOG_ERR("%s: unable to rename file: %s to %s\n", __func__, path_temporary.c_str(), path.c_str());
-            return -1;
-        }
-        if (!etag.empty()) {
-            write_etag(path, etag);
-        }
-
-        return head->status; // TODO: use actual GET status?
     }
 
+    LOG_ERR("%s: download failed after %d attempts\n", __func__, max_attempts);
     return -1; // max attempts reached
 }
 

convert_hf_to_gguf.py

@@ -570,6 +570,7 @@ class ModelBase:
                         self.match_model_tensor_name(new_name, key, bid)
                         for key in (
                             gguf.MODEL_TENSOR.FFN_GATE_INP,
+                            gguf.MODEL_TENSOR.FFN_GATE_INP_SHEXP,
                             gguf.MODEL_TENSOR.POS_EMBD,
                             gguf.MODEL_TENSOR.TOKEN_TYPES,
                             gguf.MODEL_TENSOR.SSM_CONV1D,

ggml/src/ggml-hexagon/htp/flash-attn-ops.c

@@ -17,121 +17,6 @@
 #include "htp-msg.h"
 #include "htp-ops.h"
 
-static inline HVX_Vector hvx_load_f32_to_f16(const HVX_Vector * restrict src, const HVX_Vector zero) {
-    HVX_Vector y0_qf = Q6_Vqf32_vsub_VsfVsf(src[0], zero);  // 32 elements
-    HVX_Vector y1_qf = Q6_Vqf32_vsub_VsfVsf(src[1], zero);  // 32 elements
-    return Q6_Vh_vdeal_Vh(Q6_Vhf_equals_Wqf32(Q6_W_vcombine_VV(y1_qf, y0_qf)));
-}
-
-// Dot product of FP32 and FP16 vectors, accumulating to float
-static inline void hvx_dot_f32_f16_aa(float * restrict r, const void * restrict y, const void * restrict x, unsigned int n, float s) {
-    const HVX_Vector * restrict vy = (const HVX_Vector * restrict) y; // fp32
-    const HVX_Vector * restrict vx = (const HVX_Vector * restrict) x; // fp16
-
-    uint32_t nvec = n / VLEN_FP16; // num full fp16 hvx vectors
-    uint32_t nloe = n % VLEN_FP16; // leftover elements
-
-    const HVX_Vector zero = Q6_V_vsplat_R(0);
-    HVX_Vector       rsum = Q6_V_vsplat_R(0);
-
-    uint32_t i = 0;
-
-    #pragma unroll(4)
-    for (i = 0; i < nvec; i++) {
-        // Load y (fp32) and convert into fp16
-        HVX_Vector y_hf  = hvx_load_f32_to_f16(&vy[i*2], zero);
-
-        // Load x (fp16)
-        HVX_Vector x_hf  = vx[i];
-
-        HVX_VectorPair xy_qf = Q6_Wqf32_vmpy_VhfVhf(x_hf, y_hf);
-
-        rsum = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_lo_W(xy_qf), Q6_V_hi_W(xy_qf)), rsum));
-    }
-
-    if (nloe) {
-        // Load y (fp32) and convert into fp16
-        HVX_Vector y_hf  = hvx_load_f32_to_f16(&vy[i*2], zero);
-
-        // Load x (fp16)
-        HVX_Vector x_hf  = vx[i];
-
-        // Zero-out unused elements
-        // Note that we need to clear both x and y because they may contain NANs
-        HVX_VectorPred bmask = Q6_Q_vsetq_R(nloe * 2);
-        x_hf = Q6_V_vand_QV(bmask, x_hf);
-        y_hf = Q6_V_vand_QV(bmask, y_hf);
-
-        HVX_VectorPair xy_qf = Q6_Wqf32_vmpy_VhfVhf(x_hf, y_hf);
-
-        rsum = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_lo_W(xy_qf), Q6_V_hi_W(xy_qf)), rsum));
-    }
-
-    rsum = Q6_Vqf32_vmpy_VsfVsf(hvx_vec_splat_f32(s), hvx_vec_reduce_sum_f32(rsum));
-    hvx_vec_store_u(r, 4, Q6_Vsf_equals_Vqf32(rsum));
-}
-
-// Dot product of FP32 and FP16 vectors, accumulating to float
-static inline void hvx_dot_f32_f16_aa_rx2(float * restrict r,
-                                          const void * restrict y,
-                                          const void * restrict x0,
-                                          const void * restrict x1,
-                                          unsigned int n,
-                                          float        s) {
-    const HVX_Vector * restrict vy  = (const HVX_Vector * restrict) y;   // fp32
-    const HVX_Vector * restrict vx0 = (const HVX_Vector * restrict) x0;  // fp16
-    const HVX_Vector * restrict vx1 = (const HVX_Vector * restrict) x1;  // fp16
-
-    uint32_t nvec = n / VLEN_FP16;                                       // num full fp16 hvx vectors
-    uint32_t nloe = n % VLEN_FP16;                                       // leftover elements
-
-    const HVX_Vector zero  = Q6_V_vsplat_R(0);
-    HVX_Vector       rsum0 = Q6_V_vsplat_R(0);
-    HVX_Vector       rsum1 = Q6_V_vsplat_R(0);
-
-    uint32_t i = 0;
-
-    #pragma unroll(2)
-    for (i = 0; i < nvec; i++) {
-        // Load y (fp32) and convert into fp16
-        HVX_Vector y_hf  = hvx_load_f32_to_f16(&vy[i*2], zero);
-        // Load x (fp16)
-        HVX_Vector x0_hf = vx0[i];
-        HVX_Vector x1_hf = vx1[i];
-
-        HVX_VectorPair xy0_qf = Q6_Wqf32_vmpy_VhfVhf(x0_hf, y_hf);
-        HVX_VectorPair xy1_qf = Q6_Wqf32_vmpy_VhfVhf(x1_hf, y_hf);
-
-        rsum0 = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_lo_W(xy0_qf), Q6_V_hi_W(xy0_qf)), rsum0));
-        rsum1 = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_lo_W(xy1_qf), Q6_V_hi_W(xy1_qf)), rsum1));
-    }
-
-    if (nloe) {
-        // Load y (fp32) and convert into fp16
-        HVX_Vector y_hf  = hvx_load_f32_to_f16(&vy[i*2], zero);
-
-        // Load x (fp16)
-        HVX_Vector x0_hf = vx0[i];
-        HVX_Vector x1_hf = vx1[i];
-
-        // Zero-out unused elements
-        // Note that we need to clear both x and y because they may contain NANs
-        HVX_VectorPred bmask = Q6_Q_vsetq_R(nloe * 2);
-        x0_hf                = Q6_V_vand_QV(bmask, x0_hf);
-        x1_hf                = Q6_V_vand_QV(bmask, x1_hf);
-        y_hf                 = Q6_V_vand_QV(bmask, y_hf);
-
-        HVX_VectorPair xy0_qf = Q6_Wqf32_vmpy_VhfVhf(x0_hf, y_hf);
-        HVX_VectorPair xy1_qf = Q6_Wqf32_vmpy_VhfVhf(x1_hf, y_hf);
-
-        rsum0 = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_lo_W(xy0_qf), Q6_V_hi_W(xy0_qf)), rsum0));
-        rsum1 = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_lo_W(xy1_qf), Q6_V_hi_W(xy1_qf)), rsum1));
-    }
-
-    HVX_Vector rsum = Q6_Vqf32_vmpy_VsfVsf(hvx_vec_splat_f32(s), hvx_vec_reduce_sum_f32x2(rsum0, rsum1));
-    hvx_vec_store_u(r, 8, Q6_Vsf_equals_Vqf32(rsum));
-}
-
 // Dot product of two F16 vectors, accumulating to float
 static inline void hvx_dot_f16_f16_aa(float * restrict r, const void * restrict x, const void * restrict y, unsigned int n, float s) {
     const HVX_Vector * restrict vx = (const HVX_Vector * restrict) x; // fp16
@@ -140,8 +25,7 @@ static inline void hvx_dot_f16_f16_aa(float * restrict r, const void * restrict
     uint32_t nvec = n / VLEN_FP16; // num full fp16 hvx vectors
     uint32_t nloe = n % VLEN_FP16; // leftover elements
 
-    const HVX_Vector zero = Q6_V_vsplat_R(0);
-    HVX_Vector       rsum = Q6_V_vsplat_R(0);
+    HVX_Vector rsum = Q6_V_vsplat_R(0);
 
     uint32_t i = 0;
 
@@ -156,11 +40,10 @@ static inline void hvx_dot_f16_f16_aa(float * restrict r, const void * restrict
     }
 
     if (nloe) {
-        HVX_Vector y_hf = vy[i];
-
         // Load x (fp16) and zero-out unused elements
         HVX_VectorPred bmask = Q6_Q_vsetq_R(nloe * 2);
-        HVX_Vector      x_hf = Q6_V_vand_QV(bmask, vx[i]);
+        HVX_Vector y_hf = Q6_V_vand_QV(bmask, vy[i]);
+        HVX_Vector x_hf = Q6_V_vand_QV(bmask, vx[i]);
 
         HVX_VectorPair xy_qf = Q6_Wqf32_vmpy_VhfVhf(x_hf, y_hf);
 
@@ -181,12 +64,11 @@ static inline void hvx_dot_f16_f16_aa_rx2(float * restrict r,
     const HVX_Vector * restrict vx1 = (const HVX_Vector * restrict) x1;  // fp16
     const HVX_Vector * restrict vy  = (const HVX_Vector * restrict) y;   // fp16
 
-    uint32_t nvec = n / VLEN_FP16;                                       // num full fp16 hvx vectors
-    uint32_t nloe = n % VLEN_FP16;                                       // leftover elements
+    uint32_t nvec = n / VLEN_FP16;  // num full fp16 hvx vectors
+    uint32_t nloe = n % VLEN_FP16;  // leftover elements
 
-    const HVX_Vector zero  = Q6_V_vsplat_R(0);
-    HVX_Vector       rsum0 = Q6_V_vsplat_R(0);
-    HVX_Vector       rsum1 = Q6_V_vsplat_R(0);
+    HVX_Vector rsum0 = Q6_V_vsplat_R(0);
+    HVX_Vector rsum1 = Q6_V_vsplat_R(0);
 
     uint32_t i = 0;
 
@@ -204,12 +86,11 @@ static inline void hvx_dot_f16_f16_aa_rx2(float * restrict r,
     }
 
     if (nloe) {
-        HVX_Vector y_hf = vy[i];
-
         // Load x (fp16) and zero-out unused elements
         HVX_VectorPred bmask = Q6_Q_vsetq_R(nloe * 2);
-        HVX_Vector     x0_hf = Q6_V_vand_QV(bmask, vx0[i]);
-        HVX_Vector     x1_hf = Q6_V_vand_QV(bmask, vx1[i]);
+        HVX_Vector x0_hf = Q6_V_vand_QV(bmask, vx0[i]);
+        HVX_Vector x1_hf = Q6_V_vand_QV(bmask, vx1[i]);
+        HVX_Vector y_hf  = Q6_V_vand_QV(bmask, vy[i]);
 
         HVX_VectorPair xy0_qf = Q6_Wqf32_vmpy_VhfVhf(x0_hf, y_hf);
         HVX_VectorPair xy1_qf = Q6_Wqf32_vmpy_VhfVhf(x1_hf, y_hf);
@@ -222,7 +103,7 @@ static inline void hvx_dot_f16_f16_aa_rx2(float * restrict r,
     hvx_vec_store_u(r, 8, Q6_Vsf_equals_Vqf32(rsum));
 }
 
-// MAD: y (F32) += x (F16) * s (float)
+// MAD: y (F32) += x (F16) * s (F32)
 static inline void hvx_mad_f32_f16_aa(float * restrict y, const void * restrict x, int n, float s) {
     const HVX_Vector * restrict ptr_x = (const HVX_Vector *) x;
     HVX_Vector * restrict ptr_y = (HVX_Vector *) y;
@@ -259,15 +140,125 @@ static inline void hvx_mad_f32_f16_aa(float * restrict y, const void * restrict
     }
 }
 
+// MAD: y (F32) += x0 (F16) * s0 (F32) + x1 (F16) * s1 (F32)
+static inline void hvx_mad_f32_f16_aa_rx2(float * restrict y,
+                                          const void * restrict x0,
+                                          const void * restrict x1,
+                                          float s0,
+                                          float s1,
+                                          int   n) {
+    const HVX_Vector * restrict ptr_x0 = (const HVX_Vector *) x0;
+    const HVX_Vector * restrict ptr_x1 = (const HVX_Vector *) x1;
+    HVX_Vector * restrict ptr_y        = (HVX_Vector *) y;
+
+    uint32_t nvec = n / VLEN_FP16;  // num full fp16 hvx vectors
+    uint32_t nloe = n % VLEN_FP16;  // leftover elements
+
+    HVX_Vector S0 = hvx_vec_splat_f16(s0);
+    HVX_Vector S1 = hvx_vec_splat_f16(s1);
+
+    uint32_t i = 0;
+    #pragma unroll(2)
+    for (i = 0; i < nvec; ++i) {
+        // Multiply x * s -> pair of F32 vectors
+        HVX_VectorPair xs0_p = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(ptr_x0[i]), S0);
+        HVX_VectorPair xs1_p = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(ptr_x1[i]), S1);
+
+        HVX_Vector xs_p_lo = Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_lo_W(xs0_p), Q6_V_lo_W(xs1_p));
+        HVX_Vector xs_p_hi = Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_hi_W(xs0_p), Q6_V_hi_W(xs1_p));
+
+        ptr_y[i * 2]     = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(xs_p_lo, ptr_y[i * 2]));
+        ptr_y[i * 2 + 1] = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(xs_p_hi, ptr_y[i * 2 + 1]));
+    }
+
+    if (nloe) {
+        HVX_VectorPair xs0_p = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(ptr_x0[i]), S0);
+        HVX_VectorPair xs1_p = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(ptr_x1[i]), S1);
+
+        HVX_Vector xs_p_lo = Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_lo_W(xs0_p), Q6_V_lo_W(xs1_p));
+        HVX_Vector xs      = xs_p_lo;
+        i = 2 * i;  // index for ptr_y
+
+        if (nloe >= 32) {
+            ptr_y[i] = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(xs, ptr_y[i]));
+            nloe -= 32; ++i;
+            xs = Q6_Vqf32_vadd_Vqf32Vqf32(Q6_V_hi_W(xs0_p), Q6_V_hi_W(xs1_p));
+        }
+
+        if (nloe) {
+            HVX_Vector xy = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_Vqf32Vsf(xs, ptr_y[i]));
+            hvx_vec_store_a(&ptr_y[i], nloe * 4, xy);
+        }
+    }
+}
+
 #define FLASH_ATTN_BLOCK_SIZE 128
 
-static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, int nth) {
+struct htp_fa_context {
+    const struct htp_ops_context * octx;
+
+    struct fastdiv_values src0_div21;
+    struct fastdiv_values src0_div1;
+
+    struct fastdiv_values broadcast_rk2;
+    struct fastdiv_values broadcast_rk3;
+    struct fastdiv_values broadcast_rv2;
+    struct fastdiv_values broadcast_rv3;
+
+    struct fastdiv_values src3_div2;
+    struct fastdiv_values src3_div3;
+
+    float scale;
+    float max_bias;
+    float logit_softcap;
+
+    uint32_t n_head_log2;
+    float m0;
+    float m1;
+
+    uint32_t n_blocks;
+
+    size_t size_q_row_padded;
+    size_t size_k_row_padded;
+    size_t size_v_row_padded;
+
+    size_t size_k_block;
+    size_t size_v_block;
+    size_t size_m_block;
+
+    bool is_q_fp32;
+};
+
+static inline void hvx_scale_vec_f32_aa(uint8_t * restrict dst, const uint8_t * restrict src, const int n, HVX_Vector vs) {
+    assert((size_t) dst % 128 == 0);
+    assert((size_t) src % 128 == 0);
+
+    const HVX_Vector * restrict vsrc = (const HVX_Vector * restrict) src;
+    HVX_Vector * restrict vdst       = (HVX_Vector * restrict) dst;
+
+    const uint32_t nvec = n / VLEN_FP32;
+    const uint32_t nloe = n % VLEN_FP32;
+
+    uint32_t i = 0;
+    #pragma unroll(4)
+    for (; i < nvec; ++i) {
+        vdst[i] = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vmpy_VsfVsf(vsrc[i], vs));
+    }
+    if (nloe) {
+        HVX_Vector v = Q6_Vqf32_vmpy_VsfVsf(vsrc[i], vs);
+        hvx_vec_store_a(&vdst[i], nloe * sizeof(float), Q6_Vsf_equals_Vqf32(v));
+    }
+}
+
+static void flash_attn_ext_f16_thread(unsigned int nth, unsigned int ith, void * data) {
+    struct htp_fa_context * factx = (struct htp_fa_context *) data;
+    const struct htp_ops_context * octx = factx->octx;
     const struct htp_tensor * q = &octx->src0;
     const struct htp_tensor * k = &octx->src1;
     const struct htp_tensor * v = &octx->src2;
     const struct htp_tensor * mask  = (octx->src3.data) ? &octx->src3 : NULL;
     const struct htp_tensor * sinks = (octx->src4.data) ? &octx->src4 : NULL;
-    struct htp_tensor * dst = &octx->dst;
+    const struct htp_tensor * dst = &octx->dst;
 
     const uint32_t neq0 = q->ne[0];
     const uint32_t neq1 = q->ne[1];
@@ -304,18 +295,6 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
     const uint32_t nb2 = dst->nb[2];
     const uint32_t nb3 = dst->nb[3];
 
-    float scale         = 1.0f;
-    float max_bias      = 0.0f;
-    float logit_softcap = 0.0f;
-
-    memcpy(&scale,         (float *) octx->op_params + 0, sizeof(float));
-    memcpy(&max_bias,      (float *) octx->op_params + 1, sizeof(float));
-    memcpy(&logit_softcap, (float *) octx->op_params + 2, sizeof(float));
-
-    if (logit_softcap != 0) {
-        scale /= logit_softcap;
-    }
-
     // total rows in q
     const uint32_t nr = neq1*neq2*neq3;
 
@@ -331,18 +310,8 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
     const uint32_t DV = nev0;
 
     const size_t size_q_row = DK * ((q->type == HTP_TYPE_F32) ? 4 : 2);
-    const size_t size_q_row_padded = hex_round_up(size_q_row, 128);
-
     const size_t size_k_row = DK * sizeof(__fp16);
     const size_t size_v_row = DV * sizeof(__fp16);
-    const size_t size_m_row = FLASH_ATTN_BLOCK_SIZE * sizeof(__fp16); // Treat block as one row for mask
-
-    const size_t size_k_row_padded = hex_round_up(size_k_row, 128);
-    const size_t size_v_row_padded = hex_round_up(size_v_row, 128);
-
-    const size_t size_k_block = size_k_row_padded * FLASH_ATTN_BLOCK_SIZE;
-    const size_t size_v_block = size_v_row_padded * FLASH_ATTN_BLOCK_SIZE;
-    const size_t size_m_block = hex_round_up(FLASH_ATTN_BLOCK_SIZE * sizeof(__fp16), 128);
 
     // Scratchpad buffers for Q, K, V, Mask, and VKQ32 accumulator
     uint8_t * spad_q = octx->src0_spad.data + octx->src0_spad.size_per_thread * ith;
@@ -351,31 +320,28 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
     uint8_t * spad_m = octx->src3_spad.data + octx->src3_spad.size_per_thread * ith;
     uint8_t * spad_a = octx->dst_spad.data  + octx->dst_spad.size_per_thread  * ith;
 
-    const uint32_t n_head = neq2;
-    const uint32_t n_head_log2 = 1u << (uint32_t) floor(log2(n_head));
-    const float m0 = powf(2.0f, -(max_bias       ) / n_head_log2);
-    const float m1 = powf(2.0f, -(max_bias / 2.0f) / n_head_log2);
+    const HVX_Vector logit_cap = hvx_vec_splat_f32(factx->logit_softcap);
 
     for (uint32_t ir = ir0; ir < ir1; ++ir) {
-        const uint32_t iq3 = fastdiv(ir, &octx->src0_div21);
-        const uint32_t iq2 = fastdiv(ir - iq3*neq2*neq1, &octx->src0_div1);
+        const uint32_t iq3 = fastdiv(ir, &factx->src0_div21);
+        const uint32_t iq2 = fastdiv(ir - iq3*neq2*neq1, &factx->src0_div1);
         const uint32_t iq1 = (ir - iq3*neq2*neq1 - iq2 * neq1);
 
-        const uint32_t ik3 = fastdiv(iq3, &octx->broadcast_rk3);
-        const uint32_t ik2 = fastdiv(iq2, &octx->broadcast_rk2);
+        const uint32_t ik3 = fastdiv(iq3, &factx->broadcast_rk3);
+        const uint32_t ik2 = fastdiv(iq2, &factx->broadcast_rk2);
 
-        const uint32_t iv3 = fastdiv(iq3, &octx->broadcast_rv3);
-        const uint32_t iv2 = fastdiv(iq2, &octx->broadcast_rv2);
+        const uint32_t iv3 = fastdiv(iq3, &factx->broadcast_rv3);
+        const uint32_t iv2 = fastdiv(iq2, &factx->broadcast_rv2);
 
         // Fetch Q row
         const uint8_t * q_row_ptr = (const uint8_t *) q->data + (iq1*nbq1 + iq2*nbq2 + iq3*nbq3);
-        dma_queue_push(dma, dma_make_ptr(spad_q, q_row_ptr), size_q_row_padded, nbq1, size_q_row, 1);
+        dma_queue_push(dma, dma_make_ptr(spad_q, q_row_ptr), factx->size_q_row_padded, nbq1, size_q_row, 1);
 
         const uint32_t h = iq2; // head index
-        const float slope = (max_bias > 0.0f) ? (h < n_head_log2 ? powf(m0, h + 1) : powf(m1, 2*(h - n_head_log2) + 1)) : 1.0f;
+        const float slope = (factx->max_bias > 0.0f) ? (h < factx->n_head_log2 ? powf(factx->m0, h + 1) : powf(factx->m1, 2*(h - factx->n_head_log2) + 1)) : 1.0f;
 
-        float S = 0.0f;      // sum
-        float M = -INFINITY; // maximum KQ value
+        HVX_Vector S_vec = hvx_vec_splat_f32(0.0f);
+        HVX_Vector M_vec = hvx_vec_splat_f32(-INFINITY);
 
         // Clear accumulator
         hvx_splat_f32_a(spad_a, 0, DV);
@@ -383,40 +349,42 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
 
         const __fp16 * mp_base = NULL;
         if (mask) {
-            const uint32_t im2 = fastmodulo(iq2, mask->ne[2], &octx->src3_div2);
-            const uint32_t im3 = fastmodulo(iq3, mask->ne[3], &octx->src3_div3);
+            const uint32_t im2 = fastmodulo(iq2, mask->ne[2], &factx->src3_div2);
+            const uint32_t im3 = fastmodulo(iq3, mask->ne[3], &factx->src3_div3);
             mp_base = (const __fp16 *) ((const uint8_t *) mask->data + iq1*mask->nb[1] + im2*mask->nb[2] + im3*mask->nb[3]);
         }
 
-        const uint32_t n_blocks = (nek1 + FLASH_ATTN_BLOCK_SIZE - 1) / FLASH_ATTN_BLOCK_SIZE;
-
         // Prefetch first two blocks
-        for (uint32_t ib = 0; ib < MIN(n_blocks, 2); ++ib) {
+        for (uint32_t ib = 0; ib < MIN(factx->n_blocks, 2); ++ib) {
             const uint32_t ic_start = ib * FLASH_ATTN_BLOCK_SIZE;
             const uint32_t current_block_size = MIN(FLASH_ATTN_BLOCK_SIZE, nek1 - ic_start);
 
             // K
             const uint8_t * k_src = (const uint8_t *) k->data + (ic_start*nbk1 + ik2*nbk2 + ik3*nbk3);
-            uint8_t * k_dst = spad_k + (ib % 2) * size_k_block;
-            dma_queue_push(dma, dma_make_ptr(k_dst, k_src), size_k_row_padded, nbk1, size_k_row, current_block_size);
+            uint8_t * k_dst = spad_k + (ib % 2) * factx->size_k_block;
+            dma_queue_push(dma, dma_make_ptr(k_dst, k_src), factx->size_k_row_padded, nbk1, size_k_row, current_block_size);
 
             // V
             const uint8_t * v_src = (const uint8_t *) v->data + (ic_start*nbv1 + iv2*nbv2 + iv3*nbv3);
-            uint8_t * v_dst = spad_v + (ib % 2) * size_v_block;
-            dma_queue_push(dma, dma_make_ptr(v_dst, v_src), size_v_row_padded, nbv1, size_v_row, current_block_size);
+            uint8_t * v_dst = spad_v + (ib % 2) * factx->size_v_block;
+            dma_queue_push(dma, dma_make_ptr(v_dst, v_src), factx->size_v_row_padded, nbv1, size_v_row, current_block_size);
 
             // Mask
             if (mask) {
                 const uint8_t * m_src = (const uint8_t *) (mp_base + ic_start);
-                uint8_t * m_dst = spad_m + (ib % 2) * size_m_block;
+                uint8_t * m_dst = spad_m + (ib % 2) * factx->size_m_block;
                 // Mask is 1D contiguous for this row
                 dma_queue_push(dma, dma_make_ptr(m_dst, m_src), current_block_size * 2, current_block_size * 2, current_block_size * 2, 1);
             }
         }
 
-        const uint8_t * q_ptr_vtcm = dma_queue_pop(dma).dst;
+        uint8_t * q_ptr_vtcm = dma_queue_pop(dma).dst;
+        if (factx->is_q_fp32) {
+            hvx_copy_f16_f32_aa(q_ptr_vtcm, q_ptr_vtcm, DK);  // inplace convert f32 to f16
+        }
 
-        for (uint32_t ib = 0; ib < n_blocks; ++ib) {
+        const HVX_Vector slope_vec = hvx_vec_splat_f16(slope);
+        for (uint32_t ib = 0; ib < factx->n_blocks; ++ib) {
             const uint32_t ic_start = ib * FLASH_ATTN_BLOCK_SIZE;
             const uint32_t current_block_size = MIN(FLASH_ATTN_BLOCK_SIZE, nek1 - ic_start);
 
@@ -428,8 +396,6 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
             // Inner loop processing the block from VTCM
             uint32_t ic = 0;
 
-            const bool is_q_fp32 = (q->type == HTP_TYPE_F32);
-
             // Process in blocks of 32 (VLEN_FP32)
             static_assert(FLASH_ATTN_BLOCK_SIZE / VLEN_FP32 <= 4, "FLASH_ATTN_BLOCK_SIZE changed, fix HVX_Vector_x4 usage");
             HVX_Vector_x4 scores_x4;
@@ -437,22 +403,18 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
             for (uint32_t iv = 0; ic + VLEN_FP32 <= current_block_size; ic += VLEN_FP32, ++iv) {
                 // 1. Compute scores
                 float __attribute__((aligned(VLEN))) scores_arr[VLEN_FP32];
-                for (int j = 0; j < VLEN_FP32; j += 2) {
+                for (uint32_t j = 0; j < VLEN_FP32; j += 2) {
                     const uint32_t cur_ic = ic + j;
-                    const uint8_t * k_ptr = k_base + cur_ic * size_k_row_padded;
-                    if (is_q_fp32) {
-                        hvx_dot_f32_f16_aa_rx2(&scores_arr[j], q_ptr_vtcm, k_ptr, k_ptr + size_k_row_padded, DK, scale);
-                    } else {
-                        hvx_dot_f16_f16_aa_rx2(&scores_arr[j], q_ptr_vtcm, k_ptr, k_ptr + size_k_row_padded, DK, scale);
-                    }
+                    const uint8_t * k_ptr = k_base + cur_ic * factx->size_k_row_padded;
+                    hvx_dot_f16_f16_aa_rx2(&scores_arr[j], q_ptr_vtcm, k_ptr, k_ptr + factx->size_k_row_padded, DK, factx->scale);
                 }
 
                 HVX_Vector scores = *(HVX_Vector *) scores_arr;
 
                 // 2. Softcap
-                if (logit_softcap != 0.0f) {
+                if (factx->logit_softcap != 0.0f) {
                     scores = hvx_vec_tanh_f32(scores);
-                    scores = Q6_Vqf32_vmpy_VsfVsf(scores, hvx_vec_splat_f32(logit_softcap));
+                    scores = Q6_Vqf32_vmpy_VsfVsf(scores, logit_cap);
                     scores = Q6_Vsf_equals_Vqf32(scores);
                 }
 
@@ -460,70 +422,59 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
                 if (mask) {
                     const __fp16 * mp = m_base + ic;
                     HVX_Vector m_vals_f16 = *(const HVX_UVector *) mp;
-
-                    HVX_Vector one_f16 = Q6_Vh_vsplat_R(0x3c00);
-                    HVX_VectorPair m_vals_f32_pair = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(m_vals_f16), one_f16);
-
-                    HVX_Vector m_vals_f32 = Q6_Vsf_equals_Vqf32(Q6_V_lo_W(m_vals_f32_pair));
-
-                    HVX_Vector slope_vec = hvx_vec_splat_f32(slope);
-                    HVX_Vector add_val = Q6_Vqf32_vmpy_VsfVsf(m_vals_f32, slope_vec);
-                    scores = Q6_Vqf32_vadd_VsfVsf(scores, Q6_Vsf_equals_Vqf32(add_val));
+                    HVX_VectorPair m_vals_f32_pair = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(m_vals_f16), slope_vec);
+                    HVX_Vector add_val = Q6_V_lo_W(m_vals_f32_pair);
+                    scores = Q6_Vqf32_vadd_Vqf32Vsf(add_val, scores);
                     scores = Q6_Vsf_equals_Vqf32(scores);
                 }
 
                 scores_x4.v[iv] = scores;
-                v_max = Q6_Vsf_vmax_VsfVsf(scores, v_max);
+                v_max = hvx_vec_reduce_max2_f32(scores, v_max); // All lanes have block max
             }
 
             {
                 // 4. Online Softmax Update
-                v_max = hvx_vec_reduce_max_f32(v_max);
-                float m_block = hvx_vec_get_f32(v_max);
-                float M_old = M;
-                float M_new = (m_block > M) ? m_block : M;
-                M = M_new;
+                HVX_Vector M_new_vec = Q6_Vsf_vmax_VsfVsf(v_max, M_vec);
+                HVX_Vector diff_vec  = Q6_Vqf32_vsub_VsfVsf(M_vec, M_new_vec);
+                HVX_Vector ms_vec    = hvx_vec_exp_f32(Q6_Vsf_equals_Vqf32(diff_vec));
+                M_vec = M_new_vec;
 
-                const float ms = expf(M_old - M_new);
-                hvx_scale_f32_aa((uint8_t *) VKQ32, (const uint8_t *) VKQ32, DV, ms);
+                hvx_scale_vec_f32_aa((uint8_t *) VKQ32, (const uint8_t *) VKQ32, DV, ms_vec);
 
-                HVX_Vector M_new_vec = hvx_vec_splat_f32(M_new);
                 HVX_Vector p_sum_vec = hvx_vec_splat_f32(0.0f);
                 for (uint32_t ic2 = 0, iv = 0; ic2 + VLEN_FP32 <= current_block_size; ic2 += VLEN_FP32, ++iv) {
                     HVX_Vector scores = scores_x4.v[iv];
-                    HVX_Vector scores_shifted = Q6_Vqf32_vsub_VsfVsf(scores, M_new_vec);
+                    HVX_Vector scores_shifted = Q6_Vqf32_vsub_VsfVsf(scores, M_vec);
                     HVX_Vector P = hvx_vec_exp_f32(Q6_Vsf_equals_Vqf32(scores_shifted));
 
                     p_sum_vec = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_VsfVsf(p_sum_vec, P));
 
                     // 5. Accumulate V
                     float __attribute__((aligned(VLEN))) p_arr[VLEN_FP32];
-                    *(HVX_Vector*)p_arr = P;
+                    *(HVX_Vector *) p_arr = P;
 
