Merge origin/master into feat/darwin-privacy-filter

Keep both the privacy-filter and merged localvqe includeDarwin entries.

Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
Assisted-by: Claude:opus-4.8 [Claude Code]

.docker/llama-cpp-compile.sh

@@ -17,19 +17,29 @@ if [[ -n "${CUDA_DOCKER_ARCH:-}" ]]; then
   rm -rf /LocalAI/backend/cpp/llama-cpp-*-build
 fi
 
-if [ "${TARGETARCH}" = "arm64" ] || [ "${BUILD_TYPE}" = "hipblas" ]; then
-  cd /LocalAI/backend/cpp/llama-cpp
-  make llama-cpp-fallback
-  make llama-cpp-grpc
-  make llama-cpp-rpc-server
+cd /LocalAI/backend/cpp/llama-cpp
+if [ -z "${BUILD_TYPE:-}" ]; then
+  # Pure CPU image (BUILD_TYPE empty): one build with ggml CPU_ALL_VARIANTS replaces the
+  # per-microarch binaries (x86: avx/avx2/avx512/fallback; arm64: armv8.x/armv9.x). ggml
+  # dlopens the best libggml-cpu-*.so at runtime by probing host CPU features.
+  #
+  # arm64: the CPU_ALL_VARIANTS table includes armv9.2 SME variants whose -march=...+sme is
+  # rejected by the Ubuntu 24.04 default gcc-13. gcc-14 accepts it, so build the arm64
+  # variants with it (the host never *selects* SME unless it has it, but every variant must
+  # still compile).
+  if [ "${TARGETARCH}" = "arm64" ]; then
+    apt-get update -qq && apt-get install -y -qq gcc-14 g++-14
+    export CC=gcc-14 CXX=g++-14
+  fi
+  make llama-cpp-cpu-all
 else
-  cd /LocalAI/backend/cpp/llama-cpp
-  make llama-cpp-avx
-  make llama-cpp-avx2
-  make llama-cpp-avx512
+  # GPU build (cublas/hipblas/sycl/vulkan/...): the accelerator does the compute, so a
+  # single fallback CPU build is enough - no per-microarch CPU variants needed. (This also
+  # keeps the heavy GPU backend compile from also building the whole CPU variant matrix,
+  # and avoids the gcc-14 apt step on GPU base images such as nvidia l4t.)
   make llama-cpp-fallback
-  make llama-cpp-grpc
-  make llama-cpp-rpc-server
 fi
+make llama-cpp-grpc
+make llama-cpp-rpc-server
 
 ccache -s || true

.docker/turboquant-compile.sh

@@ -19,17 +19,21 @@ fi
 
 cd /LocalAI/backend/cpp/turboquant
 
-if [ "${TARGETARCH}" = "arm64" ] || [ "${BUILD_TYPE}" = "hipblas" ]; then
-  make turboquant-fallback
-  make turboquant-grpc
-  make turboquant-rpc-server
+if [ -z "${BUILD_TYPE:-}" ]; then
+  # Pure CPU image: one ggml CPU_ALL_VARIANTS build replaces the per-microarch binaries.
+  # arm64: the armv9.2 SME variants need gcc-14 (gcc-13 rejects +sme).
+  if [ "${TARGETARCH}" = "arm64" ]; then
+    apt-get update -qq && apt-get install -y -qq gcc-14 g++-14
+    export CC=gcc-14 CXX=g++-14
+  fi
+  make turboquant-cpu-all
 else
-  make turboquant-avx
-  make turboquant-avx2
-  make turboquant-avx512
+  # GPU build (cublas/hipblas/sycl/vulkan/...): single fallback CPU build, the accelerator
+  # does the compute. Keeps the GPU compile from also building the CPU variant matrix and
+  # avoids the gcc-14 apt step on GPU base images such as nvidia l4t.
   make turboquant-fallback
-  make turboquant-grpc
-  make turboquant-rpc-server
 fi
+make turboquant-grpc
+make turboquant-rpc-server
 
 ccache -s || true

.github/backend-matrix.yml

@@ -4922,6 +4922,37 @@ includeDarwin:
     tag-suffix: "-metal-darwin-arm64-vibevoice-cpp"
     build-type: "metal"
     lang: "go"
+  # Vision/utility C++/ggml backends (go+cgo). Their Makefiles already carry a
+  # Darwin/Metal path (GGML_METAL=ON when build-type=metal); this just builds and
+  # publishes the metal image so Apple Silicon can install them.
+  - backend: "depth-anything-cpp"
+    tag-suffix: "-metal-darwin-arm64-depth-anything-cpp"
+    build-type: "metal"
+    lang: "go"
+  - backend: "locate-anything-cpp"
+    tag-suffix: "-metal-darwin-arm64-locate-anything-cpp"
+    build-type: "metal"
+    lang: "go"
+  - backend: "rfdetr-cpp"
+    tag-suffix: "-metal-darwin-arm64-rfdetr-cpp"
+    build-type: "metal"
+    lang: "go"
+  - backend: "sam3-cpp"
+    tag-suffix: "-metal-darwin-arm64-sam3-cpp"
+    build-type: "metal"
+    lang: "go"
+  # privacy-filter (PII/NER) is a C++/ggml backend built by a bespoke darwin
+  # script (make backends/privacy-filter-darwin); ggml defaults Metal ON on Apple
+  # so the build is Metal-enabled. lang=go drives runner/toolchain selection only.
+  - backend: "privacy-filter"
+    tag-suffix: "-metal-darwin-arm64-privacy-filter"
+    lang: "go"
+  # LocalVQE has no Metal path; on Apple Silicon it builds CPU-only (GGML_METAL
+  # OFF) but is still a native arm64 image. Uses the darwin/metal build profile.
+  - backend: "localvqe"
+    tag-suffix: "-metal-darwin-arm64-localvqe"
+    build-type: "metal"
+    lang: "go"
   - backend: "voxtral"
     tag-suffix: "-metal-darwin-arm64-voxtral"
     build-type: "metal"
@@ -4974,6 +5005,19 @@ includeDarwin:
   - backend: "kitten-tts"
     tag-suffix: "-metal-darwin-arm64-kitten-tts"
     build-type: "mps"
+  # vLLM on Apple Silicon via vllm-metal (MLX). The install is custom
+  # (backend/python/vllm/install.sh has a darwin branch); lang stays python so
+  # backend_build_darwin.yml drives it through build-darwin-python-backend ->
+  # scripts/build/python-darwin.sh, which runs the backend's install.sh.
+  - backend: "vllm"
+    tag-suffix: "-metal-darwin-arm64-vllm"
+    build-type: "mps"
+  - backend: "trl"
+    tag-suffix: "-metal-darwin-arm64-trl"
+    build-type: "mps"
+  - backend: "liquid-audio"
+    tag-suffix: "-metal-darwin-arm64-liquid-audio"
+    build-type: "mps"
   - backend: "piper"
     tag-suffix: "-metal-darwin-arm64-piper"
     build-type: "metal"
@@ -4990,6 +5034,10 @@ includeDarwin:
     tag-suffix: "-metal-darwin-arm64-sherpa-onnx"
     build-type: "metal"
     lang: "go"
+  - backend: "supertonic"
+    tag-suffix: "-metal-darwin-arm64-supertonic"
+    build-type: "metal"
+    lang: "go"
   - backend: "local-store"
     tag-suffix: "-metal-darwin-arm64-local-store"
     build-type: "metal"

.github/bump_vllm_metal.sh

@@ -0,0 +1,55 @@
+#!/bin/bash
+# Bump the single vllm-metal pin (VLLM_METAL_VERSION) in the vLLM backend's
+# darwin (Apple Silicon) install path. The macOS/Metal build
+# (backend/python/vllm/install.sh, Darwin branch) installs vllm-metal, which is
+# version-locked to a specific vLLM source release. install.sh derives that vLLM
+# version at build time from vllm-metal's own installer (`vllm_v=`) at the pinned
+# tag, so there is only ONE value to bump here -- mirroring bump_vllm_wheel.sh,
+# which bumps the Linux cu130 wheel pin.
+#
+# This deliberately tracks vllm-project/vllm-metal, NOT vllm-project/vllm: the
+# darwin build can only use the exact vLLM version vllm-metal supports, so it may
+# lag the Linux pin (requirements-cublas13-after.txt) until vllm-metal catches up.
+set -xe
+REPO=$1   # vllm-project/vllm-metal
+FILE=$2   # backend/python/vllm/install.sh
+VAR=$3    # VLLM_METAL_VERSION (used for the workflow's output file names)
+
+if [ -z "$FILE" ] || [ -z "$REPO" ] || [ -z "$VAR" ]; then
+    echo "usage: $0 <repo> <install-file> <var-name>" >&2
+    exit 1
+fi
+
+# vllm-metal ships frequent dev releases, all flagged as non-prerelease, so
+# /releases/latest returns the newest one (with its cp312 wheel asset).
+LATEST_TAG=$(curl -sS -H "Accept: application/vnd.github+json" \
+    "https://api.github.com/repos/$REPO/releases/latest" \
+    | python3 -c "import json,sys; print(json.load(sys.stdin)['tag_name'])")
+
+# The coupled vLLM source version lives in vllm-metal's installer at that tag.
+NEW_VLLM_VERSION=$(curl -fsSL \
+    "https://raw.githubusercontent.com/$REPO/$LATEST_TAG/install.sh" \
+    | grep -oE 'vllm_v="[0-9]+\.[0-9]+\.[0-9]+"' | head -1 | cut -d'"' -f2)
+
+if [ -z "$LATEST_TAG" ] || [ -z "$NEW_VLLM_VERSION" ]; then
+    echo "Could not resolve vllm-metal tag ($LATEST_TAG) or its vllm_v ($NEW_VLLM_VERSION)." >&2
+    exit 1
+fi
+
+set +e
+CURRENT_TAG=$(grep -oE 'VLLM_METAL_VERSION="[^"]*"' "$FILE" | head -1 | cut -d'"' -f2)
+set -e
+
+# Rewrite the single pin. install.sh derives VLLM_VERSION from this tag at build
+# time, so there is nothing else to touch. peter-evans/create-pull-request opens
+# no PR on a clean tree, so a no-op rewrite (already current) is safe.
+sed -i "$FILE" \
+    -e "s|VLLM_METAL_VERSION=\"[^\"]*\"|VLLM_METAL_VERSION=\"$LATEST_TAG\"|"
+
+if [ -z "$CURRENT_TAG" ]; then
+    echo "Could not find VLLM_METAL_VERSION=\"...\" in $FILE." >&2
+    exit 0
+fi
+
+echo "vllm-metal ${CURRENT_TAG} -> ${LATEST_TAG} (builds vLLM ${NEW_VLLM_VERSION}): https://github.com/$REPO/releases/tag/${LATEST_TAG}" >> "${VAR}_message.txt"
+echo "${LATEST_TAG}" >> "${VAR}_commit.txt"

.github/workflows/backend_build_darwin.yml

@@ -228,8 +228,17 @@ jobs:
         run: |
           make backends/ds4-darwin
 
+      # privacy-filter is a C++/ggml backend like ds4 - a single grpc-server with
+      # otool dylib bundling - so it gets its own bespoke darwin script rather than
+      # the generic build-darwin-go-backend path.
+      - name: Build privacy-filter backend (Darwin Metal)
+        if: inputs.backend == 'privacy-filter'
+        run: |
+          make protogen-go
+          make backends/privacy-filter-darwin
+
       - name: Build ${{ inputs.backend }}-darwin
-        if: inputs.backend != 'llama-cpp' && inputs.backend != 'ds4'
+        if: inputs.backend != 'llama-cpp' && inputs.backend != 'ds4' && inputs.backend != 'privacy-filter'
         run: |
           make protogen-go
           BACKEND=${{ inputs.backend }} BUILD_TYPE=${{ inputs.build-type }} USE_PIP=${{ inputs.use-pip }} make build-darwin-${{ inputs.lang }}-backend

.github/workflows/bump_deps.yaml

@@ -154,3 +154,39 @@ jobs:
           branch: "update/VLLM_VERSION"
           body: ${{ steps.bump.outputs.message }}
           signoff: true
+
+  bump-vllm-metal:
+    # The darwin (Apple Silicon) vLLM build installs vllm-metal, which is locked
+    # to a specific vLLM source release. install.sh pins both VLLM_METAL_VERSION
+    # (the wheel release) and VLLM_VERSION (the vLLM it builds against); this job
+    # tracks vllm-project/vllm-metal and rewrites both atomically. Separate from
+    # bump-vllm-wheel because darwin follows vllm-metal, not vllm/vllm latest.
+    if: github.repository == 'mudler/LocalAI'
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v7
+      - name: Bump vllm-metal pin 🔧
+        id: bump
+        run: |
+          bash .github/bump_vllm_metal.sh vllm-project/vllm-metal backend/python/vllm/install.sh VLLM_METAL_VERSION
+          {
+            echo 'message<<EOF'
+            cat "VLLM_METAL_VERSION_message.txt"
+            echo EOF
+          } >> "$GITHUB_OUTPUT"
+          {
+            echo 'commit<<EOF'
+            cat "VLLM_METAL_VERSION_commit.txt"
+            echo EOF
+          } >> "$GITHUB_OUTPUT"
+          rm -rfv VLLM_METAL_VERSION_message.txt VLLM_METAL_VERSION_commit.txt
+      - name: Create Pull Request
+        uses: peter-evans/create-pull-request@v8
+        with:
+          token: ${{ secrets.UPDATE_BOT_TOKEN }}
+          push-to-fork: ci-forks/LocalAI
+          commit-message: ':arrow_up: Update vllm-project/vllm-metal (darwin)'
+          title: 'chore: :arrow_up: Update vllm-metal (darwin) to `${{ steps.bump.outputs.commit }}`'
+          branch: "update/VLLM_METAL_VERSION"
+          body: ${{ steps.bump.outputs.message }}
+          signoff: true

Makefile

@@ -1,5 +1,5 @@
 # Disable parallel execution for backend builds
-.NOTPARALLEL: backends/diffusers backends/llama-cpp backends/turboquant backends/outetts backends/piper backends/stablediffusion-ggml backends/whisper backends/crispasr backends/parakeet-cpp backends/faster-whisper backends/silero-vad backends/local-store backends/huggingface backends/rfdetr backends/rfdetr-cpp backends/insightface backends/speaker-recognition backends/kitten-tts backends/kokoro backends/chatterbox backends/llama-cpp-darwin backends/neutts build-darwin-python-backend build-darwin-go-backend backends/mlx backends/diffuser-darwin backends/mlx-vlm backends/mlx-audio backends/mlx-distributed backends/stablediffusion-ggml-darwin backends/vllm backends/vllm-omni backends/sglang backends/moonshine backends/pocket-tts backends/qwen-tts backends/faster-qwen3-tts backends/qwen-asr backends/nemo backends/voxcpm backends/whisperx backends/ace-step backends/acestep-cpp backends/fish-speech backends/voxtral backends/opus backends/trl backends/llama-cpp-quantization backends/kokoros backends/sam3-cpp backends/qwen3-tts-cpp backends/omnivoice-cpp backends/vibevoice-cpp backends/localvqe backends/tinygrad backends/sherpa-onnx backends/ds4 backends/ds4-darwin backends/liquid-audio backends/supertonic backends/depth-anything-cpp backends/privacy-filter
+.NOTPARALLEL: backends/diffusers backends/llama-cpp backends/turboquant backends/outetts backends/piper backends/stablediffusion-ggml backends/whisper backends/crispasr backends/parakeet-cpp backends/faster-whisper backends/silero-vad backends/local-store backends/huggingface backends/rfdetr backends/rfdetr-cpp backends/insightface backends/speaker-recognition backends/kitten-tts backends/kokoro backends/chatterbox backends/llama-cpp-darwin backends/neutts build-darwin-python-backend build-darwin-go-backend backends/mlx backends/diffuser-darwin backends/mlx-vlm backends/mlx-audio backends/mlx-distributed backends/stablediffusion-ggml-darwin backends/vllm backends/vllm-omni backends/sglang backends/moonshine backends/pocket-tts backends/qwen-tts backends/faster-qwen3-tts backends/qwen-asr backends/nemo backends/voxcpm backends/whisperx backends/ace-step backends/acestep-cpp backends/fish-speech backends/voxtral backends/opus backends/trl backends/llama-cpp-quantization backends/kokoros backends/sam3-cpp backends/qwen3-tts-cpp backends/omnivoice-cpp backends/vibevoice-cpp backends/localvqe backends/tinygrad backends/sherpa-onnx backends/ds4 backends/ds4-darwin backends/liquid-audio backends/supertonic backends/depth-anything-cpp backends/privacy-filter backends/privacy-filter-darwin
 
 GOCMD=go
 GOTEST=$(GOCMD) test
@@ -1129,6 +1129,10 @@ backends/ds4-darwin: build
 	bash ./scripts/build/ds4-darwin.sh
 	./local-ai backends install "ocifile://$(abspath ./backend-images/ds4.tar)"
 
+backends/privacy-filter-darwin: build
+	bash ./scripts/build/privacy-filter-darwin.sh
+	./local-ai backends install "ocifile://$(abspath ./backend-images/privacy-filter.tar)"
+
 build-darwin-python-backend: build
 	bash ./scripts/build/python-darwin.sh
 

backend/cpp/ik-llama-cpp/Makefile

@@ -1,5 +1,5 @@
 
-IK_LLAMA_VERSION?=6c00e87ac84404af588ad2e65935bd6f079c696f
+IK_LLAMA_VERSION?=d5507e33ae7ee2b7b41475f08044d3bde3b839ee
 LLAMA_REPO?=https://github.com/ikawrakow/ik_llama.cpp
 
 CMAKE_ARGS?=

backend/cpp/llama-cpp/CMakeLists.txt

@@ -50,8 +50,13 @@ add_custom_command(
         "${hw_proto}"
       DEPENDS "${hw_proto}")
 
-# hw_grpc_proto
-add_library(hw_grpc_proto
+# hw_grpc_proto: force STATIC. Under the CPU_ALL_VARIANTS build BUILD_SHARED_LIBS=ON
+# (ggml/llama become shared), which would otherwise make this glue library a DSO. As a
+# DSO it references the hidden-visibility symbols in the static libprotobuf.a, which the
+# linker cannot satisfy ("hidden symbol ... in libprotobuf.a is referenced by DSO").
+# Keeping it STATIC links protobuf/gRPC directly into the grpc-server executable while
+# only ggml/llama stay shared. No effect on the static variants (already BUILD_SHARED_LIBS=OFF).
+add_library(hw_grpc_proto STATIC
   ${hw_grpc_srcs}
   ${hw_grpc_hdrs}
   ${hw_proto_srcs}

backend/cpp/llama-cpp/Makefile

@@ -1,5 +1,5 @@
 
-LLAMA_VERSION?=73618f27a801c0b8614ceaf3547d3c2a99baae14
+LLAMA_VERSION?=8be759e6f70d629638a7eb70db3824cbdcea370b
 LLAMA_REPO?=https://github.com/ggerganov/llama.cpp
 
 CMAKE_ARGS?=
@@ -10,8 +10,16 @@ TARGET?=--target grpc-server
 JOBS?=$(shell nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null || echo 1)
 ARCH?=$(shell uname -m)
 