-                    for (int j = 0; j < VLEN_FP32; ++j) {
-                        const uint32_t cur_ic = ic2 + j;
-                        const uint8_t * v_ptr = v_base + cur_ic * size_v_row_padded;
-                        hvx_mad_f32_f16_aa(VKQ32, v_ptr, DV, p_arr[j]);
+                    for (uint32_t j = 0; j < VLEN_FP32; j += 2) {
+                        const uint32_t  cur_ic = ic2 + j;
+                        const uint8_t * v_ptr  = v_base + cur_ic * factx->size_v_row_padded;
+                        hvx_mad_f32_f16_aa_rx2(VKQ32, v_ptr, v_ptr + factx->size_v_row_padded, p_arr[j], p_arr[j + 1], DV);
                     }
                 }
 
                 p_sum_vec = hvx_vec_reduce_sum_f32(p_sum_vec);
-                S = S * ms + hvx_vec_get_f32(p_sum_vec);
+                S_vec = Q6_Vsf_equals_Vqf32(Q6_Vqf32_vadd_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_Vqf32_vmpy_VsfVsf(S_vec, ms_vec)), p_sum_vec));
             }
 
+            // Sync scalars for leftover/next block if needed
+            float M = hvx_vec_get_f32(M_vec);
+            float S = hvx_vec_get_f32(S_vec);
+
             // Leftover
             for (; ic < current_block_size; ++ic) {
                 float s_val;
-                const uint8_t * k_ptr = k_base + ic * size_k_row_padded;
-
-                if (is_q_fp32) {
-                    hvx_dot_f32_f16_aa(&s_val, q_ptr_vtcm, k_ptr, DK, scale);
-                } else {
-                    hvx_dot_f16_f16_aa(&s_val, q_ptr_vtcm, k_ptr, DK, scale);
-                }
-
-                if (logit_softcap != 0.0f) {
-                    s_val = logit_softcap * tanhf(s_val);
+                const uint8_t * k_ptr = k_base + ic * factx->size_k_row_padded;
+                hvx_dot_f16_f16_aa(&s_val, q_ptr_vtcm, k_ptr, DK, factx->scale);
+                if (factx->logit_softcap != 0.0f) {
+                    s_val = factx->logit_softcap * tanhf(s_val);
                 }
 
                 if (mask) {
@@ -532,37 +483,42 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
                 }
 
                 const float Mold = M;
-                float ms = 1.0f;
                 float vs = 1.0f;
 
                 if (s_val > M) {
                     M = s_val;
-                    ms = expf(Mold - M);
-                    hvx_scale_f32_aa((uint8_t *) VKQ32, (const uint8_t *) VKQ32, DV, ms);
+                    HVX_Vector diff_vec = hvx_vec_splat_f32(Mold - M);
+                    HVX_Vector ms_vec   = hvx_vec_exp_f32(diff_vec);
+                    hvx_scale_vec_f32_aa((uint8_t *) VKQ32, (const uint8_t *) VKQ32, DV, ms_vec);
+
+                    float ms = hvx_vec_get_f32(ms_vec);
+                    S = S * ms + vs;
                 } else {
-                    vs = expf(s_val - M);
+                    HVX_Vector diff_vec = hvx_vec_splat_f32(s_val - M);
+                    vs = hvx_vec_get_f32(hvx_vec_exp_f32(diff_vec));
+                    S += vs;
                 }
 
-                const uint8_t * v_ptr = v_base + ic * size_v_row_padded;
+                const uint8_t * v_ptr = v_base + ic * factx->size_v_row_padded;
 
                 hvx_mad_f32_f16_aa(VKQ32, v_ptr, DV, vs);
-
-                S = S * ms + vs;
             }
+            M_vec = hvx_vec_splat_f32(M);
+            S_vec = hvx_vec_splat_f32(S);
 
             // Issue DMA for next+1 block (if exists)
-            if (ib + 2 < n_blocks) {
+            if (ib + 2 < factx->n_blocks) {
                 const uint32_t next_ib = ib + 2;
                 const uint32_t next_ic_start = next_ib * FLASH_ATTN_BLOCK_SIZE;
                 const uint32_t next_block_size = MIN(FLASH_ATTN_BLOCK_SIZE, nek1 - next_ic_start);
 
                 // K
                 const uint8_t * k_src = (const uint8_t *) k->data + (next_ic_start*nbk1 + ik2*nbk2 + ik3*nbk3);
-                dma_queue_push(dma, dma_make_ptr(k_base, k_src), size_k_row_padded, nbk1, size_k_row, next_block_size);
+                dma_queue_push(dma, dma_make_ptr(k_base, k_src), factx->size_k_row_padded, nbk1, size_k_row, next_block_size);
 
                 // V
                 const uint8_t * v_src = (const uint8_t *) v->data + (next_ic_start*nbv1 + iv2*nbv2 + iv3*nbv3);
-                dma_queue_push(dma, dma_make_ptr(v_base, v_src), size_v_row_padded, nbv1, size_v_row, next_block_size);
+                dma_queue_push(dma, dma_make_ptr(v_base, v_src), factx->size_v_row_padded, nbv1, size_v_row, next_block_size);
 
                 // Mask
                 if (mask) {
@@ -573,20 +529,26 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
         }
 
         // sinks
+        float M = hvx_vec_get_f32(M_vec);
+        float S = hvx_vec_get_f32(S_vec);
+
         if (sinks) {
             const float s = ((float *)((char *) sinks->data))[h];
 
-            float ms = 1.0f;
             float vs = 1.0f;
 
             if (s > M) {
-                ms = expf(M - s);
-                hvx_scale_f32_aa((uint8_t *) VKQ32, (const uint8_t *) VKQ32, DV, ms);
-            } else {
-                vs = expf(s - M);
-            }
+                HVX_Vector diff_vec = hvx_vec_splat_f32(M - s);
+                HVX_Vector ms_vec   = hvx_vec_exp_f32(diff_vec);
+                hvx_scale_vec_f32_aa((uint8_t *) VKQ32, (const uint8_t *) VKQ32, DV, ms_vec);
 
-            S = S * ms + vs;
+                float ms = hvx_vec_get_f32(ms_vec);
+                S = S * ms + vs;
+            } else {
+                HVX_Vector diff_vec = hvx_vec_splat_f32(s - M);
+                vs = hvx_vec_get_f32(hvx_vec_exp_f32(diff_vec));
+                S += vs;
+            }
         }
 
         const float S_inv = S == 0.0f ? 0.0f : 1.0f/S;
@@ -609,53 +571,73 @@ static void flash_attn_ext_f16_thread(struct htp_ops_context * octx, int ith, in
     }
 }
 
-static void htp_flash_attn_ext_job(unsigned int n, unsigned int i, void * data) {
-    struct htp_ops_context * octx = data;
-    flash_attn_ext_f16_thread(octx, i, n);
-}
-
 int op_flash_attn_ext(struct htp_ops_context * octx) {
     const struct htp_tensor * q = &octx->src0;
     const struct htp_tensor * k = &octx->src1;
     const struct htp_tensor * v = &octx->src2;
-    const struct htp_tensor * mask = (octx->src3.type != HTP_TYPE_COUNT) ? &octx->src3 : NULL;
-    struct htp_tensor * dst = &octx->dst;
+    const struct htp_tensor * mask = (octx->src3.data) ? &octx->src3 : NULL;
+    const struct htp_tensor * dst = &octx->dst;
 
     // Check support
-    if ((q->type != HTP_TYPE_F16 && q->type != HTP_TYPE_F32) ||
-        k->type != HTP_TYPE_F16 ||
-        v->type != HTP_TYPE_F16) {
+    if ((q->type != HTP_TYPE_F16 && q->type != HTP_TYPE_F32) || k->type != HTP_TYPE_F16 || v->type != HTP_TYPE_F16) {
         return HTP_STATUS_NO_SUPPORT;
     }
 
-    octx->src0_div21 = init_fastdiv_values(q->ne[2] * q->ne[1]);
-    octx->src0_div1  = init_fastdiv_values(q->ne[1]);
+    struct htp_fa_context factx;
+    factx.octx = octx;
 
-    octx->broadcast_rk2 = init_fastdiv_values(q->ne[2]/k->ne[2]);
-    octx->broadcast_rk3 = init_fastdiv_values(q->ne[3]/k->ne[3]);
-    octx->broadcast_rv2 = init_fastdiv_values(q->ne[2]/v->ne[2]);
-    octx->broadcast_rv3 = init_fastdiv_values(q->ne[3]/v->ne[3]);
+    factx.src0_div21 = init_fastdiv_values(q->ne[2] * q->ne[1]);
+    factx.src0_div1  = init_fastdiv_values(q->ne[1]);
+
+    factx.broadcast_rk2 = init_fastdiv_values(q->ne[2]/k->ne[2]);
+    factx.broadcast_rk3 = init_fastdiv_values(q->ne[3]/k->ne[3]);
+    factx.broadcast_rv2 = init_fastdiv_values(q->ne[2]/v->ne[2]);
+    factx.broadcast_rv3 = init_fastdiv_values(q->ne[3]/v->ne[3]);
 
     if (mask) {
-        octx->src3_div2 = init_fastdiv_values(mask->ne[2]);
-        octx->src3_div3 = init_fastdiv_values(mask->ne[3]);
+        factx.src3_div2 = init_fastdiv_values(mask->ne[2]);
+        factx.src3_div3 = init_fastdiv_values(mask->ne[3]);
     }
 
-    size_t size_q_row_padded = hex_round_up(q->ne[0] * (q->type == HTP_TYPE_F32 ? 4 : 2), 128);
-    size_t size_k_row_padded = hex_round_up(k->ne[0] * sizeof(__fp16), 128);
-    size_t size_v_row_padded = hex_round_up(v->ne[0] * sizeof(__fp16), 128);
+    factx.is_q_fp32 = (q->type == HTP_TYPE_F32);
+    factx.size_q_row_padded = hex_round_up(q->ne[0] * (factx.is_q_fp32 ? 4 : 2), 128);
+    factx.size_k_row_padded = hex_round_up(k->ne[0] * sizeof(__fp16), 128);
+    factx.size_v_row_padded = hex_round_up(v->ne[0] * sizeof(__fp16), 128);
 
-    size_t size_q_block = size_q_row_padded * 1; // single row for now
-    size_t size_k_block = size_k_row_padded * FLASH_ATTN_BLOCK_SIZE;
-    size_t size_v_block = size_v_row_padded * FLASH_ATTN_BLOCK_SIZE;
-    size_t size_m_block = hex_round_up(FLASH_ATTN_BLOCK_SIZE * sizeof(__fp16), 128);
+    size_t size_q_block = factx.size_q_row_padded * 1; // single row for now
+    factx.size_k_block = factx.size_k_row_padded * FLASH_ATTN_BLOCK_SIZE;
+    factx.size_v_block = factx.size_v_row_padded * FLASH_ATTN_BLOCK_SIZE;
+    factx.size_m_block = hex_round_up(FLASH_ATTN_BLOCK_SIZE * sizeof(__fp16), 128);
+
+    factx.n_blocks = (k->ne[1] + FLASH_ATTN_BLOCK_SIZE - 1) / FLASH_ATTN_BLOCK_SIZE;
+
+    float scale         = 1.0f;
+    float max_bias      = 0.0f;
+    float logit_softcap = 0.0f;
+
+    memcpy(&scale,         (float *) octx->op_params + 0, sizeof(float));
+    memcpy(&max_bias,      (float *) octx->op_params + 1, sizeof(float));
+    memcpy(&logit_softcap, (float *) octx->op_params + 2, sizeof(float));
+
+    if (logit_softcap != 0.0f) {
+        scale /= logit_softcap;
+    }
+
+    factx.scale = scale;
+    factx.max_bias = max_bias;
+    factx.logit_softcap = logit_softcap;
+
+    uint32_t n_head = q->ne[2];
+    factx.n_head_log2 = 1u << (uint32_t) floor(log2(n_head));
+    factx.m0 = powf(2.0f, -(max_bias       ) / factx.n_head_log2);
+    factx.m1 = powf(2.0f, -(max_bias / 2.0f) / factx.n_head_log2);
 
     size_t size_vkq_acc = hex_round_up(v->ne[0] * sizeof(float), 128); // VKQ32
 
     octx->src0_spad.size_per_thread = size_q_block * 1;
-    octx->src1_spad.size_per_thread = size_k_block * 2;
-    octx->src2_spad.size_per_thread = size_v_block * 2;
-    octx->src3_spad.size_per_thread = mask ? size_m_block * 2 : 0;
+    octx->src1_spad.size_per_thread = factx.size_k_block * 2;
+    octx->src2_spad.size_per_thread = factx.size_v_block * 2;
+    octx->src3_spad.size_per_thread = mask ? factx.size_m_block * 2 : 0;
     octx->dst_spad.size_per_thread  = size_vkq_acc;
 
     octx->src0_spad.size = octx->src0_spad.size_per_thread * octx->n_threads;
@@ -677,7 +659,7 @@ int op_flash_attn_ext(struct htp_ops_context * octx) {
     octx->dst_spad.data  = octx->src3_spad.data + octx->src3_spad.size;
 
     if (!(octx->flags & HTP_OPFLAGS_SKIP_COMPUTE)) {
-        worker_pool_run_func(octx->ctx->worker_pool, htp_flash_attn_ext_job, octx, octx->n_threads);
+        worker_pool_run_func(octx->ctx->worker_pool, flash_attn_ext_f16_thread, &factx, octx->n_threads);
     }
 
     return HTP_STATUS_OK;

ggml/src/ggml-metal/ggml-metal-device.m

@@ -1067,8 +1067,8 @@ bool ggml_metal_device_supports_op(ggml_metal_device_t dev, const struct ggml_te
         case GGML_OP_MUL:
         case GGML_OP_DIV:
         case GGML_OP_ADD_ID:
-            return ggml_is_contiguous_rows(op->src[0]) && ggml_is_contiguous_rows(op->src[1]) && op->src[0]->type == GGML_TYPE_F32;
         case GGML_OP_ACC:
+            return ggml_is_contiguous_rows(op->src[0]) && ggml_is_contiguous_rows(op->src[1]) && op->src[0]->type == GGML_TYPE_F32;
         case GGML_OP_REPEAT:
         case GGML_OP_CONV_TRANSPOSE_1D:
             return true;

ggml/src/ggml-metal/ggml-metal-ops.cpp

@@ -620,8 +620,8 @@ int ggml_metal_op_acc(ggml_metal_op_t ctx, int idx) {
     GGML_ASSERT(op->src[1]->type == GGML_TYPE_F32);
     GGML_ASSERT(op->type         == GGML_TYPE_F32);
 
-    GGML_ASSERT(ggml_is_contiguous(op->src[0]));
-    GGML_ASSERT(ggml_is_contiguous(op->src[1]));
+    GGML_ASSERT(ggml_is_contiguous_rows(op->src[0]));
+    GGML_ASSERT(ggml_is_contiguous_rows(op->src[1]));
 
     const size_t pnb1 = ((const int32_t *) op->op_params)[0];
     const size_t pnb2 = ((const int32_t *) op->op_params)[1];
@@ -671,10 +671,10 @@ int ggml_metal_op_acc(ggml_metal_op_t ctx, int idx) {
     }
 
     ggml_metal_kargs_bin args = {
-        /*.ne00 =*/ ne00,
-        /*.ne01 =*/ ne01,
-        /*.ne02 =*/ ne02,
-        /*.ne03 =*/ ne03,
+        /*.ne00 =*/ ne10,
+        /*.ne01 =*/ ne11,
+        /*.ne02 =*/ ne12,
+        /*.ne03 =*/ ne13,
         /*.nb00 =*/ nb00,
         /*.nb01 =*/ pnb1,
         /*.nb02 =*/ pnb2,
@@ -687,10 +687,10 @@ int ggml_metal_op_acc(ggml_metal_op_t ctx, int idx) {
         /*.nb11 =*/ nb11,
         /*.nb12 =*/ nb12,
         /*.nb13 =*/ nb13,
-        /*.ne0  =*/ ne0,
-        /*.ne1  =*/ ne1,
-        /*.ne2  =*/ ne2,
-        /*.ne3  =*/ ne3,
+        /*.ne0  =*/ ne10,
+        /*.ne1  =*/ ne11,
+        /*.ne2  =*/ ne12,
+        /*.ne3  =*/ ne13,
         /*.nb0  =*/ nb0,
         /*.nb1  =*/ pnb1,
         /*.nb2  =*/ pnb2,
@@ -707,7 +707,13 @@ int ggml_metal_op_acc(ggml_metal_op_t ctx, int idx) {
     ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op->src[1]), 2);
     ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op),         3);
 
-    const int nth = std::min(ggml_metal_pipeline_max_theads_per_threadgroup(pipeline), ne00);
+    const int nth_max = MIN(256, ggml_metal_pipeline_max_theads_per_threadgroup(pipeline));
+
+    int nth = 1;
+
+    while (2*nth < args.ne0 && nth < nth_max) {
+        nth *= 2;
+    }
 
     ggml_metal_encoder_dispatch_threadgroups(enc, ne11, ne12, ne13, nth, 1, 1);
 

ggml/src/ggml-vulkan/ggml-vulkan.cpp

@@ -92,6 +92,7 @@ static bool is_pow2(uint32_t x) { return x > 1 && (x & (x-1)) == 0; }
 #define VK_VENDOR_ID_APPLE 0x106b
 #define VK_VENDOR_ID_INTEL 0x8086
 #define VK_VENDOR_ID_NVIDIA 0x10de
+#define VK_VENDOR_ID_QUALCOMM 0x5143
 
 #define VK_DEVICE_DESCRIPTOR_POOL_SIZE 256
 
@@ -687,6 +688,7 @@ struct vk_device_struct {
     vk_pipeline pipeline_get_rows[GGML_TYPE_COUNT];
     vk_pipeline pipeline_get_rows_f32[GGML_TYPE_COUNT];
     vk_pipeline pipeline_acc_f32;
+    vk_pipeline pipeline_set_f32;
 
     // [src0 0=fp32,1=fp16][src1 0=fp32,1=fp16][dst 0=fp32,1=fp16]
     vk_pipeline pipeline_add[2][2][2];
@@ -4080,7 +4082,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
     }
 
     ggml_vk_create_pipeline(device, device->pipeline_rms_norm_back_f32, "rms_norm_back_f32", rms_norm_back_f32_len, rms_norm_back_f32_data, "main", 3, sizeof(vk_op_push_constants), {1, 1, 1}, {}, 1);
-    ggml_vk_create_pipeline(device, device->pipeline_l2_norm_f32, "l2_norm_f32", l2_norm_f32_len, l2_norm_f32_data, "main", 2, sizeof(vk_op_push_constants), {1, 1, 1}, {}, 1);
+    ggml_vk_create_pipeline(device, device->pipeline_l2_norm_f32, "l2_norm_f32", l2_norm_f32_len, l2_norm_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {1, 1, 1}, {}, 1);
 
     ggml_vk_create_pipeline(device, device->pipeline_cpy_f32_f32, "cpy_f32_f32", cpy_f32_f32_len, cpy_f32_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_cpy_f32_f16, "cpy_f32_f16", cpy_f32_f16_len, cpy_f32_f16_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
@@ -4181,7 +4183,8 @@ static void ggml_vk_load_shaders(vk_device& device) {
 
     ggml_vk_create_pipeline(device, device->pipeline_add_id_f32, "add_id_f32", add_id_f32_len, add_id_f32_data, "main", 4, sizeof(vk_op_add_id_push_constants), {1, 1, 1}, {}, 1);
 
-    ggml_vk_create_pipeline(device, device->pipeline_acc_f32, "acc_f32", acc_f32_len, acc_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {512, 1, 1}, {}, 1);
+    ggml_vk_create_pipeline(device, device->pipeline_acc_f32, "acc_f32", acc_f32_len, acc_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {512, 1, 1}, {0, 1}, 1);
+    ggml_vk_create_pipeline(device, device->pipeline_set_f32, "set_f32", acc_f32_len, acc_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {512, 1, 1}, {0, 0}, 1);
 
     ggml_vk_create_pipeline(device, device->pipeline_concat_f32, "concat_f32", concat_f32_len, concat_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {512, 1, 1}, {}, 1);
     ggml_vk_create_pipeline(device, device->pipeline_concat_f16, "concat_f16", concat_f16_len, concat_f16_data, "main", 3, sizeof(vk_op_binary_push_constants), {512, 1, 1}, {}, 1);
@@ -5641,6 +5644,10 @@ static void ggml_vk_instance_init() {
                             driver_priorities[vk::DriverId::eMesaNvk] = 2;
 #endif
                             break;
+                        case VK_VENDOR_ID_QUALCOMM:
+                            driver_priorities[vk::DriverId::eQualcommProprietary] = 1;
+                            driver_priorities[vk::DriverId::eMesaTurnip] = 2;
+                            break;
                     }
                     driver_priorities[vk::DriverId::eMesaDozen] = 100;
 
@@ -8422,6 +8429,8 @@ static bool ggml_vk_flash_attn_coopmat_shmem_support(const vk_device& device, co
     const uint32_t acctype = f32acc ? 4 : 2;
     const uint32_t f16vec4 = 8;
 
+    const uint32_t tmpsh = (Bc / MatBc) * sizeof(float);
+
     const uint32_t qstride = hsk_pad / 4 + 2;
     const uint32_t Qf = Br * qstride * f16vec4;
 
@@ -8438,7 +8447,7 @@ static bool ggml_vk_flash_attn_coopmat_shmem_support(const vk_device& device, co
 
     const uint32_t slope = Br * acctype;
 
-    const uint32_t total_size = Qf + Psh + sfsh + ksh + slope;
+    const uint32_t total_size = tmpsh + Qf + Psh + sfsh + ksh + slope;
     const bool supported = total_size <= device->properties.limits.maxComputeSharedMemorySize;
 
     VK_LOG_DEBUG("ggml_vk_flash_attn_coopmat_shmem_support(HSK=" << hsk << ", HSV=" << hsv << ", f32acc=" << f32acc << ", kv_type=" << kv_type << ", total_size=" << total_size << ", supported=" << supported);
@@ -8815,6 +8824,12 @@ static vk_pipeline ggml_vk_op_get_pipeline(ggml_backend_vk_context * ctx, const
             return ctx->device->pipeline_acc_f32;
         }
         return nullptr;
+    case GGML_OP_SET:
+        if (src0->type == src1->type && src0->type == dst->type &&
+            (src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_I32)) {
+            return ctx->device->pipeline_set_f32;
+        }
+        return nullptr;
     case GGML_OP_ADD:
     case GGML_OP_SUB:
     case GGML_OP_MUL:
@@ -9806,7 +9821,7 @@ static void ggml_vk_acc(ggml_backend_vk_context * ctx, vk_context& subctx, const
     int nb3 = dst->op_params[2] / src0_type_size; // 4 bytes of float32
     int offset = dst->op_params[3] / src0_type_size; // offset in bytes
 
-    ggml_vk_op_f32<vk_op_binary_push_constants>(ctx, subctx, src0, src1, nullptr, nullptr, dst, GGML_OP_ACC, {
+    ggml_vk_op_f32<vk_op_binary_push_constants>(ctx, subctx, src0, src1, nullptr, nullptr, dst, dst->op, {
         (uint32_t)ggml_nelements(src0),
         (uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2],(uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)nb1, (uint32_t)nb2, (uint32_t)nb3,
         (uint32_t)src1->ne[0], (uint32_t)src1->ne[1], (uint32_t)src1->ne[2],(uint32_t)src1->ne[3], (uint32_t)src1->nb[0] / src1_type_size, (uint32_t)src1->nb[1] / src1_type_size, (uint32_t)src1->nb[2] / src1_type_size, (uint32_t)src1->nb[3] / src1_type_size,
@@ -10624,8 +10639,10 @@ static void ggml_vk_rms_norm_back(ggml_backend_vk_context * ctx, vk_context& sub
 }
 
 static void ggml_vk_l2_norm(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst) {
-    float * op_params = (float *)dst->op_params;
-    ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, nullptr, dst, GGML_OP_L2_NORM, { (uint32_t)src0->ne[0], (uint32_t)src0->ne[1], op_params[0], 0.0f, 0.0f, 0.0f });
+    const float * op_params = (const float *)dst->op_params;
+    vk_op_unary_push_constants p = vk_op_unary_push_constants_init(src0, dst);
+    p.param1 = op_params[0];
+    ggml_vk_op_f32<vk_op_unary_push_constants>(ctx, subctx, src0, nullptr, nullptr, nullptr, dst, GGML_OP_L2_NORM, std::move(p));
 }
 
 static void ggml_vk_unary(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst) {
@@ -12500,6 +12517,7 @@ static bool ggml_vk_build_graph(ggml_backend_vk_context * ctx, ggml_cgraph * cgr
 
         break;
     case GGML_OP_ACC:
+    case GGML_OP_SET:
         ggml_vk_acc(ctx, compute_ctx, src0, src1, node);
 
         break;
@@ -14896,8 +14914,10 @@ static bool ggml_backend_vk_device_supports_op(ggml_backend_dev_t dev, const ggm
             return true;
         case GGML_OP_NORM:
         case GGML_OP_GROUP_NORM:
-        case GGML_OP_L2_NORM:
             return ggml_is_contiguous(op->src[0]);
+        case GGML_OP_L2_NORM:
+            return ggml_is_contiguous_rows(op->src[0]) &&
+                   op->src[0]->type == GGML_TYPE_F32 && op->type == GGML_TYPE_F32;
         case GGML_OP_ADD:
         case GGML_OP_SUB:
         case GGML_OP_MUL:
@@ -14960,7 +14980,10 @@ static bool ggml_backend_vk_device_supports_op(ggml_backend_dev_t dev, const ggm
             }
             return op->src[0]->type == GGML_TYPE_F32;
         case GGML_OP_ACC:
-            return op->src[0]->type == GGML_TYPE_F32;
+            return op->src[0]->type == GGML_TYPE_F32 && op->src[1]->type == GGML_TYPE_F32 && op->type == GGML_TYPE_F32;
+        case GGML_OP_SET:
+            return op->src[0]->type == op->src[1]->type && op->src[0]->type == op->type &&
+                   (op->src[0]->type == GGML_TYPE_F32 || op->src[0]->type == GGML_TYPE_I32);
         case GGML_OP_CONCAT:
             return ggml_type_size(op->src[0]->type) == ggml_type_size(GGML_TYPE_F32);
         case GGML_OP_ADD1:
@@ -15611,6 +15634,8 @@ static void ggml_vk_check_results_0(ggml_backend_vk_context * ctx, ggml_cgraph *
             tensor_clone = ggml_add(ggml_ctx, src_clone[0], src_clone[1]);
         } else if (tensor->op == GGML_OP_ACC) {
             tensor_clone = ggml_acc(ggml_ctx, src_clone[0], src_clone[1], tensor->op_params[0], tensor->op_params[1], tensor->op_params[2], tensor->op_params[3]);
+        } else if (tensor->op == GGML_OP_SET) {
+            tensor_clone = ggml_set(ggml_ctx, src_clone[0], src_clone[1], tensor->op_params[0], tensor->op_params[1], tensor->op_params[2], tensor->op_params[3]);
         } else if (tensor->op == GGML_OP_NORM) {
             tensor_clone = ggml_norm(ggml_ctx, src_clone[0], *(float *)tensor->op_params);
         } else if (tensor->op == GGML_OP_GROUP_NORM) {

ggml/src/ggml-vulkan/vulkan-shaders/acc.comp

@@ -3,6 +3,9 @@
 #include "types.glsl"
 #include "generic_binary_head.glsl"
 
+// false for SET, true for ACC
+layout(constant_id = 1) const bool ACC = true;
+
 layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
 
 void main() {
@@ -23,7 +26,11 @@ void main() {
     uint i00, i01, i02, i03;
 
     if (i0 < p.ne10 && i1 < p.ne11 && i2 < p.ne12 && i3 < p.ne13) {
-        data_d[get_doffset() + idx] = D_TYPE(FLOAT_TYPE(data_a[get_aoffset() + idx]) + FLOAT_TYPE(data_b[get_boffset() + src1_idx(i0, i1, i2, i3)]));
+        if (ACC) {
+            data_d[get_doffset() + idx] = D_TYPE(FLOAT_TYPE(data_a[get_aoffset() + idx]) + FLOAT_TYPE(data_b[get_boffset() + src1_idx(i0, i1, i2, i3)]));
+        } else {
+            data_d[get_doffset() + idx] = D_TYPE(FLOAT_TYPE(data_b[get_boffset() + src1_idx(i0, i1, i2, i3)]));
+        }
     } else {
         data_d[get_doffset() + idx] = D_TYPE(FLOAT_TYPE(data_a[get_aoffset() + idx]));
     }

ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp

@@ -130,6 +130,7 @@ void main() {
         if (MASK_ENABLE && mask_opt_bits != MASK_OPT_ALL_ZERO) {
             bool nem1_bounds_check = !(p.gqa_ratio > 1) && (p.nem1 % Br) != 0;
 
+            float max_mask = NEG_FLT_MAX_OVER_2;
             [[unroll]] for (uint32_t idx = 0; idx < Bc * Br; idx += gl_WorkGroupSize.x) {
                 uint32_t c = (idx + tid) % Bc;
                 uint32_t r = (idx + tid) / Bc;
@@ -137,12 +138,25 @@ void main() {
                     if ((!KV_bounds_check || j * Bc + c < KV) && (!nem1_bounds_check || i * Br + r < p.nem1)) {
                         float m = float(data_m[m_offset + (i * Br + r) * m_stride + (j * Bc + c)]);
                         masksh[c][r] = m;
+                        max_mask = max(max_mask, m);
                     } else {
                         masksh[c][r] = float(0);
                     }
                 }
             }
+            // skip the block if the mask is entirely -inf
+            bool all_less = subgroupAll(max_mask <= NEG_FLT_MAX_OVER_2);
             barrier();
+            if (gl_SubgroupInvocationID == 0) {
+                tmpsh[gl_SubgroupID] = all_less ? NEG_FLT_MAX_OVER_2 : 0.0f;
+            }
+            barrier();
+            [[unroll]] for (uint s = 0; s < gl_NumSubgroups; ++s) {
+                max_mask = max(max_mask, tmpsh[s]);
+            }
+            if (max_mask <= NEG_FLT_MAX_OVER_2) {
+                continue;
+            }
         }
 
         float Sf[Br][cols_per_thread];
@@ -260,6 +274,9 @@ void main() {
         barrier();
     }
 
+    // prevent race on tmpsh
+    barrier();
+
     // reduce across threads
 
     [[unroll]] for (uint32_t r = 0; r < Br; ++r) {

ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp

@@ -42,6 +42,8 @@ D_TYPE perElemOpGqaStore(const in uint32_t r, const in uint32_t c, const in D_TY
     return elem;
 }
 
+shared float tmpsh[row_split];
+
 const uint32_t qstride = HSK_pad / 4 + 2; // in units of f16vec4
 shared f16vec4 Qf[Br * qstride];
 
@@ -213,6 +215,19 @@ void main() {
                         }
                     }
                 }
+                // skip the block if the mask is entirely -inf
+                bool all_less = subgroupAll(max_mask <= NEG_FLT_MAX_OVER_2);
+                barrier();
+                if (gl_SubgroupInvocationID == 0) {
+                    tmpsh[gl_SubgroupID] = all_less ? NEG_FLT_MAX_OVER_2 : 0.0f;
+                }
+                barrier();
+                [[unroll]] for (uint s = 0; s < gl_NumSubgroups; ++s) {
+                    max_mask = max(max_mask, tmpsh[s]);
+                }
+                if (max_mask <= NEG_FLT_MAX_OVER_2) {
+                    continue;
+                }
             }
         }
 

ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp

@@ -176,7 +176,14 @@ void main() {
                     tensorLayoutM = setTensorLayoutStrideNV(tensorLayoutM, m_stride, 1);
                     tensorLayoutM = setTensorLayoutClampValueNV(tensorLayoutM, 0xfc00); // -inf in float16_t
 
+                    coopmat<float16_t, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> mvmax;
+
                     coopMatLoadTensorNV(mv, data_m, m_offset, sliceTensorLayoutNV(tensorLayoutM, i * Br, Br, j * Bc, Bc));
+                    // skip the block if the mask is entirely -inf
+                    coopMatReduceNV(mvmax, mv, gl_CooperativeMatrixReduceRowAndColumnNV, maxReduceFp16);
+                    if (mvmax[0] <= NEG_FLT_MAX_OVER_2) {
+                        continue;
+                    }
                 } else {
                     tensorLayoutNV<2, Clamp> tensorLayoutM = createTensorLayoutNV(2, Clamp);
                     // Don't clamp against nem1 when GQA is enabled
@@ -184,7 +191,14 @@ void main() {
                     tensorLayoutM = setTensorLayoutDimensionNV(tensorLayoutM, m_height, KV);
                     tensorLayoutM = setTensorLayoutStrideNV(tensorLayoutM, m_stride, 1);
 
+                    coopmat<float16_t, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> mvmax;
+
                     coopMatLoadTensorNV(mv, data_m, m_offset, sliceTensorLayoutNV(tensorLayoutM, i * Br, Br, j * Bc, Bc));
+                    // skip the block if the mask is entirely -inf
+                    coopMatReduceNV(mvmax, mv, gl_CooperativeMatrixReduceRowAndColumnNV, maxReduceFp16);
+                    if (mvmax[0] <= NEG_FLT_MAX_OVER_2) {
+                        continue;
+                    }
                 }
             }
         }

ggml/src/ggml-vulkan/vulkan-shaders/l2_norm.comp

@@ -1,6 +1,6 @@
 #version 450
 
-#include "generic_head.glsl"
+#include "generic_unary_head.glsl"
 #include "types.glsl"
 
 #extension GL_EXT_control_flow_attributes : enable
@@ -8,19 +8,22 @@
 
 layout(local_size_x = BLOCK_SIZE, local_size_y = 1, local_size_z = 1) in;
 
-layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
-layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
-
 shared FLOAT_TYPE sum[BLOCK_SIZE];
 
 void main() {
     const uint row = gl_WorkGroupID.z * 262144 + gl_WorkGroupID.y * 512 + gl_WorkGroupID.x;
     const uint tid = gl_LocalInvocationID.x;
 
+    const uint i3 = row / (p.ne11 * p.ne12);
+    const uint i3_offset = i3 * p.ne12 * p.ne11;
+    const uint i2 = (row - i3_offset) / p.ne11;
+    const uint i2_offset = i2 * p.ne11;
+    const uint i1 = row - i3_offset - i2_offset;
+
     sum[tid] = FLOAT_TYPE(0.0f); // partial sum for thread in warp
 
-    [[unroll]] for (uint col = tid; col < p.KX; col += BLOCK_SIZE) {
-        const FLOAT_TYPE xi = FLOAT_TYPE(data_a[row*p.KX + col]);
+    [[unroll]] for (uint i0 = tid; i0 < p.ne00; i0 += BLOCK_SIZE) {
+        const FLOAT_TYPE xi = FLOAT_TYPE(data_a[i3*p.nb03 + i2*p.nb02 + i1*p.nb01 + i0]);
         sum[tid] += xi * xi;
     }
 
@@ -35,7 +38,7 @@ void main() {
 
     const FLOAT_TYPE scale = inversesqrt(max(sum[0], FLOAT_TYPE(p.param1)));
 
-    [[unroll]] for (uint col = tid; col < p.KX; col += BLOCK_SIZE) {
-        data_d[row*p.KX + col] = D_TYPE(scale * FLOAT_TYPE(data_a[row*p.KX + col]));
+    [[unroll]] for (uint i0 = tid; i0 < p.ne00; i0 += BLOCK_SIZE) {
+        data_d[i3*p.nb13 + i2*p.nb12 + i1*p.nb11 + i0] = D_TYPE(scale * FLOAT_TYPE(data_a[i3*p.nb03 + i2*p.nb02 + i1*p.nb01 + i0]));
     }
 }

include/llama.h

@@ -656,21 +656,12 @@ extern "C" {
 
     // The following functions operate on a llama_context, hence the naming: llama_verb_...
 
-    // Add a loaded LoRA adapter to given context
-    // This will not modify model's weight
-    LLAMA_API int32_t llama_set_adapter_lora(
+    // Set LoRa adapters on the context. Will only modify if the adapters currently in context are different.
+    LLAMA_API int32_t llama_set_adapters_lora(
             struct llama_context * ctx,
-            struct llama_adapter_lora * adapter,
-            float scale);
-
-    // Remove a specific LoRA adapter from given context
-    // Return -1 if the adapter is not present in the context
-    LLAMA_API int32_t llama_rm_adapter_lora(
-            struct llama_context * ctx,
-            struct llama_adapter_lora * adapter);
-
-    // Remove all LoRA adapters from given context
-    LLAMA_API void llama_clear_adapter_lora(struct llama_context * ctx);
+            struct llama_adapter_lora ** adapters,
+            size_t n_adapters,
+            float * scales);
 
     // Apply a loaded control vector to a llama_context, or if data is NULL, clear
     // the currently loaded vector.
@@ -678,7 +669,7 @@ extern "C" {
     // to an n_embd x n_layers buffer starting from layer 1.
     // il_start and il_end are the layer range the vector should apply to (both inclusive)
     // See llama_control_vector_load in common to load a control vector.
-    LLAMA_API int32_t llama_apply_adapter_cvec(
+    LLAMA_API int32_t llama_set_adapter_cvec(
             struct llama_context * ctx,
                      const float * data,
                           size_t   len,

scripts/sync_vendor.py

@@ -1,6 +1,9 @@
 #!/usr/bin/env python3
 
 import urllib.request
+import os
+import sys
+import subprocess
 
 HTTPLIB_VERSION = "f80864ca031932351abef49b74097c67f14719c6"
 
@@ -14,7 +17,8 @@ vendor = {
     # "https://github.com/mackron/miniaudio/raw/refs/tags/0.11.23/miniaudio.h": "vendor/miniaudio/miniaudio.h",
     "https://github.com/mackron/miniaudio/raw/669ed3e844524fcd883231b13095baee9f6de304/miniaudio.h": "vendor/miniaudio/miniaudio.h",
 
-    f"https://raw.githubusercontent.com/yhirose/cpp-httplib/{HTTPLIB_VERSION}/httplib.h": "vendor/cpp-httplib/httplib.h",
+    f"https://raw.githubusercontent.com/yhirose/cpp-httplib/{HTTPLIB_VERSION}/httplib.h": "httplib.h",
+    f"https://raw.githubusercontent.com/yhirose/cpp-httplib/{HTTPLIB_VERSION}/split.py":  "split.py",
     f"https://raw.githubusercontent.com/yhirose/cpp-httplib/{HTTPLIB_VERSION}/LICENSE":   "vendor/cpp-httplib/LICENSE",
 
     "https://raw.githubusercontent.com/sheredom/subprocess.h/b49c56e9fe214488493021017bf3954b91c7c1f5/subprocess.h": "vendor/sheredom/subprocess.h",
@@ -24,19 +28,16 @@ for url, filename in vendor.items():
     print(f"downloading {url} to {filename}") # noqa: NP100
     urllib.request.urlretrieve(url, filename)
 
-    # split cpp/h files for httplib
-    # see: https://github.com/yhirose/cpp-httplib/blob/master/split.py
-    if 'httplib.h' in filename:
-        border = '// ----------------------------------------------------------------------------'
-        with open(filename, 'r') as f:
-            content = f.read()
-        header, implementation, footer = content.split(border, 2)
-        fname_cpp = filename.replace('.h', '.cpp')
-        with open(filename, 'w') as fh:
-            fh.write(header)
-            fh.write(footer)
-        with open(fname_cpp, 'w') as fc:
-            fc.write('#include "httplib.h"\n')
-            fc.write('namespace httplib {\n')
-            fc.write(implementation.replace('\ninline ', '\n'))
-            fc.write('} // namespace httplib\n')
+print("Splitting httplib.h...") # noqa: NP100
+try:
+    subprocess.check_call([
+        sys.executable, "split.py",
+        "--extension", "cpp",
+        "--out", "vendor/cpp-httplib"
+    ])
+except Exception as e:
+    print(f"Error: {e}") # noqa: NP100
+    sys.exit(1)
+finally:
+    os.remove("split.py")
+    os.remove("httplib.h")

src/llama-context.cpp

@@ -1057,51 +1057,43 @@ bool llama_context::set_sampler(llama_seq_id seq_id, llama_sampler * sampler) {
     return true;
 }
 
-void llama_context::set_adapter_lora(
-            llama_adapter_lora * adapter,
-            float scale) {
-    LLAMA_LOG_DEBUG("%s: adapter = %p, scale = %f\n", __func__, (void *) adapter, scale);
+void llama_context::set_adapters_lora(llama_adapter_lora ** adapters, size_t n_adapters, float * scales) {
+    LLAMA_LOG_DEBUG("%s: adapters = %p\n", __func__, (void *) adapters);
 
-    if (auto it = loras.find(adapter); it != loras.end()) {
-        if (it->second == scale) {
-            return;
-        }
-    }
-
-    loras[adapter] = scale;
-
-    sched_need_reserve = true;
-}
-
-bool llama_context::rm_adapter_lora(
-            llama_adapter_lora * adapter) {
-    LLAMA_LOG_DEBUG("%s: adapter = %p\n", __func__, (void *) adapter);
-
-    auto it = loras.find(adapter);
-    if (it != loras.end()) {
-        loras.erase(it);
-
-        sched_need_reserve = true;
-
-        return true;
-    }
-
-    return false;
-}
-
-void llama_context::clear_adapter_lora() {
-    LLAMA_LOG_DEBUG("%s: call\n", __func__);
-
-    if (loras.empty()) {
+    if (adapters_lora_are_same(adapters, n_adapters, scales)) {
         return;
     }
 
     loras.clear();
 
+    for (size_t i = 0; i < n_adapters; i ++) {
+        if (scales[i] != 0.0f) {
+            loras[adapters[i]] = scales[i];
+        }
+    }
+
     sched_need_reserve = true;
 }
 
-bool llama_context::apply_adapter_cvec(
+bool llama_context::adapters_lora_are_same(llama_adapter_lora ** adapters, size_t n_adapters, float * scales) {
+    LLAMA_LOG_DEBUG("%s: adapters = %p\n", __func__, (void *) adapters);
+
+    if (n_adapters != loras.size()) {
+        return false;
+    }
+
+    for (size_t i = 0; i < n_adapters; i ++) {
+        auto it = loras.find(adapters[i]);
+
+        if (it == loras.end() || it->second != scales[i]) {
+            return false;
+        }
+    }
+
+    return true;
+}
+
+bool llama_context::set_adapter_cvec(
             const float * data,
                  size_t   len,
                 int32_t   n_embd,
@@ -3209,35 +3201,28 @@ uint32_t llama_get_sampled_probs_count_ith(llama_context * ctx, int32_t i) {
 
 // llama adapter API
 
-int32_t llama_set_adapter_lora(
+int32_t llama_set_adapters_lora(
             llama_context * ctx,
-            llama_adapter_lora * adapter,
-            float scale) {
-    ctx->set_adapter_lora(adapter, scale);
+            llama_adapter_lora ** adapters,
+            size_t n_adapters,
+            float * scales) {
+    if (adapters == nullptr || scales == nullptr) {
+        GGML_ASSERT(n_adapters == 0 && "invalid llama_set_adapters_lora call");
+    }
+
+    ctx->set_adapters_lora(adapters, n_adapters, scales);
 
     return 0;
 }
 
-int32_t llama_rm_adapter_lora(
-            llama_context * ctx,
-            llama_adapter_lora * adapter) {
-    bool res = ctx->rm_adapter_lora(adapter);
-
-    return res ? 0 : -1;
-}
-
-void llama_clear_adapter_lora(llama_context * ctx) {
-    ctx->clear_adapter_lora();
-}
-
-int32_t llama_apply_adapter_cvec(
+int32_t llama_set_adapter_cvec(
         llama_context * ctx,
-                 const float * data,
-                      size_t   len,
-                     int32_t   n_embd,
-                     int32_t   il_start,
-                     int32_t   il_end) {
-    bool res = ctx->apply_adapter_cvec(data, len, n_embd, il_start, il_end);
+          const float * data,
+               size_t   len,
+              int32_t   n_embd,
+              int32_t   il_start,
+              int32_t   il_end) {
+    bool res = ctx->set_adapter_cvec(data, len, n_embd, il_start, il_end);
 
     return res ? 0 : -1;
 }

src/llama-context.h

@@ -105,16 +105,11 @@ struct llama_context {
     void set_causal_attn(bool value);
     void set_warmup(bool value);
 
-    void set_adapter_lora(
-            llama_adapter_lora * adapter,
-            float scale);
+    void set_adapters_lora(llama_adapter_lora ** adapters, size_t n_adapters, float * scales);
 
-    bool rm_adapter_lora(
-            llama_adapter_lora * adapter);
+    bool adapters_lora_are_same(llama_adapter_lora ** adapters, size_t n_adapters, float * scales);
 
-    void clear_adapter_lora();
-
-    bool apply_adapter_cvec(
+    bool set_adapter_cvec(
             const float * data,
                  size_t   len,
                 int32_t   n_embd,

src/llama-mmap.cpp

@@ -504,6 +504,8 @@ struct llama_mmap::impl {
         }
     }
 #elif defined(_WIN32)
+    HANDLE hMapping = nullptr;
+
     impl(struct llama_file * file, size_t prefetch, bool numa) {
         GGML_UNUSED(numa);
 
@@ -511,7 +513,7 @@ struct llama_mmap::impl {
 
         HANDLE hFile = (HANDLE) _get_osfhandle(file->file_id());
 
-        HANDLE hMapping = CreateFileMappingA(hFile, NULL, PAGE_READONLY, 0, 0, NULL);
+        hMapping = CreateFileMappingA(hFile, NULL, PAGE_READONLY, 0, 0, NULL);
 
         if (hMapping == NULL) {
             DWORD error = GetLastError();
@@ -520,9 +522,9 @@ struct llama_mmap::impl {
 
         addr = MapViewOfFile(hMapping, FILE_MAP_READ, 0, 0, 0);
         DWORD error = GetLastError();
-        CloseHandle(hMapping);
 
         if (addr == NULL) {
+            CloseHandle(hMapping);
             throw std::runtime_error(format("MapViewOfFile failed: %s", llama_format_win_err(error).c_str()));
         }
 
@@ -554,9 +556,17 @@ struct llama_mmap::impl {
     }
 
     ~impl() {
-        if (!UnmapViewOfFile(addr)) {
-            LLAMA_LOG_WARN("warning: UnmapViewOfFile failed: %s\n",
-                    llama_format_win_err(GetLastError()).c_str());
+        if (hMapping) {
+            if (addr) {
+                if (!UnmapViewOfFile(addr)) {
+                    LLAMA_LOG_WARN("warning: UnmapViewOfFile failed: %s\n",
+                            llama_format_win_err(GetLastError()).c_str());
+                }
+            }
+            if (!CloseHandle(hMapping)) {
+                LLAMA_LOG_WARN("warning: CloseHandle failed: %s\n",
+                        llama_format_win_err(GetLastError()).c_str());
+            }
         }
     }
 #else

tests/test-backend-ops.cpp

@@ -5821,20 +5821,27 @@ struct test_l2_norm : public test_case {
     const ggml_type type;
     const std::array<int64_t, 4> ne;
     const float eps;
+    bool v;
 
     std::string vars() override {
-        return VARS_TO_STR2(type, ne);
+        return VARS_TO_STR4(type, ne, eps, v);
     }
 
     test_l2_norm(ggml_type type = GGML_TYPE_F32,
             std::array<int64_t, 4> ne = {64, 64, 320, 1},
-            float eps = 1e-12f)
-        : type(type), ne(ne), eps(eps) {}
+            float eps = 1e-12f,
+            bool v = false)
+        : type(type), ne(ne), eps(eps), v(v) {}
 
     ggml_tensor * build_graph(ggml_context * ctx) override {
         ggml_tensor * a = ggml_new_tensor(ctx, type, 4, ne.data());
         ggml_set_name(a, "a");
 
+        if (v) {
+            a = ggml_view_4d(ctx, a, a->ne[0]/2, a->ne[1]/2, a->ne[2]/2, a->ne[3]/2, a->nb[1], a->nb[2], a->nb[3], 0);
+            ggml_set_name(a, "view of a");
+        }
+
         ggml_tensor * out = ggml_l2_norm(ctx, a, eps);
         ggml_set_name(out, "out");
 
@@ -7596,7 +7603,8 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
                 test_cases.emplace_back(new test_rms_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, v, eps));
             }
             test_cases.emplace_back(new test_rms_norm_back(GGML_TYPE_F32, { n, 5, 4, 3 }, eps));
-            test_cases.emplace_back(new test_l2_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, eps));
+            test_cases.emplace_back(new test_l2_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, eps, false));
+            test_cases.emplace_back(new test_l2_norm(GGML_TYPE_F32, { n, 5, 4, 3 }, eps, true));
         }
     }
 

tools/rpc/rpc-server.cpp

@@ -132,7 +132,8 @@ static std::string fs_get_cache_directory() {
     if (getenv("LLAMA_CACHE")) {
         cache_directory = std::getenv("LLAMA_CACHE");
     } else {
-#if defined(__linux__) || defined(__FreeBSD__) || defined(_AIX) || defined(__OpenBSD__)
+#if defined(__linux__) || defined(__FreeBSD__) || defined(_AIX) || \
+    defined(__OpenBSD__) || defined(__NetBSD__)
         if (std::getenv("XDG_CACHE_HOME")) {
             cache_directory = std::getenv("XDG_CACHE_HOME");
         } else if (std::getenv("HOME")) {

tools/server/webui/.storybook/main.ts

@@ -1,17 +1,24 @@
 import type { StorybookConfig } from '@storybook/sveltekit';
+import { dirname, resolve } from 'path';
+import { fileURLToPath } from 'url';
+
+const __dirname = dirname(fileURLToPath(import.meta.url));
 
 const config: StorybookConfig = {
 	stories: ['../tests/stories/**/*.mdx', '../tests/stories/**/*.stories.@(js|ts|svelte)'],
 	addons: [
 		'@storybook/addon-svelte-csf',
 		'@chromatic-com/storybook',
-		'@storybook/addon-docs',
+		'@storybook/addon-vitest',
 		'@storybook/addon-a11y',
-		'@storybook/addon-vitest'
+		'@storybook/addon-docs'
 	],
-	framework: {
-		name: '@storybook/sveltekit',
-		options: {}
+	framework: '@storybook/sveltekit',
+	viteFinal: async (config) => {
+		config.server = config.server || {};
+		config.server.fs = config.server.fs || {};
+		config.server.fs.allow = [...(config.server.fs.allow || []), resolve(__dirname, '../tests')];
+		return config;
 	}
 };
 export default config;

tools/server/webui/.storybook/preview.ts

@@ -13,7 +13,7 @@ const preview: Preview = {
 		},
 
 		backgrounds: {
-			disable: true
+			disabled: true
 		},
 
 		a11y: {

tools/server/webui/docs/flows/settings-flow.md

@@ -49,14 +49,20 @@ sequenceDiagram
     settingsStore->>serverStore: defaultParams
     serverStore-->>settingsStore: {temperature, top_p, top_k, ...}
 
-    settingsStore->>ParamSvc: extractServerDefaults(defaultParams)
-    ParamSvc-->>settingsStore: Record<string, value>
+    loop each SYNCABLE_PARAMETER
+        alt key NOT in userOverrides
+            settingsStore->>settingsStore: config[key] = serverDefault[key]
+            Note right of settingsStore: Non-overridden params adopt server default
+        else key in userOverrides
+            Note right of settingsStore: Keep user value, skip server default
+        end
+    end
 
-    settingsStore->>ParamSvc: mergeWithServerDefaults(config, serverDefaults)
-    Note right of ParamSvc: For each syncable parameter:<br/>- If NOT in userOverrides → use server default<br/>- If in userOverrides → keep user value
-    ParamSvc-->>settingsStore: mergedConfig
+    alt serverStore.props has webuiSettings
+        settingsStore->>settingsStore: Apply webuiSettings from server
+        Note right of settingsStore: Server-provided UI settings<br/>(e.g. showRawOutputSwitch)
+    end
 
-    settingsStore->>settingsStore: config = mergedConfig
     settingsStore->>settingsStore: saveConfig()
     deactivate settingsStore
 
@@ -67,11 +73,18 @@ sequenceDiagram
     UI->>settingsStore: updateConfig(key, value)
     activate settingsStore
     settingsStore->>settingsStore: config[key] = value
-    settingsStore->>settingsStore: userOverrides.add(key)
-    Note right of settingsStore: Mark as user-modified (won't be overwritten by server)
+
+    alt value matches server default for key
+        settingsStore->>settingsStore: userOverrides.delete(key)
+        Note right of settingsStore: Matches server default, remove override
+    else value differs from server default
+        settingsStore->>settingsStore: userOverrides.add(key)
+        Note right of settingsStore: Mark as user-modified (won't be overwritten)
+    end
+
     settingsStore->>settingsStore: saveConfig()
-    settingsStore->>LS: set("llama-config", config)
-    settingsStore->>LS: set("llama-userOverrides", [...userOverrides])
+    settingsStore->>LS: set(CONFIG_LOCALSTORAGE_KEY, config)
+    settingsStore->>LS: set(USER_OVERRIDES_LOCALSTORAGE_KEY, [...userOverrides])
     deactivate settingsStore
 
     UI->>settingsStore: updateMultipleConfig({key1: val1, key2: val2})
@@ -88,10 +101,9 @@ sequenceDiagram
 
     UI->>settingsStore: resetConfig()
     activate settingsStore
-    settingsStore->>settingsStore: config = SETTING_CONFIG_DEFAULT
+    settingsStore->>settingsStore: config = {...SETTING_CONFIG_DEFAULT}
     settingsStore->>settingsStore: userOverrides.clear()
-    settingsStore->>settingsStore: syncWithServerDefaults()
-    Note right of settingsStore: Apply server defaults for syncable params
+    Note right of settingsStore: All params reset to defaults<br/>Next syncWithServerDefaults will adopt server values
     settingsStore->>settingsStore: saveConfig()
     deactivate settingsStore
 

tools/server/webui/src/lib/components/app/actions/ActionIconsCodeBlock.svelte

@@ -1,6 +1,6 @@
 <script lang="ts">
 	import { Eye } from '@lucide/svelte';
-	import ActionIconCopyToClipboard from '$lib/components/app/actions/ActionIconCopyToClipboard.svelte';
+	import { ActionIconCopyToClipboard } from '$lib/components/app';
 	import { FileTypeText } from '$lib/enums';
 
 	interface Props {

tools/server/webui/src/lib/components/app/chat/ChatForm/ChatForm.svelte

@@ -57,13 +57,13 @@
 	let currentConfig = $derived(config());
 	let fileInputRef: ChatFormFileInputInvisible | undefined = $state(undefined);
 	let isRecording = $state(false);
-	let message = $state(initialMessage);
+	let message = $derived(initialMessage);
 	let pasteLongTextToFileLength = $derived.by(() => {
 		const n = Number(currentConfig.pasteLongTextToFileLen);
 		return Number.isNaN(n) ? Number(SETTING_CONFIG_DEFAULT.pasteLongTextToFileLen) : n;
 	});
-	let previousIsLoading = $state(isLoading);
-	let previousInitialMessage = $state(initialMessage);
+	let previousIsLoading = $derived(isLoading);
+	let previousInitialMessage = $derived(initialMessage);
 	let recordingSupported = $state(false);
 	let textareaRef: ChatFormTextarea | undefined = $state(undefined);
 
@@ -289,7 +289,7 @@
 
 <form
 	onsubmit={handleSubmit}
-	class="{INPUT_CLASSES} border-radius-bottom-none mx-auto max-w-[48rem] overflow-hidden rounded-3xl backdrop-blur-md {disabled
+	class="relative {INPUT_CLASSES} border-radius-bottom-none mx-auto max-w-[48rem] overflow-hidden rounded-3xl backdrop-blur-md {disabled
 		? 'cursor-not-allowed opacity-60'
 		: ''} {className}"
 	data-slot="chat-form"
@@ -304,10 +304,11 @@
 	/>
 
 	<div
-		class="flex-column relative min-h-[48px] items-center rounded-3xl px-5 py-3 shadow-sm transition-all focus-within:shadow-md"
+		class="flex-column relative min-h-[48px] items-center rounded-3xl py-2 pb-2.25 shadow-sm transition-all focus-within:shadow-md md:!py-3"
 		onpaste={handlePaste}
 	>
 		<ChatFormTextarea
+			class="px-5 py-1.5 md:pt-0"
 			bind:this={textareaRef}
 			bind:value={message}
 			onKeydown={handleKeydown}
@@ -315,6 +316,7 @@
 		/>
 
 		<ChatFormActions
+			class="px-3"
 			bind:this={chatFormActionsRef}
 			canSend={message.trim().length > 0 || uploadedFiles.length > 0}
 			hasText={message.trim().length > 0}

tools/server/webui/src/lib/components/app/chat/ChatForm/ChatFormActions/ChatFormActionAttachmentsDropdown.svelte

@@ -0,0 +1,189 @@
+<script lang="ts">
+	import { page } from '$app/state';
+	import { MessageSquare, Plus } from '@lucide/svelte';
+	import { Button } from '$lib/components/ui/button';
+	import * as DropdownMenu from '$lib/components/ui/dropdown-menu';
+	import * as Tooltip from '$lib/components/ui/tooltip';
+	import { FILE_TYPE_ICONS } from '$lib/constants/icons';
+	import { TOOLTIP_DELAY_DURATION } from '$lib/constants/tooltip-config';
+
+	interface Props {
+		class?: string;
+		disabled?: boolean;
+		hasAudioModality?: boolean;
+		hasVisionModality?: boolean;
+		onFileUpload?: () => void;
+		onSystemPromptClick?: () => void;
+	}
+
+	type AttachmentActionId = 'images' | 'audio' | 'text' | 'pdf' | 'system';
+
+	interface AttachmentAction {
+		id: AttachmentActionId;
+		label: string;
+		disabled?: boolean;
+		disabledReason?: string;
+		tooltip?: string;
+	}
+
+	let {
+		class: className = '',
+		disabled = false,
+		hasAudioModality = false,
+		hasVisionModality = false,
+		onFileUpload,
+		onSystemPromptClick
+	}: Props = $props();
+
+	let isNewChat = $derived(!page.params.id);
+	let systemMessageTooltip = $derived(
+		isNewChat
+			? 'Add custom system message for a new conversation'
+			: 'Inject custom system message at the beginning of the conversation'
+	);
+
+	let actions = $derived.by<AttachmentAction[]>(() => [
+		{
+			id: 'images',
+			label: 'Images',
+			disabled: !hasVisionModality,
+			disabledReason: !hasVisionModality
+				? 'Images require vision models to be processed'
+				: undefined
+		},
+		{
+			id: 'audio',
+			label: 'Audio Files',
+			disabled: !hasAudioModality,
+			disabledReason: !hasAudioModality
+				? 'Audio files require audio models to be processed'
+				: undefined
+		},
+		{
+			id: 'text',
+			label: 'Text Files'
+		},
+		{
+			id: 'pdf',
+			label: 'PDF Files',
+			tooltip: !hasVisionModality
+				? 'PDFs will be converted to text. Image-based PDFs may not work properly.'
+				: undefined
+		},
+		{
+			id: 'system',
+			label: 'System Message',
+			tooltip: systemMessageTooltip
+		}
+	]);
+
+	function handleActionClick(id: AttachmentActionId) {
+		if (id === 'system') {
+			onSystemPromptClick?.();
+			return;
+		}
+
+		onFileUpload?.();
+	}
+
+	const triggerTooltipText = 'Add files or system message';
+	const itemClass = 'flex cursor-pointer items-center gap-2';
+</script>
+
+<div class="flex items-center gap-1 {className}">
+	<DropdownMenu.Root>
+		<DropdownMenu.Trigger name="Attach files" {disabled}>
+			<Tooltip.Root>
+				<Tooltip.Trigger class="w-full">
+					<Button
+						class="file-upload-button h-8 w-8 rounded-full p-0"
+						{disabled}
+						variant="secondary"
+						type="button"
+					>
+						<span class="sr-only">{triggerTooltipText}</span>
+
+						<Plus class="h-4 w-4" />
+					</Button>
+				</Tooltip.Trigger>
+
+				<Tooltip.Content>
+					<p>{triggerTooltipText}</p>
+				</Tooltip.Content>
+			</Tooltip.Root>
+		</DropdownMenu.Trigger>
+
+		<DropdownMenu.Content align="start" class="w-56">
+			{#each actions as item (item.id)}
+				{@const hasDisabledTooltip = !!item.disabled && !!item.disabledReason}
+				{@const hasEnabledTooltip = !item.disabled && !!item.tooltip}
+
+				{#if hasDisabledTooltip}
+					<Tooltip.Root delayDuration={TOOLTIP_DELAY_DURATION}>
+						<Tooltip.Trigger class="w-full">
+							<DropdownMenu.Item class={itemClass} disabled>
+								{#if item.id === 'images'}
+									<FILE_TYPE_ICONS.image class="h-4 w-4" />
+								{:else if item.id === 'audio'}
+									<FILE_TYPE_ICONS.audio class="h-4 w-4" />
+								{:else if item.id === 'text'}
+									<FILE_TYPE_ICONS.text class="h-4 w-4" />
+								{:else if item.id === 'pdf'}
+									<FILE_TYPE_ICONS.pdf class="h-4 w-4" />
+								{:else}
+									<MessageSquare class="h-4 w-4" />
+								{/if}
+
+								<span>{item.label}</span>
+							</DropdownMenu.Item>
+						</Tooltip.Trigger>
+
+						<Tooltip.Content side="right">
+							<p>{item.disabledReason}</p>
+						</Tooltip.Content>
+					</Tooltip.Root>
+				{:else if hasEnabledTooltip}
+					<Tooltip.Root delayDuration={TOOLTIP_DELAY_DURATION}>
+						<Tooltip.Trigger class="w-full">
+							<DropdownMenu.Item class={itemClass} onclick={() => handleActionClick(item.id)}>
+								{#if item.id === 'images'}
+									<FILE_TYPE_ICONS.image class="h-4 w-4" />
+								{:else if item.id === 'audio'}
+									<FILE_TYPE_ICONS.audio class="h-4 w-4" />
+								{:else if item.id === 'text'}
+									<FILE_TYPE_ICONS.text class="h-4 w-4" />
+								{:else if item.id === 'pdf'}
+									<FILE_TYPE_ICONS.pdf class="h-4 w-4" />
+								{:else}
+									<MessageSquare class="h-4 w-4" />
+								{/if}
+
+								<span>{item.label}</span>
+							</DropdownMenu.Item>
+						</Tooltip.Trigger>
+
+						<Tooltip.Content side="right">
+							<p>{item.tooltip}</p>
+						</Tooltip.Content>
+					</Tooltip.Root>
+				{:else}
+					<DropdownMenu.Item class={itemClass} onclick={() => handleActionClick(item.id)}>
+						{#if item.id === 'images'}
+							<FILE_TYPE_ICONS.image class="h-4 w-4" />
+						{:else if item.id === 'audio'}
+							<FILE_TYPE_ICONS.audio class="h-4 w-4" />
+						{:else if item.id === 'text'}
+							<FILE_TYPE_ICONS.text class="h-4 w-4" />
+						{:else if item.id === 'pdf'}
+							<FILE_TYPE_ICONS.pdf class="h-4 w-4" />
+						{:else}
+							<MessageSquare class="h-4 w-4" />
+						{/if}
+
+						<span>{item.label}</span>
+					</DropdownMenu.Item>
+				{/if}
+			{/each}
+		</DropdownMenu.Content>
+	</DropdownMenu.Root>
+</div>

tools/server/webui/src/lib/components/app/chat/ChatForm/ChatFormActions/ChatFormActions.svelte

@@ -2,7 +2,7 @@
 	import { Square } from '@lucide/svelte';
 	import { Button } from '$lib/components/ui/button';
 	import {
-		ChatFormActionFileAttachments,
+		ChatFormActionAttachmentsDropdown,
 		ChatFormActionRecord,
 		ChatFormActionSubmit,
 		ModelsSelector
@@ -157,7 +157,7 @@
 
 	const { handleModelChange } = useModelChangeValidation({
 		getRequiredModalities: () => usedModalities(),
-		onValidationFailure: async (previousModelId) => {
+		onValidationFailure: async (previousModelId: string | null) => {
 			if (previousModelId) {
 				await modelsStore.selectModelById(previousModelId);
 			}
@@ -166,32 +166,39 @@
 </script>
 
 <div class="flex w-full items-center gap-3 {className}" style="container-type: inline-size">
-	<ChatFormActionFileAttachments
-		class="mr-auto"
-		{disabled}
-		{hasAudioModality}
-		{hasVisionModality}
-		{onFileUpload}
-		{onSystemPromptClick}
-	/>
+	<div class="mr-auto flex items-center gap-2">
+		<ChatFormActionAttachmentsDropdown
+			{disabled}
+			{hasAudioModality}
+			{hasVisionModality}
+			{onFileUpload}
+			{onSystemPromptClick}
+		/>
+	</div>
 