-# Disable Shared libs as we are linking on static gRPC and we can't mix shared and static
-CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF -DLLAMA_CURL=OFF
+# Shared libs default to OFF: we link static gRPC and the avx/avx2/avx512/fallback
+# variants are fully static. The CPU_ALL_VARIANTS build flips SHARED_LIBS=ON (ggml/llama
+# become shared so the dynamic CPU backends work; gRPC stays static via its imported
+# targets). SHARED_LIBS is a make variable, not an appended -D, so it survives the
+# recursive sub-make into the VARIANT build dir (which re-parses this Makefile) instead
+# of being re-clobbered by a second -DBUILD_SHARED_LIBS=OFF. EXTRA_CMAKE_ARGS is the hook
+# the CPU_ALL_VARIANTS target uses to inject -DGGML_BACKEND_DL/-DGGML_CPU_ALL_VARIANTS.
+SHARED_LIBS?=OFF
+EXTRA_CMAKE_ARGS?=
+CMAKE_ARGS+=-DBUILD_SHARED_LIBS=$(SHARED_LIBS) -DLLAMA_CURL=OFF $(EXTRA_CMAKE_ARGS)
 
 CURRENT_MAKEFILE_DIR := $(dir $(abspath $(lastword $(MAKEFILE_LIST))))
 ifeq ($(NATIVE),false)
@@ -120,6 +128,30 @@ llama-cpp-fallback: llama.cpp
 	CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) VARIANT="llama-cpp-fallback-build" build-llama-cpp-grpc-server
 	cp -rfv $(CURRENT_MAKEFILE_DIR)/../llama-cpp-fallback-build/grpc-server llama-cpp-fallback
 
+# Single-build CPU backend using ggml's CPU_ALL_VARIANTS. Produces ONE grpc-server
+# plus a set of dlopen-able libggml-cpu-*.so (sandybridge/haswell/skylakex/...) that
+# ggml's backend registry selects from at runtime by probing host CPU features.
+# Replaces the avx/avx2/avx512/fallback multi-binary build on x86.
+#
+# CPU_ALL_VARIANTS requires GGML_BACKEND_DL, which requires BUILD_SHARED_LIBS=ON, so we
+# pass SHARED_LIBS=ON and the DL flags as make variables (NOT pre-expanded into the
+# CMAKE_ARGS env string): command-line make variables propagate through every recursive
+# sub-make, so the deepest VARIANT-dir build computes BUILD_SHARED_LIBS=ON consistently.
+# Only ggml/llama go shared - gRPC is found via its static imported targets, so the
+# grpc-server binary keeps static gRPC and only dynamically links ggml.
+#
+# TARGET adds "ggml": the per-microarch backends are runtime-dlopened, not link deps of
+# grpc-server, so they only build because each is an add_dependencies() of the ggml target.
+llama-cpp-cpu-all: llama.cpp
+	cp -rf $(CURRENT_MAKEFILE_DIR)/../llama-cpp $(CURRENT_MAKEFILE_DIR)/../llama-cpp-cpu-all-build
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../llama-cpp-cpu-all-build purge
+	$(info ${GREEN}I llama-cpp build info:cpu-all-variants${RESET})
+	$(MAKE) SHARED_LIBS=ON EXTRA_CMAKE_ARGS="-DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON" TARGET="--target grpc-server --target ggml" VARIANT="llama-cpp-cpu-all-build" build-llama-cpp-grpc-server
+	cp -rfv $(CURRENT_MAKEFILE_DIR)/../llama-cpp-cpu-all-build/grpc-server llama-cpp-cpu-all
+	rm -rf ggml-shared-libs && mkdir -p ggml-shared-libs
+	find $(CURRENT_MAKEFILE_DIR)/../llama-cpp-cpu-all-build/llama.cpp/build \( -name '*.so*' -o -name '*.dylib' \) -exec cp -av {} ggml-shared-libs/ \;
+	@echo "Collected ggml shared backends:" && ls -la ggml-shared-libs/
+
 llama-cpp-grpc: llama.cpp
 	cp -rf $(CURRENT_MAKEFILE_DIR)/../llama-cpp $(CURRENT_MAKEFILE_DIR)/../llama-cpp-grpc-build
 	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../llama-cpp-grpc-build purge

backend/cpp/llama-cpp/grpc-server.cpp

@@ -37,6 +37,7 @@
 #include "backend.pb.h"
 #include "backend.grpc.pb.h"
 #include "common.h"
+#include "arg.h"
 #include "chat-auto-parser.h"
 #include <getopt.h>
 #include <grpcpp/ext/proto_server_reflection_plugin.h>
@@ -592,6 +593,10 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
     params.checkpoint_min_step = 256;
 #endif
 
+    // Raw upstream llama-server flags collected from any option entry that
+    // starts with '-'. Applied once after the loop via common_params_parse.
+    std::vector<std::string> extra_argv;
+
      // decode options. Options are in form optname:optvale, or if booleans only optname.
     for (int i = 0; i < request->options_size(); i++) {
         std::string opt = request->options(i);
@@ -1080,6 +1085,31 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
                 } catch (...) {}
             }
 
+        // --- main model MoE on CPU (upstream --cpu-moe / --n-cpu-moe) ---
+        } else if (!strcmp(optname, "cpu_moe")) {
+            // Bool-style flag: keep all MoE expert weights on CPU.
+            const bool enable = (optval == NULL) ||
+                optval_str == "true" || optval_str == "1" || optval_str == "yes" ||
+                optval_str == "on" || optval_str == "enabled";
+            if (enable) {
+                params.tensor_buft_overrides.push_back(llm_ffn_exps_cpu_override());
+            }
+        } else if (!strcmp(optname, "n_cpu_moe")) {
+            if (optval != NULL) {
+                try {
+                    int n = std::stoi(optval_str);
+                    if (n < 0) n = 0;
+                    // Keep override-name storage alive for the lifetime of the
+                    // params struct (mirrors upstream arg.cpp's function-local static).
+                    static std::list<std::string> buft_overrides_main;
+                    for (int i = 0; i < n; ++i) {
+                        buft_overrides_main.push_back(llm_ffn_exps_block_regex(i));
+                        params.tensor_buft_overrides.push_back(
+                            {buft_overrides_main.back().c_str(), ggml_backend_cpu_buffer_type()});
+                    }
+                } catch (...) {}
+            }
+
         // --- draft model tensor buffer overrides (upstream --spec-draft-override-tensor) ---
         } else if (!strcmp(optname, "draft_override_tensor") || !strcmp(optname, "spec_draft_override_tensor")) {
             // Format: <tensor regex>=<buffer type>,<tensor regex>=<buffer type>,...
@@ -1111,6 +1141,30 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
                 else { cur.push_back(c); }
             }
             if (!cur.empty()) flush(cur);
+
+        // --- generic passthrough: any entry starting with '-' is a raw
+        //     upstream llama-server flag, forwarded verbatim to the parser. ---
+        } else if (optname[0] == '-') {
+            std::string flag = optname;
+            // These flags make upstream's parser exit() (printing usage /
+            // completion), which would kill the backend process. Skip them.
+            if (flag == "-h" || flag == "--help" || flag == "--usage" ||
+                flag == "--version" || flag == "--license" ||
+                flag == "--list-devices" || flag == "-cl" ||
+                flag == "--cache-list" ||
+                flag.rfind("--completion", 0) == 0) {
+                fprintf(stderr,
+                    "[llama-cpp] ignoring passthrough flag that would exit: %s\n",
+                    flag.c_str());
+            } else {
+                extra_argv.push_back(flag);
+                // Preserve the whole value after the first ':' so embedded
+                // colons (e.g. host:port) survive strtok's truncation of optval.
+                auto colon = opt.find(':');
+                if (colon != std::string::npos) {
+                    extra_argv.push_back(opt.substr(colon + 1));
+                }
+            }
         }
     }
 
@@ -1146,27 +1200,6 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
         }
     }
 
-    if (!params.kv_overrides.empty()) {
-        params.kv_overrides.emplace_back();
-        params.kv_overrides.back().key[0] = 0;
-    }
-
-    // tensor_buft_overrides sentinel termination (mirrors upstream common/arg.cpp).
-    // Real entries are pushed during option parsing; here we pad/terminate so the
-    // model loader sees back().pattern == nullptr (GGML_ASSERT at common.cpp:1543)
-    // and so llama_params_fit has the placeholder slots it requires.
-    {
-        const size_t ntbo = llama_max_tensor_buft_overrides();
-        while (params.tensor_buft_overrides.size() < ntbo) {
-            params.tensor_buft_overrides.push_back({nullptr, nullptr});
-        }
-    }
-    // Terminate the draft tensor_buft_overrides list with a sentinel, mirroring
-    // the main-model handling above.
-    if (!params.speculative.draft.tensor_buft_overrides.empty()) {
-        params.speculative.draft.tensor_buft_overrides.push_back({nullptr, nullptr});
-    }
-
     // TODO: Add yarn
 
     if (!request->tensorsplit().empty()) {
@@ -1259,6 +1292,69 @@ static void params_parse(server_context& /*ctx_server*/, const backend::ModelOpt
             params.sampling.grammar_triggers.push_back(std::move(trigger));
         }
     }
+
+    // Apply any raw upstream flags last so an explicit passthrough flag wins
+    // over the LocalAI-resolved field it maps to (e.g. --ctx-size beats
+    // context_size). This is the same parser llama-server itself uses.
+    if (!extra_argv.empty()) {
+        // common_params_parser_init resets a few fields for the SERVER example
+        // (n_parallel -> -1, use_color). Snapshot n_parallel so an unrelated
+        // passthrough flag can't silently clobber LocalAI's resolved value.
+        const int saved_n_parallel = params.n_parallel;
+
+        std::vector<char *> argv;
+        std::string prog = "llama-server";
+        argv.push_back(prog.data());
+        for (auto & a : extra_argv) {
+            argv.push_back(a.data());
+        }
+
+        // ctx_arg.params is a reference, so this overlays the given flags onto
+        // `params` in place. Returns false on a recoverable parse error (and
+        // self-restores params); may exit() on a hard error, exactly as
+        // passing the same bad flag to llama-server would.
+        if (!common_params_parse((int)argv.size(), argv.data(), params,
+                                 LLAMA_EXAMPLE_SERVER)) {
+            fprintf(stderr,
+                "[llama-cpp] failed to parse passthrough options; ignoring them\n");
+        }
+
+        // Restore n_parallel unless a passthrough flag explicitly set it
+        // (parser_init's reset sentinel for SERVER is -1).
+        if (params.n_parallel == -1) {
+            params.n_parallel = saved_n_parallel;
+        }
+    }
+
+    // Terminate/pad the override vectors only after BOTH the named-option loop
+    // and the generic passthrough (common_params_parse above) have pushed their
+    // real entries, so back() is the null sentinel the model loader asserts on.
+    // Running these before the passthrough let a passthrough flag (--cpu-moe,
+    // --override-tensor, --override-kv, ...) append a real entry after the
+    // sentinel: a GGML_ASSERT crash for tensor_buft_overrides, a silent drop for
+    // kv_overrides. Double-termination is harmless (the while is a no-op if the
+    // passthrough parse already padded; an extra trailing null is ignored).
+
+    if (!params.kv_overrides.empty()) {
+        params.kv_overrides.emplace_back();
+        params.kv_overrides.back().key[0] = 0;
+    }
+
+    // tensor_buft_overrides sentinel termination (mirrors upstream common/arg.cpp).
+    // Real entries are pushed during option parsing; here we pad/terminate so the
+    // model loader sees back().pattern == nullptr (GGML_ASSERT at common.cpp:1543)
+    // and so llama_params_fit has the placeholder slots it requires.
+    {
+        const size_t ntbo = llama_max_tensor_buft_overrides();
+        while (params.tensor_buft_overrides.size() < ntbo) {
+            params.tensor_buft_overrides.push_back({nullptr, nullptr});
+        }
+    }
+    // Terminate the draft tensor_buft_overrides list with a sentinel, mirroring
+    // the main-model handling above.
+    if (!params.speculative.draft.tensor_buft_overrides.empty()) {
+        params.speculative.draft.tensor_buft_overrides.push_back({nullptr, nullptr});
+    }
 }
 
 

backend/cpp/llama-cpp/package.sh

@@ -14,6 +14,22 @@ mkdir -p $CURDIR/package/lib
 cp -avrf $CURDIR/llama-cpp-* $CURDIR/package/
 cp -rfv $CURDIR/run.sh $CURDIR/package/
 
+# Bundle the ggml shared backends produced by the CPU_ALL_VARIANTS build (libggml-base.so,
+# libggml.so, libllama.so and the per-microarch libggml-cpu-*.so), all into package/lib.
+#
+# Two distinct resolution mechanisms both land here:
+#   - NEEDED deps (libggml-base/libggml/libllama): resolved by the dynamic linker via the
+#     LD_LIBRARY_PATH=$CURDIR/lib that run.sh exports.
+#   - The per-microarch libggml-cpu-*.so are NOT linked; ggml *discovers* them at runtime by
+#     scanning the executable's own directory (readlink /proc/self/exe). run.sh launches via
+#     the bundled $CURDIR/lib/ld.so, so /proc/self/exe -> .../lib/ld.so and ggml scans lib/.
+#     That is why the variants must sit in lib/ (next to ld.so), not just on the link path.
+# No-op on builds (arm64/darwin) that don't produce the all-variants set.
+if [ -d "$CURDIR/ggml-shared-libs" ]; then
+    echo "Bundling ggml shared backends (CPU_ALL_VARIANTS)..."
+    cp -avf $CURDIR/ggml-shared-libs/*.so* $CURDIR/package/lib/
+fi
+
 # Detect architecture and copy appropriate libraries
 if [ -f "/lib64/ld-linux-x86-64.so.2" ]; then
     # x86_64 architecture

backend/cpp/llama-cpp/run.sh

@@ -12,26 +12,12 @@ grep -e "flags" /proc/cpuinfo | head -1
 
 BINARY=llama-cpp-fallback
 
-if grep -q -e "\savx\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX    found OK"
-	if [ -e $CURDIR/llama-cpp-avx ]; then
-		BINARY=llama-cpp-avx
-	fi
-fi
-
-if grep -q -e "\savx2\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX2   found OK"
-	if [ -e $CURDIR/llama-cpp-avx2 ]; then
-		BINARY=llama-cpp-avx2
-	fi
-fi
-
-# Check avx 512
-if grep -q -e "\savx512f\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX512F found OK"
-	if [ -e $CURDIR/llama-cpp-avx512 ]; then
-		BINARY=llama-cpp-avx512
-	fi
+# CPU images (x86, arm64, darwin) ship a single llama-cpp-cpu-all built with ggml
+# CPU_ALL_VARIANTS: ggml's backend registry dlopens the best libggml-cpu-*.so for this
+# host, so no shell-side AVX probing. GPU images (cublas/sycl/vulkan/hipblas) ship only
+# llama-cpp-fallback (the accelerator does the compute), so fall back to it when absent.
+if [ -e $CURDIR/llama-cpp-cpu-all ]; then
+	BINARY=llama-cpp-cpu-all
 fi
 
 if [ -n "$LLAMACPP_GRPC_SERVERS" ]; then

backend/cpp/privacy-filter/CMakeLists.txt

@@ -51,6 +51,14 @@ add_library(hw_grpc_proto STATIC
     ${HW_GRPC_SRCS} ${HW_GRPC_HDRS}
     ${HW_PROTO_SRCS} ${HW_PROTO_HDRS})
 target_include_directories(hw_grpc_proto PUBLIC ${CMAKE_CURRENT_BINARY_DIR})
+# The generated proto/grpc sources include protobuf and grpc++ headers, so this
+# library must see their include dirs. Linking the imported targets propagates
+# them. On Linux the apt headers live in /usr/include (default search path) so
+# this was a no-op; on macOS the Homebrew headers are under /opt/homebrew and
+# would otherwise be missed (runtime_version.h not found).
+target_link_libraries(hw_grpc_proto PUBLIC
+    protobuf::libprotobuf
+    gRPC::grpc++)
 
 # Build only the pf static lib (+ ggml) from the engine tree — no CLI/bench/tests.
 # PF_VULKAN is honored when passed on the cmake command line (it lands in the

backend/cpp/privacy-filter/run.sh

@@ -2,7 +2,13 @@
 # Entry point for the privacy-filter backend image / BACKEND_BINARY mode.
 set -e
 CURDIR=$(dirname "$(realpath "$0")")
-export LD_LIBRARY_PATH="$CURDIR/lib:$LD_LIBRARY_PATH"
+# macOS has no bundled ld.so; the darwin package ships only dylibs under lib/,
+# resolved via DYLD_LIBRARY_PATH (the ld.so branch below is skipped there).
+if [ "$(uname)" = "Darwin" ]; then
+    export DYLD_LIBRARY_PATH="$CURDIR/lib:$DYLD_LIBRARY_PATH"
+else
+    export LD_LIBRARY_PATH="$CURDIR/lib:$LD_LIBRARY_PATH"
+fi
 if [ -f "$CURDIR/lib/ld.so" ]; then
     exec "$CURDIR/lib/ld.so" "$CURDIR/grpc-server" "$@"
 fi

backend/cpp/turboquant/Makefile

@@ -65,6 +65,29 @@ turboquant-avx:
 turboquant-fallback:
 	$(call turboquant-build,fallback,-DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off,--target grpc-server)
 
+# Single-build CPU backend via ggml CPU_ALL_VARIANTS (mirrors llama-cpp-cpu-all).
+# turboquant reuses backend/cpp/llama-cpp's CMakeLists.txt (hw_grpc_proto STATIC) and
+# Makefile (SHARED_LIBS make-var + EXTRA_CMAKE_ARGS), so this passes the same overrides
+# through to the copied build: SHARED_LIBS=ON, the DL flags, and --target ggml (which
+# pulls in the per-microarch libggml-cpu-*.so via ggml's add_dependencies). The .so set
+# is collected for package.sh to bundle into package/lib.
+turboquant-cpu-all:
+	rm -rf $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build
+	cp -rf $(LLAMA_CPP_DIR) $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build purge
+	bash $(CURRENT_MAKEFILE_DIR)/patch-grpc-server.sh $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build/grpc-server.cpp
+	$(info $(GREEN)I turboquant build info:cpu-all-variants$(RESET))
+	LLAMA_REPO=$(LLAMA_REPO) LLAMA_VERSION=$(TURBOQUANT_VERSION) \
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build llama.cpp
+	bash $(CURRENT_MAKEFILE_DIR)/apply-patches.sh $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build/llama.cpp $(PATCHES_DIR)
+	SHARED_LIBS=ON EXTRA_CMAKE_ARGS="-DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON" TARGET="--target grpc-server --target ggml" \
+	LLAMA_REPO=$(LLAMA_REPO) LLAMA_VERSION=$(TURBOQUANT_VERSION) \
+	$(MAKE) -C $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build grpc-server
+	cp -rfv $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build/grpc-server turboquant-cpu-all
+	rm -rf ggml-shared-libs && mkdir -p ggml-shared-libs
+	find $(CURRENT_MAKEFILE_DIR)/../turboquant-cpu-all-build/llama.cpp/build \( -name '*.so*' -o -name '*.dylib' \) -exec cp -av {} ggml-shared-libs/ \;
+	@echo "Collected ggml shared backends:" && ls -la ggml-shared-libs/
+
 turboquant-grpc:
 	$(call turboquant-build,grpc,-DGGML_RPC=ON -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off,--target grpc-server --target rpc-server)
 

backend/cpp/turboquant/package.sh

@@ -14,6 +14,15 @@ mkdir -p $CURDIR/package/lib
 cp -avrf $CURDIR/turboquant-* $CURDIR/package/
 cp -rfv $CURDIR/run.sh $CURDIR/package/
 