-	<ModelsSelector
-		{disabled}
-		bind:this={selectorModelRef}
-		currentModel={conversationModel}
-		forceForegroundText={true}
-		useGlobalSelection={true}
-		onModelChange={handleModelChange}
-	/>
+	<div class="ml-auto flex items-center gap-1.5">
+		<ModelsSelector
+			{disabled}
+			bind:this={selectorModelRef}
+			currentModel={conversationModel}
+			forceForegroundText={true}
+			useGlobalSelection={true}
+			onModelChange={handleModelChange}
+		/>
+	</div>
 
 	{#if isLoading}
 		<Button
 			type="button"
+			variant="secondary"
 			onclick={onStop}
-			class="h-8 w-8 bg-transparent p-0 hover:bg-destructive/20"
+			class="group h-8 w-8 rounded-full p-0 hover:bg-destructive/10!"
 		>
 			<span class="sr-only">Stop</span>
-			<Square class="h-8 w-8 fill-destructive stroke-destructive" />
+
+			<Square
+				class="h-8 w-8 fill-muted-foreground stroke-muted-foreground group-hover:fill-destructive group-hover:stroke-destructive hover:fill-destructive hover:stroke-destructive"
+			/>
 		</Button>
 	{:else if shouldShowRecordButton}
 		<ChatFormActionRecord {disabled} {hasAudioModality} {isLoading} {isRecording} {onMicClick} />

tools/server/webui/src/lib/components/app/chat/ChatMessages/ChatMessage.svelte

@@ -62,8 +62,8 @@
 		assistantMessages: number;
 		messageTypes: string[];
 	} | null>(null);
-	let editedContent = $state(message.content);
-	let editedExtras = $state<DatabaseMessageExtra[]>(message.extra ? [...message.extra] : []);
+	let editedContent = $derived(message.content);
+	let editedExtras = $derived<DatabaseMessageExtra[]>(message.extra ? [...message.extra] : []);
 	let editedUploadedFiles = $state<ChatUploadedFile[]>([]);
 	let isEditing = $state(false);
 	let showDeleteDialog = $state(false);

tools/server/webui/src/lib/components/app/chat/ChatMessages/ChatMessageAssistant.svelte

@@ -105,7 +105,7 @@
 
 	const { handleModelChange } = useModelChangeValidation({
 		getRequiredModalities: () => conversationsStore.getModalitiesUpToMessage(message.id),
-		onSuccess: (modelName) => onRegenerate(modelName)
+		onSuccess: (modelName: string) => onRegenerate(modelName)
 	});
 
 	function handleCopyModel() {

tools/server/webui/src/lib/components/app/chat/ChatMessages/ChatMessageEditForm.svelte

@@ -133,7 +133,7 @@
 
 	const { handleModelChange } = useModelChangeValidation({
 		getRequiredModalities,
-		onValidationFailure: async (previousModelId) => {
+		onValidationFailure: async (previousModelId: string | null) => {
 			if (previousModelId) {
 				await modelsStore.selectModelById(previousModelId);
 			}

tools/server/webui/src/lib/components/app/chat/ChatMessages/ChatMessageStatistics.svelte

@@ -28,7 +28,7 @@
 		initialView = ChatMessageStatsView.GENERATION
 	}: Props = $props();
 
-	let activeView: ChatMessageStatsView = $state(initialView);
+	let activeView: ChatMessageStatsView = $derived(initialView);
 	let hasAutoSwitchedToGeneration = $state(false);
 
 	// In live mode: auto-switch to GENERATION tab when prompt processing completes

tools/server/webui/src/lib/components/app/chat/ChatScreen/ChatScreen.svelte

@@ -35,6 +35,7 @@
 	import { modelsStore, modelOptions, selectedModelId } from '$lib/stores/models.svelte';
 	import { isFileTypeSupported, filterFilesByModalities } from '$lib/utils';
 	import { parseFilesToMessageExtras, processFilesToChatUploaded } from '$lib/utils/browser-only';
+	import { ErrorDialogType } from '$lib/enums';
 	import { onMount } from 'svelte';
 	import { fade, fly, slide } from 'svelte/transition';
 	import { Trash2, AlertTriangle, RefreshCw } from '@lucide/svelte';
@@ -616,7 +617,7 @@
 	contextInfo={activeErrorDialog?.contextInfo}
 	onOpenChange={handleErrorDialogOpenChange}
 	open={Boolean(activeErrorDialog)}
-	type={activeErrorDialog?.type ?? 'server'}
+	type={(activeErrorDialog?.type as ErrorDialogType) ?? ErrorDialogType.SERVER}
 />
 
 <style>

tools/server/webui/src/lib/components/app/chat/ChatScreen/ChatScreenForm.svelte

@@ -0,0 +1,47 @@
+<script lang="ts">
+	import ChatForm from '$lib/components/app/chat/ChatForm/ChatForm.svelte';
+
+	interface Props {
+		class?: string;
+		disabled?: boolean;
+		initialMessage?: string;
+		isLoading?: boolean;
+		onFileRemove?: (fileId: string) => void;
+		onFileUpload?: (files: File[]) => void;
+		onSend?: (message: string, files?: ChatUploadedFile[]) => Promise<boolean>;
+		onStop?: () => void;
+		onSystemPromptAdd?: (draft: { message: string; files: ChatUploadedFile[] }) => void;
+		showHelperText?: boolean;
+		uploadedFiles?: ChatUploadedFile[];
+	}
+
+	let {
+		class: className,
+		disabled = false,
+		initialMessage = '',
+		isLoading = false,
+		onFileRemove,
+		onFileUpload,
+		onSend,
+		onStop,
+		onSystemPromptAdd,
+		showHelperText = true,
+		uploadedFiles = $bindable([])
+	}: Props = $props();
+</script>
+
+<div class="relative mx-auto max-w-[48rem]">
+	<ChatForm
+		class={className}
+		{disabled}
+		{initialMessage}
+		{isLoading}
+		{onFileRemove}
+		{onFileUpload}
+		{onSend}
+		{onStop}
+		{onSystemPromptAdd}
+		{showHelperText}
+		bind:uploadedFiles
+	/>
+</div>

tools/server/webui/src/lib/components/app/chat/ChatSettings/ChatSettings.svelte

@@ -18,19 +18,24 @@
 	} from '$lib/components/app';
 	import { ScrollArea } from '$lib/components/ui/scroll-area';
 	import { config, settingsStore } from '$lib/stores/settings.svelte';
+	import {
+		SETTINGS_SECTION_TITLES,
+		type SettingsSectionTitle
+	} from '$lib/constants/settings-sections';
 	import { setMode } from 'mode-watcher';
 	import type { Component } from 'svelte';
 
 	interface Props {
 		onSave?: () => void;
+		initialSection?: SettingsSectionTitle;
 	}
 
-	let { onSave }: Props = $props();
+	let { onSave, initialSection }: Props = $props();
 
 	const settingSections: Array<{
 		fields: SettingsFieldConfig[];
 		icon: Component;
-		title: string;
+		title: SettingsSectionTitle;
 	}> = [
 		{
 			title: 'General',
@@ -285,7 +290,9 @@
 		// }
 	];
 
-	let activeSection = $state('General');
+	let activeSection = $derived<SettingsSectionTitle>(
+		initialSection ?? SETTINGS_SECTION_TITLES.GENERAL
+	);
 	let currentSection = $derived(
 		settingSections.find((section) => section.title === activeSection) || settingSections[0]
 	);
@@ -295,6 +302,16 @@
 	let canScrollRight = $state(false);
 	let scrollContainer: HTMLDivElement | undefined = $state();
 
+	$effect(() => {
+		if (!initialSection) {
+			return;
+		}
+
+		if (settingSections.some((section) => section.title === initialSection)) {
+			activeSection = initialSection;
+		}
+	});
+
 	function handleThemeChange(newTheme: string) {
 		localConfig.theme = newTheme;
 

tools/server/webui/src/lib/components/app/chat/ChatSidebar/ChatSidebarConversationItem.svelte

@@ -142,7 +142,7 @@
 					{
 						icon: Download,
 						label: 'Export',
-						onclick: (e) => {
+						onclick: (e: Event) => {
 							e.stopPropagation();
 							conversationsStore.downloadConversation(conversation.id);
 						},

tools/server/webui/src/lib/components/app/content/MarkdownContent.svelte

@@ -15,6 +15,7 @@
 	import { rehypeRestoreTableHtml } from '$lib/markdown/table-html-restorer';
 	import { rehypeEnhanceLinks } from '$lib/markdown/enhance-links';
 	import { rehypeEnhanceCodeBlocks } from '$lib/markdown/enhance-code-blocks';
+	import { rehypeResolveAttachmentImages } from '$lib/markdown/resolve-attachment-images';
 	import { remarkLiteralHtml } from '$lib/markdown/literal-html';
 	import { copyCodeToClipboard, preprocessLaTeX, getImageErrorFallbackHtml } from '$lib/utils';
 	import {
@@ -23,6 +24,7 @@
 		DATA_ERROR_HANDLED_ATTR,
 		BOOL_TRUE_STRING
 	} from '$lib/constants/markdown';
+	import { UrlPrefix } from '$lib/enums';
 	import { FileTypeText } from '$lib/enums/files';
 	import {
 		highlightCode,
@@ -33,8 +35,7 @@
 	import githubDarkCss from 'highlight.js/styles/github-dark.css?inline';
 	import githubLightCss from 'highlight.js/styles/github.css?inline';
 	import { mode } from 'mode-watcher';
-	import ActionIconsCodeBlock from '$lib/components/app/actions/ActionIconsCodeBlock.svelte';
-	import DialogCodePreview from '$lib/components/app/misc/CodePreviewDialog.svelte';
+	import { ActionIconsCodeBlock, DialogCodePreview } from '$lib/components/app';
 	import { createAutoScrollController } from '$lib/hooks/use-auto-scroll.svelte';
 	import type { DatabaseMessageExtra } from '$lib/types/database';
 
@@ -100,6 +101,7 @@
 			.use(rehypeRestoreTableHtml) // Restore limited HTML (e.g., <br>, <ul>) inside Markdown tables
 			.use(rehypeEnhanceLinks) // Add target="_blank" to links
 			.use(rehypeEnhanceCodeBlocks) // Wrap code blocks with header and actions
+			.use(rehypeResolveAttachmentImages, { attachments })
 			.use(rehypeStringify, { allowDangerousHtml: true }); // Convert to HTML string
 	});
 
@@ -500,7 +502,10 @@
 		if (!img || !img.src) return;
 
 		// Don't handle data URLs or already-handled images
-		if (img.src.startsWith('data:') || img.dataset[DATA_ERROR_HANDLED_ATTR] === BOOL_TRUE_STRING)
+		if (
+			img.src.startsWith(UrlPrefix.DATA) ||
+			img.dataset[DATA_ERROR_HANDLED_ATTR] === BOOL_TRUE_STRING
+		)
 			return;
 		img.dataset[DATA_ERROR_HANDLED_ATTR] = BOOL_TRUE_STRING;
 

tools/server/webui/src/lib/components/app/dialogs/DialogChatError.svelte

@@ -1,10 +1,11 @@
 <script lang="ts">
 	import * as AlertDialog from '$lib/components/ui/alert-dialog';
 	import { AlertTriangle, TimerOff } from '@lucide/svelte';
+	import { ErrorDialogType } from '$lib/enums';
 
 	interface Props {
 		open: boolean;
-		type: 'timeout' | 'server';
+		type: ErrorDialogType;
 		message: string;
 		contextInfo?: { n_prompt_tokens: number; n_ctx: number };
 		onOpenChange?: (open: boolean) => void;
@@ -12,7 +13,7 @@
 
 	let { open = $bindable(), type, message, contextInfo, onOpenChange }: Props = $props();
 
-	const isTimeout = $derived(type === 'timeout');
+	const isTimeout = $derived(type === ErrorDialogType.TIMEOUT);
 	const title = $derived(isTimeout ? 'TCP Timeout' : 'Server Error');
 	const description = $derived(
 		isTimeout
@@ -58,7 +59,12 @@
 						<span class="font-medium">Prompt tokens:</span>
 						{contextInfo.n_prompt_tokens.toLocaleString()}
 					</p>
-					<p><span class="font-medium">Context size:</span> {contextInfo.n_ctx.toLocaleString()}</p>
+					{#if contextInfo.n_ctx}
+						<p>
+							<span class="font-medium">Context size:</span>
+							{contextInfo.n_ctx.toLocaleString()}
+						</p>
+					{/if}
 				</div>
 			{/if}
 		</div>

tools/server/webui/src/lib/components/app/dialogs/DialogChatSettings.svelte

@@ -1,13 +1,15 @@
 <script lang="ts">
 	import * as Dialog from '$lib/components/ui/dialog';
 	import { ChatSettings } from '$lib/components/app';
+	import type { SettingsSectionTitle } from '$lib/constants/settings-sections';
 
 	interface Props {
 		onOpenChange?: (open: boolean) => void;
 		open?: boolean;
+		initialSection?: SettingsSectionTitle;
 	}
 
-	let { onOpenChange, open = false }: Props = $props();
+	let { onOpenChange, open = false, initialSection }: Props = $props();
 
 	let chatSettingsRef: ChatSettings | undefined = $state();
 
@@ -28,10 +30,9 @@
 
 <Dialog.Root {open} onOpenChange={handleClose}>
 	<Dialog.Content
-		class="z-999999 flex h-[100dvh] max-h-[100dvh] min-h-[100dvh] flex-col gap-0 rounded-none p-0
-			md:h-[64vh] md:max-h-[64vh] md:min-h-0 md:rounded-lg"
-		style="max-width: 48rem;"
+		class="z-999999 flex h-[100dvh] max-h-[100dvh] min-h-[100dvh] max-w-4xl! flex-col gap-0 rounded-none
+			p-0 md:h-[64vh] md:max-h-[64vh] md:min-h-0 md:rounded-lg"
 	>
-		<ChatSettings bind:this={chatSettingsRef} onSave={handleSave} />
+		<ChatSettings bind:this={chatSettingsRef} onSave={handleSave} {initialSection} />
 	</Dialog.Content>
 </Dialog.Root>

tools/server/webui/src/lib/components/app/dialogs/DialogCodePreview.svelte

@@ -37,7 +37,7 @@
 			<iframe
 				bind:this={iframeRef}
 				title="Preview {language}"
-				sandbox="allow-scripts"
+				sandbox="allow-scripts allow-same-origin"
 				class="code-preview-iframe"
 			></iframe>
 

tools/server/webui/src/lib/components/app/dialogs/DialogConfirmation.svelte

@@ -1,6 +1,7 @@
 <script lang="ts">
 	import * as AlertDialog from '$lib/components/ui/alert-dialog';
 	import type { Component } from 'svelte';
+	import { KeyboardKey } from '$lib/enums';
 
 	interface Props {
 		open: boolean;
@@ -29,7 +30,7 @@
 	}: Props = $props();
 
 	function handleKeydown(event: KeyboardEvent) {
-		if (event.key === 'Enter') {
+		if (event.key === KeyboardKey.ENTER) {
 			event.preventDefault();
 			onConfirm();
 		}

tools/server/webui/src/lib/components/app/dialogs/DialogModelInformation.svelte

@@ -1,7 +1,7 @@
 <script lang="ts">
 	import * as Dialog from '$lib/components/ui/dialog';
 	import * as Table from '$lib/components/ui/table';
-	import { BadgeModality, CopyToClipboardIcon } from '$lib/components/app';
+	import { BadgeModality, ActionIconCopyToClipboard } from '$lib/components/app';
 	import { serverStore } from '$lib/stores/server.svelte';
 	import { modelsStore, modelOptions, modelsLoading } from '$lib/stores/models.svelte';
 	import { formatFileSize, formatParameters, formatNumber } from '$lib/utils';
@@ -47,6 +47,7 @@
 
 		<Dialog.Header>
 			<Dialog.Title>Model Information</Dialog.Title>
+
 			<Dialog.Description>Current model details and capabilities</Dialog.Description>
 		</Dialog.Header>
 
@@ -73,7 +74,7 @@
 											{modelName}
 										</span>
 
-										<CopyToClipboardIcon
+										<ActionIconCopyToClipboard
 											text={modelName || ''}
 											canCopy={!!modelName}
 											ariaLabel="Copy model name to clipboard"
@@ -97,7 +98,7 @@
 										{serverProps.model_path}
 									</span>
 
-									<CopyToClipboardIcon
+									<ActionIconCopyToClipboard
 										text={serverProps.model_path}
 										ariaLabel="Copy model path to clipboard"
 									/>
@@ -105,17 +106,29 @@
 							</Table.Row>
 
 							<!-- Context Size -->
-							<Table.Row>
-								<Table.Cell class="h-10 align-middle font-medium">Context Size</Table.Cell>
-								<Table.Cell
-									>{formatNumber(serverProps.default_generation_settings.n_ctx)} tokens</Table.Cell
-								>
-							</Table.Row>
+							{#if serverProps?.default_generation_settings?.n_ctx}
+								<Table.Row>
+									<Table.Cell class="h-10 align-middle font-medium">Context Size</Table.Cell>
+
+									<Table.Cell
+										>{formatNumber(serverProps.default_generation_settings.n_ctx)} tokens</Table.Cell
+									>
+								</Table.Row>
+							{:else}
+								<Table.Row>
+									<Table.Cell class="h-10 align-middle font-medium text-red-500"
+										>Context Size</Table.Cell
+									>
+
+									<Table.Cell class="text-red-500">Not available</Table.Cell>
+								</Table.Row>
+							{/if}
 
 							<!-- Training Context -->
 							{#if modelMeta?.n_ctx_train}
 								<Table.Row>
 									<Table.Cell class="h-10 align-middle font-medium">Training Context</Table.Cell>
+
 									<Table.Cell>{formatNumber(modelMeta.n_ctx_train)} tokens</Table.Cell>
 								</Table.Row>
 							{/if}
@@ -124,6 +137,7 @@
 							{#if modelMeta?.size}
 								<Table.Row>
 									<Table.Cell class="h-10 align-middle font-medium">Model Size</Table.Cell>
+
 									<Table.Cell>{formatFileSize(modelMeta.size)}</Table.Cell>
 								</Table.Row>
 							{/if}
@@ -132,6 +146,7 @@
 							{#if modelMeta?.n_params}
 								<Table.Row>
 									<Table.Cell class="h-10 align-middle font-medium">Parameters</Table.Cell>
+
 									<Table.Cell>{formatParameters(modelMeta.n_params)}</Table.Cell>
 								</Table.Row>
 							{/if}
@@ -140,6 +155,7 @@
 							{#if modelMeta?.n_embd}
 								<Table.Row>
 									<Table.Cell class="align-middle font-medium">Embedding Size</Table.Cell>
+
 									<Table.Cell>{formatNumber(modelMeta.n_embd)}</Table.Cell>
 								</Table.Row>
 							{/if}
@@ -148,6 +164,7 @@
 							{#if modelMeta?.n_vocab}
 								<Table.Row>
 									<Table.Cell class="align-middle font-medium">Vocabulary Size</Table.Cell>
+
 									<Table.Cell>{formatNumber(modelMeta.n_vocab)} tokens</Table.Cell>
 								</Table.Row>
 							{/if}
@@ -163,6 +180,7 @@
 							<!-- Total Slots -->
 							<Table.Row>
 								<Table.Cell class="align-middle font-medium">Parallel Slots</Table.Cell>
+
 								<Table.Cell>{serverProps.total_slots}</Table.Cell>
 							</Table.Row>
 
@@ -170,6 +188,7 @@
 							{#if modalities.length > 0}
 								<Table.Row>
 									<Table.Cell class="align-middle font-medium">Modalities</Table.Cell>
+
 									<Table.Cell>
 										<div class="flex flex-wrap gap-1">
 											<BadgeModality {modalities} />
@@ -181,6 +200,7 @@
 							<!-- Build Info -->
 							<Table.Row>
 								<Table.Cell class="align-middle font-medium">Build Info</Table.Cell>
+
 								<Table.Cell class="align-middle font-mono text-xs"
 									>{serverProps.build_info}</Table.Cell
 								>
@@ -190,6 +210,7 @@
 							{#if serverProps.chat_template}
 								<Table.Row>
 									<Table.Cell class="align-middle font-medium">Chat Template</Table.Cell>
+
 									<Table.Cell class="py-10">
 										<div class="max-h-120 overflow-y-auto rounded-md bg-muted p-4">
 											<pre

tools/server/webui/src/lib/components/app/forms/KeyValuePairs.svelte

@@ -0,0 +1,110 @@
+<script lang="ts">
+	import { Plus, Trash2 } from '@lucide/svelte';
+	import { Input } from '$lib/components/ui/input';
+	import { autoResizeTextarea } from '$lib/utils';
+	import type { KeyValuePair } from '$lib/types';
+
+	interface Props {
+		class?: string;
+		pairs: KeyValuePair[];
+		onPairsChange: (pairs: KeyValuePair[]) => void;
+		keyPlaceholder?: string;
+		valuePlaceholder?: string;
+		addButtonLabel?: string;
+		emptyMessage?: string;
+		sectionLabel?: string;
+		sectionLabelOptional?: boolean;
+	}
+
+	let {
+		class: className = '',
+		pairs,
+		onPairsChange,
+		keyPlaceholder = 'Key',
+		valuePlaceholder = 'Value',
+		addButtonLabel = 'Add',
+		emptyMessage = 'No items configured.',
+		sectionLabel,
+		sectionLabelOptional = true
+	}: Props = $props();
+
+	function addPair() {
+		onPairsChange([...pairs, { key: '', value: '' }]);
+	}
+
+	function removePair(index: number) {
+		onPairsChange(pairs.filter((_, i) => i !== index));
+	}
+
+	function updatePairKey(index: number, key: string) {
+		const newPairs = [...pairs];
+		newPairs[index] = { ...newPairs[index], key };
+		onPairsChange(newPairs);
+	}
+
+	function updatePairValue(index: number, value: string) {
+		const newPairs = [...pairs];
+		newPairs[index] = { ...newPairs[index], value };
+		onPairsChange(newPairs);
+	}
+</script>
+
+<div class={className}>
+	<div class="mb-2 flex items-center justify-between">
+		{#if sectionLabel}
+			<span class="text-xs font-medium">
+				{sectionLabel}
+				{#if sectionLabelOptional}
+					<span class="text-muted-foreground">(optional)</span>
+				{/if}
+			</span>
+		{/if}
+
+		<button
+			type="button"
+			class="inline-flex cursor-pointer items-center gap-1 rounded-md px-1.5 py-1 text-xs text-muted-foreground hover:bg-muted hover:text-foreground"
+			onclick={addPair}
+		>
+			<Plus class="h-3 w-3" />
+			{addButtonLabel}
+		</button>
+	</div>
+	{#if pairs.length > 0}
+		<div class="space-y-3">
+			{#each pairs as pair, index (index)}
+				<div class="flex items-start gap-2">
+					<Input
+						type="text"
+						placeholder={keyPlaceholder}
+						value={pair.key}
+						oninput={(e) => updatePairKey(index, e.currentTarget.value)}
+						class="flex-1"
+					/>
+
+					<textarea
+						use:autoResizeTextarea
+						placeholder={valuePlaceholder}
+						value={pair.value}
+						oninput={(e) => {
+							updatePairValue(index, e.currentTarget.value);
+							autoResizeTextarea(e.currentTarget);
+						}}
+						class="flex-1 resize-none rounded-md border border-input bg-transparent px-3 py-2 text-sm leading-5 placeholder:text-muted-foreground focus-visible:ring-1 focus-visible:ring-ring focus-visible:outline-none"
+						rows="1"
+					></textarea>
+
+					<button
+						type="button"
+						class="mt-1.5 shrink-0 cursor-pointer rounded-md p-1 text-muted-foreground hover:bg-destructive/10 hover:text-destructive"
+						onclick={() => removePair(index)}
+						aria-label="Remove item"
+					>
+						<Trash2 class="h-3.5 w-3.5" />
+					</button>
+				</div>
+			{/each}
+		</div>
+	{:else}
+		<p class="text-xs text-muted-foreground">{emptyMessage}</p>
+	{/if}
+</div>

tools/server/webui/src/lib/components/app/forms/SearchInput.svelte

@@ -46,7 +46,7 @@
 
 <div class="relative {className}">
 	<Search
-		class="absolute top-1/2 left-3 h-4 w-4 -translate-y-1/2 transform text-muted-foreground"
+		class="absolute top-1/2 left-3 z-10 h-4 w-4 -translate-y-1/2 transform text-muted-foreground"
 	/>
 
 	<Input

tools/server/webui/src/lib/components/app/forms/index.ts

@@ -0,0 +1,30 @@
+/**
+ *
+ * FORMS & INPUTS
+ *
+ * Form-related utility components.
+ *
+ */
+
+/**
+ * **SearchInput** - Search field with clear button
+ *
+ * Input field optimized for search with clear button and keyboard handling.
+ * Supports placeholder, autofocus, and change callbacks.
+ */
+export { default as SearchInput } from './SearchInput.svelte';
+
+/**
+ * **KeyValuePairs** - Editable key-value list
+ *
+ * Dynamic list of key-value pairs with add/remove functionality.
+ * Used for HTTP headers, metadata, and configuration.
+ *
+ * **Features:**
+ * - Add new pairs with button
+ * - Remove individual pairs
+ * - Customizable placeholders and labels
+ * - Empty state message
+ * - Auto-resize value textarea
+ */
+export { default as KeyValuePairs } from './KeyValuePairs.svelte';

tools/server/webui/src/lib/components/app/index.ts

@@ -1,12 +1,20 @@
-// Chat
+export * from './actions';
+export * from './badges';
+export * from './content';
+export * from './forms';
+export * from './misc';
+export * from './models';
+export * from './navigation';
+export * from './server';
 
+// Chat
 export { default as ChatAttachmentPreview } from './chat/ChatAttachments/ChatAttachmentPreview.svelte';
 export { default as ChatAttachmentThumbnailFile } from './chat/ChatAttachments/ChatAttachmentThumbnailFile.svelte';
 export { default as ChatAttachmentThumbnailImage } from './chat/ChatAttachments/ChatAttachmentThumbnailImage.svelte';
 export { default as ChatAttachmentsList } from './chat/ChatAttachments/ChatAttachmentsList.svelte';
 export { default as ChatAttachmentsViewAll } from './chat/ChatAttachments/ChatAttachmentsViewAll.svelte';
-
 export { default as ChatForm } from './chat/ChatForm/ChatForm.svelte';
+export { default as ChatFormActionAttachmentsDropdown } from './chat/ChatForm/ChatFormActions/ChatFormActionAttachmentsDropdown.svelte';
 export { default as ChatFormActionFileAttachments } from './chat/ChatForm/ChatFormActions/ChatFormActionFileAttachments.svelte';
 export { default as ChatFormActionRecord } from './chat/ChatForm/ChatFormActions/ChatFormActionRecord.svelte';
 export { default as ChatFormActions } from './chat/ChatForm/ChatFormActions/ChatFormActions.svelte';
@@ -14,36 +22,38 @@ export { default as ChatFormActionSubmit } from './chat/ChatForm/ChatFormActions
 export { default as ChatFormFileInputInvisible } from './chat/ChatForm/ChatFormFileInputInvisible.svelte';
 export { default as ChatFormHelperText } from './chat/ChatForm/ChatFormHelperText.svelte';
 export { default as ChatFormTextarea } from './chat/ChatForm/ChatFormTextarea.svelte';
-
 export { default as ChatMessage } from './chat/ChatMessages/ChatMessage.svelte';
 export { default as ChatMessageActions } from './chat/ChatMessages/ChatMessageActions.svelte';
+export { default as ChatMessageAssistant } from './chat/ChatMessages/ChatMessageAssistant.svelte';
 export { default as ChatMessageBranchingControls } from './chat/ChatMessages/ChatMessageBranchingControls.svelte';
+export { default as ChatMessageEditForm } from './chat/ChatMessages/ChatMessageEditForm.svelte';
 export { default as ChatMessageStatistics } from './chat/ChatMessages/ChatMessageStatistics.svelte';
 export { default as ChatMessageSystem } from './chat/ChatMessages/ChatMessageSystem.svelte';
 export { default as ChatMessageThinkingBlock } from './chat/ChatMessages/ChatMessageThinkingBlock.svelte';
+export { default as ChatMessageUser } from './chat/ChatMessages/ChatMessageUser.svelte';
 export { default as ChatMessages } from './chat/ChatMessages/ChatMessages.svelte';
 export { default as MessageBranchingControls } from './chat/ChatMessages/ChatMessageBranchingControls.svelte';
-
 export { default as ChatScreen } from './chat/ChatScreen/ChatScreen.svelte';
+export { default as ChatScreenDragOverlay } from './chat/ChatScreen/ChatScreenDragOverlay.svelte';
+export { default as ChatScreenForm } from './chat/ChatScreen/ChatScreenForm.svelte';
 export { default as ChatScreenHeader } from './chat/ChatScreen/ChatScreenHeader.svelte';
 export { default as ChatScreenProcessingInfo } from './chat/ChatScreen/ChatScreenProcessingInfo.svelte';
-
 export { default as ChatSettings } from './chat/ChatSettings/ChatSettings.svelte';
 export { default as ChatSettingsFooter } from './chat/ChatSettings/ChatSettingsFooter.svelte';
 export { default as ChatSettingsFields } from './chat/ChatSettings/ChatSettingsFields.svelte';
 export { default as ChatSettingsImportExportTab } from './chat/ChatSettings/ChatSettingsImportExportTab.svelte';
 export { default as ChatSettingsParameterSourceIndicator } from './chat/ChatSettings/ChatSettingsParameterSourceIndicator.svelte';
-
 export { default as ChatSidebar } from './chat/ChatSidebar/ChatSidebar.svelte';
+export { default as ChatSidebarActions } from './chat/ChatSidebar/ChatSidebarActions.svelte';
 export { default as ChatSidebarConversationItem } from './chat/ChatSidebar/ChatSidebarConversationItem.svelte';
 export { default as ChatSidebarSearch } from './chat/ChatSidebar/ChatSidebarSearch.svelte';
 
 // Dialogs
-
 export { default as DialogChatAttachmentPreview } from './dialogs/DialogChatAttachmentPreview.svelte';
 export { default as DialogChatAttachmentsViewAll } from './dialogs/DialogChatAttachmentsViewAll.svelte';
 export { default as DialogChatError } from './dialogs/DialogChatError.svelte';
 export { default as DialogChatSettings } from './dialogs/DialogChatSettings.svelte';
+export { default as DialogCodePreview } from './dialogs/DialogCodePreview.svelte';
 export { default as DialogConfirmation } from './dialogs/DialogConfirmation.svelte';
 export { default as DialogConversationSelection } from './dialogs/DialogConversationSelection.svelte';
 export { default as DialogConversationTitleUpdate } from './dialogs/DialogConversationTitleUpdate.svelte';
@@ -51,25 +61,8 @@ export { default as DialogEmptyFileAlert } from './dialogs/DialogEmptyFileAlert.
 export { default as DialogModelInformation } from './dialogs/DialogModelInformation.svelte';
 export { default as DialogModelNotAvailable } from './dialogs/DialogModelNotAvailable.svelte';
 
-// Miscellanous
-
-export { default as ActionButton } from './misc/ActionButton.svelte';
-export { default as ActionDropdown } from './misc/ActionDropdown.svelte';
-export { default as BadgeChatStatistic } from './misc/BadgeChatStatistic.svelte';
-export { default as BadgeInfo } from './misc/BadgeInfo.svelte';
-export { default as ModelBadge } from './models/ModelBadge.svelte';
-export { default as BadgeModality } from './misc/BadgeModality.svelte';
-export { default as ConversationSelection } from './misc/ConversationSelection.svelte';
-export { default as CopyToClipboardIcon } from './misc/CopyToClipboardIcon.svelte';
-export { default as KeyboardShortcutInfo } from './misc/KeyboardShortcutInfo.svelte';
-export { default as MarkdownContent } from './misc/MarkdownContent.svelte';
-export { default as RemoveButton } from './misc/RemoveButton.svelte';
-export { default as SearchInput } from './misc/SearchInput.svelte';
-export { default as SyntaxHighlightedCode } from './misc/SyntaxHighlightedCode.svelte';
-export { default as ModelsSelector } from './models/ModelsSelector.svelte';
-
-// Server
-
-export { default as ServerStatus } from './server/ServerStatus.svelte';
-export { default as ServerErrorSplash } from './server/ServerErrorSplash.svelte';
-export { default as ServerLoadingSplash } from './server/ServerLoadingSplash.svelte';
+// Compatibility aliases
+export { default as ActionButton } from './actions/ActionIcon.svelte';
+export { default as ActionDropdown } from './navigation/DropdownMenuActions.svelte';
+export { default as CopyToClipboardIcon } from './actions/ActionIconCopyToClipboard.svelte';
+export { default as RemoveButton } from './actions/ActionIconRemove.svelte';

tools/server/webui/src/lib/components/app/misc/ActionButton.svelte

@@ -1,47 +0,0 @@
-<script lang="ts">
-	import { Button } from '$lib/components/ui/button';
-	import * as Tooltip from '$lib/components/ui/tooltip';
-	import type { Component } from 'svelte';
-
-	interface Props {
-		icon: Component;
-		tooltip: string;
-		variant?: 'default' | 'destructive' | 'outline' | 'secondary' | 'ghost' | 'link';
-		size?: 'default' | 'sm' | 'lg' | 'icon';
-		class?: string;
-		disabled?: boolean;
-		onclick: () => void;
-		'aria-label'?: string;
-	}
-
-	let {
-		icon,
-		tooltip,
-		variant = 'ghost',
-		size = 'sm',
-		class: className = '',
-		disabled = false,
-		onclick,
-		'aria-label': ariaLabel
-	}: Props = $props();
-</script>
-
-<Tooltip.Root>
-	<Tooltip.Trigger>
-		<Button
-			{variant}
-			{size}
-			{disabled}
-			{onclick}
-			class="h-6 w-6 p-0 {className} flex"
-			aria-label={ariaLabel || tooltip}
-		>
-			{@const IconComponent = icon}
-			<IconComponent class="h-3 w-3" />
-		</Button>
-	</Tooltip.Trigger>
-
-	<Tooltip.Content>
-		<p>{tooltip}</p>
-	</Tooltip.Content>
-</Tooltip.Root>

tools/server/webui/src/lib/components/app/misc/ActionDropdown.svelte

@@ -1,86 +0,0 @@
-<script lang="ts">
-	import * as DropdownMenu from '$lib/components/ui/dropdown-menu';
-	import * as Tooltip from '$lib/components/ui/tooltip';
-	import { KeyboardShortcutInfo } from '$lib/components/app';
-	import type { Component } from 'svelte';
-
-	interface ActionItem {
-		icon: Component;
-		label: string;
-		onclick: (event: Event) => void;
-		variant?: 'default' | 'destructive';
-		disabled?: boolean;
-		shortcut?: string[];
-		separator?: boolean;
-	}
-
-	interface Props {
-		triggerIcon: Component;
-		triggerTooltip?: string;
-		triggerClass?: string;
-		actions: ActionItem[];
-		align?: 'start' | 'center' | 'end';
-		open?: boolean;
-	}
-
-	let {
-		triggerIcon,
-		triggerTooltip,
-		triggerClass = '',
-		actions,
-		align = 'end',
-		open = $bindable(false)
-	}: Props = $props();
-</script>
-
-<DropdownMenu.Root bind:open>
-	<DropdownMenu.Trigger
-		class="flex h-6 w-6 cursor-pointer items-center justify-center rounded-md p-0 text-sm font-medium transition-colors hover:bg-accent hover:text-accent-foreground focus:bg-accent focus:text-accent-foreground focus:outline-none disabled:pointer-events-none disabled:opacity-50 data-[state=open]:bg-accent data-[state=open]:text-accent-foreground {triggerClass}"
-		onclick={(e) => e.stopPropagation()}
-	>
-		{#if triggerTooltip}
-			<Tooltip.Root>
-				<Tooltip.Trigger>
-					{@render iconComponent(triggerIcon, 'h-3 w-3')}
-					<span class="sr-only">{triggerTooltip}</span>
-				</Tooltip.Trigger>
-				<Tooltip.Content>
-					<p>{triggerTooltip}</p>
-				</Tooltip.Content>
-			</Tooltip.Root>
-		{:else}
-			{@render iconComponent(triggerIcon, 'h-3 w-3')}
-		{/if}
-	</DropdownMenu.Trigger>
-
-	<DropdownMenu.Content {align} class="z-[999999] w-48">
-		{#each actions as action, index (action.label)}
-			{#if action.separator && index > 0}
-				<DropdownMenu.Separator />
-			{/if}
-
-			<DropdownMenu.Item
-				onclick={action.onclick}
-				variant={action.variant}
-				disabled={action.disabled}
-				class="flex items-center justify-between hover:[&>kbd]:opacity-100"
-			>
-				<div class="flex items-center gap-2">
-					{@render iconComponent(
-						action.icon,
-						`h-4 w-4 ${action.variant === 'destructive' ? 'text-destructive' : ''}`
-					)}
-					{action.label}
-				</div>
-
-				{#if action.shortcut}
-					<KeyboardShortcutInfo keys={action.shortcut} variant={action.variant} />
-				{/if}
-			</DropdownMenu.Item>
-		{/each}
-	</DropdownMenu.Content>
-</DropdownMenu.Root>
-
-{#snippet iconComponent(IconComponent: Component, className: string)}
-	<IconComponent class={className} />
-{/snippet}

tools/server/webui/src/lib/components/app/misc/BadgeChatStatistic.svelte

@@ -1,44 +0,0 @@
-<script lang="ts">
-	import { BadgeInfo } from '$lib/components/app';
-	import * as Tooltip from '$lib/components/ui/tooltip';
-	import { copyToClipboard } from '$lib/utils';
-	import type { Component } from 'svelte';
-
-	interface Props {
-		class?: string;
-		icon: Component;
-		value: string | number;
-		tooltipLabel?: string;
-	}
-
-	let { class: className = '', icon: Icon, value, tooltipLabel }: Props = $props();
-
-	function handleClick() {
-		void copyToClipboard(String(value));
-	}
-</script>
-
-{#if tooltipLabel}
-	<Tooltip.Root>
-		<Tooltip.Trigger>
-			<BadgeInfo class={className} onclick={handleClick}>
-				{#snippet icon()}
-					<Icon class="h-3 w-3" />
-				{/snippet}
-
-				{value}
-			</BadgeInfo>
-		</Tooltip.Trigger>
-		<Tooltip.Content>
-			<p>{tooltipLabel}</p>
-		</Tooltip.Content>
-	</Tooltip.Root>
-{:else}
-	<BadgeInfo class={className} onclick={handleClick}>
-		{#snippet icon()}
-			<Icon class="h-3 w-3" />
-		{/snippet}
-
-		{value}
-	</BadgeInfo>
-{/if}

tools/server/webui/src/lib/components/app/misc/BadgeInfo.svelte

@@ -1,27 +0,0 @@
-<script lang="ts">
-	import { cn } from '$lib/components/ui/utils';
-	import type { Snippet } from 'svelte';
-
-	interface Props {
-		children: Snippet;
-		class?: string;
-		icon?: Snippet;
-		onclick?: () => void;
-	}
-
-	let { children, class: className = '', icon, onclick }: Props = $props();
-</script>
-
-<button
-	class={cn(
-		'inline-flex cursor-pointer items-center gap-1 rounded-sm bg-muted-foreground/15 px-1.5 py-0.75',
-		className
-	)}
-	{onclick}
->
-	{#if icon}
-		{@render icon()}
-	{/if}
-
-	{@render children()}
-</button>

tools/server/webui/src/lib/components/app/misc/BadgeModality.svelte

@@ -1,39 +0,0 @@
-<script lang="ts">
-	import { ModelModality } from '$lib/enums';
-	import { MODALITY_ICONS, MODALITY_LABELS } from '$lib/constants/icons';
-	import { cn } from '$lib/components/ui/utils';
-
-	type DisplayableModality = ModelModality.VISION | ModelModality.AUDIO;
-
-	interface Props {
-		modalities: ModelModality[];
-		class?: string;
-	}
-
-	let { modalities, class: className = '' }: Props = $props();
-
-	// Filter to only modalities that have icons (VISION, AUDIO)
-	const displayableModalities = $derived(
-		modalities.filter(
-			(m): m is DisplayableModality => m === ModelModality.VISION || m === ModelModality.AUDIO
-		)
-	);
-</script>
-
-{#each displayableModalities as modality, index (index)}
-	{@const IconComponent = MODALITY_ICONS[modality]}
-	{@const label = MODALITY_LABELS[modality]}
-
-	<span
-		class={cn(
-			'inline-flex items-center gap-1 rounded-md bg-muted px-2 py-1 text-xs font-medium',
-			className
-		)}
-	>
-		{#if IconComponent}
-			<IconComponent class="h-3 w-3" />
-		{/if}
-
-		{label}
-	</span>
-{/each}

tools/server/webui/src/lib/components/app/misc/CopyToClipboardIcon.svelte

@@ -1,18 +0,0 @@
-<script lang="ts">
-	import { Copy } from '@lucide/svelte';
-	import { copyToClipboard } from '$lib/utils';
-
-	interface Props {
-		ariaLabel?: string;
-		canCopy?: boolean;
-		text: string;
-	}
-
-	let { ariaLabel = 'Copy to clipboard', canCopy = true, text }: Props = $props();
-</script>
-
-<Copy
-	class="h-3 w-3 flex-shrink-0 cursor-{canCopy ? 'pointer' : 'not-allowed'}"
-	aria-label={ariaLabel}
-	onclick={() => canCopy && copyToClipboard(text)}
-/>

tools/server/webui/src/lib/components/app/misc/DropdownMenuSearchable.svelte

@@ -1,88 +0,0 @@
-<script lang="ts">
-	import type { Snippet } from 'svelte';
-	import * as DropdownMenu from '$lib/components/ui/dropdown-menu';
-	import { cn } from '$lib/components/ui/utils';
-	import { SearchInput } from '$lib/components/app';
-
-	interface Props {
-		open?: boolean;
-		onOpenChange?: (open: boolean) => void;
-		placeholder?: string;
-		searchValue?: string;
-		onSearchChange?: (value: string) => void;
-		onSearchKeyDown?: (event: KeyboardEvent) => void;
-		align?: 'start' | 'center' | 'end';
-		contentClass?: string;
-		emptyMessage?: string;
-		isEmpty?: boolean;
-		disabled?: boolean;
-		trigger: Snippet;
-		children: Snippet;
-		footer?: Snippet;
-	}
-
-	let {
-		open = $bindable(false),
-		onOpenChange,
-		placeholder = 'Search...',
-		searchValue = $bindable(''),
-		onSearchChange,
-		onSearchKeyDown,
-		align = 'start',
-		contentClass = 'w-72',
-		emptyMessage = 'No items found',
-		isEmpty = false,
-		disabled = false,
-		trigger,
-		children,
-		footer
-	}: Props = $props();
-
-	function handleOpenChange(newOpen: boolean) {
-		open = newOpen;
-
-		if (!newOpen) {
-			searchValue = '';
-			onSearchChange?.('');
-		}
-
-		onOpenChange?.(newOpen);
-	}
-</script>
-
-<DropdownMenu.Root bind:open onOpenChange={handleOpenChange}>
-	<DropdownMenu.Trigger
-		{disabled}
-		onclick={(e) => {
-			e.preventDefault();
-			e.stopPropagation();
-		}}
-	>
-		{@render trigger()}
-	</DropdownMenu.Trigger>
-
-	<DropdownMenu.Content {align} class={cn(contentClass, 'pt-0')}>
-		<div class="sticky top-0 z-10 mb-2 bg-popover p-1 pt-2">
-			<SearchInput
-				{placeholder}
-				bind:value={searchValue}
-				onInput={onSearchChange}
-				onKeyDown={onSearchKeyDown}
-			/>
-		</div>
-
-		<div class={cn('overflow-y-auto')}>
-			{@render children()}
-
-			{#if isEmpty}
-				<div class="px-2 py-3 text-center text-sm text-muted-foreground">{emptyMessage}</div>
-			{/if}
-		</div>
-
-		{#if footer}
-			<DropdownMenu.Separator />
-
-			{@render footer()}
-		{/if}
-	</DropdownMenu.Content>
-</DropdownMenu.Root>

tools/server/webui/src/lib/components/app/misc/MarkdownContent.svelte

@@ -1,872 +0,0 @@
-<script lang="ts">
-	import { remark } from 'remark';
-	import remarkBreaks from 'remark-breaks';
-	import remarkGfm from 'remark-gfm';
-	import remarkMath from 'remark-math';
-	import rehypeHighlight from 'rehype-highlight';
-	import remarkRehype from 'remark-rehype';
-	import rehypeKatex from 'rehype-katex';
-	import rehypeStringify from 'rehype-stringify';
-	import type { Root as HastRoot, RootContent as HastRootContent } from 'hast';
-	import type { Root as MdastRoot } from 'mdast';
-	import { browser } from '$app/environment';
-	import { onDestroy, tick } from 'svelte';
-	import { rehypeRestoreTableHtml } from '$lib/markdown/table-html-restorer';
-	import { rehypeEnhanceLinks } from '$lib/markdown/enhance-links';
-	import { rehypeEnhanceCodeBlocks } from '$lib/markdown/enhance-code-blocks';
-	import { remarkLiteralHtml } from '$lib/markdown/literal-html';
-	import { copyCodeToClipboard, preprocessLaTeX } from '$lib/utils';
-	import '$styles/katex-custom.scss';
-	import githubDarkCss from 'highlight.js/styles/github-dark.css?inline';
-	import githubLightCss from 'highlight.js/styles/github.css?inline';
-	import { mode } from 'mode-watcher';
-	import CodePreviewDialog from './CodePreviewDialog.svelte';
-
-	interface Props {
-		content: string;
-		class?: string;
-	}
-
-	interface MarkdownBlock {
-		id: string;
-		html: string;
-	}
-
-	let { content, class: className = '' }: Props = $props();
-
-	let containerRef = $state<HTMLDivElement>();
-	let renderedBlocks = $state<MarkdownBlock[]>([]);
-	let unstableBlockHtml = $state('');
-	let previewDialogOpen = $state(false);
-	let previewCode = $state('');
-	let previewLanguage = $state('text');
-
-	let pendingMarkdown: string | null = null;
-	let isProcessing = false;
-
-	const themeStyleId = `highlight-theme-${(window.idxThemeStyle = (window.idxThemeStyle ?? 0) + 1)}`;
-
-	let processor = $derived(() => {
-		return remark()
-			.use(remarkGfm) // GitHub Flavored Markdown
-			.use(remarkMath) // Parse $inline$ and $$block$$ math
-			.use(remarkBreaks) // Convert line breaks to <br>
-			.use(remarkLiteralHtml) // Treat raw HTML as literal text with preserved indentation
-			.use(remarkRehype) // Convert Markdown AST to rehype
-			.use(rehypeKatex) // Render math using KaTeX
-			.use(rehypeHighlight) // Add syntax highlighting
-			.use(rehypeRestoreTableHtml) // Restore limited HTML (e.g., <br>, <ul>) inside Markdown tables
-			.use(rehypeEnhanceLinks) // Add target="_blank" to links
-			.use(rehypeEnhanceCodeBlocks) // Wrap code blocks with header and actions
-			.use(rehypeStringify, { allowDangerousHtml: true }); // Convert to HTML string
-	});
-
-	/**
-	 * Removes click event listeners from copy and preview buttons.
-	 * Called on component destroy.
-	 */
-	function cleanupEventListeners() {
-		if (!containerRef) return;
-
-		const copyButtons = containerRef.querySelectorAll<HTMLButtonElement>('.copy-code-btn');
-		const previewButtons = containerRef.querySelectorAll<HTMLButtonElement>('.preview-code-btn');
-
-		for (const button of copyButtons) {
-			button.removeEventListener('click', handleCopyClick);
-		}
-
-		for (const button of previewButtons) {
-			button.removeEventListener('click', handlePreviewClick);
-		}
-	}
-
-	/**
-	 * Removes this component's highlight.js theme style from the document head.
-	 * Called on component destroy to clean up injected styles.
-	 */
-	function cleanupHighlightTheme() {
-		if (!browser) return;
-
-		const existingTheme = document.getElementById(themeStyleId);
-		existingTheme?.remove();
-	}
-
-	/**
-	 * Loads the appropriate highlight.js theme based on dark/light mode.
-	 * Injects a scoped style element into the document head.
-	 * @param isDark - Whether to load the dark theme (true) or light theme (false)
-	 */
-	function loadHighlightTheme(isDark: boolean) {
-		if (!browser) return;
-
-		const existingTheme = document.getElementById(themeStyleId);
-		existingTheme?.remove();
-
-		const style = document.createElement('style');
-		style.id = themeStyleId;
-		style.textContent = isDark ? githubDarkCss : githubLightCss;
-
-		document.head.appendChild(style);
-	}
-
-	/**
-	 * Extracts code information from a button click target within a code block.
-	 * @param target - The clicked button element
-	 * @returns Object with rawCode and language, or null if extraction fails
-	 */
-	function getCodeInfoFromTarget(target: HTMLElement) {
-		const wrapper = target.closest('.code-block-wrapper');
-
-		if (!wrapper) {
-			console.error('No wrapper found');
-			return null;
-		}
-
-		const codeElement = wrapper.querySelector<HTMLElement>('code[data-code-id]');
-
-		if (!codeElement) {
-			console.error('No code element found in wrapper');
-			return null;
-		}
-
-		const rawCode = codeElement.textContent ?? '';
-
-		const languageLabel = wrapper.querySelector<HTMLElement>('.code-language');
-		const language = languageLabel?.textContent?.trim() || 'text';
-
-		return { rawCode, language };
-	}
-
-	/**
-	 * Generates a unique identifier for a HAST node based on its position.
-	 * Used for stable block identification during incremental rendering.
-	 * @param node - The HAST root content node
-	 * @param indexFallback - Fallback index if position is unavailable
-	 * @returns Unique string identifier for the node
-	 */
-	function getHastNodeId(node: HastRootContent, indexFallback: number): string {
-		const position = node.position;
-
-		if (position?.start?.offset != null && position?.end?.offset != null) {
-			return `hast-${position.start.offset}-${position.end.offset}`;
-		}
-
-		return `${node.type}-${indexFallback}`;
-	}
-
-	/**
-	 * Handles click events on copy buttons within code blocks.
-	 * Copies the raw code content to the clipboard.
-	 * @param event - The click event from the copy button
-	 */
-	async function handleCopyClick(event: Event) {
-		event.preventDefault();
-		event.stopPropagation();
-
-		const target = event.currentTarget as HTMLButtonElement | null;
-
-		if (!target) {
-			return;
-		}
-
-		const info = getCodeInfoFromTarget(target);
-
-		if (!info) {
-			return;
-		}
-
-		try {
-			await copyCodeToClipboard(info.rawCode);
-		} catch (error) {
-			console.error('Failed to copy code:', error);
-		}
-	}
-
-	/**
-	 * Handles preview dialog open state changes.
-	 * Clears preview content when dialog is closed.
-	 * @param open - Whether the dialog is being opened or closed
-	 */
-	function handlePreviewDialogOpenChange(open: boolean) {
-		previewDialogOpen = open;
-
-		if (!open) {
-			previewCode = '';
-			previewLanguage = 'text';
-		}
-	}
-
-	/**
-	 * Handles click events on preview buttons within HTML code blocks.
-	 * Opens a preview dialog with the rendered HTML content.
-	 * @param event - The click event from the preview button
-	 */
-	function handlePreviewClick(event: Event) {
-		event.preventDefault();
-		event.stopPropagation();
-
-		const target = event.currentTarget as HTMLButtonElement | null;
-
-		if (!target) {
-			return;
-		}
-
-		const info = getCodeInfoFromTarget(target);
-
-		if (!info) {
-			return;
-		}
-
-		previewCode = info.rawCode;
-		previewLanguage = info.language;
-		previewDialogOpen = true;
-	}
-
-	/**
-	 * Processes markdown content into stable and unstable HTML blocks.
-	 * Uses incremental rendering: stable blocks are cached, unstable block is re-rendered.
-	 * @param markdown - The raw markdown string to process
-	 */
-	async function processMarkdown(markdown: string) {
-		if (!markdown) {
-			renderedBlocks = [];
-			unstableBlockHtml = '';
-			return;
-		}
-
-		const normalized = preprocessLaTeX(markdown);
-		const processorInstance = processor();
-		const ast = processorInstance.parse(normalized) as MdastRoot;
-		const processedRoot = (await processorInstance.run(ast)) as HastRoot;
-		const processedChildren = processedRoot.children ?? [];
-		const stableCount = Math.max(processedChildren.length - 1, 0);
-		const nextBlocks: MarkdownBlock[] = [];
-
-		for (let index = 0; index < stableCount; index++) {
-			const hastChild = processedChildren[index];
-			const id = getHastNodeId(hastChild, index);
-			const existing = renderedBlocks[index];
-
-			if (existing && existing.id === id) {
-				nextBlocks.push(existing);
-				continue;
-			}
-
-			const html = stringifyProcessedNode(
-				processorInstance,
-				processedRoot,
-				processedChildren[index]
-			);
-
-			nextBlocks.push({ id, html });
-		}
-
-		let unstableHtml = '';
-
-		if (processedChildren.length > stableCount) {
-			const unstableChild = processedChildren[stableCount];
-			unstableHtml = stringifyProcessedNode(processorInstance, processedRoot, unstableChild);
-		}
-
-		renderedBlocks = nextBlocks;
-		await tick(); // Force DOM sync before updating unstable HTML block
-		unstableBlockHtml = unstableHtml;
-	}
-
-	/**
-	 * Attaches click event listeners to copy and preview buttons in code blocks.
-	 * Uses data-listener-bound attribute to prevent duplicate bindings.
-	 */
-	function setupCodeBlockActions() {
-		if (!containerRef) return;
-
-		const wrappers = containerRef.querySelectorAll<HTMLElement>('.code-block-wrapper');
-
-		for (const wrapper of wrappers) {
-			const copyButton = wrapper.querySelector<HTMLButtonElement>('.copy-code-btn');
-			const previewButton = wrapper.querySelector<HTMLButtonElement>('.preview-code-btn');
-
-			if (copyButton && copyButton.dataset.listenerBound !== 'true') {
-				copyButton.dataset.listenerBound = 'true';
-				copyButton.addEventListener('click', handleCopyClick);
-			}
-
-			if (previewButton && previewButton.dataset.listenerBound !== 'true') {
-				previewButton.dataset.listenerBound = 'true';
-				previewButton.addEventListener('click', handlePreviewClick);
-			}
-		}
-	}
-
-	/**
-	 * Converts a single HAST node to an enhanced HTML string.
-	 * Applies link and code block enhancements to the output.
-	 * @param processorInstance - The remark/rehype processor instance
-	 * @param processedRoot - The full processed HAST root (for context)
-	 * @param child - The specific HAST child node to stringify
-	 * @returns Enhanced HTML string representation of the node
-	 */
-	function stringifyProcessedNode(
-		processorInstance: ReturnType<typeof processor>,
-		processedRoot: HastRoot,
-		child: unknown
-	) {
-		const root: HastRoot = {
-			...(processedRoot as HastRoot),
-			children: [child as never]
-		};
-
-		return processorInstance.stringify(root);
-	}
-
-	/**
-	 * Queues markdown for processing with coalescing support.
-	 * Only processes the latest markdown when multiple updates arrive quickly.
-	 * @param markdown - The markdown content to render
-	 */
-	async function updateRenderedBlocks(markdown: string) {
-		pendingMarkdown = markdown;
-
-		if (isProcessing) {
-			return;
-		}
-
-		isProcessing = true;
-
-		try {
-			while (pendingMarkdown !== null) {
-				const nextMarkdown = pendingMarkdown;
-				pendingMarkdown = null;
-
-				await processMarkdown(nextMarkdown);
-			}
-		} catch (error) {
-			console.error('Failed to process markdown:', error);
-			renderedBlocks = [];
-			unstableBlockHtml = markdown.replace(/\n/g, '<br>');
-		} finally {
-			isProcessing = false;
-		}
-	}
-
-	$effect(() => {
-		const currentMode = mode.current;
-		const isDark = currentMode === 'dark';
-
-		loadHighlightTheme(isDark);
-	});
-
-	$effect(() => {
-		updateRenderedBlocks(content);
-	});
-
-	$effect(() => {
-		const hasRenderedBlocks = renderedBlocks.length > 0;
-		const hasUnstableBlock = Boolean(unstableBlockHtml);
-
-		if ((hasRenderedBlocks || hasUnstableBlock) && containerRef) {
-			setupCodeBlockActions();
-		}
-	});
-
-	onDestroy(() => {
-		cleanupEventListeners();
-		cleanupHighlightTheme();
-	});
-</script>
-
-<div bind:this={containerRef} class={className}>
-	{#each renderedBlocks as block (block.id)}
-		<div class="markdown-block" data-block-id={block.id}>
-			<!-- eslint-disable-next-line no-at-html-tags -->
-			{@html block.html}
-		</div>
-	{/each}
-
-	{#if unstableBlockHtml}
-		<div class="markdown-block markdown-block--unstable" data-block-id="unstable">
-			<!-- eslint-disable-next-line no-at-html-tags -->
-			{@html unstableBlockHtml}
-		</div>
-	{/if}
-</div>
-
-<CodePreviewDialog
-	open={previewDialogOpen}
-	code={previewCode}
-	language={previewLanguage}
-	onOpenChange={handlePreviewDialogOpenChange}
-/>
-
-<style>
-	.markdown-block,
-	.markdown-block--unstable {
-		display: contents;
-	}
-
-	/* Base typography styles */
-	div :global(p:not(:last-child)) {
-		margin-bottom: 1rem;
-		line-height: 1.75;
-	}
-
-	div :global(:is(h1, h2, h3, h4, h5, h6):first-child) {
-		margin-top: 0;
-	}
-
-	/* Headers with consistent spacing */
-	div :global(h1) {
-		font-size: 1.875rem;
-		font-weight: 700;
-		line-height: 1.2;
-		margin: 1.5rem 0 0.75rem 0;
-	}
-
-	div :global(h2) {
-		font-size: 1.5rem;
-		font-weight: 600;
-		line-height: 1.3;
-		margin: 1.25rem 0 0.5rem 0;
-	}
-
-	div :global(h3) {
-		font-size: 1.25rem;
-		font-weight: 600;
-		margin: 1.5rem 0 0.5rem 0;
-		line-height: 1.4;
-	}
-
-	div :global(h4) {
-		font-size: 1.125rem;
-		font-weight: 600;
-		margin: 0.75rem 0 0.25rem 0;
-	}
-
-	div :global(h5) {
-		font-size: 1rem;
-		font-weight: 600;
-		margin: 0.5rem 0 0.25rem 0;
-	}
-
-	div :global(h6) {
-		font-size: 0.875rem;
-		font-weight: 600;
-		margin: 0.5rem 0 0.25rem 0;
-	}
-
-	/* Text formatting */
-	div :global(strong) {
-		font-weight: 600;
-	}
-
-	div :global(em) {
-		font-style: italic;
-	}
-
-	div :global(del) {
-		text-decoration: line-through;
-		opacity: 0.7;
-	}
-
-	/* Inline code */
-	div :global(code:not(pre code)) {
-		background: var(--muted);
-		color: var(--muted-foreground);
-		padding: 0.125rem 0.375rem;
-		border-radius: 0.375rem;
-		font-size: 0.875rem;
-		font-family:
-			ui-monospace, SFMono-Regular, 'SF Mono', Monaco, 'Cascadia Code', 'Roboto Mono', Consolas,
-			'Liberation Mono', Menlo, monospace;
-	}
-
-	/* Links */
-	div :global(a) {
-		color: var(--primary);
-		text-decoration: underline;
-		text-underline-offset: 2px;
-		transition: color 0.2s ease;
-		overflow-wrap: anywhere;
-		word-break: break-all;
-	}
-
-	div :global(a:hover) {
-		color: var(--primary);
-	}
-
-	/* Lists */
-	div :global(ul) {
-		list-style-type: disc;
-		margin-left: 1.5rem;
-		margin-bottom: 1rem;
-	}
-
-	div :global(ol) {
-		list-style-type: decimal;
-		margin-left: 1.5rem;
-		margin-bottom: 1rem;
-	}
-
-	div :global(li) {
-		margin-bottom: 0.25rem;
-		padding-left: 0.5rem;
-	}
-
-	div :global(li::marker) {
-		color: var(--muted-foreground);
-	}
-
-	/* Nested lists */
-	div :global(ul ul) {
-		list-style-type: circle;
-		margin-top: 0.25rem;
-		margin-bottom: 0.25rem;
-	}
-
-	div :global(ol ol) {
-		list-style-type: lower-alpha;
-		margin-top: 0.25rem;
-		margin-bottom: 0.25rem;
-	}
-
-	/* Task lists */
-	div :global(.task-list-item) {
-		list-style: none;
-		margin-left: 0;
-		padding-left: 0;
-	}
-
-	div :global(.task-list-item-checkbox) {
-		margin-right: 0.5rem;
-		margin-top: 0.125rem;
-	}
-
-	/* Blockquotes */
-	div :global(blockquote) {
-		border-left: 4px solid var(--border);
-		padding: 0.5rem 1rem;
-		margin: 1.5rem 0;
-		font-style: italic;
-		color: var(--muted-foreground);
-		background: var(--muted);
-		border-radius: 0 0.375rem 0.375rem 0;
-	}
-
-	/* Tables */
-	div :global(table) {
-		width: 100%;
-		margin: 1.5rem 0;
-		border-collapse: collapse;
-		border: 1px solid var(--border);
-		border-radius: 0.375rem;
-		overflow: hidden;
-	}
-
-	div :global(th) {
-		background: hsl(var(--muted) / 0.3);
-		border: 1px solid var(--border);
-		padding: 0.5rem 0.75rem;
-		text-align: left;
-		font-weight: 600;
-	}
-
-	div :global(td) {
-		border: 1px solid var(--border);
-		padding: 0.5rem 0.75rem;
-	}
-
-	div :global(tr:nth-child(even)) {
-		background: hsl(var(--muted) / 0.1);
-	}
-
-	/* User message markdown should keep table borders visible on light primary backgrounds */
-	div.markdown-user-content :global(table),
-	div.markdown-user-content :global(th),
-	div.markdown-user-content :global(td),
-	div.markdown-user-content :global(.table-wrapper) {
-		border-color: currentColor;
-	}
-
-	/* Horizontal rules */
-	div :global(hr) {
-		border: none;
-		border-top: 1px solid var(--border);
-		margin: 1.5rem 0;
-	}
-
-	/* Images */
-	div :global(img) {
-		border-radius: 0.5rem;
-		box-shadow:
-			0 1px 3px 0 rgb(0 0 0 / 0.1),
-			0 1px 2px -1px rgb(0 0 0 / 0.1);
-		margin: 1.5rem 0;
-		max-width: 100%;
-		height: auto;
-	}
-
-	/* Code blocks */
-
-	div :global(.code-block-wrapper) {
-		margin: 1.5rem 0;
-		border-radius: 0.75rem;
-		overflow: hidden;
-		border: 1px solid var(--border);
-		background: var(--code-background);
-	}
-
-	div :global(.code-block-header) {
-		display: flex;
-		justify-content: space-between;
-		align-items: center;
-		padding: 0.5rem 1rem;
-		background: hsl(var(--muted) / 0.5);
-		border-bottom: 1px solid var(--border);
-		font-size: 0.875rem;
-	}
-
-	div :global(.code-language) {
-		color: var(--code-foreground);
-		font-weight: 500;
-		font-family:
-			ui-monospace, SFMono-Regular, 'SF Mono', Monaco, 'Cascadia Code', 'Roboto Mono', Consolas,
-			'Liberation Mono', Menlo, monospace;
-		text-transform: uppercase;
-		font-size: 0.75rem;
-		letter-spacing: 0.05em;
-	}
-
-	div :global(.code-block-actions) {
-		display: flex;
-		align-items: center;
-		gap: 0.5rem;
-	}
-
-	div :global(.copy-code-btn),
-	div :global(.preview-code-btn) {
-		display: flex;
-		align-items: center;
-		justify-content: center;
-		padding: 0;
-		background: transparent;
-		color: var(--code-foreground);
-		cursor: pointer;
-		transition: all 0.2s ease;
-	}
-
-	div :global(.copy-code-btn:hover),
-	div :global(.preview-code-btn:hover) {
-		transform: scale(1.05);
-	}
-
-	div :global(.copy-code-btn:active),
-	div :global(.preview-code-btn:active) {
-		transform: scale(0.95);
-	}
-
-	div :global(.code-block-wrapper pre) {
-		background: transparent;
-		padding: 1rem;
-		margin: 0;
-		overflow-x: auto;
-		border-radius: 0;
-		border: none;
-		font-size: 0.875rem;
-		line-height: 1.5;
-	}
-
-	div :global(pre) {
-		background: var(--muted);
-		margin: 1.5rem 0;
-		overflow-x: auto;
-		border-radius: 1rem;
-		border: none;
-	}
-
-	div :global(code) {
-		background: transparent;
-		color: var(--code-foreground);
-	}
-
-	/* Mentions and hashtags */
-	div :global(.mention) {
-		color: hsl(var(--primary));
-		font-weight: 500;
-		text-decoration: none;
-	}
-
-	div :global(.mention:hover) {
-		text-decoration: underline;
-	}
-
-	div :global(.hashtag) {
-		color: hsl(var(--primary));
-		font-weight: 500;
-		text-decoration: none;
-	}
-
-	div :global(.hashtag:hover) {
-		text-decoration: underline;
-	}
-
-	/* Advanced table enhancements */
-	div :global(table) {
-		transition: all 0.2s ease;
-	}
-
-	div :global(table:hover) {
-		box-shadow:
-			0 4px 6px -1px rgb(0 0 0 / 0.1),
-			0 2px 4px -2px rgb(0 0 0 / 0.1);
-	}
-
-	div :global(th:hover),
-	div :global(td:hover) {
-		background: var(--muted);
-	}
-
-	/* Disable hover effects when rendering user messages */
-	.markdown-user-content :global(a),
-	.markdown-user-content :global(a:hover) {
-		color: var(--primary-foreground);
-	}
-
-	.markdown-user-content :global(table:hover) {
-		box-shadow: none;
-	}
-
-	.markdown-user-content :global(th:hover),
-	.markdown-user-content :global(td:hover) {
-		background: inherit;
-	}
-
-	/* Enhanced blockquotes */
-	div :global(blockquote) {
-		transition: all 0.2s ease;
-		position: relative;
-	}
-
-	div :global(blockquote:hover) {
-		border-left-width: 6px;
-		background: var(--muted);
-		transform: translateX(2px);
-	}
-
-	div :global(blockquote::before) {
-		content: '"';
-		position: absolute;
-		top: -0.5rem;
-		left: 0.5rem;
-		font-size: 3rem;
-		color: var(--muted-foreground);
-		font-family: serif;
-		line-height: 1;
-	}
-
-	/* Enhanced images */
-	div :global(img) {
-		transition: all 0.3s ease;
-		cursor: pointer;
-	}
-
-	div :global(img:hover) {
-		transform: scale(1.02);
-		box-shadow:
-			0 10px 15px -3px rgb(0 0 0 / 0.1),
-			0 4px 6px -4px rgb(0 0 0 / 0.1);
-	}
-
-	/* Image zoom overlay */
-	div :global(.image-zoom-overlay) {
-		position: fixed;
-		top: 0;
-		left: 0;
-		right: 0;
-		bottom: 0;
-		background: rgba(0, 0, 0, 0.8);
-		display: flex;
-		align-items: center;
-		justify-content: center;
-		z-index: 1000;
-		cursor: pointer;
-	}
-
-	div :global(.image-zoom-overlay img) {
-		max-width: 90vw;
-		max-height: 90vh;
-		border-radius: 0.5rem;
-		box-shadow: 0 25px 50px -12px rgb(0 0 0 / 0.25);
-	}
-
-	/* Enhanced horizontal rules */
-	div :global(hr) {
-		border: none;
-		height: 2px;
-		background: linear-gradient(to right, transparent, var(--border), transparent);
-		margin: 2rem 0;
-		position: relative;
-	}
-
-	div :global(hr::after) {
-		content: '';
-		position: absolute;
-		top: 50%;
-		left: 50%;
-		transform: translate(-50%, -50%);
-		width: 1rem;
-		height: 1rem;
-		background: var(--border);
-		border-radius: 50%;
-	}
-
-	/* Scrollable tables */
-	div :global(.table-wrapper) {
-		overflow-x: auto;
-		margin: 1.5rem 0;
-		border-radius: 0.5rem;
-		border: 1px solid var(--border);
-	}
-
-	div :global(.table-wrapper table) {
-		margin: 0;
-		border: none;
-	}
-
-	/* Responsive adjustments */
-	@media (max-width: 640px) {
-		div :global(h1) {
-			font-size: 1.5rem;
-		}
-
-		div :global(h2) {
-			font-size: 1.25rem;
-		}
-
-		div :global(h3) {
-			font-size: 1.125rem;
-		}
-
-		div :global(table) {
-			font-size: 0.875rem;
-		}
-
-		div :global(th),
-		div :global(td) {
-			padding: 0.375rem 0.5rem;
-		}
-
-		div :global(.table-wrapper) {
-			margin: 0.5rem -1rem;
-			border-radius: 0;
-			border-left: none;
-			border-right: none;
-		}
-	}
-
-	/* Dark mode adjustments */
-	@media (prefers-color-scheme: dark) {
-		div :global(blockquote:hover) {
-			background: var(--muted);
-		}
-	}
-</style>