+# Bundle the ggml shared backends from the CPU_ALL_VARIANTS build into package/lib. ggml
+# discovers the per-microarch libggml-cpu-*.so by scanning the executable directory, which
+# (via the bundled lib/ld.so that run.sh launches through) resolves to lib/. See the
+# matching comment in backend/cpp/llama-cpp/package.sh. No-op on the fallback/ROCm builds.
+if [ -d "$CURDIR/ggml-shared-libs" ]; then
+    echo "Bundling ggml shared backends (CPU_ALL_VARIANTS)..."
+    cp -avf $CURDIR/ggml-shared-libs/*.so* $CURDIR/package/lib/
+fi
+
 # Detect architecture and copy appropriate libraries
 if [ -f "/lib64/ld-linux-x86-64.so.2" ]; then
     # x86_64 architecture

backend/cpp/turboquant/run.sh

@@ -12,26 +12,11 @@ grep -e "flags" /proc/cpuinfo | head -1
 
 BINARY=turboquant-fallback
 
-if grep -q -e "\savx\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX    found OK"
-	if [ -e $CURDIR/turboquant-avx ]; then
-		BINARY=turboquant-avx
-	fi
-fi
-
-if grep -q -e "\savx2\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX2   found OK"
-	if [ -e $CURDIR/turboquant-avx2 ]; then
-		BINARY=turboquant-avx2
-	fi
-fi
-
-# Check avx 512
-if grep -q -e "\savx512f\s" /proc/cpuinfo ; then
-	echo "CPU:    AVX512F found OK"
-	if [ -e $CURDIR/turboquant-avx512 ]; then
-		BINARY=turboquant-avx512
-	fi
+# x86/arm64 ship a single turboquant-cpu-all built with ggml CPU_ALL_VARIANTS: ggml's
+# backend registry dlopens the best libggml-cpu-*.so for this host, so no shell-side
+# probing. ROCm ships only turboquant-fallback, so fall back to it when cpu-all is absent.
+if [ -e $CURDIR/turboquant-cpu-all ]; then
+	BINARY=turboquant-cpu-all
 fi
 
 if [ -n "$LLAMACPP_GRPC_SERVERS" ]; then

backend/go/acestep-cpp/Makefile

@@ -117,7 +117,8 @@ libgoacestepcpp-custom: CMakeLists.txt cpp/goacestepcpp.cpp cpp/goacestepcpp.h
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) --target goacestepcpp && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libgoacestepcpp.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/libgoacestepcpp.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/libgoacestepcpp.dylib ./$(SO_TARGET) 2>/dev/null)
 
 test: acestep-cpp
 	@echo "Running acestep-cpp tests..."

backend/go/acestep-cpp/main.go

@@ -4,6 +4,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -22,7 +23,11 @@ func main() {
 	// Get library name from environment variable, default to fallback
 	libName := os.Getenv("ACESTEP_LIBRARY")
 	if libName == "" {
-		libName = "./libgoacestepcpp-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libgoacestepcpp-fallback.dylib"
+		} else {
+			libName = "./libgoacestepcpp-fallback.so"
+		}
 	}
 
 	gosd, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/acestep-cpp/package.sh

@@ -13,6 +13,7 @@ mkdir -p $CURDIR/package/lib
 
 cp -avf $CURDIR/acestep-cpp $CURDIR/package/
 cp -fv $CURDIR/libgoacestepcpp-*.so $CURDIR/package/
+cp -fv $CURDIR/libgoacestepcpp-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -fv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries

backend/go/acestep-cpp/run.sh

@@ -12,9 +12,19 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libgoacestepcpp-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single library variant (Metal or Accelerate). The goacestepcpp
+	# target is built as a CMake MODULE, which emits a .dylib for a SHARED
+	# build but a .so for a MODULE build on Apple, so prefer .dylib and fall
+	# back to .so.
+	LIBRARY="$CURDIR/libgoacestepcpp-fallback.dylib"
+	if [ ! -e "$LIBRARY" ]; then
+		LIBRARY="$CURDIR/libgoacestepcpp-fallback.so"
+	fi
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libgoacestepcpp-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libgoacestepcpp-avx.so ]; then
@@ -36,9 +46,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libgoacestepcpp-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export ACESTEP_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/go/ced/Makefile

@@ -57,6 +57,7 @@ libced.so: sources/ced.cpp
 	cmake -B sources/ced.cpp/build-shared -S sources/ced.cpp $(CMAKE_ARGS)
 	cmake --build sources/ced.cpp/build-shared --config Release -j$(JOBS)
 	cp -fv sources/ced.cpp/build-shared/libced.so* ./ 2>/dev/null || true
+	cp -fv sources/ced.cpp/build-shared/libced.dylib ./ 2>/dev/null || true
 	cp -fv sources/ced.cpp/include/ced_capi.h ./
 
 ced-grpc: libced.so main.go goced.go

backend/go/ced/main.go

@@ -12,6 +12,7 @@ import (
 	"flag"
 	"fmt"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -27,7 +28,11 @@ type libFunc struct {
 func main() {
 	libName := os.Getenv("CED_LIBRARY")
 	if libName == "" {
-		libName = "libced.so"
+		if runtime.GOOS == "darwin" {
+			libName = "libced.dylib"
+		} else {
+			libName = "libced.so"
+		}
 	}
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)
 	if err != nil {

backend/go/ced/package.sh

@@ -15,10 +15,12 @@ mkdir -p "$CURDIR/package/lib"
 cp -avf "$CURDIR/ced-grpc" "$CURDIR/package/"
 cp -avf "$CURDIR/run.sh" "$CURDIR/package/"
 
-cp -avf "$CURDIR"/libced.so* "$CURDIR/package/lib/" 2>/dev/null || {
-	echo "ERROR: libced.so not found in $CURDIR, run 'make' first" >&2
+cp -avf "$CURDIR"/libced.so* "$CURDIR/package/lib/" 2>/dev/null || true
+cp -avf "$CURDIR"/libced.dylib "$CURDIR/package/lib/" 2>/dev/null || true
+if ! ls "$CURDIR"/package/lib/libced.* >/dev/null 2>&1; then
+	echo "ERROR: libced shared library not found in $CURDIR, run 'make' first" >&2
 	exit 1
-}
+fi
 
 if [ -f "/lib64/ld-linux-x86-64.so.2" ]; then
     echo "Detected x86_64 architecture, copying x86_64 libraries..."

backend/go/ced/run.sh

@@ -3,7 +3,12 @@ set -e
 
 CURDIR=$(dirname "$(realpath "$0")")
 
-export LD_LIBRARY_PATH="$CURDIR/lib:$CURDIR:${LD_LIBRARY_PATH:-}"
+if [ "$(uname)" = "Darwin" ]; then
+	export DYLD_LIBRARY_PATH="$CURDIR/lib:$CURDIR:${DYLD_LIBRARY_PATH:-}"
+	export CED_LIBRARY="$CURDIR/lib/libced.dylib"
+else
+	export LD_LIBRARY_PATH="$CURDIR/lib:$CURDIR:${LD_LIBRARY_PATH:-}"
+fi
 
 # If a self-contained ld.so was packaged, route through it so the packaged
 # libc / libstdc++ are used instead of the host's (matches the sibling backends).

backend/go/crispasr/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # CrispASR version (release tag)
 CRISPASR_REPO?=https://github.com/CrispStrobe/CrispASR
-CRISPASR_VERSION?=63b57289255267edf66e43e33bc3911e04a2e92d
+CRISPASR_VERSION?=96b2a6ee31d30389fed8a7ef1a54239b75231ddc
 SO_TARGET?=libgocrispasr.so
 
 CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
@@ -75,7 +75,8 @@ UNAME_S := $(shell uname -s)
 ifeq ($(UNAME_S),Linux)
 	VARIANT_TARGETS = libgocrispasr-avx.so libgocrispasr-avx2.so libgocrispasr-avx512.so libgocrispasr-fallback.so
 else
-	VARIANT_TARGETS = libgocrispasr-fallback.so
+	# On non-Linux (e.g., Darwin), build only fallback variant (as a dylib)
+	VARIANT_TARGETS = libgocrispasr-fallback.dylib
 endif
 
 crispasr: main.go gocrispasr.go $(VARIANT_TARGETS)
@@ -87,7 +88,7 @@ package: crispasr
 build: package
 
 clean: purge
-	rm -rf libgocrispasr*.so package sources/CrispASR crispasr
+	rm -rf libgocrispasr*.so libgocrispasr*.dylib package sources/CrispASR crispasr
 
 purge:
 	rm -rf build*
@@ -118,13 +119,21 @@ libgocrispasr-fallback.so: sources/CrispASR
 	SO_TARGET=libgocrispasr-fallback.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgocrispasr-custom
 	rm -rfv build*
 
+# Build fallback variant as a dylib (Darwin)
+libgocrispasr-fallback.dylib: sources/CrispASR
+	$(MAKE) purge
+	$(info ${GREEN}I crispasr build info:fallback (dylib)${RESET})
+	SO_TARGET=libgocrispasr-fallback.dylib CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgocrispasr-custom
+	rm -rfv build*
+
 libgocrispasr-custom: CMakeLists.txt cpp/crispasr_shim.cpp cpp/crispasr_shim.h
 	mkdir -p build-$(SO_TARGET) && \
 	cd build-$(SO_TARGET) && \
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libgocrispasr.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/libgocrispasr.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/libgocrispasr.dylib ./$(SO_TARGET) 2>/dev/null)
 
 test: crispasr
 	CGO_ENABLED=0 $(GOCMD) test -v ./...

backend/go/crispasr/main.go

@@ -4,6 +4,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -21,7 +22,11 @@ type LibFuncs struct {
 func main() {
 	libName := os.Getenv("CRISPASR_LIBRARY")
 	if libName == "" {
-		libName = "./libgocrispasr-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libgocrispasr-fallback.dylib"
+		} else {
+			libName = "./libgocrispasr-fallback.so"
+		}
 	}
 
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/crispasr/package.sh

@@ -12,7 +12,8 @@ REPO_ROOT="${CURDIR}/../../.."
 mkdir -p $CURDIR/package/lib
 
 cp -avf $CURDIR/crispasr $CURDIR/package/
-cp -fv $CURDIR/libgocrispasr-*.so $CURDIR/package/
+cp -fv $CURDIR/libgocrispasr-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/libgocrispasr-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -fv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries

backend/go/crispasr/run.sh

@@ -12,9 +12,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libgocrispasr-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/libgocrispasr-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libgocrispasr-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libgocrispasr-avx.so ]; then
@@ -36,9 +40,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libgocrispasr-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export CRISPASR_LIBRARY=$LIBRARY
 
 # Point piper's espeak-ng phonemizer at the bundled voice data. The variable

backend/go/depth-anything-cpp/Makefile

@@ -40,6 +40,8 @@ else ifeq ($(BUILD_TYPE),hipblas)
 else ifeq ($(BUILD_TYPE),vulkan)
 	CMAKE_ARGS+=-DGGML_VULKAN=ON -DDA_GGML_VULKAN=ON
 else ifeq ($(OS),Darwin)
+	# macOS/Metal: built + published as an OCI image by CI (includeDarwin in
+	# .github/backend-matrix.yml) so Apple Silicon users can install this backend.
 	ifneq ($(BUILD_TYPE),metal)
 		CMAKE_ARGS+=-DGGML_METAL=OFF
 	else
@@ -77,7 +79,7 @@ ifeq ($(UNAME_S),Linux)
 	VARIANT_TARGETS = libdepthanythingcpp-avx.so libdepthanythingcpp-avx2.so libdepthanythingcpp-avx512.so libdepthanythingcpp-fallback.so
 else
 	# On non-Linux (e.g., Darwin), build only fallback variant
-	VARIANT_TARGETS = libdepthanythingcpp-fallback.so
+	VARIANT_TARGETS = libdepthanythingcpp-fallback.dylib
 endif
 
 depth-anything-cpp: main.go godepthanythingcpp.go $(VARIANT_TARGETS)
@@ -89,7 +91,7 @@ package: depth-anything-cpp
 build: package
 
 clean: purge
-	rm -rf libdepthanythingcpp*.so depth-anything-cpp package sources
+	rm -rf libdepthanythingcpp*.so libdepthanythingcpp*.dylib depth-anything-cpp package sources
 
 purge:
 	rm -rf build*
@@ -116,11 +118,19 @@ libdepthanythingcpp-avx512.so: sources/depth-anything.cpp
 endif
 
 # Build fallback variant (all platforms)
+ifeq ($(UNAME_S),Darwin)
+libdepthanythingcpp-fallback.dylib: sources/depth-anything.cpp
+	rm -rfv build-$@
+	$(info ${GREEN}I depth-anything-cpp build info:fallback${RESET})
+	SO_TARGET=$@ CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libdepthanythingcpp-custom
+	rm -rfv build-$@
+else
 libdepthanythingcpp-fallback.so: sources/depth-anything.cpp
 	rm -rfv build-$@
 	$(info ${GREEN}I depth-anything-cpp build info:fallback${RESET})
 	SO_TARGET=$@ CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libdepthanythingcpp-custom
 	rm -rfv build-$@
+endif
 
 libdepthanythingcpp-custom: CMakeLists.txt
 	mkdir -p build-$(SO_TARGET) && \
@@ -128,7 +138,8 @@ libdepthanythingcpp-custom: CMakeLists.txt
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libdepthanything.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/libdepthanything.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/libdepthanything.dylib ./$(SO_TARGET) 2>/dev/null)
 
 all: depth-anything-cpp package
 

backend/go/depth-anything-cpp/main.go

@@ -9,6 +9,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -27,7 +28,11 @@ func main() {
 	// Get library name from environment variable, default to fallback
 	libName := os.Getenv("DEPTHANYTHING_LIBRARY")
 	if libName == "" {
-		libName = "./libdepthanythingcpp-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libdepthanythingcpp-fallback.dylib"
+		} else {
+			libName = "./libdepthanythingcpp-fallback.so"
+		}
 	}
 
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/depth-anything-cpp/package.sh

@@ -10,7 +10,8 @@ REPO_ROOT="${CURDIR}/../../.."
 # Create lib directory
 mkdir -p $CURDIR/package/lib
 
-cp -avf $CURDIR/libdepthanythingcpp-*.so $CURDIR/package/
+cp -fv $CURDIR/libdepthanythingcpp-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/libdepthanythingcpp-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -avf $CURDIR/depth-anything-cpp $CURDIR/package/
 cp -fv $CURDIR/run.sh $CURDIR/package/
 

backend/go/depth-anything-cpp/run.sh

@@ -12,9 +12,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libdepthanythingcpp-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/libdepthanythingcpp-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libdepthanythingcpp-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libdepthanythingcpp-avx.so ]; then
@@ -36,9 +40,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libdepthanythingcpp-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export DEPTHANYTHING_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/go/localvqe/Makefile

@@ -32,6 +32,8 @@ endif
 ifeq ($(BUILD_TYPE),vulkan)
 	CMAKE_ARGS+=-DGGML_VULKAN=ON -DLOCALVQE_VULKAN=ON
 else ifeq ($(OS),Darwin)
+	# Apple Silicon: CPU-only (no Metal upstream); built + published as an arm64
+	# image by CI (includeDarwin in .github/backend-matrix.yml) for macOS install.
 	CMAKE_ARGS+=-DGGML_METAL=OFF
 endif
 
@@ -67,8 +69,9 @@ $(LIB_SENTINEL): sources/LocalVQE
 	# that the loader picks at runtime. We must build every target — the
 	# default `--target localvqe_shared` drops these. CMAKE_LIBRARY_OUTPUT_DIRECTORY
 	# routes all of them into build/bin; copy them out next to the binary.
-	cp -P build/bin/liblocalvqe.so* . 2>/dev/null || cp -P build/liblocalvqe.so* .
+	cp -P build/bin/liblocalvqe.so* . 2>/dev/null || cp -P build/bin/liblocalvqe.dylib . 2>/dev/null || cp -P build/liblocalvqe.so* . 2>/dev/null || cp -P build/liblocalvqe.dylib .
 	cp -P build/bin/libggml*.so* . 2>/dev/null || true
+	cp -P build/bin/libggml*.dylib . 2>/dev/null || true
 	touch $(LIB_SENTINEL)
 
 liblocalvqe.so: $(LIB_SENTINEL)

backend/go/localvqe/main.go

@@ -4,6 +4,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -21,7 +22,11 @@ type LibFuncs struct {
 func main() {
 	libName := os.Getenv("LOCALVQE_LIBRARY")
 	if libName == "" {
-		libName = "./liblocalvqe.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./liblocalvqe.dylib"
+		} else {
+			libName = "./liblocalvqe.so"
+		}
 	}
 
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/localvqe/package.sh

@@ -15,7 +15,9 @@ cp -avf $CURDIR/localvqe $CURDIR/package/
 # liblocalvqe.so* (with SOVERSION symlinks) and the libggml-*.so runtime
 # variants — LocalVQE picks the matching CPU variant at load time.
 cp -P $CURDIR/liblocalvqe.so* $CURDIR/package/ 2>/dev/null || true
+cp -P $CURDIR/liblocalvqe.dylib $CURDIR/package/ 2>/dev/null || true
 cp -P $CURDIR/libggml*.so* $CURDIR/package/ 2>/dev/null || true
+cp -P $CURDIR/libggml*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -fv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries

backend/go/localvqe/run.sh

@@ -10,8 +10,19 @@ CURDIR=$(dirname "$(realpath $0)")
 # exec'ing the binary.
 cd "$CURDIR"
 