tools/server/webui/src/lib/components/app/misc/RemoveButton.svelte

@@ -1,26 +0,0 @@
-<script lang="ts">
-	import { X } from '@lucide/svelte';
-	import { Button } from '$lib/components/ui/button';
-
-	interface Props {
-		id: string;
-		onRemove?: (id: string) => void;
-		class?: string;
-	}
-
-	let { id, onRemove, class: className = '' }: Props = $props();
-</script>
-
-<Button
-	type="button"
-	variant="ghost"
-	size="sm"
-	class="h-6 w-6 bg-white/20 p-0 hover:bg-white/30 {className}"
-	onclick={(e) => {
-		e.stopPropagation();
-		onRemove?.(id);
-	}}
-	aria-label="Remove file"
->
-	<X class="h-3 w-3" />
-</Button>

tools/server/webui/src/lib/components/app/misc/SyntaxHighlightedCode.svelte

@@ -1,97 +0,0 @@
-<script lang="ts">
-	import hljs from 'highlight.js';
-	import { browser } from '$app/environment';
-	import { mode } from 'mode-watcher';
-
-	import githubDarkCss from 'highlight.js/styles/github-dark.css?inline';
-	import githubLightCss from 'highlight.js/styles/github.css?inline';
-
-	interface Props {
-		code: string;
-		language?: string;
-		class?: string;
-		maxHeight?: string;
-		maxWidth?: string;
-	}
-
-	let {
-		code,
-		language = 'text',
-		class: className = '',
-		maxHeight = '60vh',
-		maxWidth = ''
-	}: Props = $props();
-
-	let highlightedHtml = $state('');
-
-	function loadHighlightTheme(isDark: boolean) {
-		if (!browser) return;
-
-		const existingThemes = document.querySelectorAll('style[data-highlight-theme-preview]');
-		existingThemes.forEach((style) => style.remove());
-
-		const style = document.createElement('style');
-		style.setAttribute('data-highlight-theme-preview', 'true');
-		style.textContent = isDark ? githubDarkCss : githubLightCss;
-
-		document.head.appendChild(style);
-	}
-
-	$effect(() => {
-		const currentMode = mode.current;
-		const isDark = currentMode === 'dark';
-
-		loadHighlightTheme(isDark);
-	});
-
-	$effect(() => {
-		if (!code) {
-			highlightedHtml = '';
-			return;
-		}
-
-		try {
-			// Check if the language is supported
-			const lang = language.toLowerCase();
-			const isSupported = hljs.getLanguage(lang);
-
-			if (isSupported) {
-				const result = hljs.highlight(code, { language: lang });
-				highlightedHtml = result.value;
-			} else {
-				// Try auto-detection or fallback to plain text
-				const result = hljs.highlightAuto(code);
-				highlightedHtml = result.value;
-			}
-		} catch {
-			// Fallback to escaped plain text
-			highlightedHtml = code.replace(/&/g, '&amp;').replace(/</g, '&lt;').replace(/>/g, '&gt;');
-		}
-	});
-</script>
-
-<div
-	class="code-preview-wrapper overflow-auto rounded-lg border border-border bg-muted {className}"
-	style="max-height: {maxHeight}; max-width: {maxWidth};"
->
-	<!-- Needs to be formatted as single line for proper rendering -->
-	<pre class="m-0 overflow-x-auto p-4"><code class="hljs text-sm leading-relaxed"
-			>{@html highlightedHtml}</code
-		></pre>
-</div>
-
-<style>
-	.code-preview-wrapper {
-		font-family:
-			ui-monospace, SFMono-Regular, 'SF Mono', Monaco, 'Cascadia Code', 'Roboto Mono', Consolas,
-			'Liberation Mono', Menlo, monospace;
-	}
-
-	.code-preview-wrapper pre {
-		background: transparent;
-	}
-
-	.code-preview-wrapper code {
-		background: transparent;
-	}
-</style>

tools/server/webui/src/lib/components/app/models/ModelBadge.svelte

@@ -1,6 +1,6 @@
 <script lang="ts">
 	import { Package } from '@lucide/svelte';
-	import { BadgeInfo, CopyToClipboardIcon } from '$lib/components/app';
+	import { BadgeInfo, ActionIconCopyToClipboard } from '$lib/components/app';
 	import { modelsStore } from '$lib/stores/models.svelte';
 	import { serverStore } from '$lib/stores/server.svelte';
 	import * as Tooltip from '$lib/components/ui/tooltip';
@@ -34,7 +34,7 @@
 		{model}
 
 		{#if showCopyIcon}
-			<CopyToClipboardIcon text={model || ''} ariaLabel="Copy model name" />
+			<ActionIconCopyToClipboard text={model || ''} ariaLabel="Copy model name" />
 		{/if}
 	</BadgeInfo>
 {/snippet}

tools/server/webui/src/lib/components/app/models/ModelsSelector.svelte

@@ -1,8 +1,8 @@
 <script lang="ts">
-	import { onMount, tick } from 'svelte';
-	import { ChevronDown, EyeOff, Loader2, MicOff, Package, Power } from '@lucide/svelte';
+	import { onMount } from 'svelte';
+	import { ChevronDown, Loader2, Package, Power } from '@lucide/svelte';
+	import * as DropdownMenu from '$lib/components/ui/dropdown-menu';
 	import * as Tooltip from '$lib/components/ui/tooltip';
-	import * as Popover from '$lib/components/ui/popover';
 	import { cn } from '$lib/components/ui/utils';
 	import {
 		modelsStore,
@@ -11,13 +11,15 @@
 		modelsUpdating,
 		selectedModelId,
 		routerModels,
-		propsCacheVersion,
 		singleModelName
 	} from '$lib/stores/models.svelte';
-	import { usedModalities, conversationsStore } from '$lib/stores/conversations.svelte';
-	import { ServerModelStatus } from '$lib/enums';
+	import { KeyboardKey, ServerModelStatus } from '$lib/enums';
 	import { isRouterMode } from '$lib/stores/server.svelte';
-	import { DialogModelInformation, SearchInput } from '$lib/components/app';
+	import {
+		DialogModelInformation,
+		DropdownMenuSearchable,
+		TruncatedText
+	} from '$lib/components/app';
 	import type { ModelOption } from '$lib/types/models';
 
 	interface Props {
@@ -29,11 +31,7 @@
 		forceForegroundText?: boolean;
 		/** When true, user's global selection takes priority over currentModel (for form selector) */
 		useGlobalSelection?: boolean;
-		/**
-		 * When provided, only consider modalities from messages BEFORE this message.
-		 * Used for regeneration - allows selecting models that don't support modalities
-		 * used in later messages.
-		 */
+		/** Optional compatibility prop for context-aware selectors. */
 		upToMessageId?: string;
 	}
 
@@ -44,7 +42,8 @@
 		disabled = false,
 		forceForegroundText = false,
 		useGlobalSelection = false,
-		upToMessageId
+		// eslint-disable-next-line @typescript-eslint/no-unused-vars
+		upToMessageId: _upToMessageId = undefined
 	}: Props = $props();
 
 	let options = $derived(modelOptions());
@@ -57,74 +56,11 @@
 	// Reactive router models state - needed for proper reactivity of status checks
 	let currentRouterModels = $derived(routerModels());
 
-	let requiredModalities = $derived(
-		upToMessageId ? conversationsStore.getModalitiesUpToMessage(upToMessageId) : usedModalities()
-	);
-
 	function getModelStatus(modelId: string): ServerModelStatus | null {
 		const model = currentRouterModels.find((m) => m.id === modelId);
 		return (model?.status?.value as ServerModelStatus) ?? null;
 	}
 
-	/**
-	 * Checks if a model supports all modalities used in the conversation.
-	 * Returns true if the model can be selected, false if it should be disabled.
-	 */
-	function isModelCompatible(option: ModelOption): boolean {
-		void propsCacheVersion();
-
-		const modelModalities = modelsStore.getModelModalities(option.model);
-
-		if (!modelModalities) {
-			const status = getModelStatus(option.model);
-
-			if (status === ServerModelStatus.LOADED) {
-				if (requiredModalities.vision || requiredModalities.audio) return false;
-			}
-
-			return true;
-		}
-
-		if (requiredModalities.vision && !modelModalities.vision) return false;
-		if (requiredModalities.audio && !modelModalities.audio) return false;
-
-		return true;
-	}
-
-	/**
-	 * Gets missing modalities for a model.
-	 * Returns object with vision/audio booleans indicating what's missing.
-	 */
-	function getMissingModalities(option: ModelOption): { vision: boolean; audio: boolean } | null {
-		void propsCacheVersion();
-
-		const modelModalities = modelsStore.getModelModalities(option.model);
-
-		if (!modelModalities) {
-			const status = getModelStatus(option.model);
-
-			if (status === ServerModelStatus.LOADED) {
-				const missing = {
-					vision: requiredModalities.vision,
-					audio: requiredModalities.audio
-				};
-
-				if (missing.vision || missing.audio) return missing;
-			}
-
-			return null;
-		}
-
-		const missing = {
-			vision: requiredModalities.vision && !modelModalities.vision,
-			audio: requiredModalities.audio && !modelModalities.audio
-		};
-
-		if (!missing.vision && !missing.audio) return null;
-
-		return missing;
-	}
-
 	let isHighlightedCurrentModelActive = $derived(
 		!isRouter || !currentModel
 			? false
@@ -142,7 +78,6 @@
 	});
 
 	let searchTerm = $state('');
-	let searchInputRef = $state<HTMLInputElement | null>(null);
 	let highlightedIndex = $state<number>(-1);
 
 	let filteredOptions: ModelOption[] = $derived(
@@ -157,13 +92,6 @@
 		})()
 	);
 
-	// Get indices of compatible options for keyboard navigation
-	let compatibleIndices = $derived(
-		filteredOptions
-			.map((option, index) => (isModelCompatible(option) ? index : -1))
-			.filter((i) => i !== -1)
-	);
-
 	// Reset highlighted index when search term changes
 	$effect(() => {
 		void searchTerm;
@@ -179,7 +107,7 @@
 		});
 	});
 
-	// Handle changes to the model selector pop-down or the model dialog, depending on if the server is in
+	// Handle changes to the model selector dropdown or the model dialog, depending on if the server is in
 	// router mode or not.
 	function handleOpenChange(open: boolean) {
 		if (loading || updating) return;
@@ -190,11 +118,6 @@
 				searchTerm = '';
 				highlightedIndex = -1;
 
-				// Focus search input after popover opens
-				tick().then(() => {
-					requestAnimationFrame(() => searchInputRef?.focus());
-				});
-
 				modelsStore.fetchRouterModels().then(() => {
 					modelsStore.fetchModalitiesForLoadedModels();
 				});
@@ -215,36 +138,32 @@
 	function handleSearchKeyDown(event: KeyboardEvent) {
 		if (event.isComposing) return;
 
-		if (event.key === 'ArrowDown') {
+		if (event.key === KeyboardKey.ARROW_DOWN) {
 			event.preventDefault();
-			if (compatibleIndices.length === 0) return;
+			if (filteredOptions.length === 0) return;
 
-			const currentPos = compatibleIndices.indexOf(highlightedIndex);
-			if (currentPos === -1 || currentPos === compatibleIndices.length - 1) {
-				highlightedIndex = compatibleIndices[0];
+			if (highlightedIndex === -1 || highlightedIndex === filteredOptions.length - 1) {
+				highlightedIndex = 0;
 			} else {
-				highlightedIndex = compatibleIndices[currentPos + 1];
+				highlightedIndex += 1;
 			}
-		} else if (event.key === 'ArrowUp') {
+		} else if (event.key === KeyboardKey.ARROW_UP) {
 			event.preventDefault();
-			if (compatibleIndices.length === 0) return;
+			if (filteredOptions.length === 0) return;
 
-			const currentPos = compatibleIndices.indexOf(highlightedIndex);
-			if (currentPos === -1 || currentPos === 0) {
-				highlightedIndex = compatibleIndices[compatibleIndices.length - 1];
+			if (highlightedIndex === -1 || highlightedIndex === 0) {
+				highlightedIndex = filteredOptions.length - 1;
 			} else {
-				highlightedIndex = compatibleIndices[currentPos - 1];
+				highlightedIndex -= 1;
 			}
-		} else if (event.key === 'Enter') {
+		} else if (event.key === KeyboardKey.ENTER) {
 			event.preventDefault();
 			if (highlightedIndex >= 0 && highlightedIndex < filteredOptions.length) {
 				const option = filteredOptions[highlightedIndex];
-				if (isModelCompatible(option)) {
-					handleSelect(option.id);
-				}
-			} else if (compatibleIndices.length > 0) {
-				// No selection - highlight first compatible option
-				highlightedIndex = compatibleIndices[0];
+				handleSelect(option.id);
+			} else if (filteredOptions.length > 0) {
+				// No selection - highlight first option
+				highlightedIndex = 0;
 			}
 		}
 	}
@@ -347,68 +266,72 @@
 		{@const selectedOption = getDisplayOption()}
 
 		{#if isRouter}
-			<Popover.Root bind:open={isOpen} onOpenChange={handleOpenChange}>
-				<Popover.Trigger
-					class={cn(
-						`inline-flex cursor-pointer items-center gap-1.5 rounded-sm bg-muted-foreground/10 px-1.5 py-1 text-xs transition hover:text-foreground focus:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-2 disabled:cursor-not-allowed disabled:opacity-60`,
-						!isCurrentModelInCache()
-							? 'bg-red-400/10 !text-red-400 hover:bg-red-400/20 hover:text-red-400'
-							: forceForegroundText
-								? 'text-foreground'
-								: isHighlightedCurrentModelActive
-									? 'text-foreground'
-									: 'text-muted-foreground',
-						isOpen ? 'text-foreground' : ''
-					)}
-					style="max-width: min(calc(100cqw - 6.5rem), 32rem)"
+			<DropdownMenu.Root bind:open={isOpen} onOpenChange={handleOpenChange}>
+				<DropdownMenu.Trigger
 					disabled={disabled || updating}
+					onclick={(e) => {
+						e.preventDefault();
+						e.stopPropagation();
+					}}
 				>
-					<Package class="h-3.5 w-3.5" />
+					<button
+						type="button"
+						class={cn(
+							`inline-grid cursor-pointer grid-cols-[1fr_auto_1fr] items-center gap-1.5 rounded-sm bg-muted-foreground/10 px-1.5 py-1 text-xs transition hover:text-foreground focus:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-2 disabled:cursor-not-allowed disabled:opacity-60`,
+							!isCurrentModelInCache()
+								? 'bg-red-400/10 !text-red-400 hover:bg-red-400/20 hover:text-red-400'
+								: forceForegroundText
+									? 'text-foreground'
+									: isHighlightedCurrentModelActive
+										? 'text-foreground'
+										: 'text-muted-foreground',
+							isOpen ? 'text-foreground' : ''
+						)}
+						style="max-width: min(calc(100cqw - 9rem), 20rem)"
+						disabled={disabled || updating}
+					>
+						<Package class="h-3.5 w-3.5" />
 
-					<span class="truncate font-medium">
-						{selectedOption?.model || 'Select model'}
-					</span>
+						<TruncatedText
+							text={selectedOption?.model || 'Select model'}
+							class="min-w-0 font-medium"
+						/>
 
-					{#if updating}
-						<Loader2 class="h-3 w-3.5 animate-spin" />
-					{:else}
-						<ChevronDown class="h-3 w-3.5" />
-					{/if}
-				</Popover.Trigger>
+						{#if updating}
+							<Loader2 class="h-3 w-3.5 animate-spin" />
+						{:else}
+							<ChevronDown class="h-3 w-3.5" />
+						{/if}
+					</button>
+				</DropdownMenu.Trigger>
 
-				<Popover.Content
-					class="group/popover-content w-96 max-w-[calc(100vw-2rem)] p-0"
+				<DropdownMenu.Content
 					align="end"
-					sideOffset={8}
-					collisionPadding={16}
+					class="w-full max-w-[100vw] pt-0 sm:w-max sm:max-w-[calc(100vw-2rem)]"
 				>
-					<div class="flex max-h-[50dvh] flex-col overflow-hidden">
-						<div
-							class="order-1 shrink-0 border-b p-4 group-data-[side=top]/popover-content:order-2 group-data-[side=top]/popover-content:border-t group-data-[side=top]/popover-content:border-b-0"
-						>
-							<SearchInput
-								id="model-search"
-								placeholder="Search models..."
-								bind:value={searchTerm}
-								bind:ref={searchInputRef}
-								onClose={() => handleOpenChange(false)}
-								onKeyDown={handleSearchKeyDown}
-							/>
-						</div>
-						<div
-							class="models-list order-2 min-h-0 flex-1 overflow-y-auto group-data-[side=top]/popover-content:order-1"
-						>
+					<DropdownMenuSearchable
+						bind:searchValue={searchTerm}
+						placeholder="Search models..."
+						onSearchKeyDown={handleSearchKeyDown}
+						emptyMessage="No models found."
+						isEmpty={filteredOptions.length === 0 && isCurrentModelInCache()}
+					>
+						<div class="models-list">
 							{#if !isCurrentModelInCache() && currentModel}
 								<!-- Show unavailable model as first option (disabled) -->
 								<button
 									type="button"
-									class="flex w-full cursor-not-allowed items-center bg-red-400/10 px-4 py-2 text-left text-sm text-red-400"
+									class="flex w-full cursor-not-allowed items-center bg-red-400/10 p-2 text-left text-sm text-red-400"
 									role="option"
 									aria-selected="true"
 									aria-disabled="true"
 									disabled
 								>
-									<span class="truncate">{selectedOption?.name || currentModel}</span>
+									<span
+										class="min-w-0 flex-1 truncate text-left sm:overflow-visible sm:text-clip sm:whitespace-nowrap"
+									>
+										{selectedOption?.name || currentModel}
+									</span>
 									<span class="ml-2 text-xs whitespace-nowrap opacity-70">(not available)</span>
 								</button>
 								<div class="my-1 h-px bg-border"></div>
@@ -421,104 +344,78 @@
 								{@const isLoaded = status === ServerModelStatus.LOADED}
 								{@const isLoading = status === ServerModelStatus.LOADING}
 								{@const isSelected = currentModel === option.model || activeId === option.id}
-								{@const isCompatible = isModelCompatible(option)}
 								{@const isHighlighted = index === highlightedIndex}
-								{@const missingModalities = getMissingModalities(option)}
 
 								<div
 									class={cn(
-										'group flex w-full items-center gap-2 px-4 py-2 text-left text-sm transition focus:outline-none',
-										isCompatible
-											? 'cursor-pointer hover:bg-muted focus:bg-muted'
-											: 'cursor-not-allowed opacity-50',
+										'group flex w-full items-center gap-2 rounded-sm p-2 text-left text-sm transition focus:outline-none',
+										'cursor-pointer hover:bg-muted focus:bg-muted',
 										isSelected || isHighlighted
 											? 'bg-accent text-accent-foreground'
-											: isCompatible
-												? 'hover:bg-accent hover:text-accent-foreground'
-												: '',
+											: 'hover:bg-accent hover:text-accent-foreground',
 										isLoaded ? 'text-popover-foreground' : 'text-muted-foreground'
 									)}
 									role="option"
 									aria-selected={isSelected || isHighlighted}
-									aria-disabled={!isCompatible}
-									tabindex={isCompatible ? 0 : -1}
-									onclick={() => isCompatible && handleSelect(option.id)}
+									tabindex="0"
+									onclick={() => handleSelect(option.id)}
 									onmouseenter={() => (highlightedIndex = index)}
 									onkeydown={(e) => {
-										if (isCompatible && (e.key === 'Enter' || e.key === ' ')) {
+										if (e.key === 'Enter' || e.key === ' ') {
 											e.preventDefault();
 											handleSelect(option.id);
 										}
 									}}
 								>
-									<span class="min-w-0 flex-1 truncate">{option.model}</span>
+									<span
+										class="min-w-0 flex-1 truncate text-left sm:overflow-visible sm:pr-2 sm:text-clip sm:whitespace-nowrap"
+									>
+										{option.model}
+									</span>
 
-									{#if missingModalities}
-										<span class="flex shrink-0 items-center gap-1 text-muted-foreground/70">
-											{#if missingModalities.vision}
-												<Tooltip.Root>
-													<Tooltip.Trigger>
-														<EyeOff class="h-3.5 w-3.5" />
-													</Tooltip.Trigger>
-													<Tooltip.Content class="z-[9999]">
-														<p>No vision support</p>
-													</Tooltip.Content>
-												</Tooltip.Root>
-											{/if}
-											{#if missingModalities.audio}
-												<Tooltip.Root>
-													<Tooltip.Trigger>
-														<MicOff class="h-3.5 w-3.5" />
-													</Tooltip.Trigger>
-													<Tooltip.Content class="z-[9999]">
-														<p>No audio support</p>
-													</Tooltip.Content>
-												</Tooltip.Root>
-											{/if}
-										</span>
-									{/if}
-
-									{#if isLoading}
-										<Tooltip.Root>
-											<Tooltip.Trigger>
-												<Loader2 class="h-4 w-4 shrink-0 animate-spin text-muted-foreground" />
-											</Tooltip.Trigger>
-											<Tooltip.Content class="z-[9999]">
-												<p>Loading model...</p>
-											</Tooltip.Content>
-										</Tooltip.Root>
-									{:else if isLoaded}
-										<Tooltip.Root>
-											<Tooltip.Trigger>
-												<button
-													type="button"
-													class="relative ml-2 flex h-4 w-4 shrink-0 items-center justify-center"
-													onclick={(e) => {
-														e.stopPropagation();
-														modelsStore.unloadModel(option.model);
-													}}
-												>
-													<span
-														class="mr-2 h-2 w-2 rounded-full bg-green-500 transition-opacity group-hover:opacity-0"
-													></span>
-													<Power
-														class="absolute mr-2 h-4 w-4 text-red-500 opacity-0 transition-opacity group-hover:opacity-100 hover:text-red-600"
-													/>
-												</button>
-											</Tooltip.Trigger>
-											<Tooltip.Content class="z-[9999]">
-												<p>Unload model</p>
-											</Tooltip.Content>
-										</Tooltip.Root>
-									{:else}
-										<span class="mx-2 h-2 w-2 rounded-full bg-muted-foreground/50"></span>
-									{/if}
+									<div class="flex w-6 shrink-0 justify-center">
+										{#if isLoading}
+											<Tooltip.Root>
+												<Tooltip.Trigger>
+													<Loader2 class="h-4 w-4 animate-spin text-muted-foreground" />
+												</Tooltip.Trigger>
+												<Tooltip.Content class="z-[9999]">
+													<p>Loading model...</p>
+												</Tooltip.Content>
+											</Tooltip.Root>
+										{:else if isLoaded}
+											<Tooltip.Root>
+												<Tooltip.Trigger>
+													<button
+														type="button"
+														class="relative flex h-4 w-4 items-center justify-center"
+														onclick={(e) => {
+															e.stopPropagation();
+															modelsStore.unloadModel(option.model);
+														}}
+													>
+														<span
+															class="h-2 w-2 rounded-full bg-green-500 transition-opacity group-hover:opacity-0"
+														></span>
+														<Power
+															class="absolute h-4 w-4 text-red-500 opacity-0 transition-opacity group-hover:opacity-100 hover:text-red-600"
+														/>
+													</button>
+												</Tooltip.Trigger>
+												<Tooltip.Content class="z-[9999]">
+													<p>Unload model</p>
+												</Tooltip.Content>
+											</Tooltip.Root>
+										{:else}
+											<span class="h-2 w-2 rounded-full bg-muted-foreground/50"></span>
+										{/if}
+									</div>
 								</div>
 							{/each}
 						</div>
-					</div>
-				</Popover.Content>
-			</Popover.Root>
+					</DropdownMenuSearchable>
+				</DropdownMenu.Content>
+			</DropdownMenu.Root>
 		{:else}
 			<button
 				class={cn(
@@ -538,9 +435,7 @@
 			>
 				<Package class="h-3.5 w-3.5" />
 
-				<span class="truncate font-medium">
-					{selectedOption?.model}
-				</span>
+				<TruncatedText text={selectedOption?.model || ''} class="min-w-0 font-medium" />
 
 				{#if updating}
 					<Loader2 class="h-3 w-3.5 animate-spin" />

tools/server/webui/src/lib/components/app/models/index.ts

@@ -0,0 +1,73 @@
+/**
+ *
+ * MODELS
+ *
+ * Components for model selection and display. Supports two server modes:
+ * - **Single model mode**: Server runs with one model, selector shows model info
+ * - **Router mode**: Server runs with multiple models, selector enables switching
+ *
+ * Integrates with modelsStore for model data and serverStore for mode detection.
+ *
+ */
+
+/**
+ * **ModelsSelector** - Model selection dropdown
+ *
+ * Dropdown for selecting AI models with status indicators,
+ * search, and model information display. Adapts UI based on server mode.
+ *
+ * **Architecture:**
+ * - Uses DropdownMenuSearchable for model list
+ * - Integrates with modelsStore for model options and selection
+ * - Detects router vs single mode from serverStore
+ * - Opens DialogModelInformation for model details
+ *
+ * **Features:**
+ * - Searchable model list with keyboard navigation
+ * - Model status indicators (loading/ready/error/updating)
+ * - Model capabilities badges (vision, tools, etc.)
+ * - Current/active model highlighting
+ * - Model information dialog on info button click
+ * - Router mode: shows all available models with status
+ * - Single mode: shows current model name only
+ * - Loading/updating skeleton states
+ * - Global selection support for form integration
+ *
+ * @example
+ * ```svelte
+ * <ModelsSelector
+ *   currentModel={conversation.modelId}
+ *   onModelChange={(id, name) => updateModel(id)}
+ *   useGlobalSelection
+ * />
+ * ```
+ */
+export { default as ModelsSelector } from './ModelsSelector.svelte';
+
+/**
+ * **ModelBadge** - Model name display badge
+ *
+ * Compact badge showing current model name with package icon.
+ * Only visible in single model mode. Supports tooltip and copy functionality.
+ *
+ * **Architecture:**
+ * - Reads model name from modelsStore or prop
+ * - Checks server mode from serverStore
+ * - Uses BadgeInfo for consistent styling
+ *
+ * **Features:**
+ * - Optional copy to clipboard button
+ * - Optional tooltip with model details
+ * - Click handler for model info dialog
+ * - Only renders in model mode (not router)
+ *
+ * @example
+ * ```svelte
+ * <ModelBadge
+ *   onclick={() => showModelInfo = true}
+ *   showTooltip
+ *   showCopyIcon
+ * />
+ * ```
+ */
+export { default as ModelBadge } from './ModelBadge.svelte';

tools/server/webui/src/lib/components/app/server/ServerErrorSplash.svelte

@@ -8,7 +8,7 @@
 	import { serverStore, serverLoading } from '$lib/stores/server.svelte';
 	import { config, settingsStore } from '$lib/stores/settings.svelte';
 	import { fade, fly, scale } from 'svelte/transition';
-	import { KeyboardKey } from '$lib/enums/keyboard';
+	import { KeyboardKey } from '$lib/enums';
 
 	interface Props {
 		class?: string;

tools/server/webui/src/lib/constants/binary-detection.ts

@@ -1,8 +1,4 @@
-export interface BinaryDetectionOptions {
-	prefixLength: number;
-	suspiciousCharThresholdRatio: number;
-	maxAbsoluteNullBytes: number;
-}
+import type { BinaryDetectionOptions } from '$lib/types';
 
 export const DEFAULT_BINARY_DETECTION_OPTIONS: BinaryDetectionOptions = {
 	prefixLength: 1024 * 10, // Check the first 10KB of the string

tools/server/webui/src/lib/constants/cache.ts

@@ -0,0 +1,33 @@
+/**
+ * Cache configuration constants
+ */
+
+/**
+ * Default TTL (Time-To-Live) for cache entries in milliseconds.
+ */
+export const DEFAULT_CACHE_TTL_MS = 5 * 60 * 1000;
+
+/**
+ * Default maximum number of entries in a cache.
+ */
+export const DEFAULT_CACHE_MAX_ENTRIES = 100;
+
+/**
+ * TTL for model props cache in milliseconds.
+ */
+export const MODEL_PROPS_CACHE_TTL_MS = 10 * 60 * 1000;
+
+/**
+ * Maximum number of model props to cache.
+ */
+export const MODEL_PROPS_CACHE_MAX_ENTRIES = 50;
+
+/**
+ * Maximum number of inactive conversation states to keep in memory.
+ */
+export const MAX_INACTIVE_CONVERSATION_STATES = 10;
+
+/**
+ * Maximum age (in ms) for inactive conversation states before cleanup.
+ */
+export const INACTIVE_CONVERSATION_STATE_MAX_AGE_MS = 30 * 60 * 1000;

tools/server/webui/src/lib/constants/input-classes.ts

@@ -1,6 +1 @@
-export const INPUT_CLASSES = `
-    bg-muted/70 dark:bg-muted/85
-    border border-border/30 focus-within:border-border  dark:border-border/20 dark:focus-within:border-border
-    outline-none
-    text-foreground
-`;
+export { INPUT_CLASSES } from './css-classes';

tools/server/webui/src/lib/constants/settings-sections.ts

@@ -0,0 +1,14 @@
+/**
+ * Settings section titles constants for ChatSettings component.
+ */
+export const SETTINGS_SECTION_TITLES = {
+	GENERAL: 'General',
+	DISPLAY: 'Display',
+	SAMPLING: 'Sampling',
+	PENALTIES: 'Penalties',
+	IMPORT_EXPORT: 'Import/Export',
+	DEVELOPER: 'Developer'
+} as const;
+
+export type SettingsSectionTitle =
+	(typeof SETTINGS_SECTION_TITLES)[keyof typeof SETTINGS_SECTION_TITLES];

tools/server/webui/src/lib/enums/index.ts

@@ -1,6 +1,13 @@
 export { AttachmentType } from './attachment';
 