-export LD_LIBRARY_PATH=$CURDIR:$CURDIR/lib:$LD_LIBRARY_PATH
-export LOCALVQE_LIBRARY=$CURDIR/liblocalvqe.so
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: LocalVQE is built as a SHARED library, so dyld needs the .dylib +
+	# DYLD_LIBRARY_PATH. Prefer .dylib and fall back to .so just in case.
+	export DYLD_LIBRARY_PATH=$CURDIR:$CURDIR/lib:$DYLD_LIBRARY_PATH
+	LOCALVQE_LIBRARY=$CURDIR/liblocalvqe.dylib
+	if [ ! -e "$LOCALVQE_LIBRARY" ]; then
+		LOCALVQE_LIBRARY=$CURDIR/liblocalvqe.so
+	fi
+	export LOCALVQE_LIBRARY
+else
+	export LD_LIBRARY_PATH=$CURDIR:$CURDIR/lib:$LD_LIBRARY_PATH
+	export LOCALVQE_LIBRARY=$CURDIR/liblocalvqe.so
+fi
 
 if [ -f $CURDIR/lib/ld.so ]; then
 	echo "Using lib/ld.so"

backend/go/locate-anything-cpp/Makefile

@@ -33,6 +33,8 @@ else ifeq ($(BUILD_TYPE),hipblas)
 else ifeq ($(BUILD_TYPE),vulkan)
 	CMAKE_ARGS+=-DGGML_VULKAN=ON -DLA_GGML_VULKAN=ON
 else ifeq ($(OS),Darwin)
+	# macOS/Metal: built + published as an OCI image by CI (includeDarwin in
+	# .github/backend-matrix.yml) so Apple Silicon users can install this backend.
 	ifneq ($(BUILD_TYPE),metal)
 		CMAKE_ARGS+=-DGGML_METAL=OFF
 	else
@@ -70,7 +72,7 @@ ifeq ($(UNAME_S),Linux)
 	VARIANT_TARGETS = liblocateanythingcpp-avx.so liblocateanythingcpp-avx2.so liblocateanythingcpp-avx512.so liblocateanythingcpp-fallback.so
 else
 	# On non-Linux (e.g., Darwin), build only fallback variant
-	VARIANT_TARGETS = liblocateanythingcpp-fallback.so
+	VARIANT_TARGETS = liblocateanythingcpp-fallback.dylib
 endif
 
 locate-anything-cpp: main.go golocateanythingcpp.go $(VARIANT_TARGETS)
@@ -82,7 +84,7 @@ package: locate-anything-cpp
 build: package
 
 clean: purge
-	rm -rf liblocateanythingcpp*.so locate-anything-cpp package sources
+	rm -rf liblocateanythingcpp*.so liblocateanythingcpp*.dylib locate-anything-cpp package sources
 
 purge:
 	rm -rf build*
@@ -109,11 +111,19 @@ liblocateanythingcpp-avx512.so: sources/locate-anything.cpp
 endif
 
 # Build fallback variant (all platforms)
+ifeq ($(UNAME_S),Darwin)
+liblocateanythingcpp-fallback.dylib: sources/locate-anything.cpp
+	rm -rfv build-$@
+	$(info ${GREEN}I locate-anything-cpp build info:fallback${RESET})
+	SO_TARGET=$@ CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) liblocateanythingcpp-custom
+	rm -rfv build-$@
+else
 liblocateanythingcpp-fallback.so: sources/locate-anything.cpp
 	rm -rfv build-$@
 	$(info ${GREEN}I locate-anything-cpp build info:fallback${RESET})
 	SO_TARGET=$@ CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) liblocateanythingcpp-custom
 	rm -rfv build-$@
+endif
 
 liblocateanythingcpp-custom: CMakeLists.txt
 	mkdir -p build-$(SO_TARGET) && \
@@ -121,7 +131,8 @@ liblocateanythingcpp-custom: CMakeLists.txt
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) && \
 	cd .. && \
-	mv build-$(SO_TARGET)/liblocateanythingcpp.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/liblocateanythingcpp.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/liblocateanythingcpp.dylib ./$(SO_TARGET) 2>/dev/null)
 
 all: locate-anything-cpp package
 

backend/go/locate-anything-cpp/main.go

@@ -9,6 +9,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -27,7 +28,11 @@ func main() {
 	// Get library name from environment variable, default to fallback
 	libName := os.Getenv("LOCATEANYTHING_LIBRARY")
 	if libName == "" {
-		libName = "./liblocateanythingcpp-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./liblocateanythingcpp-fallback.dylib"
+		} else {
+			libName = "./liblocateanythingcpp-fallback.so"
+		}
 	}
 
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/locate-anything-cpp/package.sh

@@ -10,7 +10,8 @@ REPO_ROOT="${CURDIR}/../../.."
 # Create lib directory
 mkdir -p $CURDIR/package/lib
 
-cp -avf $CURDIR/liblocateanythingcpp-*.so $CURDIR/package/
+cp -fv $CURDIR/liblocateanythingcpp-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/liblocateanythingcpp-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -avf $CURDIR/locate-anything-cpp $CURDIR/package/
 cp -fv $CURDIR/run.sh $CURDIR/package/
 

backend/go/locate-anything-cpp/run.sh

@@ -12,9 +12,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/liblocateanythingcpp-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/liblocateanythingcpp-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/liblocateanythingcpp-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/liblocateanythingcpp-avx.so ]; then
@@ -36,9 +40,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/liblocateanythingcpp-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export LOCATEANYTHING_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/go/omnivoice-cpp/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # omnivoice.cpp version
 OMNIVOICE_REPO?=https://github.com/ServeurpersoCom/omnivoice.cpp
-OMNIVOICE_VERSION?=96d30169afd5e6bb3fd6a0e9be0eb505bfe81fcd
+OMNIVOICE_VERSION?=0f37401bebe9b20c0160a888e592108fc1d17607
 SO_TARGET?=libgomnivoicecpp.so
 
 CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
@@ -65,7 +65,8 @@ UNAME_S := $(shell uname -s)
 ifeq ($(UNAME_S),Linux)
 	VARIANT_TARGETS = libgomnivoicecpp-avx.so libgomnivoicecpp-avx2.so libgomnivoicecpp-avx512.so libgomnivoicecpp-fallback.so
 else
-	VARIANT_TARGETS = libgomnivoicecpp-fallback.so
+	# On non-Linux (e.g., Darwin), build only fallback variant (as a dylib)
+	VARIANT_TARGETS = libgomnivoicecpp-fallback.dylib
 endif
 
 omnivoice-cpp: main.go gomnivoicecpp.go $(VARIANT_TARGETS)
@@ -77,7 +78,7 @@ package: omnivoice-cpp
 build: package
 
 clean: purge
-	rm -rf libgomnivoicecpp*.so package sources/omnivoice.cpp omnivoice-cpp
+	rm -rf libgomnivoicecpp*.so libgomnivoicecpp*.dylib package sources/omnivoice.cpp omnivoice-cpp
 
 purge:
 	rm -rf build*
@@ -106,13 +107,20 @@ libgomnivoicecpp-fallback.so: sources/omnivoice.cpp
 	SO_TARGET=libgomnivoicecpp-fallback.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgomnivoicecpp-custom
 	rm -rf build-libgomnivoicecpp-fallback.so
 
+# Build fallback variant as a dylib (Darwin)
+libgomnivoicecpp-fallback.dylib: sources/omnivoice.cpp
+	$(info ${GREEN}I omnivoice-cpp build info:fallback (dylib)${RESET})
+	SO_TARGET=libgomnivoicecpp-fallback.dylib CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgomnivoicecpp-custom
+	rm -rf build-libgomnivoicecpp-fallback.dylib
+
 libgomnivoicecpp-custom: CMakeLists.txt cpp/gomnivoicecpp.cpp cpp/gomnivoicecpp.h
 	mkdir -p build-$(SO_TARGET) && \
 	cd build-$(SO_TARGET) && \
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) --target gomnivoicecpp && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libgomnivoicecpp.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/libgomnivoicecpp.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/libgomnivoicecpp.dylib ./$(SO_TARGET) 2>/dev/null)
 
 test: omnivoice-cpp
 	@echo "Running omnivoice-cpp tests..."

backend/go/omnivoice-cpp/main.go

@@ -4,6 +4,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -21,7 +22,11 @@ type LibFuncs struct {
 func main() {
 	libName := os.Getenv("OMNIVOICE_LIBRARY")
 	if libName == "" {
-		libName = "./libgomnivoicecpp-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libgomnivoicecpp-fallback.dylib"
+		} else {
+			libName = "./libgomnivoicecpp-fallback.so"
+		}
 	}
 
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/omnivoice-cpp/package.sh

@@ -12,7 +12,8 @@ REPO_ROOT="${CURDIR}/../../.."
 mkdir -p $CURDIR/package/lib
 
 cp -avf $CURDIR/omnivoice-cpp $CURDIR/package/
-cp -fv $CURDIR/libgomnivoicecpp-*.so $CURDIR/package/
+cp -fv $CURDIR/libgomnivoicecpp-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/libgomnivoicecpp-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -fv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries

backend/go/omnivoice-cpp/run.sh

@@ -12,9 +12,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libgomnivoicecpp-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/libgomnivoicecpp-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libgomnivoicecpp-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libgomnivoicecpp-avx.so ]; then
@@ -36,9 +40,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libgomnivoicecpp-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export OMNIVOICE_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/go/parakeet-cpp/Makefile

@@ -1,6 +1,6 @@
 # parakeet-cpp backend Makefile.
 #
-# Upstream pin lives below as PARAKEET_VERSION?=db755a78d39f789bb7d4e3935158a9e8105dbe36
+# Upstream pin lives below as PARAKEET_VERSION?=89f5e2977b4d8bccd45e7bcc6f2ef7c4ed49e89a
 # (.github/bump_deps.sh) can find and update it - matches the
 # whisper.cpp / ds4 / vibevoice-cpp convention.
 #
@@ -15,7 +15,7 @@
 # That's what the L0 smoke test uses. The default target below does the
 # proper clone-at-pin + cmake build so CI doesn't need a side-checkout.
 
-PARAKEET_VERSION?=db755a78d39f789bb7d4e3935158a9e8105dbe36
+PARAKEET_VERSION?=89f5e2977b4d8bccd45e7bcc6f2ef7c4ed49e89a
 PARAKEET_REPO?=https://github.com/mudler/parakeet.cpp
 
 GOCMD?=go
@@ -74,6 +74,7 @@ libparakeet.so: sources/parakeet.cpp
 	cmake -B sources/parakeet.cpp/build-shared -S sources/parakeet.cpp $(CMAKE_ARGS)
 	cmake --build sources/parakeet.cpp/build-shared --config Release -j$(JOBS)
 	cp -fv sources/parakeet.cpp/build-shared/libparakeet.so* ./ 2>/dev/null || true
+	cp -fv sources/parakeet.cpp/build-shared/libparakeet.dylib ./ 2>/dev/null || true
 	cp -fv sources/parakeet.cpp/include/parakeet_capi.h ./
 
 parakeet-cpp-grpc: libparakeet.so main.go goparakeetcpp.go

backend/go/parakeet-cpp/main.go

@@ -2,15 +2,17 @@ package main
 
 // Started internally by LocalAI - one gRPC server per loaded model.
 //
-// Loads libparakeet.so via purego and registers the flat C-API entry
-// points declared in parakeet_capi.h. The library name can be overridden
-// with PARAKEET_LIBRARY (mirrors the WHISPER_LIBRARY / VIBEVOICECPP_LIBRARY
-// convention in the sibling backends); the default looks for the .so next
-// to this binary.
+// Loads the parakeet shared library via purego and registers the flat
+// C-API entry points declared in parakeet_capi.h. The library name can be
+// overridden with PARAKEET_LIBRARY (mirrors the WHISPER_LIBRARY /
+// VIBEVOICECPP_LIBRARY convention in the sibling backends); the default
+// looks next to this binary for libparakeet.so on Linux and
+// libparakeet.dylib on macOS.
 import (
 	"flag"
 	"fmt"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -28,7 +30,11 @@ type LibFuncs struct {
 func main() {
 	libName := os.Getenv("PARAKEET_LIBRARY")
 	if libName == "" {
-		libName = "libparakeet.so"
+		if runtime.GOOS == "darwin" {
+			libName = "libparakeet.dylib"
+		} else {
+			libName = "libparakeet.so"
+		}
 	}
 
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/parakeet-cpp/package.sh

@@ -16,12 +16,15 @@ mkdir -p "$CURDIR/package/lib"
 cp -avf "$CURDIR/parakeet-cpp-grpc" "$CURDIR/package/"
 cp -avf "$CURDIR/run.sh" "$CURDIR/package/"
 
-# libparakeet.so + any soname symlinks (libparakeet.so.X[.Y]). purego.Dlopen
-# resolves it via LD_LIBRARY_PATH, which run.sh points at lib/.
-cp -avf "$CURDIR"/libparakeet.so* "$CURDIR/package/lib/" 2>/dev/null || {
-	echo "ERROR: libparakeet.so not found in $CURDIR, run 'make' first" >&2
+# libparakeet shared lib + any soname symlinks. On Linux this is
+# libparakeet.so[.X.Y]; on macOS it is libparakeet.dylib. purego.Dlopen
+# resolves it via the *_LIBRARY_PATH that run.sh points at lib/.
+cp -avf "$CURDIR"/libparakeet.so* "$CURDIR/package/lib/" 2>/dev/null || true
+cp -avf "$CURDIR"/libparakeet.dylib "$CURDIR/package/lib/" 2>/dev/null || true
+if ! ls "$CURDIR"/package/lib/libparakeet.* >/dev/null 2>&1; then
+	echo "ERROR: libparakeet shared library not found in $CURDIR, run 'make' first" >&2
 	exit 1
-}
+fi
 
 # Detect architecture and copy the core runtime libs libparakeet.so links
 # against, plus the matching dynamic loader as lib/ld.so.
@@ -48,7 +51,7 @@ elif [ -f "/lib/ld-linux-aarch64.so.1" ]; then
     cp -arfLv /lib/aarch64-linux-gnu/librt.so.1 "$CURDIR/package/lib/librt.so.1"
     cp -arfLv /lib/aarch64-linux-gnu/libpthread.so.0 "$CURDIR/package/lib/libpthread.so.0"
 elif [ "$(uname -s)" = "Darwin" ]; then
-    echo "Detected Darwin"
+    echo "Detected Darwin — system frameworks linked dynamically, no bundled libs needed"
 else
     echo "Error: Could not detect architecture"
     exit 1

backend/go/parakeet-cpp/run.sh

@@ -3,11 +3,17 @@ set -e
 
 CURDIR=$(dirname "$(realpath "$0")")
 
-export LD_LIBRARY_PATH="$CURDIR/lib:$CURDIR:${LD_LIBRARY_PATH:-}"
+if [ "$(uname)" = "Darwin" ]; then
+	export DYLD_LIBRARY_PATH="$CURDIR/lib:$CURDIR:${DYLD_LIBRARY_PATH:-}"
+	export PARAKEET_LIBRARY="$CURDIR/lib/libparakeet.dylib"
+else
+	export LD_LIBRARY_PATH="$CURDIR/lib:$CURDIR:${LD_LIBRARY_PATH:-}"
+	export PARAKEET_LIBRARY="$CURDIR/lib/libparakeet.so"
+fi
 
 # If a self-contained ld.so was packaged, route through it so the
 # packaged libc / libstdc++ are used instead of the host's (matches the
-# whisper backend's runtime layout).
+# whisper backend's runtime layout). Linux only.
 if [ -f "$CURDIR/lib/ld.so" ]; then
 	echo "Using lib/ld.so"
 	exec "$CURDIR/lib/ld.so" "$CURDIR/parakeet-cpp-grpc" "$@"

backend/go/qwen3-tts-cpp/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # qwentts.cpp version
 QWEN3TTS_REPO?=https://github.com/ServeurpersoCom/qwentts.cpp
-QWEN3TTS_CPP_VERSION?=4536dcdce27c3764a93a06d6bf64026b124962f5
+QWEN3TTS_CPP_VERSION?=9dbe7ea26a01b30fccb117ae5e86807c1dc23d42
 SO_TARGET?=libgoqwen3ttscpp.so
 
 CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
@@ -65,8 +65,8 @@ UNAME_S := $(shell uname -s)
 ifeq ($(UNAME_S),Linux)
 	VARIANT_TARGETS = libgoqwen3ttscpp-avx.so libgoqwen3ttscpp-avx2.so libgoqwen3ttscpp-avx512.so libgoqwen3ttscpp-fallback.so
 else
-	# On non-Linux (e.g., Darwin), build only fallback variant
-	VARIANT_TARGETS = libgoqwen3ttscpp-fallback.so
+	# On non-Linux (e.g., Darwin), build only fallback variant (as a dylib)
+	VARIANT_TARGETS = libgoqwen3ttscpp-fallback.dylib
 endif
 
 qwen3-tts-cpp: main.go goqwen3ttscpp.go $(VARIANT_TARGETS)
@@ -78,7 +78,7 @@ package: qwen3-tts-cpp
 build: package
 
 clean: purge
-	rm -rf libgoqwen3ttscpp*.so package sources/qwentts.cpp qwen3-tts-cpp
+	rm -rf libgoqwen3ttscpp*.so libgoqwen3ttscpp*.dylib package sources/qwentts.cpp qwen3-tts-cpp
 
 purge:
 	rm -rf build*
@@ -110,13 +110,20 @@ libgoqwen3ttscpp-fallback.so: sources/qwentts.cpp
 	SO_TARGET=libgoqwen3ttscpp-fallback.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgoqwen3ttscpp-custom
 	rm -rf build-libgoqwen3ttscpp-fallback.so
 
+# Build fallback variant as a dylib (Darwin)
+libgoqwen3ttscpp-fallback.dylib: sources/qwentts.cpp
+	$(info ${GREEN}I qwen3-tts-cpp build info:fallback (dylib)${RESET})
+	SO_TARGET=libgoqwen3ttscpp-fallback.dylib CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgoqwen3ttscpp-custom
+	rm -rf build-libgoqwen3ttscpp-fallback.dylib
+
 libgoqwen3ttscpp-custom: CMakeLists.txt cpp/goqwen3ttscpp.cpp cpp/goqwen3ttscpp.h
 	mkdir -p build-$(SO_TARGET) && \
 	cd build-$(SO_TARGET) && \
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) --target goqwen3ttscpp && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libgoqwen3ttscpp.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/libgoqwen3ttscpp.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/libgoqwen3ttscpp.dylib ./$(SO_TARGET) 2>/dev/null)
 
 test: qwen3-tts-cpp
 	@echo "Running qwen3-tts-cpp tests..."

backend/go/qwen3-tts-cpp/main.go

@@ -4,6 +4,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -21,7 +22,11 @@ type LibFuncs struct {
 func main() {
 	libName := os.Getenv("QWEN3TTS_LIBRARY")
 	if libName == "" {
-		libName = "./libgoqwen3ttscpp-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libgoqwen3ttscpp-fallback.dylib"
+		} else {
+			libName = "./libgoqwen3ttscpp-fallback.so"
+		}
 	}
 
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/qwen3-tts-cpp/package.sh

@@ -12,7 +12,8 @@ REPO_ROOT="${CURDIR}/../../.."
 mkdir -p $CURDIR/package/lib
 
 cp -avf $CURDIR/qwen3-tts-cpp $CURDIR/package/
-cp -fv $CURDIR/libgoqwen3ttscpp-*.so $CURDIR/package/
+cp -fv $CURDIR/libgoqwen3ttscpp-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/libgoqwen3ttscpp-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -fv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries

backend/go/qwen3-tts-cpp/run.sh

@@ -12,9 +12,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libgoqwen3ttscpp-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/libgoqwen3ttscpp-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libgoqwen3ttscpp-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libgoqwen3ttscpp-avx.so ]; then
@@ -36,9 +40,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libgoqwen3ttscpp-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export QWEN3TTS_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/go/rfdetr-cpp/Makefile

@@ -34,6 +34,8 @@ else ifeq ($(BUILD_TYPE),hipblas)
 else ifeq ($(BUILD_TYPE),vulkan)
 	CMAKE_ARGS+=-DGGML_VULKAN=ON -DRFDETR_GGML_VULKAN=ON
 else ifeq ($(OS),Darwin)
+	# macOS/Metal: built + published as an OCI image by CI (includeDarwin in
+	# .github/backend-matrix.yml) so Apple Silicon users can install this backend.
 	ifneq ($(BUILD_TYPE),metal)
 		CMAKE_ARGS+=-DGGML_METAL=OFF
 	else
@@ -71,7 +73,7 @@ ifeq ($(UNAME_S),Linux)
 	VARIANT_TARGETS = librfdetrcpp-avx.so librfdetrcpp-avx2.so librfdetrcpp-avx512.so librfdetrcpp-fallback.so
 else
 	# On non-Linux (e.g., Darwin), build only fallback variant
-	VARIANT_TARGETS = librfdetrcpp-fallback.so
+	VARIANT_TARGETS = librfdetrcpp-fallback.dylib
 endif
 
 rfdetr-cpp: main.go gorfdetrcpp.go $(VARIANT_TARGETS)
@@ -83,7 +85,7 @@ package: rfdetr-cpp
 build: package
 
 clean: purge
-	rm -rf librfdetrcpp*.so rfdetr-cpp package sources
+	rm -rf librfdetrcpp*.so librfdetrcpp*.dylib rfdetr-cpp package sources
 
 purge:
 	rm -rf build*
@@ -110,11 +112,19 @@ librfdetrcpp-avx512.so: sources/rt-detr.cpp
 endif
 
 # Build fallback variant (all platforms)
+ifeq ($(UNAME_S),Darwin)
+librfdetrcpp-fallback.dylib: sources/rt-detr.cpp
+	rm -rfv build-$@
+	$(info ${GREEN}I rfdetr-cpp build info:fallback${RESET})
+	SO_TARGET=$@ CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) librfdetrcpp-custom
+	rm -rfv build-$@
+else
 librfdetrcpp-fallback.so: sources/rt-detr.cpp
 	rm -rfv build-$@
 	$(info ${GREEN}I rfdetr-cpp build info:fallback${RESET})
 	SO_TARGET=$@ CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) librfdetrcpp-custom
 	rm -rfv build-$@
+endif
 
 librfdetrcpp-custom: CMakeLists.txt
 	mkdir -p build-$(SO_TARGET) && \
@@ -122,7 +132,8 @@ librfdetrcpp-custom: CMakeLists.txt
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) && \
 	cd .. && \
-	mv build-$(SO_TARGET)/librfdetrcpp.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/librfdetrcpp.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/librfdetrcpp.dylib ./$(SO_TARGET) 2>/dev/null)
 
 all: rfdetr-cpp package
 

backend/go/rfdetr-cpp/main.go

@@ -9,6 +9,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -27,7 +28,11 @@ func main() {
 	// Get library name from environment variable, default to fallback
 	libName := os.Getenv("RFDETR_LIBRARY")
 	if libName == "" {
-		libName = "./librfdetrcpp-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./librfdetrcpp-fallback.dylib"
+		} else {
+			libName = "./librfdetrcpp-fallback.so"
+		}
 	}
 
 	rfdetrLib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/rfdetr-cpp/package.sh

@@ -10,7 +10,8 @@ REPO_ROOT="${CURDIR}/../../.."
 # Create lib directory
 mkdir -p $CURDIR/package/lib
 
-cp -avf $CURDIR/librfdetrcpp-*.so $CURDIR/package/
+cp -fv $CURDIR/librfdetrcpp-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/librfdetrcpp-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -avf $CURDIR/rfdetr-cpp $CURDIR/package/
 cp -fv $CURDIR/run.sh $CURDIR/package/
 

backend/go/rfdetr-cpp/run.sh

@@ -12,9 +12,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/librfdetrcpp-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/librfdetrcpp-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/librfdetrcpp-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/librfdetrcpp-avx.so ]; then
@@ -36,9 +40,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/librfdetrcpp-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export RFDETR_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/go/sam3-cpp/Makefile

@@ -31,6 +31,8 @@ else ifeq ($(BUILD_TYPE),hipblas)
 else ifeq ($(BUILD_TYPE),vulkan)
 	CMAKE_ARGS+=-DGGML_VULKAN=ON
 else ifeq ($(OS),Darwin)
+	# macOS/Metal: built + published as an OCI image by CI (includeDarwin in
+	# .github/backend-matrix.yml) so Apple Silicon users can install this backend.
 	ifneq ($(BUILD_TYPE),metal)
 		CMAKE_ARGS+=-DGGML_METAL=OFF
 	else
@@ -66,7 +68,7 @@ ifeq ($(UNAME_S),Linux)
 	VARIANT_TARGETS = libgosam3-avx.so libgosam3-avx2.so libgosam3-avx512.so libgosam3-fallback.so
 else
 	# On non-Linux (e.g., Darwin), build only fallback variant
-	VARIANT_TARGETS = libgosam3-fallback.so
+	VARIANT_TARGETS = libgosam3-fallback.dylib
 endif
 
 sam3-cpp: main.go gosam3.go $(VARIANT_TARGETS)
@@ -78,7 +80,7 @@ package: sam3-cpp
 build: package
 
 clean: purge
-	rm -rf libgosam3*.so sam3-cpp package sources
+	rm -rf libgosam3*.so libgosam3*.dylib sam3-cpp package sources
 
 purge:
 	rm -rf build*
@@ -105,11 +107,19 @@ libgosam3-avx512.so: sources/sam3.cpp
 endif
 
 # Build fallback variant (all platforms)
+ifeq ($(UNAME_S),Darwin)
+libgosam3-fallback.dylib: sources/sam3.cpp
+	$(MAKE) purge
+	$(info ${GREEN}I sam3-cpp build info:fallback${RESET})
+	SO_TARGET=libgosam3-fallback.dylib CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgosam3-custom
+	rm -rfv build*
+else
 libgosam3-fallback.so: sources/sam3.cpp
 	$(MAKE) purge
 	$(info ${GREEN}I sam3-cpp build info:fallback${RESET})
 	SO_TARGET=libgosam3-fallback.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgosam3-custom
 	rm -rfv build*
+endif
 
 libgosam3-custom: CMakeLists.txt cpp/gosam3.cpp cpp/gosam3.h
 	mkdir -p build-$(SO_TARGET) && \
@@ -117,6 +127,7 @@ libgosam3-custom: CMakeLists.txt cpp/gosam3.cpp cpp/gosam3.h
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libgosam3.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/libgosam3.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/libgosam3.dylib ./$(SO_TARGET) 2>/dev/null)
 
 all: sam3-cpp package

backend/go/sam3-cpp/main.go

@@ -3,6 +3,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -21,7 +22,11 @@ func main() {
 	// Get library name from environment variable, default to fallback
 	libName := os.Getenv("SAM3_LIBRARY")
 	if libName == "" {
-		libName = "./libgosam3-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libgosam3-fallback.dylib"
+		} else {
+			libName = "./libgosam3-fallback.so"
+		}
 	}
 
 	gosamLib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/sam3-cpp/package.sh

@@ -10,7 +10,8 @@ REPO_ROOT="${CURDIR}/../../.."
 # Create lib directory
 mkdir -p $CURDIR/package/lib
 
-cp -avf $CURDIR/libgosam3-*.so $CURDIR/package/
+cp -fv $CURDIR/libgosam3-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/libgosam3-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -avf $CURDIR/sam3-cpp $CURDIR/package/
 cp -fv $CURDIR/run.sh $CURDIR/package/
 

backend/go/sam3-cpp/run.sh

@@ -12,9 +12,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libgosam3-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/libgosam3-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libgosam3-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libgosam3-avx.so ]; then
@@ -36,9 +40,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libgosam3-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export SAM3_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/go/sherpa-onnx/backend.go

@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"runtime"
 	"strconv"
 	"strings"
 	"sync"
@@ -238,11 +239,19 @@ func loadSherpaLibs() error {
 func loadSherpaLibsOnce() error {
 	shimLib := os.Getenv("SHERPA_SHIM_LIBRARY")
 	if shimLib == "" {
-		shimLib = "libsherpa-shim.so"
+		if runtime.GOOS == "darwin" {
+			shimLib = "libsherpa-shim.dylib"
+		} else {
+			shimLib = "libsherpa-shim.so"
+		}
 	}
 	capiLib := os.Getenv("SHERPA_ONNX_LIBRARY")
 	if capiLib == "" {
-		capiLib = "libsherpa-onnx-c-api.so"
+		if runtime.GOOS == "darwin" {
+			capiLib = "libsherpa-onnx-c-api.dylib"
+		} else {
+			capiLib = "libsherpa-onnx-c-api.so"
+		}
 	}
 
 	shim, err := purego.Dlopen(shimLib, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/sherpa-onnx/run.sh

@@ -3,7 +3,13 @@ set -ex
 
 CURDIR=$(dirname "$(realpath $0)")
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
+if [ "$(uname)" = "Darwin" ]; then
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+	export SHERPA_SHIM_LIBRARY=$CURDIR/lib/libsherpa-shim.dylib
+	export SHERPA_ONNX_LIBRARY=$CURDIR/lib/libsherpa-onnx-c-api.dylib
+else
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
+fi
 
 if [ -f $CURDIR/lib/ld.so ]; then
 	echo "Using lib/ld.so"

backend/go/stablediffusion-ggml/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # stablediffusion.cpp (ggml)
 STABLEDIFFUSION_GGML_REPO?=https://github.com/leejet/stable-diffusion.cpp
-STABLEDIFFUSION_GGML_VERSION?=f440ad9c29dd8bc34e5d1f4b863832b96d6ea05f
+STABLEDIFFUSION_GGML_VERSION?=8caa3f908ae6d4a4bef531e73b9a969f266a3d1f
 
 CMAKE_ARGS+=-DGGML_MAX_NAME=128
 
@@ -131,6 +131,7 @@ libgosd-custom: CMakeLists.txt cpp/gosd.cpp cpp/gosd.h
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libgosd.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/libgosd.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/libgosd.dylib ./$(SO_TARGET) 2>/dev/null)
 
 all: stablediffusion-ggml package

backend/go/stablediffusion-ggml/main.go

@@ -3,6 +3,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -21,7 +22,11 @@ func main() {
 	// Get library name from environment variable, default to fallback
 	libName := os.Getenv("SD_LIBRARY")
 	if libName == "" {
-		libName = "./libgosd-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libgosd-fallback.dylib"
+		} else {
+			libName = "./libgosd-fallback.so"
+		}
 	}
 
 	gosd, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/stablediffusion-ggml/package.sh

@@ -12,6 +12,7 @@ REPO_ROOT="${CURDIR}/../../.."
 mkdir -p $CURDIR/package/lib
 
 cp -avf $CURDIR/libgosd-*.so $CURDIR/package/
+cp -fv $CURDIR/libgosd-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -avf $CURDIR/stablediffusion-ggml $CURDIR/package/
 cp -fv $CURDIR/run.sh $CURDIR/package/
 

backend/go/stablediffusion-ggml/run.sh

@@ -12,9 +12,18 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libgosd-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single library variant (Metal or Accelerate). The gosd target is
+	# built as a CMake MODULE, which emits a .dylib for a SHARED build but a
+	# .so for a MODULE build on Apple, so prefer .dylib and fall back to .so.
+	LIBRARY="$CURDIR/libgosd-fallback.dylib"
+	if [ ! -e "$LIBRARY" ]; then
+		LIBRARY="$CURDIR/libgosd-fallback.so"
+	fi
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libgosd-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libgosd-avx.so ]; then
@@ -36,9 +45,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libgosd-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export SD_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/go/supertonic/helper.go

@@ -16,6 +16,7 @@ import (
 	"os"
 	"path/filepath"
 	"regexp"
+	"runtime"
 	"strings"
 	"time"
 	"unicode"
@@ -943,7 +944,13 @@ func InitializeONNXRuntime() error {
 			}
 		}
 		if libPath == "" {
-			libPath = "/usr/local/lib/libonnxruntime.so"
+			// LocalAI: default to the platform-native shared library
+			// extension when nothing else is found (dyld vs ld.so).
+			if runtime.GOOS == "darwin" {
+				libPath = "/usr/local/lib/libonnxruntime.dylib"
+			} else {
+				libPath = "/usr/local/lib/libonnxruntime.so"
+			}
 		}
 	}
 	ort.SetSharedLibraryPath(libPath)

backend/go/supertonic/package.sh

@@ -32,6 +32,10 @@ elif [ -f "/lib/ld-linux-aarch64.so.1" ]; then
     cp -arfLv /lib/aarch64-linux-gnu/libdl.so.2 $CURDIR/package/lib/libdl.so.2
     cp -arfLv /lib/aarch64-linux-gnu/librt.so.1 $CURDIR/package/lib/librt.so.1
     cp -arfLv /lib/aarch64-linux-gnu/libpthread.so.0 $CURDIR/package/lib/libpthread.so.0
+elif [ $(uname -s) = "Darwin" ]; then
+    # macOS: dyld resolves the bundled .dylib via DYLD_LIBRARY_PATH (set in
+    # run.sh); there is no ld.so loader nor glibc to bundle.
+    echo "Detected Darwin"
 else
     echo "Error: Could not detect architecture"
     exit 1

backend/go/supertonic/run.sh

@@ -3,12 +3,19 @@ set -ex
 
 CURDIR=$(dirname "$(realpath $0)")
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
-export ONNXRUNTIME_LIB_PATH=$CURDIR/lib/libonnxruntime.so
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS uses dyld: there is no ld.so loader, and the search path env
+	# var is DYLD_LIBRARY_PATH. ONNX Runtime ships as a .dylib here.
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+	export ONNXRUNTIME_LIB_PATH=$CURDIR/lib/libonnxruntime.dylib
+else
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
+	export ONNXRUNTIME_LIB_PATH=$CURDIR/lib/libonnxruntime.so
 
-if [ -f $CURDIR/lib/ld.so ]; then
-	echo "Using lib/ld.so"
-	exec $CURDIR/lib/ld.so $CURDIR/supertonic "$@"
+	if [ -f $CURDIR/lib/ld.so ]; then
+		echo "Using lib/ld.so"
+		exec $CURDIR/lib/ld.so $CURDIR/supertonic "$@"
+	fi
 fi
 
 exec $CURDIR/supertonic "$@"

backend/go/vibevoice-cpp/Makefile

@@ -70,8 +70,8 @@ UNAME_S := $(shell uname -s)
 ifeq ($(UNAME_S),Linux)
 	VARIANT_TARGETS = libgovibevoicecpp-avx.so libgovibevoicecpp-avx2.so libgovibevoicecpp-avx512.so libgovibevoicecpp-fallback.so
 else
-	# On non-Linux (e.g., Darwin), build only fallback variant
-	VARIANT_TARGETS = libgovibevoicecpp-fallback.so
+	# On non-Linux (e.g., Darwin), build only fallback variant (as a dylib)
+	VARIANT_TARGETS = libgovibevoicecpp-fallback.dylib
 endif
 
 vibevoice-cpp: main.go govibevoicecpp.go $(VARIANT_TARGETS)
@@ -83,7 +83,7 @@ package: vibevoice-cpp
 build: package
 
 clean: purge
-	rm -rf libgovibevoicecpp*.so package sources/vibevoice.cpp vibevoice-cpp
+	rm -rf libgovibevoicecpp*.so libgovibevoicecpp*.dylib package sources/vibevoice.cpp vibevoice-cpp
 
 purge:
 	rm -rf build*
@@ -119,13 +119,21 @@ libgovibevoicecpp-fallback.so: sources/vibevoice.cpp
 	SO_TARGET=libgovibevoicecpp-fallback.so CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgovibevoicecpp-custom
 	rm -rfv build*
 
+# Build fallback variant as a dylib (Darwin)
+libgovibevoicecpp-fallback.dylib: sources/vibevoice.cpp
+	$(MAKE) purge
+	$(info ${GREEN}I vibevoice-cpp build info:fallback (dylib)${RESET})
+	SO_TARGET=libgovibevoicecpp-fallback.dylib CMAKE_ARGS="$(CMAKE_ARGS) -DGGML_AVX=off -DGGML_AVX2=off -DGGML_AVX512=off -DGGML_FMA=off -DGGML_F16C=off -DGGML_BMI2=off" $(MAKE) libgovibevoicecpp-custom
+	rm -rfv build*
+
 libgovibevoicecpp-custom: CMakeLists.txt cpp/govibevoicecpp.cpp cpp/govibevoicecpp.h
 	mkdir -p build-$(SO_TARGET) && \
 	cd build-$(SO_TARGET) && \
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) --target govibevoicecpp && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libgovibevoicecpp.so ./$(SO_TARGET)
+	(mv build-$(SO_TARGET)/libgovibevoicecpp.so ./$(SO_TARGET) 2>/dev/null || \
+	 mv build-$(SO_TARGET)/libgovibevoicecpp.dylib ./$(SO_TARGET) 2>/dev/null)
 
 test: vibevoice-cpp
 	@echo "Running vibevoice-cpp tests..."

backend/go/vibevoice-cpp/main.go

@@ -4,6 +4,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -21,7 +22,11 @@ type LibFuncs struct {
 func main() {
 	libName := os.Getenv("VIBEVOICECPP_LIBRARY")
 	if libName == "" {
-		libName = "./libgovibevoicecpp-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libgovibevoicecpp-fallback.dylib"
+		} else {
+			libName = "./libgovibevoicecpp-fallback.so"
+		}
 	}
 
 	lib, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/vibevoice-cpp/package.sh

@@ -12,7 +12,8 @@ REPO_ROOT="${CURDIR}/../../.."
 mkdir -p $CURDIR/package/lib
 
 cp -avf $CURDIR/vibevoice-cpp $CURDIR/package/
-cp -fv $CURDIR/libgovibevoicecpp-*.so $CURDIR/package/
+cp -fv $CURDIR/libgovibevoicecpp-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/libgovibevoicecpp-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -fv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries

backend/go/vibevoice-cpp/run.sh

@@ -11,9 +11,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libgovibevoicecpp-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/libgovibevoicecpp-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libgovibevoicecpp-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libgovibevoicecpp-avx.so ]; then
@@ -34,9 +38,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libgovibevoicecpp-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export VIBEVOICECPP_LIBRARY=$LIBRARY
 
 if [ -f $CURDIR/lib/ld.so ]; then

backend/go/whisper/Makefile

@@ -8,7 +8,7 @@ JOBS?=$(shell nproc --ignore=1)
 
 # whisper.cpp version
 WHISPER_REPO?=https://github.com/ggml-org/whisper.cpp
-WHISPER_CPP_VERSION?=bae6bc02b1940bbfb87b6a0299c565e563b916d1
+WHISPER_CPP_VERSION?=43d78af5be58f41d6ffbc227d608f104577741ea
 SO_TARGET?=libgowhisper.so
 