-export { ChatMessageStatsView } from './chat';
+export {
+	ChatMessageStatsView,
+	ReasoningFormat,
+	MessageRole,
+	MessageType,
+	ContentPartType,
+	ErrorDialogType
+} from './chat';
 
 export {
 	FileTypeCategory,
@@ -21,3 +28,9 @@ export {
 export { ModelModality } from './model';
 
 export { ServerRole, ServerModelStatus } from './server';
+
+export { ParameterSource, SyncableParameterType, SettingsFieldType } from './settings';
+
+export { KeyboardKey } from './keyboard';
+
+export { UrlPrefix } from './ui';

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

@@ -0,0 +1,10 @@
+/**
+ * URL prefixes for protocol detection.
+ */
+export enum UrlPrefix {
+	DATA = 'data:',
+	HTTP = 'http://',
+	HTTPS = 'https://',
+	WEBSOCKET = 'ws://',
+	WEBSOCKET_SECURE = 'wss://'
+}

tools/server/webui/src/lib/hooks/use-model-change-validation.svelte.ts

@@ -1,26 +1,21 @@
 import { modelsStore } from '$lib/stores/models.svelte';
 import { isRouterMode } from '$lib/stores/server.svelte';
 import { toast } from 'svelte-sonner';
+import type { ModelModalities } from '$lib/types';
 
 interface UseModelChangeValidationOptions {
 	/**
 	 * Function to get required modalities for validation.
-	 * For ChatForm: () => usedModalities() - all messages
-	 * For ChatMessageAssistant: () => getModalitiesUpToMessage(messageId) - messages before
 	 */
 	getRequiredModalities: () => ModelModalities;
 
 	/**
 	 * Optional callback to execute after successful validation.
-	 * For ChatForm: undefined - just select model
-	 * For ChatMessageAssistant: (modelName) => onRegenerate(modelName)
 	 */
 	onSuccess?: (modelName: string) => void;
 
 	/**
 	 * Optional callback for rollback on validation failure.
-	 * For ChatForm: (previousId) => selectModelById(previousId)
-	 * For ChatMessageAssistant: undefined - no rollback needed
 	 */
 	onValidationFailure?: (previousModelId: string | null) => Promise<void>;
 }
@@ -33,12 +28,10 @@ export function useModelChangeValidation(options: UseModelChangeValidationOption
 
 	async function handleModelChange(modelId: string, modelName: string): Promise<boolean> {
 		try {
-			// Store previous selection for potential rollback
 			if (onValidationFailure) {
 				previousSelectedModelId = modelsStore.selectedModelId;
 			}
 
-			// Load model if not already loaded (router mode only)
 			let hasLoadedModel = false;
 			const isModelLoadedBefore = modelsStore.isModelLoaded(modelName);
 
@@ -52,13 +45,11 @@ export function useModelChangeValidation(options: UseModelChangeValidationOption
 				}
 			}
 
-			// Fetch model props to validate modalities
 			const props = await modelsStore.fetchModelProps(modelName);
 
 			if (props?.modalities) {
 				const requiredModalities = getRequiredModalities();
 
-				// Check if model supports required modalities
 				const missingModalities: string[] = [];
 				if (requiredModalities.vision && !props.modalities.vision) {
 					missingModalities.push('vision');
@@ -72,7 +63,6 @@ export function useModelChangeValidation(options: UseModelChangeValidationOption
 						`Model "${modelName}" doesn't support required modalities: ${missingModalities.join(', ')}. Please select a different model.`
 					);
 
-					// Unload the model if we just loaded it
 					if (isRouter && hasLoadedModel) {
 						try {
 							await modelsStore.unloadModel(modelName);
@@ -81,7 +71,6 @@ export function useModelChangeValidation(options: UseModelChangeValidationOption
 						}
 					}
 
-					// Execute rollback callback if provided
 					if (onValidationFailure && previousSelectedModelId) {
 						await onValidationFailure(previousSelectedModelId);
 					}
@@ -90,10 +79,8 @@ export function useModelChangeValidation(options: UseModelChangeValidationOption
 				}
 			}
 
-			// Select the model (validation passed)
 			await modelsStore.selectModelById(modelId);
 
-			// Execute success callback if provided
 			if (onSuccess) {
 				onSuccess(modelName);
 			}
@@ -103,7 +90,6 @@ export function useModelChangeValidation(options: UseModelChangeValidationOption
 			console.error('Failed to change model:', error);
 			toast.error('Failed to validate model capabilities');
 
-			// Execute rollback callback on error if provided
 			if (onValidationFailure && previousSelectedModelId) {
 				await onValidationFailure(previousSelectedModelId);
 			}

tools/server/webui/src/lib/hooks/use-processing-state.svelte.ts

@@ -1,21 +1,7 @@
 import { activeProcessingState } from '$lib/stores/chat.svelte';
 import { config } from '$lib/stores/settings.svelte';
 import { STATS_UNITS } from '$lib/constants/processing-info';
-import type { ApiProcessingState } from '$lib/types';
-
-interface LiveProcessingStats {
-	tokensProcessed: number;
-	totalTokens: number;
-	timeMs: number;
-	tokensPerSecond: number;
-	etaSecs?: number;
-}
-
-interface LiveGenerationStats {
-	tokensGenerated: number;
-	timeMs: number;
-	tokensPerSecond: number;
-}
+import type { ApiProcessingState, LiveProcessingStats, LiveGenerationStats } from '$lib/types';
 
 export interface UseProcessingStateReturn {
 	readonly processingState: ApiProcessingState | null;

tools/server/webui/src/lib/markdown/resolve-attachment-images.ts

@@ -0,0 +1,31 @@
+import type { Root as HastRoot } from 'hast';
+import { visit } from 'unist-util-visit';
+import type { DatabaseMessageExtra, DatabaseMessageExtraImageFile } from '$lib/types/database';
+import { AttachmentType, UrlPrefix } from '$lib/enums';
+
+/**
+ * Rehype plugin to resolve attachment image sources.
+ * Converts attachment names to base64 data URLs.
+ */
+export function rehypeResolveAttachmentImages(options: { attachments?: DatabaseMessageExtra[] }) {
+	return (tree: HastRoot) => {
+		visit(tree, 'element', (node) => {
+			if (node.tagName === 'img' && node.properties?.src) {
+				const src = String(node.properties.src);
+
+				if (src.startsWith(UrlPrefix.DATA) || src.startsWith(UrlPrefix.HTTP)) {
+					return;
+				}
+
+				const attachment = options.attachments?.find(
+					(a): a is DatabaseMessageExtraImageFile =>
+						a.type === AttachmentType.IMAGE && a.name === src
+				);
+
+				if (attachment?.base64Url) {
+					node.properties.src = attachment.base64Url;
+				}
+			}
+		});
+	};
+}

tools/server/webui/src/lib/services/database.service.ts

@@ -17,7 +17,7 @@ class LlamacppDatabase extends Dexie {
 
 const db = new LlamacppDatabase();
 import { v4 as uuid } from 'uuid';
-import { MessageRole } from '$lib/enums/chat';
+import { MessageRole } from '$lib/enums';
 
 export class DatabaseService {
 	/**

tools/server/webui/src/lib/services/database.ts

@@ -1,400 +0,0 @@
-import Dexie, { type EntityTable } from 'dexie';
-import { findDescendantMessages } from '$lib/utils';
-
-class LlamacppDatabase extends Dexie {
-	conversations!: EntityTable<DatabaseConversation, string>;
-	messages!: EntityTable<DatabaseMessage, string>;
-
-	constructor() {
-		super('LlamacppWebui');
-
-		this.version(1).stores({
-			conversations: 'id, lastModified, currNode, name',
-			messages: 'id, convId, type, role, timestamp, parent, children'
-		});
-	}
-}
-
-const db = new LlamacppDatabase();
-import { v4 as uuid } from 'uuid';
-
-/**
- * DatabaseService - Stateless IndexedDB communication layer
- *
- * **Terminology - Chat vs Conversation:**
- * - **Chat**: The active interaction space with the Chat Completions API (ephemeral, runtime).
- * - **Conversation**: The persistent database entity storing all messages and metadata.
- *   This service handles raw database operations for conversations - the lowest layer
- *   in the persistence stack.
- *
- * This service provides a stateless data access layer built on IndexedDB using Dexie ORM.
- * It handles all low-level storage operations for conversations and messages with support
- * for complex branching and message threading. All methods are static - no instance state.
- *
- * **Architecture & Relationships (bottom to top):**
- * - **DatabaseService** (this class): Stateless IndexedDB operations
- *   - Lowest layer - direct Dexie/IndexedDB communication
- *   - Pure CRUD operations without business logic
- *   - Handles branching tree structure (parent-child relationships)
- *   - Provides transaction safety for multi-table operations
- *
- * - **ConversationsService**: Stateless business logic layer
- *   - Uses DatabaseService for all persistence operations
- *   - Adds import/export, navigation, and higher-level operations
- *
- * - **conversationsStore**: Reactive state management for conversations
- *   - Uses ConversationsService for database operations
- *   - Manages conversation list, active conversation, and messages in memory
- *
- * - **chatStore**: Active AI interaction management
- *   - Uses conversationsStore for conversation context
- *   - Directly uses DatabaseService for message CRUD during streaming
- *
- * **Key Features:**
- * - **Conversation CRUD**: Create, read, update, delete conversations
- * - **Message CRUD**: Add, update, delete messages with branching support
- * - **Branch Operations**: Create branches, find descendants, cascade deletions
- * - **Transaction Safety**: Atomic operations for data consistency
- *
- * **Database Schema:**
- * - `conversations`: id, lastModified, currNode, name
- * - `messages`: id, convId, type, role, timestamp, parent, children
- *
- * **Branching Model:**
- * Messages form a tree structure where each message can have multiple children,
- * enabling conversation branching and alternative response paths. The conversation's
- * `currNode` tracks the currently active branch endpoint.
- */
-export class DatabaseService {
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Conversations
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Creates a new conversation.
-	 *
-	 * @param name - Name of the conversation
-	 * @returns The created conversation
-	 */
-	static async createConversation(name: string): Promise<DatabaseConversation> {
-		const conversation: DatabaseConversation = {
-			id: uuid(),
-			name,
-			lastModified: Date.now(),
-			currNode: ''
-		};
-
-		await db.conversations.add(conversation);
-		return conversation;
-	}
-
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Messages
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Creates a new message branch by adding a message and updating parent/child relationships.
-	 * Also updates the conversation's currNode to point to the new message.
-	 *
-	 * @param message - Message to add (without id)
-	 * @param parentId - Parent message ID to attach to
-	 * @returns The created message
-	 */
-	static async createMessageBranch(
-		message: Omit<DatabaseMessage, 'id'>,
-		parentId: string | null
-	): Promise<DatabaseMessage> {
-		return await db.transaction('rw', [db.conversations, db.messages], async () => {
-			// Handle null parent (root message case)
-			if (parentId !== null) {
-				const parentMessage = await db.messages.get(parentId);
-				if (!parentMessage) {
-					throw new Error(`Parent message ${parentId} not found`);
-				}
-			}
-
-			const newMessage: DatabaseMessage = {
-				...message,
-				id: uuid(),
-				parent: parentId,
-				toolCalls: message.toolCalls ?? '',
-				children: []
-			};
-
-			await db.messages.add(newMessage);
-
-			// Update parent's children array if parent exists
-			if (parentId !== null) {
-				const parentMessage = await db.messages.get(parentId);
-				if (parentMessage) {
-					await db.messages.update(parentId, {
-						children: [...parentMessage.children, newMessage.id]
-					});
-				}
-			}
-
-			await this.updateConversation(message.convId, {
-				currNode: newMessage.id
-			});
-
-			return newMessage;
-		});
-	}
-
-	/**
-	 * Creates a root message for a new conversation.
-	 * Root messages are not displayed but serve as the tree root for branching.
-	 *
-	 * @param convId - Conversation ID
-	 * @returns The created root message
-	 */
-	static async createRootMessage(convId: string): Promise<string> {
-		const rootMessage: DatabaseMessage = {
-			id: uuid(),
-			convId,
-			type: 'root',
-			timestamp: Date.now(),
-			role: 'system',
-			content: '',
-			parent: null,
-			thinking: '',
-			toolCalls: '',
-			children: []
-		};
-
-		await db.messages.add(rootMessage);
-		return rootMessage.id;
-	}
-
-	/**
-	 * Creates a system prompt message for a conversation.
-	 *
-	 * @param convId - Conversation ID
-	 * @param systemPrompt - The system prompt content (must be non-empty)
-	 * @param parentId - Parent message ID (typically the root message)
-	 * @returns The created system message
-	 * @throws Error if systemPrompt is empty
-	 */
-	static async createSystemMessage(
-		convId: string,
-		systemPrompt: string,
-		parentId: string
-	): Promise<DatabaseMessage> {
-		const trimmedPrompt = systemPrompt.trim();
-		if (!trimmedPrompt) {
-			throw new Error('Cannot create system message with empty content');
-		}
-
-		const systemMessage: DatabaseMessage = {
-			id: uuid(),
-			convId,
-			type: 'system',
-			timestamp: Date.now(),
-			role: 'system',
-			content: trimmedPrompt,
-			parent: parentId,
-			thinking: '',
-			children: []
-		};
-
-		await db.messages.add(systemMessage);
-
-		const parentMessage = await db.messages.get(parentId);
-		if (parentMessage) {
-			await db.messages.update(parentId, {
-				children: [...parentMessage.children, systemMessage.id]
-			});
-		}
-
-		return systemMessage;
-	}
-
-	/**
-	 * Deletes a conversation and all its messages.
-	 *
-	 * @param id - Conversation ID
-	 */
-	static async deleteConversation(id: string): Promise<void> {
-		await db.transaction('rw', [db.conversations, db.messages], async () => {
-			await db.conversations.delete(id);
-			await db.messages.where('convId').equals(id).delete();
-		});
-	}
-
-	/**
-	 * Deletes a message and removes it from its parent's children array.
-	 *
-	 * @param messageId - ID of the message to delete
-	 */
-	static async deleteMessage(messageId: string): Promise<void> {
-		await db.transaction('rw', db.messages, async () => {
-			const message = await db.messages.get(messageId);
-			if (!message) return;
-
-			// Remove this message from its parent's children array
-			if (message.parent) {
-				const parent = await db.messages.get(message.parent);
-				if (parent) {
-					parent.children = parent.children.filter((childId: string) => childId !== messageId);
-					await db.messages.put(parent);
-				}
-			}
-
-			// Delete the message
-			await db.messages.delete(messageId);
-		});
-	}
-
-	/**
-	 * Deletes a message and all its descendant messages (cascading deletion).
-	 * This removes the entire branch starting from the specified message.
-	 *
-	 * @param conversationId - ID of the conversation containing the message
-	 * @param messageId - ID of the root message to delete (along with all descendants)
-	 * @returns Array of all deleted message IDs
-	 */
-	static async deleteMessageCascading(
-		conversationId: string,
-		messageId: string
-	): Promise<string[]> {
-		return await db.transaction('rw', db.messages, async () => {
-			// Get all messages in the conversation to find descendants
-			const allMessages = await db.messages.where('convId').equals(conversationId).toArray();
-
-			// Find all descendant messages
-			const descendants = findDescendantMessages(allMessages, messageId);
-			const allToDelete = [messageId, ...descendants];
-
-			// Get the message to delete for parent cleanup
-			const message = await db.messages.get(messageId);
-			if (message && message.parent) {
-				const parent = await db.messages.get(message.parent);
-				if (parent) {
-					parent.children = parent.children.filter((childId: string) => childId !== messageId);
-					await db.messages.put(parent);
-				}
-			}
-
-			// Delete all messages in the branch
-			await db.messages.bulkDelete(allToDelete);
-
-			return allToDelete;
-		});
-	}
-
-	/**
-	 * Gets all conversations, sorted by last modified time (newest first).
-	 *
-	 * @returns Array of conversations
-	 */
-	static async getAllConversations(): Promise<DatabaseConversation[]> {
-		return await db.conversations.orderBy('lastModified').reverse().toArray();
-	}
-
-	/**
-	 * Gets a conversation by ID.
-	 *
-	 * @param id - Conversation ID
-	 * @returns The conversation if found, otherwise undefined
-	 */
-	static async getConversation(id: string): Promise<DatabaseConversation | undefined> {
-		return await db.conversations.get(id);
-	}
-
-	/**
-	 * Gets all messages in a conversation, sorted by timestamp (oldest first).
-	 *
-	 * @param convId - Conversation ID
-	 * @returns Array of messages in the conversation
-	 */
-	static async getConversationMessages(convId: string): Promise<DatabaseMessage[]> {
-		return await db.messages.where('convId').equals(convId).sortBy('timestamp');
-	}
-
-	/**
-	 * Updates a conversation.
-	 *
-	 * @param id - Conversation ID
-	 * @param updates - Partial updates to apply
-	 * @returns Promise that resolves when the conversation is updated
-	 */
-	static async updateConversation(
-		id: string,
-		updates: Partial<Omit<DatabaseConversation, 'id'>>
-	): Promise<void> {
-		await db.conversations.update(id, {
-			...updates,
-			lastModified: Date.now()
-		});
-	}
-
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Navigation
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Updates the conversation's current node (active branch).
-	 * This determines which conversation path is currently being viewed.
-	 *
-	 * @param convId - Conversation ID
-	 * @param nodeId - Message ID to set as current node
-	 */
-	static async updateCurrentNode(convId: string, nodeId: string): Promise<void> {
-		await this.updateConversation(convId, {
-			currNode: nodeId
-		});
-	}
-
-	/**
-	 * Updates a message.
-	 *
-	 * @param id - Message ID
-	 * @param updates - Partial updates to apply
-	 * @returns Promise that resolves when the message is updated
-	 */
-	static async updateMessage(
-		id: string,
-		updates: Partial<Omit<DatabaseMessage, 'id'>>
-	): Promise<void> {
-		await db.messages.update(id, updates);
-	}
-
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Import
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Imports multiple conversations and their messages.
-	 * Skips conversations that already exist.
-	 *
-	 * @param data - Array of { conv, messages } objects
-	 */
-	static async importConversations(
-		data: { conv: DatabaseConversation; messages: DatabaseMessage[] }[]
-	): Promise<{ imported: number; skipped: number }> {
-		let importedCount = 0;
-		let skippedCount = 0;
-
-		return await db.transaction('rw', [db.conversations, db.messages], async () => {
-			for (const item of data) {
-				const { conv, messages } = item;
-
-				const existing = await db.conversations.get(conv.id);
-				if (existing) {
-					console.warn(`Conversation "${conv.name}" already exists, skipping...`);
-					skippedCount++;
-					continue;
-				}
-
-				await db.conversations.add(conv);
-				for (const msg of messages) {
-					await db.messages.put(msg);
-				}
-
-				importedCount++;
-			}
-
-			return { imported: importedCount, skipped: skippedCount };
-		});
-	}
-}

tools/server/webui/src/lib/services/index.ts

@@ -1,5 +1,5 @@
 export { ChatService } from './chat';
-export { DatabaseService } from './database';
-export { ModelsService } from './models';
-export { PropsService } from './props';
-export { ParameterSyncService } from './parameter-sync';
+export { DatabaseService } from './database.service';
+export { ModelsService } from './models.service';
+export { PropsService } from './props.service';
+export { ParameterSyncService, SYNCABLE_PARAMETERS } from './parameter-sync.service';

tools/server/webui/src/lib/services/models.service.ts

@@ -1,5 +1,5 @@
 import { ServerModelStatus } from '$lib/enums';
-import { apiFetch, apiPost } from '$lib/utils/api-fetch';
+import { apiFetch, apiPost } from '$lib/utils';
 
 export class ModelsService {
 	/**

tools/server/webui/src/lib/services/models.ts

@@ -1,124 +0,0 @@
-import { base } from '$app/paths';
-import { ServerModelStatus } from '$lib/enums';
-import { getJsonHeaders } from '$lib/utils';
-
-/**
- * ModelsService - Stateless service for model management API communication
- *
- * This service handles communication with model-related endpoints:
- * - `/v1/models` - OpenAI-compatible model list (MODEL + ROUTER mode)
- * - `/models/load`, `/models/unload` - Router-specific model management (ROUTER mode only)
- *
- * **Responsibilities:**
- * - List available models
- * - Load/unload models (ROUTER mode)
- * - Check model status (ROUTER mode)
- *
- * **Used by:**
- * - modelsStore: Primary consumer for model state management
- */
-export class ModelsService {
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Listing
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Fetch list of models from OpenAI-compatible endpoint
-	 * Works in both MODEL and ROUTER modes
-	 */
-	static async list(): Promise<ApiModelListResponse> {
-		const response = await fetch(`${base}/v1/models`, {
-			headers: getJsonHeaders()
-		});
-
-		if (!response.ok) {
-			throw new Error(`Failed to fetch model list (status ${response.status})`);
-		}
-
-		return response.json() as Promise<ApiModelListResponse>;
-	}
-
-	/**
-	 * Fetch list of all models with detailed metadata (ROUTER mode)
-	 * Returns models with load status, paths, and other metadata
-	 */
-	static async listRouter(): Promise<ApiRouterModelsListResponse> {
-		const response = await fetch(`${base}/v1/models`, {
-			headers: getJsonHeaders()
-		});
-
-		if (!response.ok) {
-			throw new Error(`Failed to fetch router models list (status ${response.status})`);
-		}
-
-		return response.json() as Promise<ApiRouterModelsListResponse>;
-	}
-
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Load/Unload
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Load a model (ROUTER mode)
-	 * POST /models/load
-	 * @param modelId - Model identifier to load
-	 * @param extraArgs - Optional additional arguments to pass to the model instance
-	 */
-	static async load(modelId: string, extraArgs?: string[]): Promise<ApiRouterModelsLoadResponse> {
-		const payload: { model: string; extra_args?: string[] } = { model: modelId };
-		if (extraArgs && extraArgs.length > 0) {
-			payload.extra_args = extraArgs;
-		}
-
-		const response = await fetch(`${base}/models/load`, {
-			method: 'POST',
-			headers: getJsonHeaders(),
-			body: JSON.stringify(payload)
-		});
-
-		if (!response.ok) {
-			const errorData = await response.json().catch(() => ({}));
-			throw new Error(errorData.error || `Failed to load model (status ${response.status})`);
-		}
-
-		return response.json() as Promise<ApiRouterModelsLoadResponse>;
-	}
-
-	/**
-	 * Unload a model (ROUTER mode)
-	 * POST /models/unload
-	 * @param modelId - Model identifier to unload
-	 */
-	static async unload(modelId: string): Promise<ApiRouterModelsUnloadResponse> {
-		const response = await fetch(`${base}/models/unload`, {
-			method: 'POST',
-			headers: getJsonHeaders(),
-			body: JSON.stringify({ model: modelId })
-		});
-
-		if (!response.ok) {
-			const errorData = await response.json().catch(() => ({}));
-			throw new Error(errorData.error || `Failed to unload model (status ${response.status})`);
-		}
-
-		return response.json() as Promise<ApiRouterModelsUnloadResponse>;
-	}
-
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Status
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Check if a model is loaded based on its metadata
-	 */
-	static isModelLoaded(model: ApiModelDataEntry): boolean {
-		return model.status.value === ServerModelStatus.LOADED;
-	}
-
-	/**
-	 * Check if a model is currently loading
-	 */
-	static isModelLoading(model: ApiModelDataEntry): boolean {
-		return model.status.value === ServerModelStatus.LOADING;
-	}
-}

tools/server/webui/src/lib/services/parameter-sync.service.ts

@@ -1,22 +1,6 @@
 import { normalizeFloatingPoint } from '$lib/utils';
-import { SyncableParameterType, ParameterSource } from '$lib/enums/settings';
-
-type ParameterValue = string | number | boolean;
-type ParameterRecord = Record<string, ParameterValue>;
-
-interface ParameterInfo {
-	value: string | number | boolean;
-	source: ParameterSource;
-	serverDefault?: string | number | boolean;
-	userOverride?: string | number | boolean;
-}
-
-interface SyncableParameter {
-	key: string;
-	serverKey: string;
-	type: SyncableParameterType;
-	canSync: boolean;
-}
+import type { SyncableParameter, ParameterRecord, ParameterInfo, ParameterValue } from '$lib/types';
+import { SyncableParameterType, ParameterSource } from '$lib/enums';
 
 /**
  * Mapping of webui setting keys to server parameter keys.

tools/server/webui/src/lib/services/parameter-sync.spec.ts

@@ -1,148 +0,0 @@
-import { describe, it, expect } from 'vitest';
-import { ParameterSyncService } from './parameter-sync';
-
-describe('ParameterSyncService', () => {
-	describe('roundFloatingPoint', () => {
-		it('should fix JavaScript floating-point precision issues', () => {
-			// Test the specific values from the screenshot
-			const mockServerParams = {
-				top_p: 0.949999988079071,
-				min_p: 0.009999999776482582,
-				temperature: 0.800000011920929,
-				top_k: 40,
-				samplers: ['top_k', 'typ_p', 'top_p', 'min_p', 'temperature']
-			};
-
-			const result = ParameterSyncService.extractServerDefaults({
-				...mockServerParams,
-				// Add other required fields to match the API type
-				n_predict: 512,
-				seed: -1,
-				dynatemp_range: 0.0,
-				dynatemp_exponent: 1.0,
-				xtc_probability: 0.0,
-				xtc_threshold: 0.1,
-				typ_p: 1.0,
-				repeat_last_n: 64,
-				repeat_penalty: 1.0,
-				presence_penalty: 0.0,
-				frequency_penalty: 0.0,
-				dry_multiplier: 0.0,
-				dry_base: 1.75,
-				dry_allowed_length: 2,
-				dry_penalty_last_n: -1,
-				mirostat: 0,
-				mirostat_tau: 5.0,
-				mirostat_eta: 0.1,
-				stop: [],
-				max_tokens: -1,
-				n_keep: 0,
-				n_discard: 0,
-				ignore_eos: false,
-				stream: true,
-				logit_bias: [],
-				n_probs: 0,
-				min_keep: 0,
-				grammar: '',
-				grammar_lazy: false,
-				grammar_triggers: [],
-				preserved_tokens: [],
-				chat_format: '',
-				reasoning_format: '',
-				reasoning_in_content: false,
-				thinking_forced_open: false,
-				'speculative.n_max': 0,
-				'speculative.n_min': 0,
-				'speculative.p_min': 0.0,
-				timings_per_token: false,
-				post_sampling_probs: false,
-				lora: [],
-				top_n_sigma: 0.0,
-				dry_sequence_breakers: []
-			} as ApiLlamaCppServerProps['default_generation_settings']['params']);
-
-			// Check that the problematic floating-point values are rounded correctly
-			expect(result.top_p).toBe(0.95);
-			expect(result.min_p).toBe(0.01);
-			expect(result.temperature).toBe(0.8);
-			expect(result.top_k).toBe(40); // Integer should remain unchanged
-			expect(result.samplers).toBe('top_k;typ_p;top_p;min_p;temperature');
-		});
-
-		it('should preserve non-numeric values', () => {
-			const mockServerParams = {
-				samplers: ['top_k', 'temperature'],
-				max_tokens: -1,
-				temperature: 0.7
-			};
-
-			const result = ParameterSyncService.extractServerDefaults({
-				...mockServerParams,
-				// Minimal required fields
-				n_predict: 512,
-				seed: -1,
-				dynatemp_range: 0.0,
-				dynatemp_exponent: 1.0,
-				top_k: 40,
-				top_p: 0.95,
-				min_p: 0.05,
-				xtc_probability: 0.0,
-				xtc_threshold: 0.1,
-				typ_p: 1.0,
-				repeat_last_n: 64,
-				repeat_penalty: 1.0,
-				presence_penalty: 0.0,
-				frequency_penalty: 0.0,
-				dry_multiplier: 0.0,
-				dry_base: 1.75,
-				dry_allowed_length: 2,
-				dry_penalty_last_n: -1,
-				mirostat: 0,
-				mirostat_tau: 5.0,
-				mirostat_eta: 0.1,
-				stop: [],
-				n_keep: 0,
-				n_discard: 0,
-				ignore_eos: false,
-				stream: true,
-				logit_bias: [],
-				n_probs: 0,
-				min_keep: 0,
-				grammar: '',
-				grammar_lazy: false,
-				grammar_triggers: [],
-				preserved_tokens: [],
-				chat_format: '',
-				reasoning_format: '',
-				reasoning_in_content: false,
-				thinking_forced_open: false,
-				'speculative.n_max': 0,
-				'speculative.n_min': 0,
-				'speculative.p_min': 0.0,
-				timings_per_token: false,
-				post_sampling_probs: false,
-				lora: [],
-				top_n_sigma: 0.0,
-				dry_sequence_breakers: []
-			} as ApiLlamaCppServerProps['default_generation_settings']['params']);
-
-			expect(result.samplers).toBe('top_k;temperature');
-			expect(result.max_tokens).toBe(-1);
-			expect(result.temperature).toBe(0.7);
-		});
-
-		it('should merge webui settings from props when provided', () => {
-			const result = ParameterSyncService.extractServerDefaults(null, {
-				pasteLongTextToFileLen: 0,
-				pdfAsImage: true,
-				renderUserContentAsMarkdown: false,
-				theme: 'dark'
-			});
-
-			expect(result.pasteLongTextToFileLen).toBe(0);
-			expect(result.pdfAsImage).toBe(true);
-			expect(result.renderUserContentAsMarkdown).toBe(false);
-			expect(result.theme).toBeUndefined();
-		});
-	});
-});