 CMAKE_ARGS+=-DBUILD_SHARED_LIBS=OFF
@@ -117,6 +117,7 @@ libgowhisper-custom: CMakeLists.txt cpp/gowhisper.cpp cpp/gowhisper.h
 	cmake .. $(CMAKE_ARGS) && \
 	cmake --build . --config Release -j$(JOBS) && \
 	cd .. && \
-	mv build-$(SO_TARGET)/libgowhisper.so ./$(SO_TARGET)
+	mv build-$(SO_TARGET)/libgowhisper.so ./$(SO_TARGET) 2>/dev/null || \
+		mv build-$(SO_TARGET)/libgowhisper.dylib ./$(SO_TARGET:.so=.dylib)
 
 all: whisper package

backend/go/whisper/main.go

@@ -4,6 +4,7 @@ package main
 import (
 	"flag"
 	"os"
+	"runtime"
 
 	"github.com/ebitengine/purego"
 	grpc "github.com/mudler/LocalAI/pkg/grpc"
@@ -22,7 +23,11 @@ func main() {
 	// Get library name from environment variable, default to fallback
 	libName := os.Getenv("WHISPER_LIBRARY")
 	if libName == "" {
-		libName = "./libgowhisper-fallback.so"
+		if runtime.GOOS == "darwin" {
+			libName = "./libgowhisper-fallback.dylib"
+		} else {
+			libName = "./libgowhisper-fallback.so"
+		}
 	}
 
 	gosd, err := purego.Dlopen(libName, purego.RTLD_NOW|purego.RTLD_GLOBAL)

backend/go/whisper/package.sh

@@ -12,7 +12,8 @@ REPO_ROOT="${CURDIR}/../../.."
 mkdir -p $CURDIR/package/lib
 
 cp -avf $CURDIR/whisper $CURDIR/package/
-cp -fv $CURDIR/libgowhisper-*.so $CURDIR/package/
+cp -fv $CURDIR/libgowhisper-*.so $CURDIR/package/ 2>/dev/null || true
+cp -fv $CURDIR/libgowhisper-*.dylib $CURDIR/package/ 2>/dev/null || true
 cp -fv $CURDIR/run.sh $CURDIR/package/
 
 # Detect architecture and copy appropriate libraries

backend/go/whisper/run.sh

@@ -12,9 +12,13 @@ if [ "$(uname)" != "Darwin" ]; then
 	grep -e "flags" /proc/cpuinfo | head -1
 fi
 
-LIBRARY="$CURDIR/libgowhisper-fallback.so"
+if [ "$(uname)" = "Darwin" ]; then
+	# macOS: single dylib variant (Metal or Accelerate)
+	LIBRARY="$CURDIR/libgowhisper-fallback.dylib"
+	export DYLD_LIBRARY_PATH=$CURDIR/lib:$DYLD_LIBRARY_PATH
+else
+	LIBRARY="$CURDIR/libgowhisper-fallback.so"
 
-if [ "$(uname)" != "Darwin" ]; then
 	if grep -q -e "\savx\s" /proc/cpuinfo ; then
 		echo "CPU:    AVX    found OK"
 		if [ -e $CURDIR/libgowhisper-avx.so ]; then
@@ -36,9 +40,10 @@ if [ "$(uname)" != "Darwin" ]; then
 			LIBRARY="$CURDIR/libgowhisper-avx512.so"
 		fi
 	fi
+
+	export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 fi
 
-export LD_LIBRARY_PATH=$CURDIR/lib:$LD_LIBRARY_PATH
 export WHISPER_LIBRARY=$LIBRARY
 
 # If there is a lib/ld.so, use it

backend/index.yaml

@@ -340,6 +340,7 @@
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-sam3-cpp"
     intel: "intel-sycl-f32-sam3-cpp"
     vulkan: "vulkan-sam3-cpp"
+    metal: "metal-sam3-cpp"
 - &rfdetrcpp
   name: "rfdetr-cpp"
   alias: "rfdetr-cpp"
@@ -368,6 +369,7 @@
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-rfdetr-cpp"
     intel: "intel-sycl-f32-rfdetr-cpp"
     vulkan: "vulkan-rfdetr-cpp"
+    metal: "metal-rfdetr-cpp"
 - &locateanything
   name: "locate-anything"
   alias: "locate-anything"
@@ -397,6 +399,7 @@
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-locate-anything-cpp"
     intel: "intel-sycl-f32-locate-anything-cpp"
     vulkan: "vulkan-locate-anything-cpp"
+    metal: "metal-locate-anything-cpp"
 - !!merge <<: *locateanything
   name: "locate-anything-development"
   capabilities:
@@ -409,6 +412,7 @@
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-locate-anything-cpp-development"
     intel: "intel-sycl-f32-locate-anything-cpp-development"
     vulkan: "vulkan-locate-anything-cpp-development"
+    metal: "metal-locate-anything-cpp-development"
 - !!merge <<: *locateanything
   name: "cpu-locate-anything-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-locate-anything-cpp"
@@ -419,6 +423,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-locate-anything-cpp"
   mirrors:
     - localai/localai-backends:master-cpu-locate-anything-cpp
+- !!merge <<: *locateanything
+  name: "metal-locate-anything-cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-locate-anything-cpp"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-locate-anything-cpp
+- !!merge <<: *locateanything
+  name: "metal-locate-anything-cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-locate-anything-cpp"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-locate-anything-cpp
 - !!merge <<: *locateanything
   name: "cuda12-locate-anything-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-locate-anything-cpp"
@@ -517,6 +531,7 @@
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-depth-anything-cpp"
     intel: "intel-sycl-f32-depth-anything-cpp"
     vulkan: "vulkan-depth-anything-cpp"
+    metal: "metal-depth-anything-cpp"
 - !!merge <<: *depthanything
   name: "depth-anything-development"
   capabilities:
@@ -529,6 +544,7 @@
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-depth-anything-cpp-development"
     intel: "intel-sycl-f32-depth-anything-cpp-development"
     vulkan: "vulkan-depth-anything-cpp-development"
+    metal: "metal-depth-anything-cpp-development"
 - !!merge <<: *depthanything
   name: "cpu-depth-anything-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-depth-anything-cpp"
@@ -539,6 +555,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-depth-anything-cpp"
   mirrors:
     - localai/localai-backends:master-cpu-depth-anything-cpp
+- !!merge <<: *depthanything
+  name: "metal-depth-anything-cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-depth-anything-cpp"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-depth-anything-cpp
+- !!merge <<: *depthanything
+  name: "metal-depth-anything-cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-depth-anything-cpp"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-depth-anything-cpp
 - !!merge <<: *depthanything
   name: "cuda12-depth-anything-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-depth-anything-cpp"
@@ -645,6 +671,7 @@
     nvidia-cuda-13: "cuda13-vllm"
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-vllm"
     cpu: "cpu-vllm"
+    metal: "metal-vllm"
 - &sglang
   name: "sglang"
   license: apache-2.0
@@ -1030,6 +1057,8 @@
     nvidia-l4t: "vulkan-localvqe"
     nvidia-l4t-cuda-12: "vulkan-localvqe"
     nvidia-l4t-cuda-13: "vulkan-localvqe"
+    # Apple Silicon: CPU build (LocalVQE has no Metal path); still arm64-native.
+    metal: "metal-localvqe"
 - &privacyfilter
   name: "privacy-filter"
   alias: "privacy-filter"
@@ -1066,6 +1095,7 @@
     amd: "vulkan-privacy-filter"
     intel: "vulkan-privacy-filter"
     vulkan: "vulkan-privacy-filter"
+    metal: "metal-privacy-filter"
 - &faster-whisper
   icon: https://avatars.githubusercontent.com/u/1520500?s=200&v=4
   description: |
@@ -1284,6 +1314,7 @@
     nvidia-cuda-13: "cuda13-liquid-audio"
     nvidia-cuda-12: "cuda12-liquid-audio"
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-liquid-audio"
+    metal: "metal-liquid-audio"
   icon: https://cdn-avatars.huggingface.co/v1/production/uploads/61b8e2ba285851687028d395/7_6D7rWrLxp2hb6OHSV1p.png
 - &qwen-tts
   urls:
@@ -1569,6 +1600,7 @@
     - TTS
   capabilities:
     default: "cpu-supertonic"
+    metal: "metal-supertonic"
 - !!merge <<: *neutts
   name: "neutts-development"
   capabilities:
@@ -2906,6 +2938,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-vulkan-privacy-filter"
   mirrors:
     - localai/localai-backends:master-gpu-vulkan-privacy-filter
+- !!merge <<: *privacyfilter
+  name: "metal-privacy-filter"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-privacy-filter"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-privacy-filter
+- !!merge <<: *privacyfilter
+  name: "metal-privacy-filter-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-privacy-filter"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-privacy-filter
 - !!merge <<: *privacyfilter
   name: "cuda13-privacy-filter"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-13-privacy-filter"
@@ -2927,6 +2969,17 @@
     nvidia-cuda-13: "cuda13-vllm-development"
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-vllm-development"
     cpu: "cpu-vllm-development"
+    metal: "metal-vllm-development"
+- !!merge <<: *vllm
+  name: "metal-vllm"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-vllm"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-vllm
+- !!merge <<: *vllm
+  name: "metal-vllm-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-vllm"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-vllm
 - !!merge <<: *vllm
   name: "cuda12-vllm"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-vllm"
@@ -3206,6 +3259,7 @@
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-sam3-cpp-development"
     intel: "intel-sycl-f32-sam3-cpp-development"
     vulkan: "vulkan-sam3-cpp-development"
+    metal: "metal-sam3-cpp-development"
 - !!merge <<: *sam3cpp
   name: "cpu-sam3-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-sam3-cpp"
@@ -3216,6 +3270,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-sam3-cpp"
   mirrors:
     - localai/localai-backends:master-cpu-sam3-cpp
+- !!merge <<: *sam3cpp
+  name: "metal-sam3-cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-sam3-cpp"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-sam3-cpp
+- !!merge <<: *sam3cpp
+  name: "metal-sam3-cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-sam3-cpp"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-sam3-cpp
 - !!merge <<: *sam3cpp
   name: "cuda12-sam3-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-sam3-cpp"
@@ -3289,6 +3353,7 @@
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-rfdetr-cpp-development"
     intel: "intel-sycl-f32-rfdetr-cpp-development"
     vulkan: "vulkan-rfdetr-cpp-development"
+    metal: "metal-rfdetr-cpp-development"
 - !!merge <<: *rfdetrcpp
   name: "cpu-rfdetr-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-rfdetr-cpp"
@@ -3299,6 +3364,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-rfdetr-cpp"
   mirrors:
     - localai/localai-backends:master-cpu-rfdetr-cpp
+- !!merge <<: *rfdetrcpp
+  name: "metal-rfdetr-cpp"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-rfdetr-cpp"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-rfdetr-cpp
+- !!merge <<: *rfdetrcpp
+  name: "metal-rfdetr-cpp-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-rfdetr-cpp"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-rfdetr-cpp
 - !!merge <<: *rfdetrcpp
   name: "cuda12-rfdetr-cpp"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-rfdetr-cpp"
@@ -4087,6 +4162,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-vulkan-localvqe"
   mirrors:
     - localai/localai-backends:master-gpu-vulkan-localvqe
+- !!merge <<: *localvqecpp
+  name: "metal-localvqe"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-localvqe"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-localvqe
+- !!merge <<: *localvqecpp
+  name: "metal-localvqe-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-localvqe"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-localvqe
 ## kokoro
 - !!merge <<: *kokoro
   name: "kokoro-development"
@@ -4612,6 +4697,7 @@
     nvidia-cuda-13: "cuda13-liquid-audio-development"
     nvidia-cuda-12: "cuda12-liquid-audio-development"
     nvidia-l4t-cuda-13: "cuda13-nvidia-l4t-arm64-liquid-audio-development"
+    metal: "metal-liquid-audio-development"
 - !!merge <<: *liquid-audio
   name: "cpu-liquid-audio"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-liquid-audio"
@@ -4622,6 +4708,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-liquid-audio"
   mirrors:
     - localai/localai-backends:master-cpu-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "metal-liquid-audio"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-liquid-audio"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-liquid-audio
+- !!merge <<: *liquid-audio
+  name: "metal-liquid-audio-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-liquid-audio"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-liquid-audio
 - !!merge <<: *liquid-audio
   name: "cuda12-liquid-audio"
   uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-liquid-audio"
@@ -5282,6 +5378,7 @@
     nvidia: "cuda12-trl"
     nvidia-cuda-12: "cuda12-trl"
     nvidia-cuda-13: "cuda13-trl"
+    metal: "metal-trl"
 ## TRL backend images
 - !!merge <<: *trl
   name: "cpu-trl"
@@ -5313,6 +5410,16 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-13-trl"
   mirrors:
     - localai/localai-backends:master-gpu-nvidia-cuda-13-trl
+- !!merge <<: *trl
+  name: "metal-trl"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-trl"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-trl
+- !!merge <<: *trl
+  name: "metal-trl-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-trl"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-trl
 ## llama.cpp quantization backend
 - &llama-cpp-quantization
   name: "llama-cpp-quantization"
@@ -5484,6 +5591,7 @@
   name: "supertonic-development"
   capabilities:
     default: "cpu-supertonic-development"
+    metal: "metal-supertonic-development"
 - !!merge <<: *supertonic
   name: "cpu-supertonic"
   uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-supertonic"
@@ -5494,3 +5602,13 @@
   uri: "quay.io/go-skynet/local-ai-backends:master-cpu-supertonic"
   mirrors:
     - localai/localai-backends:master-cpu-supertonic
+- !!merge <<: *supertonic
+  name: "metal-supertonic"
+  uri: "quay.io/go-skynet/local-ai-backends:latest-metal-darwin-arm64-supertonic"
+  mirrors:
+    - localai/localai-backends:latest-metal-darwin-arm64-supertonic
+- !!merge <<: *supertonic
+  name: "metal-supertonic-development"
+  uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-supertonic"
+  mirrors:
+    - localai/localai-backends:master-metal-darwin-arm64-supertonic

backend/python/liquid-audio/install.sh

@@ -14,5 +14,11 @@ else
 fi
 
 # liquid-audio's torch wheels are large; allow upgrades to satisfy transitive pins
-EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
+EXTRA_PIP_INSTALL_FLAGS+=" --upgrade"
+# --index-strategy is a uv-only flag. The darwin/MPS build installs with pip
+# (USE_PIP=true in scripts/build/python-darwin.sh), which rejects it. Only add
+# it on the uv path; Linux/CUDA resolution is unchanged.
+if [ "x${USE_PIP:-}" != "xtrue" ]; then
+    EXTRA_PIP_INSTALL_FLAGS+=" --index-strategy=unsafe-first-match"
+fi
 installRequirements

backend/python/liquid-audio/requirements-mps.txt

@@ -1,3 +1,4 @@
+# MPS (Apple Silicon / Metal) build profile - installed by the darwin CI job.
 torch>=2.8.0
 torchaudio>=2.8.0
 torchcodec>=0.9.1

backend/python/trl/install.sh

@@ -8,7 +8,13 @@ else
     source $backend_dir/../common/libbackend.sh
 fi
 
-EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
+EXTRA_PIP_INSTALL_FLAGS+=" --upgrade"
+# --index-strategy is a uv-only flag. The darwin/MPS build installs with pip
+# (USE_PIP=true in scripts/build/python-darwin.sh), which rejects it. Only add
+# it when uv is the installer, keeping the Linux/CUDA resolution unchanged.
+if [ "x${USE_PIP:-}" != "xtrue" ]; then
+    EXTRA_PIP_INSTALL_FLAGS+=" --index-strategy=unsafe-first-match"
+fi
 installRequirements
 
 # Fetch convert_hf_to_gguf.py and gguf package from the same llama.cpp version

backend/python/trl/requirements-mps.txt

@@ -0,0 +1,12 @@
+torch==2.10.0
+trl
+peft
+datasets>=3.0.0
+transformers>=4.56.2
+accelerate>=1.4.0
+huggingface-hub>=1.3.0
+sentencepiece
+# Note: bitsandbytes is intentionally omitted on MPS. It is only used by the
+# CUDA (cublas) variants for 8-bit/4-bit quantization and has poor support on
+# Apple Silicon. torch here uses the plain PyPI wheels, which ship MPS support
+# on macOS arm64.

backend/python/vllm/backend.py

@@ -457,9 +457,14 @@ class BackendServicer(backend_pb2_grpc.BackendServicer):
                     except Exception:
                         pass
 
-                if last_output is None or not getattr(last_output, "prompt_logprobs", None):
-                    context.set_code(grpc.StatusCode.INTERNAL)
-                    context.set_details("vLLM did not return prompt_logprobs")
+                _pl = getattr(last_output, "prompt_logprobs", None) if last_output is not None else None
+                # Some engines accept the prompt_logprobs request but return a
+                # list of all-None entries instead of computing them (observed
+                # with vllm-metal's MLX backend on macOS). Treat that as
+                # unsupported rather than silently scoring every candidate as 0.
+                if not _pl or all(e is None for e in _pl):
+                    context.set_code(grpc.StatusCode.UNIMPLEMENTED)
+                    context.set_details("This backend did not return prompt_logprobs; scoring is unsupported on this engine (e.g. vllm-metal / MLX on macOS).")
                     return backend_pb2.ScoreResponse()
 
                 prompt_logprobs = last_output.prompt_logprobs

backend/python/vllm/install.sh

@@ -43,6 +43,24 @@ if [ "x${BUILD_PROFILE}" == "xcublas13" ]; then
     EXTRA_PIP_INSTALL_FLAGS+=" --index-strategy=unsafe-best-match"
 fi
 
+# Apple Silicon (Metal/MLX) via vllm-metal.
+# vllm-metal (github.com/vllm-project/vllm-metal) brings vLLM to macOS on Apple
+# Silicon: it registers through vLLM's platform-plugin entry point
+# (metal -> vllm_metal:register), MetalPlatform activates, and the vLLM v1
+# AsyncLLM engine runs on the GPU through MLX. LocalAI's backend.py is UNCHANGED
+# on darwin — AsyncEngineArgs(...) -> AsyncLLMEngine.from_engine_args transparently
+# resolves to the MLX engine (proven on a real M4 / macOS 26.5 against Qwen3-0.6B).
+#
+# vllm-metal REQUIRES Python 3.12, so force the portable CPython before the venv
+# is created (ensureVenv reads PYTHON_VERSION/PYTHON_PATCH/PY_STANDALONE_TAG).
+# The patch + standalone tag mirror the l4t13 cp312 pin — a known-good
+# python-build-standalone release that also ships an aarch64-apple-darwin asset.
+if [ "$(uname -s)" = "Darwin" ]; then
+    PYTHON_VERSION="3.12"
+    PYTHON_PATCH="12"
+    PY_STANDALONE_TAG="20251120"
+fi
+
 # JetPack 7 / L4T arm64 vllm + torch wheels come straight from PyPI now
 # (torch 2.11+ ships aarch64 + cu130 manylinux wheels and vllm 0.20+ ships
 # an aarch64 wheel pinned to that torch). They're cp312-only, so bump the
@@ -57,11 +75,87 @@ if [ "x${BUILD_PROFILE}" == "xl4t13" ]; then
     PY_STANDALONE_TAG="20251120"
 fi
 
+# ===================== Apple Silicon (Metal/MLX) =====================
+# Reproduce vllm-metal's upstream installer
+# (curl -fsSL https://raw.githubusercontent.com/vllm-project/vllm-metal/main/install.sh)
+# but INTO LocalAI's managed venv (ensureVenv) instead of a throwaway
+# ~/.venv-vllm-metal, so the backend integrates with LocalAI's venv lifecycle
+# (portable CPython, _makeVenvPortable relocation, runtime activation). The
+# normal CUDA/CPU installRequirements is skipped on darwin — there is no
+# macOS/arm64 vLLM wheel on PyPI; vLLM is built from source and the MLX engine
+# is layered on by the vllm-metal wheel.
+if [ "$(uname -s)" = "Darwin" ]; then
+    # Create/activate the portable 3.12 venv. On darwin USE_PIP=true and
+    # PORTABLE_PYTHON=true (set by scripts/build/python-darwin.sh), so this is a
+    # `python -m venv` based, relocatable venv.
+    ensureVenv
+
+    # vllm-metal's installer drives everything through `uv`: building vLLM from
+    # the CPU requirements needs `--index-strategy unsafe-best-match` (mixes the
+    # pytorch CPU channel with PyPI), a flag plain pip does not have. The darwin
+    # venv is pip-based, so bootstrap uv into it. uv honours $VIRTUAL_ENV (set by
+    # libbackend's _activateVenv) and installs into THIS venv — same pattern the
+    # intel branch below relies on.
+    pip install uv
+
+    # The ONLY darwin version pin -- AUTO-BUMPED by .github/bump_vllm_metal.sh,
+    # which tracks vllm-project/vllm-metal releases (NOT vllm/vllm latest). Keep
+    # it as a plain double-quoted assignment on its own line so the bumper's sed
+    # can rewrite it. Darwin therefore follows vllm-metal and can lag the Linux
+    # vllm pin (requirements-cublas13-after.txt, bumped independently against
+    # vllm/vllm) until vllm-metal supports a newer vLLM.
+    VLLM_METAL_VERSION="v0.3.0.dev20260622062346"
+
+    # The coupled vLLM source version is whatever this vllm-metal release builds
+    # against -- it declares it in its own installer as `vllm_v=`. Derive it from
+    # the PINNED tag rather than hardcoding a second value that could drift. The
+    # tag is immutable, so this stays reproducible across rebuilds.
+    VLLM_VERSION=$(curl -fsSL "https://raw.githubusercontent.com/vllm-project/vllm-metal/${VLLM_METAL_VERSION}/install.sh" \
+        | grep -oE 'vllm_v="[0-9]+\.[0-9]+\.[0-9]+"' | head -n1 | cut -d'"' -f2)
+    if [ -z "${VLLM_VERSION}" ]; then
+        echo "ERROR: could not derive the vLLM version from vllm-metal ${VLLM_METAL_VERSION}" >&2
+        exit 1
+    fi
+    echo "vllm-metal ${VLLM_METAL_VERSION} builds against vLLM ${VLLM_VERSION}"
+
+    _vllm_src=$(mktemp -d)
+    trap 'rm -rf "${_vllm_src}"' EXIT
+    pushd "${_vllm_src}"
+        # 1) Build vLLM ${VLLM_VERSION} from the release source tarball against
+        #    the CPU requirements. vllm-metal layers its MLX platform plugin on
+        #    top of this exact build.
+        curl -fsSL -o "vllm-${VLLM_VERSION}.tar.gz" \
+            "https://github.com/vllm-project/vllm/releases/download/v${VLLM_VERSION}/vllm-${VLLM_VERSION}.tar.gz"
+        tar -xzf "vllm-${VLLM_VERSION}.tar.gz"
+        pushd "vllm-${VLLM_VERSION}"
+            uv pip install -r requirements/cpu.txt --index-strategy unsafe-best-match
+            # -Wno-parentheses: clang on macOS treats one of vLLM's C++ warnings
+            # as an error without it (matches the upstream installer's CXXFLAGS).
+            CXXFLAGS="-Wno-parentheses" uv pip install .
+        popd
+    popd
+
+    # 2) Install the prebuilt vllm-metal wheel for the PINNED release. It pulls
+    #    mlx / mlx-metal as deps and registers the `metal` platform plugin that
+    #    backend.py resolves to at engine-init time. Build the release-asset URL
+    #    deterministically (tag + the cp312/arm64 wheel name) rather than querying
+    #    api.github.com, whose unauthenticated rate limit (60/hr per IP) 403s on
+    #    shared CI runners. The wheel version is the tag without its leading 'v'.
+    _metal_wheel="vllm_metal-${VLLM_METAL_VERSION#v}-cp312-cp312-macosx_11_0_arm64.whl"
+    _metal_wheel_url="https://github.com/vllm-project/vllm-metal/releases/download/${VLLM_METAL_VERSION}/${_metal_wheel}"
+    echo "Installing vllm-metal wheel: ${_metal_wheel_url}"
+    uv pip install "${_metal_wheel_url}"
+
+    # Generate the gRPC stubs (backend_pb2*). installRequirements normally does
+    # this via runProtogen at the end; we skipped installRequirements on darwin,
+    # so call it explicitly here.
+    runProtogen
+
 # Intel XPU has no upstream-published vllm wheels, so we always build vllm
 # from source against torch-xpu and replace the default triton with
 # triton-xpu (matching torch 2.11). Mirrors the upstream procedure:
 # https://github.com/vllm-project/vllm/blob/main/docs/getting_started/installation/gpu.xpu.inc.md
-if [ "x${BUILD_TYPE}" == "xintel" ]; then
+elif [ "x${BUILD_TYPE}" == "xintel" ]; then
     # Hide requirements-intel-after.txt so installRequirements doesn't
     # try `pip install vllm` (would either fail or grab a non-XPU wheel).
     _intel_after="${backend_dir}/requirements-intel-after.txt"

backend/python/vllm/requirements-cublas13-after.txt

@@ -4,4 +4,7 @@
 # instead — the cublas13 case in install.sh adds --index-strategy=unsafe-best-match
 # so uv consults this index alongside PyPI.
 --extra-index-url https://wheels.vllm.ai/0.23.0/cu130
+# VERSION COUPLING: darwin/Apple-Silicon builds use vllm-metal (see install.sh),
+# which pins this exact vLLM version. Bumping vllm here means coordinating with a
+# vllm-metal release that supports the new version, or macOS/Metal builds break.
 vllm==0.23.0

core/application/config_file_watcher.go

@@ -215,6 +215,7 @@ func readRuntimeSettingsJson(startupAppConfig config.ApplicationConfig) fileHand
 		envBackendGalleries := slices.Equal(appConfig.BackendGalleries, startupAppConfig.BackendGalleries)
 		envAutoloadGalleries := appConfig.AutoloadGalleries == startupAppConfig.AutoloadGalleries
 		envAutoloadBackendGalleries := appConfig.AutoloadBackendGalleries == startupAppConfig.AutoloadBackendGalleries
+		envPIIDefaultDetectors := slices.Equal(appConfig.PIIDefaultDetectors, startupAppConfig.PIIDefaultDetectors)
 		envAgentJobRetentionDays := appConfig.AgentJobRetentionDays == startupAppConfig.AgentJobRetentionDays
 		envForceEvictionWhenBusy := appConfig.ForceEvictionWhenBusy == startupAppConfig.ForceEvictionWhenBusy
 		envLRUEvictionMaxRetries := appConfig.LRUEvictionMaxRetries == startupAppConfig.LRUEvictionMaxRetries
@@ -335,6 +336,15 @@ func readRuntimeSettingsJson(startupAppConfig config.ApplicationConfig) fileHand
 			if settings.AutoloadBackendGalleries != nil && !envAutoloadBackendGalleries {
 				appConfig.AutoloadBackendGalleries = *settings.AutoloadBackendGalleries
 			}
+			if settings.PIIDefaultDetectors != nil && !envPIIDefaultDetectors {
+				// Request-side default redaction reads this live via
+				// ResolvePIIPolicy, so a file edit takes effect on the next chat
+				// request. The MITM listener resolves its per-host detector map
+				// once at start, so a raw file edit reaches cloud-proxy traffic
+				// only after a restart or a POST /api/settings (which rebuilds
+				// the listener) — the admin UI uses the latter.
+				appConfig.PIIDefaultDetectors = append([]string(nil), (*settings.PIIDefaultDetectors)...)
+			}
 			if settings.AutoUpgradeBackends != nil {
 				appConfig.AutoUpgradeBackends = *settings.AutoUpgradeBackends
 			}

core/application/runtime_settings_branding_test.go

@@ -109,6 +109,52 @@ var _ = Describe("loadRuntimeSettingsFromFile", func() {
 		})
 	})
 
+	// Instance-wide default PII detectors. The file is the only source (no
+	// env var), and the loader runs immediately before startMITMIfConfigured,
+	// so a regression here means the cloud-proxy MITM listener resolves an
+	// empty detector set at boot and forwards intercepted traffic unredacted —
+	// even though pii_default_detectors is on disk and the MITM model has PII
+	// enabled. It also breaks request-side default redaction the same way.
+	Describe("PII default detectors", func() {
+		It("loads pii_default_detectors from the file", func() {
+			cfg := &config.ApplicationConfig{DynamicConfigsDir: seedSettings(`{"pii_default_detectors": ["privacy-filter-nemotron", "secret-filter"]}`)}
+			loadRuntimeSettingsFromFile(cfg)
+			Expect(cfg.PIIDefaultDetectors).To(Equal([]string{"privacy-filter-nemotron", "secret-filter"}))
+		})
+
+		It("does not override an env/CLI-set value (LOCALAI_PII_DEFAULT_DETECTORS)", func() {
+			cfg := &config.ApplicationConfig{
+				DynamicConfigsDir:   seedSettings(`{"pii_default_detectors": ["from-file"]}`),
+				PIIDefaultDetectors: []string{"from-env"}, // simulate WithPIIDefaultDetectors(env)
+			}
+			loadRuntimeSettingsFromFile(cfg)
+			Expect(cfg.PIIDefaultDetectors).To(Equal([]string{"from-env"}), "env var must win over the persisted file value")
+		})
+	})
+
+	// The live file watcher applies pii_default_detectors on a runtime change
+	// the same way it handles galleries/threads/etc.: env-set values (current
+	// == startup snapshot) are left alone, otherwise the file value is applied
+	// to the live config so request-side default redaction picks it up without
+	// a restart.
+	Describe("file watcher: pii_default_detectors", func() {
+		It("applies a changed file value to the live config", func() {
+			startup := config.ApplicationConfig{} // no env baseline
+			live := &config.ApplicationConfig{PIIDefaultDetectors: []string{"old"}}
+			handler := readRuntimeSettingsJson(startup)
+			Expect(handler([]byte(`{"pii_default_detectors":["new-a","new-b"]}`), live)).To(Succeed())
+			Expect(live.PIIDefaultDetectors).To(Equal([]string{"new-a", "new-b"}))
+		})
+
+		It("leaves an env-controlled value untouched", func() {
+			startup := config.ApplicationConfig{PIIDefaultDetectors: []string{"from-env"}}
+			live := &config.ApplicationConfig{PIIDefaultDetectors: []string{"from-env"}}
+			handler := readRuntimeSettingsJson(startup)
+			Expect(handler([]byte(`{"pii_default_detectors":["from-file"]}`), live)).To(Succeed())
+			Expect(live.PIIDefaultDetectors).To(Equal([]string{"from-env"}), "env-controlled detectors must not be overwritten by the file")
+		})
+	})
+
 	// The Agent Pool block has a mix of zero and non-zero defaults
 	// (Enabled=true, EmbeddingModel="granite-...", MaxChunkingSize=400,
 	// VectorEngine="chromem", AgentHubURL="https://agenthub.localai.io").

core/application/startup.go

@@ -750,6 +750,20 @@ func loadRuntimeSettingsFromFile(options *config.ApplicationConfig) {
 		options.MITMListen = *settings.MITMListen
 	}
 
+	// Instance-wide default PII detectors. LOCALAI_PII_DEFAULT_DETECTORS (via
+	// WithPIIDefaultDetectors) wins when set; otherwise the file is the source
+	// — apply it only when the env/CLI left the value empty, mirroring the
+	// "env > file" precedence used for the other fields. This must land before
+	// startMITMIfConfigured (called right after this loader): the cloud-proxy
+	// listener resolves each intercept host's detectors once at start via
+	// ResolvePIIPolicy, and a MITM model that names no detectors of its own
+	// falls back to these defaults. Without it the listener (and request-side
+	// default redaction) starts with an empty detector set and forwards
+	// traffic unredacted even though pii_default_detectors is on disk.
+	if settings.PIIDefaultDetectors != nil && len(options.PIIDefaultDetectors) == 0 {
+		options.PIIDefaultDetectors = append([]string(nil), (*settings.PIIDefaultDetectors)...)
+	}
+
 	// Backend upgrade flags
 	if settings.AutoUpgradeBackends != nil {
 		if !options.AutoUpgradeBackends {

core/cli/run.go

@@ -140,7 +140,7 @@ type RunCMD struct {
 	OIDCIssuer           string `env:"LOCALAI_OIDC_ISSUER" help:"OIDC issuer URL for auto-discovery" group:"auth"`
 	OIDCClientID         string `env:"LOCALAI_OIDC_CLIENT_ID" help:"OIDC Client ID (auto-enables auth)" group:"auth"`
 	OIDCClientSecret     string `env:"LOCALAI_OIDC_CLIENT_SECRET" help:"OIDC Client Secret" group:"auth"`
-	AuthBaseURL          string `env:"LOCALAI_BASE_URL" help:"Base URL for OAuth callbacks (e.g. http://localhost:8080)" group:"auth"`
+	ExternalBaseURL      string `env:"LOCALAI_BASE_URL" help:"External base URL of this instance (e.g. https://localhost:8080). Used for OAuth callbacks and self-referential links (generated images/videos, job status). When unset, derived from X-Forwarded-Proto/Host or Forwarded headers." group:"api"`
 	AuthAdminEmail       string `env:"LOCALAI_ADMIN_EMAIL" help:"Email address to auto-promote to admin role" group:"auth"`
 	AuthRegistrationMode string `env:"LOCALAI_REGISTRATION_MODE" default:"open" help:"Registration mode: 'open' (default), 'approval', or 'invite' (invite code required)" group:"auth"`
 	DisableLocalAuth     bool   `env:"LOCALAI_DISABLE_LOCAL_AUTH" default:"false" help:"Disable local email/password registration and login (use with OAuth/OIDC-only setups)" group:"auth"`
@@ -181,6 +181,8 @@ type RunCMD struct {
 	// Cloud-proxy MITM listener (off by default).
 	MITMListen string `env:"LOCALAI_MITM_LISTEN" help:"Address (host:port) for the cloudproxy MITM listener. Empty = disabled. Clients set HTTPS_PROXY=http://<this>:<port>. Intercept hosts are declared per-model via the model YAML mitm.hosts: block; create one from the Add Model UI." group:"middleware"`
 	MITMCADir  string `env:"LOCALAI_MITM_CA_DIR" type:"path" help:"Directory holding the MITM proxy CA cert + key. Defaults to <data-path>/mitm-ca." group:"middleware"`
+
+	PIIDefaultDetectors []string `env:"LOCALAI_PII_DEFAULT_DETECTORS" help:"Instance-wide default PII/secret detector model names applied to any PII-enabled model (chiefly cloud-proxy / MITM models) that names no pii.detectors of its own. Comma-separated, e.g. privacy-filter-nemotron,secret-filter. Takes precedence over the value persisted via the Middleware UI." group:"middleware"`
 }
 
 func (r *RunCMD) Run(ctx *cliContext.Context) error {
@@ -243,6 +245,7 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 		config.WithAPIAddress(r.Address),
 		config.WithMITMListen(r.MITMListen),
 		config.WithMITMCADir(r.MITMCADir),
+		config.WithPIIDefaultDetectors(r.PIIDefaultDetectors),
 		config.WithAgentJobRetentionDays(r.AgentJobRetentionDays),
 		config.WithLlamaCPPTunnelCallback(func(tunnels []string) {
 			tunnelEnvVar := strings.Join(tunnels, ",")
@@ -500,9 +503,6 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 			opts = append(opts, config.WithAuthOIDCClientID(r.OIDCClientID))
 			opts = append(opts, config.WithAuthOIDCClientSecret(r.OIDCClientSecret))
 		}
-		if r.AuthBaseURL != "" {
-			opts = append(opts, config.WithAuthBaseURL(r.AuthBaseURL))
-		}
 		if r.AuthAdminEmail != "" {
 			opts = append(opts, config.WithAuthAdminEmail(r.AuthAdminEmail))
 		}
@@ -520,6 +520,12 @@ func (r *RunCMD) Run(ctx *cliContext.Context) error {
 		}
 	}
 
+	// Applied unconditionally: the external base URL governs all self-referential
+	// links (not just OAuth callbacks), so it must take effect even when auth is off.
+	if r.ExternalBaseURL != "" {
+		opts = append(opts, config.WithExternalBaseURL(r.ExternalBaseURL))
+	}
+
 	if idleWatchDog || busyWatchDog {
 		opts = append(opts, config.EnableWatchDog)
 		if idleWatchDog {

core/config/application_config.go

@@ -49,6 +49,13 @@ type ApplicationConfig struct {
 	P2PNetworkID                  string
 	Federated                     bool
 
+	// ExternalBaseURL is the externally visible base URL of this instance
+	// (scheme+host[:port]), set via LOCALAI_BASE_URL. When non-empty it is
+	// authoritative for every self-referential URL LocalAI emits (OAuth
+	// callbacks, generated image/video links, async job StatusURLs),
+	// overriding proxy-header detection. Empty = derive from request headers.
+	ExternalBaseURL string
+
 	// DisableStats turns off per-request token tracking. By default the
 	// routing module's billing recorder runs in every mode (including
 	// no-auth single-user) so dashboards and `/api/usage` are immediately
@@ -196,7 +203,6 @@ type AuthConfig struct {
 	OIDCIssuer          string // OIDC issuer URL for auto-discovery (e.g. https://accounts.google.com)
 	OIDCClientID        string
 	OIDCClientSecret    string
-	BaseURL             string // for OAuth callback URLs (e.g. "http://localhost:8080")
 	AdminEmail          string // auto-promote to admin on login
 	RegistrationMode    string // "open", "approval" (default when empty), "invite"
 	DisableLocalAuth    bool   // disable local email/password registration and login
@@ -712,6 +718,18 @@ func WithMITMCADir(dir string) AppOption {
 	}
 }
 
+// WithPIIDefaultDetectors sets the instance-wide default PII/secret detector
+// model names applied to any PII-enabled model (chiefly cloud-proxy / MITM
+// models) that names no pii.detectors of its own. CLI/env:
+// LOCALAI_PII_DEFAULT_DETECTORS. Empty leaves the value to
+// runtime_settings.json / the Middleware UI; a non-empty value takes
+// precedence over the file (env > file).
+func WithPIIDefaultDetectors(detectors []string) AppOption {
+	return func(o *ApplicationConfig) {
+		o.PIIDefaultDetectors = detectors
+	}
+}
+
 func WithDynamicConfigDir(dynamicConfigsDir string) AppOption {
 	return func(o *ApplicationConfig) {
 		o.DynamicConfigsDir = dynamicConfigsDir
@@ -938,9 +956,9 @@ func WithAuthGitHubClientSecret(clientSecret string) AppOption {
 	}
 }
 