tools/server/webui/src/lib/services/parameter-sync.ts

@@ -1,273 +0,0 @@
-/**
- * ParameterSyncService - Handles synchronization between server defaults and user settings
- *
- * This service manages the complex logic of merging server-provided default parameters
- * with user-configured overrides, ensuring the UI reflects the actual server state
- * while preserving user customizations.
- *
- * **Key Responsibilities:**
- * - Extract syncable parameters from server props
- * - Merge server defaults with user overrides
- * - Track parameter sources (server, user, default)
- * - Provide sync utilities for settings store integration
- */
-
-import { normalizeFloatingPoint } from '$lib/utils';
-
-export type ParameterSource = 'default' | 'custom';
-export type ParameterValue = string | number | boolean;
-export type ParameterRecord = Record<string, ParameterValue>;
-
-export interface ParameterInfo {
-	value: string | number | boolean;
-	source: ParameterSource;
-	serverDefault?: string | number | boolean;
-	userOverride?: string | number | boolean;
-}
-
-export interface SyncableParameter {
-	key: string;
-	serverKey: string;
-	type: 'number' | 'string' | 'boolean';
-	canSync: boolean;
-}
-
-/**
- * Mapping of webui setting keys to server parameter keys
- * Only parameters that should be synced from server are included
- */
-export const SYNCABLE_PARAMETERS: SyncableParameter[] = [
-	{ key: 'temperature', serverKey: 'temperature', type: 'number', canSync: true },
-	{ key: 'top_k', serverKey: 'top_k', type: 'number', canSync: true },
-	{ key: 'top_p', serverKey: 'top_p', type: 'number', canSync: true },
-	{ key: 'min_p', serverKey: 'min_p', type: 'number', canSync: true },
-	{ key: 'dynatemp_range', serverKey: 'dynatemp_range', type: 'number', canSync: true },
-	{ key: 'dynatemp_exponent', serverKey: 'dynatemp_exponent', type: 'number', canSync: true },
-	{ key: 'xtc_probability', serverKey: 'xtc_probability', type: 'number', canSync: true },
-	{ key: 'xtc_threshold', serverKey: 'xtc_threshold', type: 'number', canSync: true },
-	{ key: 'typ_p', serverKey: 'typ_p', type: 'number', canSync: true },
-	{ key: 'repeat_last_n', serverKey: 'repeat_last_n', type: 'number', canSync: true },
-	{ key: 'repeat_penalty', serverKey: 'repeat_penalty', type: 'number', canSync: true },
-	{ key: 'presence_penalty', serverKey: 'presence_penalty', type: 'number', canSync: true },
-	{ key: 'frequency_penalty', serverKey: 'frequency_penalty', type: 'number', canSync: true },
-	{ key: 'dry_multiplier', serverKey: 'dry_multiplier', type: 'number', canSync: true },
-	{ key: 'dry_base', serverKey: 'dry_base', type: 'number', canSync: true },
-	{ key: 'dry_allowed_length', serverKey: 'dry_allowed_length', type: 'number', canSync: true },
-	{ key: 'dry_penalty_last_n', serverKey: 'dry_penalty_last_n', type: 'number', canSync: true },
-	{ key: 'max_tokens', serverKey: 'max_tokens', type: 'number', canSync: true },
-	{ key: 'samplers', serverKey: 'samplers', type: 'string', canSync: true },
-	{
-		key: 'pasteLongTextToFileLen',
-		serverKey: 'pasteLongTextToFileLen',
-		type: 'number',
-		canSync: true
-	},
-	{ key: 'pdfAsImage', serverKey: 'pdfAsImage', type: 'boolean', canSync: true },
-	{
-		key: 'showThoughtInProgress',
-		serverKey: 'showThoughtInProgress',
-		type: 'boolean',
-		canSync: true
-	},
-	{ key: 'showToolCalls', serverKey: 'showToolCalls', type: 'boolean', canSync: true },
-	{ key: 'keepStatsVisible', serverKey: 'keepStatsVisible', type: 'boolean', canSync: true },
-	{ key: 'showMessageStats', serverKey: 'showMessageStats', type: 'boolean', canSync: true },
-	{
-		key: 'askForTitleConfirmation',
-		serverKey: 'askForTitleConfirmation',
-		type: 'boolean',
-		canSync: true
-	},
-	{ key: 'disableAutoScroll', serverKey: 'disableAutoScroll', type: 'boolean', canSync: true },
-	{
-		key: 'renderUserContentAsMarkdown',
-		serverKey: 'renderUserContentAsMarkdown',
-		type: 'boolean',
-		canSync: true
-	},
-	{ key: 'autoMicOnEmpty', serverKey: 'autoMicOnEmpty', type: 'boolean', canSync: true },
-	{
-		key: 'pyInterpreterEnabled',
-		serverKey: 'pyInterpreterEnabled',
-		type: 'boolean',
-		canSync: true
-	},
-	{
-		key: 'enableContinueGeneration',
-		serverKey: 'enableContinueGeneration',
-		type: 'boolean',
-		canSync: true
-	}
-];
-
-export class ParameterSyncService {
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Extraction
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Round floating-point numbers to avoid JavaScript precision issues
-	 */
-	private static roundFloatingPoint(value: ParameterValue): ParameterValue {
-		return normalizeFloatingPoint(value) as ParameterValue;
-	}
-
-	/**
-	 * Extract server default parameters that can be synced
-	 */
-	static extractServerDefaults(
-		serverParams: ApiLlamaCppServerProps['default_generation_settings']['params'] | null,
-		webuiSettings?: Record<string, string | number | boolean>
-	): ParameterRecord {
-		const extracted: ParameterRecord = {};
-
-		if (serverParams) {
-			for (const param of SYNCABLE_PARAMETERS) {
-				if (param.canSync && param.serverKey in serverParams) {
-					const value = (serverParams as unknown as Record<string, ParameterValue>)[
-						param.serverKey
-					];
-					if (value !== undefined) {
-						// Apply precision rounding to avoid JavaScript floating-point issues
-						extracted[param.key] = this.roundFloatingPoint(value);
-					}
-				}
-			}
-
-			// Handle samplers array conversion to string
-			if (serverParams.samplers && Array.isArray(serverParams.samplers)) {
-				extracted.samplers = serverParams.samplers.join(';');
-			}
-		}
-
-		if (webuiSettings) {
-			for (const param of SYNCABLE_PARAMETERS) {
-				if (param.canSync && param.serverKey in webuiSettings) {
-					const value = webuiSettings[param.serverKey];
-					if (value !== undefined) {
-						extracted[param.key] = this.roundFloatingPoint(value);
-					}
-				}
-			}
-		}
-
-		return extracted;
-	}
-
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Merging
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Merge server defaults with current user settings
-	 * Returns updated settings that respect user overrides while using server defaults
-	 */
-	static mergeWithServerDefaults(
-		currentSettings: ParameterRecord,
-		serverDefaults: ParameterRecord,
-		userOverrides: Set<string> = new Set()
-	): ParameterRecord {
-		const merged = { ...currentSettings };
-
-		for (const [key, serverValue] of Object.entries(serverDefaults)) {
-			// Only update if user hasn't explicitly overridden this parameter
-			if (!userOverrides.has(key)) {
-				merged[key] = this.roundFloatingPoint(serverValue);
-			}
-		}
-
-		return merged;
-	}
-
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Info
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Get parameter information including source and values
-	 */
-	static getParameterInfo(
-		key: string,
-		currentValue: ParameterValue,
-		propsDefaults: ParameterRecord,
-		userOverrides: Set<string>
-	): ParameterInfo {
-		const hasPropsDefault = propsDefaults[key] !== undefined;
-		const isUserOverride = userOverrides.has(key);
-
-		// Simple logic: either using default (from props) or custom (user override)
-		const source: ParameterSource = isUserOverride ? 'custom' : 'default';
-
-		return {
-			value: currentValue,
-			source,
-			serverDefault: hasPropsDefault ? propsDefaults[key] : undefined, // Keep same field name for compatibility
-			userOverride: isUserOverride ? currentValue : undefined
-		};
-	}
-
-	/**
-	 * Check if a parameter can be synced from server
-	 */
-	static canSyncParameter(key: string): boolean {
-		return SYNCABLE_PARAMETERS.some((param) => param.key === key && param.canSync);
-	}
-
-	/**
-	 * Get all syncable parameter keys
-	 */
-	static getSyncableParameterKeys(): string[] {
-		return SYNCABLE_PARAMETERS.filter((param) => param.canSync).map((param) => param.key);
-	}
-
-	/**
-	 * Validate server parameter value
-	 */
-	static validateServerParameter(key: string, value: ParameterValue): boolean {
-		const param = SYNCABLE_PARAMETERS.find((p) => p.key === key);
-		if (!param) return false;
-
-		switch (param.type) {
-			case 'number':
-				return typeof value === 'number' && !isNaN(value);
-			case 'string':
-				return typeof value === 'string';
-			case 'boolean':
-				return typeof value === 'boolean';
-			default:
-				return false;
-		}
-	}
-
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Diff
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Create a diff between current settings and server defaults
-	 */
-	static createParameterDiff(
-		currentSettings: ParameterRecord,
-		serverDefaults: ParameterRecord
-	): Record<string, { current: ParameterValue; server: ParameterValue; differs: boolean }> {
-		const diff: Record<
-			string,
-			{ current: ParameterValue; server: ParameterValue; differs: boolean }
-		> = {};
-
-		for (const key of this.getSyncableParameterKeys()) {
-			const currentValue = currentSettings[key];
-			const serverValue = serverDefaults[key];
-
-			if (serverValue !== undefined) {
-				diff[key] = {
-					current: currentValue,
-					server: serverValue,
-					differs: currentValue !== serverValue
-				};
-			}
-		}
-
-		return diff;
-	}
-}

tools/server/webui/src/lib/services/props.service.ts

@@ -1,4 +1,4 @@
-import { apiFetchWithParams } from '$lib/utils/api-fetch';
+import { apiFetchWithParams } from '$lib/utils';
 
 export class PropsService {
 	/**

tools/server/webui/src/lib/services/props.ts

@@ -1,77 +0,0 @@
-import { getAuthHeaders } from '$lib/utils';
-
-/**
- * PropsService - Server properties management
- *
- * This service handles communication with the /props endpoint to retrieve
- * server configuration, model information, and capabilities.
- *
- * **Responsibilities:**
- * - Fetch server properties from /props endpoint
- * - Handle API authentication
- * - Parse and validate server response
- *
- * **Used by:**
- * - serverStore: Primary consumer for server state management
- */
-export class PropsService {
-	// ─────────────────────────────────────────────────────────────────────────────
-	// Fetching
-	// ─────────────────────────────────────────────────────────────────────────────
-
-	/**
-	 * Fetches server properties from the /props endpoint
-	 *
-	 * @param autoload - If false, prevents automatic model loading (default: false)
-	 * @returns {Promise<ApiLlamaCppServerProps>} Server properties
-	 * @throws {Error} If the request fails or returns invalid data
-	 */
-	static async fetch(autoload = false): Promise<ApiLlamaCppServerProps> {
-		const url = new URL('./props', window.location.href);
-		if (!autoload) {
-			url.searchParams.set('autoload', 'false');
-		}
-
-		const response = await fetch(url.toString(), {
-			headers: getAuthHeaders()
-		});
-
-		if (!response.ok) {
-			throw new Error(
-				`Failed to fetch server properties: ${response.status} ${response.statusText}`
-			);
-		}
-
-		const data = await response.json();
-		return data as ApiLlamaCppServerProps;
-	}
-
-	/**
-	 * Fetches server properties for a specific model (ROUTER mode)
-	 *
-	 * @param modelId - The model ID to fetch properties for
-	 * @param autoload - If false, prevents automatic model loading (default: false)
-	 * @returns {Promise<ApiLlamaCppServerProps>} Server properties for the model
-	 * @throws {Error} If the request fails or returns invalid data
-	 */
-	static async fetchForModel(modelId: string, autoload = false): Promise<ApiLlamaCppServerProps> {
-		const url = new URL('./props', window.location.href);
-		url.searchParams.set('model', modelId);
-		if (!autoload) {
-			url.searchParams.set('autoload', 'false');
-		}
-
-		const response = await fetch(url.toString(), {
-			headers: getAuthHeaders()
-		});
-
-		if (!response.ok) {
-			throw new Error(
-				`Failed to fetch model properties: ${response.status} ${response.statusText}`
-			);
-		}
-
-		const data = await response.json();
-		return data as ApiLlamaCppServerProps;
-	}
-}

tools/server/webui/src/lib/stores/chat.svelte.ts

@@ -137,6 +137,7 @@ class ChatStore {
 	clearUIState(): void {
 		this.isLoading = false;
 		this.currentResponse = '';
+		this.isStreamingActive = false;
 	}
 
 	// ─────────────────────────────────────────────────────────────────────────────

tools/server/webui/src/lib/stores/conversations.svelte.ts

@@ -1,7 +1,7 @@
 import { browser } from '$app/environment';
 import { goto } from '$app/navigation';
 import { toast } from 'svelte-sonner';
-import { DatabaseService } from '$lib/services/database';
+import { DatabaseService } from '$lib/services/database.service';
 import { config } from '$lib/stores/settings.svelte';
 import { filterByLeafNodeId, findLeafNode } from '$lib/utils';
 import { AttachmentType } from '$lib/enums';
@@ -241,7 +241,9 @@ class ConversationsStore {
 
 		const leafNodeId =
 			this.activeConversation.currNode ||
-			allMessages.reduce((latest, msg) => (msg.timestamp > latest.timestamp ? msg : latest)).id;
+			allMessages.reduce((latest: DatabaseMessage, msg: DatabaseMessage) =>
+				msg.timestamp > latest.timestamp ? msg : latest
+			).id;
 
 		const currentPath = filterByLeafNodeId(allMessages, leafNodeId, false) as DatabaseMessage[];
 
@@ -341,9 +343,11 @@ class ConversationsStore {
 		if (!this.activeConversation) return;
 
 		const allMessages = await DatabaseService.getConversationMessages(this.activeConversation.id);
-		const rootMessage = allMessages.find((m) => m.type === 'root' && m.parent === null);
+		const rootMessage = allMessages.find(
+			(m: DatabaseMessage) => m.type === 'root' && m.parent === null
+		);
 		const currentFirstUserMessage = this.activeMessages.find(
-			(m) => m.role === 'user' && m.parent === rootMessage?.id
+			(m: DatabaseMessage) => m.role === 'user' && m.parent === rootMessage?.id
 		);
 
 		const currentLeafNodeId = findLeafNode(allMessages, siblingId);
@@ -355,7 +359,7 @@ class ConversationsStore {
 		// Only show title dialog if we're navigating between different first user message siblings
 		if (rootMessage && this.activeMessages.length > 0) {
 			const newFirstUserMessage = this.activeMessages.find(
-				(m) => m.role === 'user' && m.parent === rootMessage.id
+				(m: DatabaseMessage) => m.role === 'user' && m.parent === rootMessage.id
 			);
 
 			if (
@@ -452,7 +456,7 @@ class ConversationsStore {
 		}
 
 		const allData = await Promise.all(
-			allConversations.map(async (conv) => {
+			allConversations.map(async (conv: DatabaseConversation) => {
 				const messages = await DatabaseService.getConversationMessages(conv.id);
 				return { conv, messages };
 			})

tools/server/webui/src/lib/stores/models.svelte.ts

@@ -1,6 +1,6 @@
 import { SvelteSet } from 'svelte/reactivity';
-import { ModelsService } from '$lib/services/models';
-import { PropsService } from '$lib/services/props';
+import { ModelsService } from '$lib/services/models.service';
+import { PropsService } from '$lib/services/props.service';
 import { ServerModelStatus, ModelModality } from '$lib/enums';
 import { serverStore } from '$lib/stores/server.svelte';
 

tools/server/webui/src/lib/stores/server.svelte.ts

@@ -1,4 +1,4 @@
-import { PropsService } from '$lib/services/props';
+import { PropsService } from '$lib/services/props.service';
 import { ServerRole } from '$lib/enums';
 
 /**

tools/server/webui/src/lib/stores/settings.svelte.ts

@@ -33,7 +33,7 @@
 
 import { browser } from '$app/environment';
 import { SETTING_CONFIG_DEFAULT } from '$lib/constants/settings-config';
-import { ParameterSyncService } from '$lib/services/parameter-sync';
+import { ParameterSyncService } from '$lib/services/parameter-sync.service';
 import { serverStore } from '$lib/stores/server.svelte';
 import {
 	configToParameterRecord,

tools/server/webui/src/lib/types/chat.d.ts

@@ -1,3 +1,6 @@
+import type { ErrorDialogType } from '$lib/enums';
+import type { DatabaseMessage, DatabaseMessageExtra } from './database';
+
 export type ChatMessageType = 'root' | 'text' | 'think' | 'system';
 export type ChatRole = 'user' | 'assistant' | 'system';
 
@@ -9,6 +12,8 @@ export interface ChatUploadedFile {
 	file: File;
 	preview?: string;
 	textContent?: string;
+	isLoading?: boolean;
+	loadError?: string;
 }
 
 export interface ChatAttachmentDisplayItem {
@@ -17,6 +22,8 @@ export interface ChatAttachmentDisplayItem {
 	size?: number;
 	preview?: string;
 	isImage: boolean;
+	isLoading?: boolean;
+	loadError?: string;
 	uploadedFile?: ChatUploadedFile;
 	attachment?: DatabaseMessageExtra;
 	attachmentIndex?: number;
@@ -53,3 +60,49 @@ export interface ChatMessageTimings {
 	prompt_ms?: number;
 	prompt_n?: number;
 }
+
+export interface ChatStreamCallbacks {
+	onChunk?: (chunk: string) => void;
+	onReasoningChunk?: (chunk: string) => void;
+	onToolCallChunk?: (chunk: string) => void;
+	onAttachments?: (extras: DatabaseMessageExtra[]) => void;
+	onModel?: (model: string) => void;
+	onTimings?: (timings?: ChatMessageTimings, promptProgress?: ChatMessagePromptProgress) => void;
+	onComplete?: (
+		content?: string,
+		reasoningContent?: string,
+		timings?: ChatMessageTimings,
+		toolCallContent?: string
+	) => void;
+	onError?: (error: Error) => void;
+}
+
+export interface ErrorDialogState {
+	type: ErrorDialogType;
+	message: string;
+	contextInfo?: { n_prompt_tokens: number; n_ctx: number };
+}
+
+export interface LiveProcessingStats {
+	tokensProcessed: number;
+	totalTokens: number;
+	timeMs: number;
+	tokensPerSecond: number;
+	etaSecs?: number;
+}
+
+export interface LiveGenerationStats {
+	tokensGenerated: number;
+	timeMs: number;
+	tokensPerSecond: number;
+}
+
+export interface AttachmentDisplayItemsOptions {
+	uploadedFiles?: ChatUploadedFile[];
+	attachments?: DatabaseMessageExtra[];
+}
+
+export interface FileProcessingResult {
+	extras: DatabaseMessageExtra[];
+	emptyFiles: string[];
+}

tools/server/webui/src/lib/types/common.d.ts

@@ -0,0 +1,35 @@
+import type { AttachmentType } from '$lib/enums';
+
+/**
+ * Represents a key-value pair.
+ */
+export interface KeyValuePair {
+	key: string;
+	value: string;
+}
+
+/**
+ * Binary detection configuration options.
+ */
+export interface BinaryDetectionOptions {
+	prefixLength: number;
+	suspiciousCharThresholdRatio: number;
+	maxAbsoluteNullBytes: number;
+}
+
+/**
+ * Format for text attachments when copied to clipboard.
+ */
+export interface ClipboardTextAttachment {
+	type: typeof AttachmentType.TEXT;
+	name: string;
+	content: string;
+}
+
+/**
+ * Parsed result from clipboard content.
+ */
+export interface ParsedClipboardContent {
+	message: string;
+	textAttachments: ClipboardTextAttachment[];
+}

tools/server/webui/src/lib/types/index.ts

@@ -41,7 +41,13 @@ export type {
 	ChatAttachmentPreviewItem,
 	ChatMessageSiblingInfo,
 	ChatMessagePromptProgress,
-	ChatMessageTimings
+	ChatMessageTimings,
+	ChatStreamCallbacks,
+	ErrorDialogState,
+	LiveProcessingStats,
+	LiveGenerationStats,
+	AttachmentDisplayItemsOptions,
+	FileProcessingResult
 } from './chat';
 
 // Database types
@@ -59,12 +65,24 @@ export type {
 } from './database';
 
 // Model types
-export type { ModelModalities, ModelOption } from './models';
+export type { ModelModalities, ModelOption, ModalityCapabilities } from './models';
 
 // Settings types
 export type {
 	SettingsConfigValue,
 	SettingsFieldConfig,
 	SettingsChatServiceOptions,
-	SettingsConfigType
+	SettingsConfigType,
+	ParameterValue,
+	ParameterRecord,
+	ParameterInfo,
+	SyncableParameter
 } from './settings';
+
+// Common types
+export type {
+	KeyValuePair,
+	BinaryDetectionOptions,
+	ClipboardTextAttachment,
+	ParsedClipboardContent
+} from './common';

tools/server/webui/src/lib/types/settings.d.ts

@@ -1,12 +1,14 @@
 import type { SETTING_CONFIG_DEFAULT } from '$lib/constants/settings-config';
-import type { ChatMessageTimings } from './chat';
+import type { ChatMessagePromptProgress, ChatMessageTimings } from './chat';
+import type { ParameterSource, SyncableParameterType, SettingsFieldType } from '$lib/enums';
+import type { DatabaseMessageExtra } from './database';
 
 export type SettingsConfigValue = string | number | boolean;
 
 export interface SettingsFieldConfig {
 	key: string;
 	label: string;
-	type: 'input' | 'textarea' | 'checkbox' | 'select';
+	type: SettingsFieldType;
 	isExperimental?: boolean;
 	help?: string;
 	options?: Array<{ value: string; label: string; icon?: typeof import('@lucide/svelte').Icon }>;
@@ -51,6 +53,7 @@ export interface SettingsChatServiceOptions {
 	onChunk?: (chunk: string) => void;
 	onReasoningChunk?: (chunk: string) => void;
 	onToolCallChunk?: (chunk: string) => void;
+	onAttachments?: (extras: DatabaseMessageExtra[]) => void;
 	onModel?: (model: string) => void;
 	onTimings?: (timings?: ChatMessageTimings, promptProgress?: ChatMessagePromptProgress) => void;
 	onComplete?: (
@@ -65,3 +68,20 @@ export interface SettingsChatServiceOptions {
 export type SettingsConfigType = typeof SETTING_CONFIG_DEFAULT & {
 	[key: string]: SettingsConfigValue;
 };
+
+export type ParameterValue = string | number | boolean;
+export type ParameterRecord = Record<string, ParameterValue>;
+
+export interface ParameterInfo {
+	value: ParameterValue;
+	source: ParameterSource;
+	serverDefault?: ParameterValue;
+	userOverride?: ParameterValue;
+}
+
+export interface SyncableParameter {
+	key: string;
+	serverKey: string;
+	type: SyncableParameterType;
+	canSync: boolean;
+}

tools/server/webui/src/lib/utils/api-fetch.ts

@@ -1,5 +1,6 @@
 import { base } from '$app/paths';
 import { getJsonHeaders, getAuthHeaders } from './api-headers';
+import { UrlPrefix } from '$lib/enums';
 
 /**
  * API Fetch Utilities
@@ -48,7 +49,8 @@ export async function apiFetch<T>(path: string, options: ApiFetchOptions = {}):
 	const baseHeaders = authOnly ? getAuthHeaders() : getJsonHeaders();
 	const headers = { ...baseHeaders, ...customHeaders };
 
-	const url = path.startsWith('http://') || path.startsWith('https://') ? path : `${base}${path}`;
+	const url =
+		path.startsWith(UrlPrefix.HTTP) || path.startsWith(UrlPrefix.HTTPS) ? path : `${base}${path}`;
 
 	const response = await fetch(url, {
 		...fetchOptions,

tools/server/webui/src/lib/utils/branching.ts

@@ -15,7 +15,7 @@
  *        └── message 5 (assistant)
  */
 
-import { MessageRole } from '$lib/enums/chat';
+import { MessageRole } from '$lib/enums';
 
 /**
  * Filters messages to get the conversation path from root to a specific leaf node.

tools/server/webui/src/lib/utils/cache-ttl.ts

@@ -1,5 +1,4 @@
-const DEFAULT_CACHE_TTL_MS = 5 * 60 * 1000;
-const DEFAULT_CACHE_MAX_ENTRIES = 100;
+import { DEFAULT_CACHE_TTL_MS, DEFAULT_CACHE_MAX_ENTRIES } from '$lib/constants/cache';
 
 /**
  * TTL Cache - Time-To-Live cache implementation for memory optimization

tools/server/webui/src/lib/utils/modality-file-validation.ts

@@ -5,7 +5,7 @@
 
 import { getFileTypeCategory } from '$lib/utils';
 import { FileTypeCategory } from '$lib/enums';
-import type { ModalityCapabilities } from '$lib/types/models';
+import type { ModalityCapabilities } from '$lib/types';
 
 /**
  * Check if a file type is supported by the given modalities

tools/server/webui/src/lib/utils/text-files.ts

@@ -4,7 +4,7 @@
  */
 
 import { DEFAULT_BINARY_DETECTION_OPTIONS } from '$lib/constants/binary-detection';
-import type { BinaryDetectionOptions } from '$lib/constants/binary-detection';
+import type { BinaryDetectionOptions } from '$lib/types';
 import { FileExtensionText } from '$lib/enums';
 
 /**

tools/server/webui/tests/stories/ChatScreenForm.stories.svelte

@@ -1,14 +1,14 @@
 <script module lang="ts">
 	import { defineMeta } from '@storybook/addon-svelte-csf';
-	import ChatForm from '$lib/components/app/chat/ChatForm/ChatForm.svelte';
+	import ChatScreenForm from '$lib/components/app/chat/ChatScreen/ChatScreenForm.svelte';
 	import { expect } from 'storybook/test';
 	import jpgAsset from './fixtures/assets/1.jpg?url';
 	import svgAsset from './fixtures/assets/hf-logo.svg?url';
 	import pdfAsset from './fixtures/assets/example.pdf?raw';
 
 	const { Story } = defineMeta({
-		title: 'Components/ChatScreen/ChatForm',
-		component: ChatForm,
+		title: 'Components/ChatScreen/ChatScreenForm',
+		component: ChatScreenForm,
 		parameters: {
 			layout: 'centered'
 		}
@@ -44,7 +44,8 @@
 <Story
 	name="Default"
 	args={{ class: 'max-w-[56rem] w-[calc(100vw-2rem)]' }}
-	play={async ({ canvas, userEvent }) => {
+	play={async (context) => {
+		const { canvas, userEvent } = context;
 		const textarea = await canvas.findByRole('textbox');
 		const submitButton = await canvas.findByRole('button', { name: 'Send' });
 
@@ -74,7 +75,8 @@
 		class: 'max-w-[56rem] w-[calc(100vw-2rem)]',
 		uploadedFiles: fileAttachments
 	}}
-	play={async ({ canvas }) => {
+	play={async (context) => {
+		const { canvas } = context;
 		const jpgAttachment = canvas.getByAltText('1.jpg');
 		const svgAttachment = canvas.getByAltText('hf-logo.svg');
 		const pdfFileExtension = canvas.getByText('PDF');

tools/server/webui/tests/stories/MarkdownContent.stories.svelte

@@ -68,18 +68,22 @@ You can also test inline links like https://example.com or https://docs.python.o
 All links should have \`target="_blank"\` and \`rel="noopener noreferrer"\` attributes for security.`,
 		class: 'max-w-[56rem] w-[calc(100vw-2rem)]'
 	}}
-	play={async ({ canvasElement }) => {
+	play={async (context) => {
+		const { canvasElement } = context;
 		// Wait for component to render
 		await new Promise((resolve) => setTimeout(resolve, 100));
 
 		// Find all links in the rendered content
-		const links = canvasElement.querySelectorAll('a[href]');
+		const links = (canvasElement as HTMLElement).querySelectorAll(
+			'a[href]'
+		) as NodeListOf<HTMLAnchorElement>;
+		const linkList = Array.from(links) as HTMLAnchorElement[];
 
 		// Test that we have the expected number of links
 		expect(links.length).toBeGreaterThan(0);
 
 		// Test each link for proper attributes
-		links.forEach((link) => {
+		links.forEach((link: HTMLAnchorElement) => {
 			const href = link.getAttribute('href');
 
 			// Test that external links have proper security attributes
@@ -90,37 +94,35 @@ All links should have \`target="_blank"\` and \`rel="noopener noreferrer"\` attr
 		});
 
 		// Test specific links exist
-		const hugginFaceLink = Array.from(links).find(
+		const hugginFaceLink = linkList.find(
 			(link) => link.getAttribute('href') === 'https://huggingface.co'
 		);
 		expect(hugginFaceLink).toBeTruthy();
 		expect(hugginFaceLink?.textContent).toBe('Hugging Face Homepage');
 
-		const githubLink = Array.from(links).find(
+		const githubLink = linkList.find(
 			(link) => link.getAttribute('href') === 'https://github.com/ggml-org/llama.cpp'
 		);
 		expect(githubLink).toBeTruthy();
 		expect(githubLink?.textContent).toBe('GitHub Repository');
 
-		const openaiLink = Array.from(links).find(
-			(link) => link.getAttribute('href') === 'https://openai.com'
-		);
+		const openaiLink = linkList.find((link) => link.getAttribute('href') === 'https://openai.com');
 		expect(openaiLink).toBeTruthy();
 		expect(openaiLink?.textContent).toBe('OpenAI Website');
 
-		const googleLink = Array.from(links).find(
+		const googleLink = linkList.find(
 			(link) => link.getAttribute('href') === 'https://www.google.com'
 		);
 		expect(googleLink).toBeTruthy();
 		expect(googleLink?.textContent).toBe('Google Search');
 
 		// Test inline links (auto-linked URLs)
-		const exampleLink = Array.from(links).find(
+		const exampleLink = linkList.find(
 			(link) => link.getAttribute('href') === 'https://example.com'
 		);
 		expect(exampleLink).toBeTruthy();
 
-		const pythonDocsLink = Array.from(links).find(
+		const pythonDocsLink = linkList.find(
 			(link) => link.getAttribute('href') === 'https://docs.python.org'
 		);
 		expect(pythonDocsLink).toBeTruthy();

tools/server/webui/vite.config.ts

@@ -2,11 +2,15 @@ import tailwindcss from '@tailwindcss/vite';
 import { sveltekit } from '@sveltejs/kit/vite';
 import * as fflate from 'fflate';
 import { readFileSync, writeFileSync, existsSync } from 'fs';
-import { resolve } from 'path';
-import { defineConfig } from 'vite';
+import { dirname, resolve } from 'path';
+import { fileURLToPath } from 'url';
+
+import { defineConfig, searchForWorkspaceRoot } from 'vite';
 import devtoolsJson from 'vite-plugin-devtools-json';
 import { storybookTest } from '@storybook/addon-vitest/vitest-plugin';
 
+const __dirname = dirname(fileURLToPath(import.meta.url));
+
 const GUIDE_FOR_FRONTEND = `
 <!--
   This is a single file build of the frontend.
@@ -156,11 +160,15 @@ export default defineConfig({
 		proxy: {
 			'/v1': 'http://localhost:8080',
 			'/props': 'http://localhost:8080',
-			'/models': 'http://localhost:8080'
+			'/models': 'http://localhost:8080',
+			'/cors-proxy': 'http://localhost:8080'
 		},
 		headers: {
 			'Cross-Origin-Embedder-Policy': 'require-corp',
 			'Cross-Origin-Opener-Policy': 'same-origin'
+		},
+		fs: {
+			allow: [searchForWorkspaceRoot(process.cwd()), resolve(__dirname, 'tests')]
 		}
 	}
 });

vendor/cpp-httplib/httplib.cpp

@@ -1,7 +1,6 @@
 #include "httplib.h"
 namespace httplib {
 
-
 /*
  * Implementation that will be part of the .cc file if split into .h + .cc.
  */
@@ -1219,7 +1218,7 @@ int close_socket(socket_t sock) {
 #endif
 }
 
-template <typename T> inline ssize_t handle_EINTR(T fn) {
+template <typename T> ssize_t handle_EINTR(T fn) {
   ssize_t res = 0;
   while (true) {
     res = fn();

vendor/cpp-httplib/httplib.h

@@ -3336,7 +3336,6 @@ private:
 } // namespace sse
 
 
-
 } // namespace httplib
 
 #endif // CPPHTTPLIB_HTTPLIB_H