-func WithAuthBaseURL(baseURL string) AppOption {
+func WithExternalBaseURL(url string) AppOption {
 	return func(o *ApplicationConfig) {
-		o.Auth.BaseURL = baseURL
+		o.ExternalBaseURL = url
 	}
 }
 

core/config/hardware_defaults.go

@@ -2,6 +2,7 @@ package config
 
 import (
 	"fmt"
+	"os"
 	"strconv"
 	"strings"
 
@@ -9,6 +10,19 @@ import (
 	"github.com/mudler/xlog"
 )
 
+// HardwareDefaultsDisabled reports whether hardware auto-tuning is turned off via
+// LOCALAI_DISABLE_HARDWARE_DEFAULTS=true (mirrors LOCALAI_DISABLE_GUESSING). When
+// set, ApplyHardwareDefaults and the distributed router's node tuning are
+// skipped entirely, so the backend runs llama.cpp's stock batch/parallel
+// behavior — an escape hatch for users who want predictable, un-tuned defaults.
+func HardwareDefaultsDisabled() bool {
+	// Read directly like the sibling LOCALAI_DISABLE_GUESSING toggle in
+	// hooks_llamacpp.go: these config-layer heuristic switches run deep in the
+	// defaults pipeline with no ApplicationConfig in scope to plumb through.
+	//nolint:forbidigo // config-layer heuristic toggle, mirrors LOCALAI_DISABLE_GUESSING
+	return os.Getenv("LOCALAI_DISABLE_HARDWARE_DEFAULTS") == "true"
+}
+
 // Hardware-driven model-config defaults.
 //
 // This sits alongside the other config overriders (ApplyInferenceDefaults for
@@ -54,8 +68,35 @@ func (g GPU) IsNVIDIABlackwell() bool {
 	return maj >= 12
 }
 
+// Compute-buffer headroom guard for the raised physical batch.
+//
+// Raising n_ubatch grows the CUDA *compute buffer* (the scratch for the forward
+// graph), which is allocated PER DEVICE — it does not benefit from a second GPU
+// the way weights or KV (which are split across devices) do. The buffer scales
+// ~linearly with n_ubatch * n_ctx, so a large context turns the GB10-tuned
+// ub2048 into multi-GiB of extra scratch that must fit on a SINGLE card. On a
+// 16 GiB consumer Blackwell with a 200k context that overflows (issue #10485),
+// even though the GB10 it was measured on (128 GiB unified memory) had room.
+//
+// These constants size a conservative guard: only raise the batch when the
+// extra scratch fits the per-device VRAM ceiling.
+const (
+	// computeBufferBytesPerCell approximates the CUDA compute-buffer cost of one
+	// (n_ubatch * n_ctx) cell. Derived from an observed allocation (ub2048 *
+	// ctx204800 ~= 4.5 GiB => ~11 B/cell) and rounded up to 16 for margin, since
+	// the real cost also grows with model width (heads / embedding dim) which we
+	// don't know at config time.
+	computeBufferBytesPerCell = 16
+	// blackwellBatchHeadroomDivisor caps the extra compute buffer from raising the
+	// physical batch at VRAM/divisor. /4 keeps the bulk of a device for weights +
+	// KV, which already dominate VRAM use.
+	blackwellBatchHeadroomDivisor = 4
+)
+
 // PhysicalBatch returns the canonical physical batch (n_batch/n_ubatch) for the
-// given hardware, used when the model config leaves batch unset.
+// given hardware class, ignoring context/VRAM headroom. Use
+// PhysicalBatchForContext when a model context and per-device VRAM are known
+// (the load paths) so the raised batch can't overflow a single device.
 func PhysicalBatch(g GPU) int {
 	if g.IsNVIDIABlackwell() {
 		return BlackwellPhysicalBatch
@@ -63,6 +104,51 @@ func PhysicalBatch(g GPU) int {
 	return DefaultPhysicalBatch
 }
 
+// PhysicalBatchForContext is PhysicalBatch gated on per-device VRAM headroom for
+// the given context: it only raises the batch above the conservative default
+// when the extra compute buffer (which is allocated on a single device and grows
+// with n_ubatch * n_ctx) fits within blackwellBatchHeadroomDivisor of the GPU's
+// VRAM. g.VRAM must be the PER-DEVICE ceiling (the smallest device on a
+// multi-GPU host), not the summed total — the compute buffer can't be split.
+//
+// VRAM 0 (unknown) stays conservative rather than risk a per-device OOM; the
+// GB10 / unified-memory path reports system RAM, so it still clears the guard.
+func PhysicalBatchForContext(g GPU, ctx int) int {
+	if !g.IsNVIDIABlackwell() {
+		return DefaultPhysicalBatch
+	}
+	if g.VRAM == 0 {
+		return DefaultPhysicalBatch
+	}
+	if largeContextForDevice(g, ctx) {
+		return DefaultPhysicalBatch
+	}
+	return BlackwellPhysicalBatch
+}
+
+// largeContextForDevice reports whether the given context is large relative to
+// the per-device VRAM ceiling — the shared "tight single-model fit" signal that
+// suppresses BOTH throughput-oriented defaults (the Blackwell batch boost and
+// the concurrency slot count). It sizes the extra compute-buffer scratch a
+// raised batch would need at this context (which grows ~n_ubatch * n_ctx and
+// is allocated per device) and asks whether it overflows a fraction of the
+// device VRAM; when it does, the device has no headroom to spend on throughput
+// and the conservative defaults must hold (issue #10485).
+//
+// g.VRAM must be the PER-DEVICE ceiling (the smallest device on a multi-GPU
+// host). VRAM 0 (unknown) is treated as not-large so detection gaps don't
+// silently disable the defaults.
+func largeContextForDevice(g GPU, ctx int) bool {
+	if g.VRAM == 0 {
+		return false
+	}
+	if ctx <= 0 {
+		ctx = DefaultContextSize
+	}
+	extra := uint64(ctx) * uint64(BlackwellPhysicalBatch-DefaultPhysicalBatch) * computeBufferBytesPerCell
+	return extra > g.VRAM/blackwellBatchHeadroomDivisor
+}
+
 // IsManagedPhysicalBatch reports whether n is a value PhysicalBatch assigns.
 // Callers that re-tune a value chosen by an upstream host (the distributed
 // router correcting the frontend's guess) use this to avoid clobbering an
@@ -99,17 +185,50 @@ func DefaultParallelSlots(g GPU) int {
 	}
 }
 
-// EnsureParallelOption appends a VRAM-scaled "parallel:N" backend option when the
-// model doesn't already set one (and the GPU warrants concurrency). Returns the
-// possibly-extended options. Shared by the single-host config path
-// (ApplyHardwareDefaults) and the distributed router (per selected node).
-func EnsureParallelOption(opts []string, gpu GPU) []string {
-	if slots := DefaultParallelSlots(gpu); slots > 1 && !hasParallelOption(opts) {
+// ParallelSlotsForContext is DefaultParallelSlots gated on per-device VRAM
+// headroom for the given context. A large context already claims most of a
+// single device's VRAM (the KV cache plus the per-slot compute/checkpoint
+// scratch that scales with n_seq_max), so defaulting multiple slots there
+// pushes a tight single-model fit into per-device CUDA OOM (issue #10485): the
+// model loads but the final allocation (e.g. an MTP draft context's KV cache)
+// overflows the tighter card by a few hundred MiB. Returns 1 (no concurrency)
+// in that tight regime, otherwise the VRAM-scaled DefaultParallelSlots.
+//
+// g.VRAM must be the PER-DEVICE ceiling (smallest device on a multi-GPU host).
+// It shares largeContextForDevice with the batch boost so both throughput
+// defaults are suppressed together; the GB10 / unified-memory path reports
+// system RAM and so keeps full concurrency even at large contexts.
+func ParallelSlotsForContext(g GPU, ctx int) int {
+	slots := DefaultParallelSlots(g)
+	if slots <= 1 || g.VRAM == 0 {
+		return slots
+	}
+	if largeContextForDevice(g, ctx) {
+		return 1
+	}
+	return slots
+}
+
+// EnsureParallelOptionForContext appends a VRAM-scaled "parallel:N" backend
+// option when the model doesn't already set one and the GPU warrants (and has
+// headroom for) concurrency at this context. Returns the possibly-extended
+// options. Shared by the single-host config path (ApplyHardwareDefaults) and
+// the distributed router (per selected node).
+func EnsureParallelOptionForContext(opts []string, gpu GPU, ctx int) []string {
+	if slots := ParallelSlotsForContext(gpu, ctx); slots > 1 && !hasParallelOption(opts) {
 		return append(opts, fmt.Sprintf("parallel:%d", slots))
 	}
 	return opts
 }
 
+// EnsureParallelOption is EnsureParallelOptionForContext with no known context
+// (defaults to DefaultContextSize, which clears the headroom gate on any device
+// large enough to warrant concurrency). Kept for callers without a model
+// context.
+func EnsureParallelOption(opts []string, gpu GPU) []string {
+	return EnsureParallelOptionForContext(opts, gpu, 0)
+}
+
 // hasParallelOption reports whether the model already sets parallel/n_parallel
 // so we never override an explicit value (helper shared with serving_defaults.go).
 func hasParallelOption(opts []string) bool {
@@ -122,7 +241,12 @@ func hasParallelOption(opts []string) bool {
 // deterministic device — detection does a live nvidia-smi call.
 var localGPU = func() GPU {
 	vendor, _ := xsysinfo.DetectGPUVendor()
-	vram, _ := xsysinfo.TotalAvailableVRAM()
+	// Use the SMALLEST device's VRAM, not the summed total: the parallel-slot
+	// tier and the batch headroom guard both reason about what fits on a single
+	// card, and per-device compute buffers can't be split across GPUs. Summing
+	// two 16 GiB cards into "32 GiB" is what over-provisioned multi-GPU hosts
+	// into OOM (issue #10485).
+	vram, _ := xsysinfo.MinPerGPUVRAM()
 	return GPU{
 		Vendor:            vendor,
 		ComputeCapability: xsysinfo.NVIDIAComputeCapability(),
@@ -134,25 +258,36 @@ var localGPU = func() GPU {
 // and were left unset by the user. Currently: a larger physical batch on
 // Blackwell. Explicit config always wins (we only touch zero values).
 func ApplyHardwareDefaults(cfg *ModelConfig, gpu GPU) {
-	if cfg == nil {
+	if cfg == nil || HardwareDefaultsDisabled() {
 		return
 	}
-	if cfg.Batch == 0 && gpu.IsNVIDIABlackwell() {
-		cfg.Batch = BlackwellPhysicalBatch
-		xlog.Debug("[hardware_defaults] Blackwell GPU: defaulting physical batch",
-			"batch", cfg.Batch, "compute_cap", gpu.ComputeCapability)
+	// Raise the physical batch on Blackwell only when the resulting compute
+	// buffer fits the per-device VRAM at THIS model's context. Leaving Batch at 0
+	// (rather than writing the default 512) preserves the downstream single-pass
+	// sizing in core/backend.EffectiveBatchSize for embedding/score/rerank.
+	ctx := DefaultContextSize
+	if cfg.ContextSize != nil {
+		ctx = *cfg.ContextSize
+	}
+	if cfg.Batch == 0 {
+		if PhysicalBatchForContext(gpu, ctx) == BlackwellPhysicalBatch {
+			cfg.Batch = BlackwellPhysicalBatch
+			xlog.Debug("[hardware_defaults] Blackwell GPU: defaulting physical batch",
+				"batch", cfg.Batch, "compute_cap", gpu.ComputeCapability, "context", ctx, "vram_gib", gpu.VRAM>>30)
+		}
 	}
 
 	// Enable concurrent serving by default on a capable GPU: without this the
 	// llama.cpp backend runs n_parallel=1 and serializes multi-user requests
 	// (continuous batching stays off). Unified KV means the slots share the
-	// context budget, so this is concurrency without extra KV memory. Explicit
-	// parallel/n_parallel in the model options always wins.
+	// context budget, but a context large enough to fill a single device leaves
+	// no room for the per-slot scratch, so the slot count is gated on per-device
+	// headroom too (issue #10485). Explicit parallel/n_parallel always wins.
 	if before := len(cfg.Options); true {
-		cfg.Options = EnsureParallelOption(cfg.Options, gpu)
+		cfg.Options = EnsureParallelOptionForContext(cfg.Options, gpu, ctx)
 		if len(cfg.Options) > before {
 			xlog.Debug("[hardware_defaults] defaulting parallel slots for concurrent serving",
-				"option", cfg.Options[len(cfg.Options)-1], "vram_gib", gpu.VRAM>>30)
+				"option", cfg.Options[len(cfg.Options)-1], "context", ctx, "vram_gib", gpu.VRAM>>30)
 		}
 	}
 }

core/config/hardware_defaults_internal_test.go

@@ -9,26 +9,37 @@ import (
 // GPU. The detection seam (localGPU) is injected so the path is deterministic
 // without a real GPU.
 var _ = Describe("SetDefaults hardware defaults (single-instance)", func() {
+	const gib = uint64(1) << 30
+
 	var orig func() GPU
 	BeforeEach(func() { orig = localGPU })
 	AfterEach(func() { localGPU = orig })
 
-	It("sets the physical batch on a local Blackwell GPU", func() {
-		localGPU = func() GPU { return GPU{ComputeCapability: "12.1"} }
+	It("sets the physical batch on a local Blackwell GPU with headroom", func() {
+		localGPU = func() GPU { return GPU{ComputeCapability: "12.1", VRAM: 119 * gib} }
 		cfg := &ModelConfig{}
 		cfg.SetDefaults()
 		Expect(cfg.Batch).To(Equal(BlackwellPhysicalBatch))
 	})
 
+	It("leaves batch unset when a large context would overflow the device", func() {
+		// Regression guard for issue #10485: 16 GiB consumer Blackwell + ~200k ctx.
+		localGPU = func() GPU { return GPU{ComputeCapability: "12.0", VRAM: 16 * gib} }
+		ctx := 204800
+		cfg := &ModelConfig{LLMConfig: LLMConfig{ContextSize: &ctx}}
+		cfg.SetDefaults()
+		Expect(cfg.Batch).To(Equal(0))
+	})
+
 	It("leaves batch unset on a non-Blackwell local GPU", func() {
-		localGPU = func() GPU { return GPU{ComputeCapability: "8.9"} }
+		localGPU = func() GPU { return GPU{ComputeCapability: "8.9", VRAM: 119 * gib} }
 		cfg := &ModelConfig{}
 		cfg.SetDefaults()
 		Expect(cfg.Batch).To(Equal(0))
 	})
 
 	It("never overrides an explicit batch", func() {
-		localGPU = func() GPU { return GPU{ComputeCapability: "12.1"} }
+		localGPU = func() GPU { return GPU{ComputeCapability: "12.1", VRAM: 119 * gib} }
 		cfg := &ModelConfig{}
 		cfg.Batch = 1024
 		cfg.SetDefaults()

core/config/hardware_defaults_test.go

@@ -7,6 +7,8 @@ import (
 )
 
 var _ = Describe("Hardware-driven config defaults", func() {
+	const gib = uint64(1) << 30
+
 	DescribeTable("GPU.IsNVIDIABlackwell (sm_12x consumer family)",
 		func(cc string, want bool) {
 			Expect(GPU{ComputeCapability: cc}.IsNVIDIABlackwell()).To(Equal(want))
@@ -35,29 +37,69 @@ var _ = Describe("Hardware-driven config defaults", func() {
 		})
 	})
 
+	Describe("PhysicalBatchForContext (per-device VRAM headroom)", func() {
+		It("raises the batch when the compute buffer fits the device", func() {
+			// 16 GiB Blackwell with a small context: the extra scratch is tiny.
+			Expect(PhysicalBatchForContext(GPU{ComputeCapability: "12.0", VRAM: 16 * gib}, 8192)).
+				To(Equal(BlackwellPhysicalBatch))
+		})
+		It("keeps the default batch when a large context would overflow one device", func() {
+			// The issue #10485 case: 16 GiB consumer Blackwell, ~200k context.
+			Expect(PhysicalBatchForContext(GPU{ComputeCapability: "12.0", VRAM: 16 * gib}, 204800)).
+				To(Equal(DefaultPhysicalBatch))
+		})
+		It("still raises the batch on a large unified-memory device (GB10)", func() {
+			// GB10 reports system RAM (~119 GiB) as its single device's VRAM.
+			Expect(PhysicalBatchForContext(GPU{ComputeCapability: "12.1", VRAM: 119 * gib}, 204800)).
+				To(Equal(BlackwellPhysicalBatch))
+		})
+		It("stays conservative when VRAM is unknown", func() {
+			Expect(PhysicalBatchForContext(GPU{ComputeCapability: "12.1"}, 8192)).
+				To(Equal(DefaultPhysicalBatch))
+		})
+		It("never raises the batch on non-Blackwell", func() {
+			Expect(PhysicalBatchForContext(GPU{ComputeCapability: "9.0", VRAM: 80 * gib}, 8192)).
+				To(Equal(DefaultPhysicalBatch))
+		})
+	})
+
 	Describe("ApplyHardwareDefaults", func() {
-		It("raises an unset batch to 2048 on Blackwell", func() {
+		It("raises an unset batch to 2048 on Blackwell with headroom", func() {
 			cfg := &ModelConfig{}
-			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "12.1"})
+			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "12.1", VRAM: 119 * gib})
 			Expect(cfg.Batch).To(Equal(BlackwellPhysicalBatch))
 		})
+		It("leaves batch unset when a large context would overflow one device", func() {
+			// Regression guard for issue #10485: 16 GiB card + ~200k context.
+			ctx := 204800
+			cfg := &ModelConfig{LLMConfig: LLMConfig{ContextSize: &ctx}}
+			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "12.0", VRAM: 16 * gib})
+			Expect(cfg.Batch).To(Equal(0))
+		})
 		It("leaves batch unset on non-Blackwell", func() {
 			cfg := &ModelConfig{}
-			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "9.0"})
+			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "9.0", VRAM: 119 * gib})
 			Expect(cfg.Batch).To(Equal(0))
 		})
 		It("never overrides an explicit batch", func() {
 			cfg := &ModelConfig{}
 			cfg.Batch = 1024
-			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "12.1"})
+			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "12.1", VRAM: 119 * gib})
 			Expect(cfg.Batch).To(Equal(1024))
 		})
 		It("no-ops on nil", func() {
 			Expect(func() { ApplyHardwareDefaults(nil, GPU{ComputeCapability: "12.1"}) }).ToNot(Panic())
 		})
-	})
 
-	const gib = uint64(1) << 30
+		It("applies nothing when hardware defaults are disabled via env", func() {
+			GinkgoT().Setenv("LOCALAI_DISABLE_HARDWARE_DEFAULTS", "true")
+			Expect(HardwareDefaultsDisabled()).To(BeTrue())
+			cfg := &ModelConfig{}
+			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "12.1", VRAM: 119 * gib})
+			Expect(cfg.Batch).To(Equal(0))
+			Expect(cfg.Options).To(BeEmpty())
+		})
+	})
 
 	DescribeTable("DefaultParallelSlots (by VRAM)",
 		func(vramGiB uint64, want int) {
@@ -72,12 +114,46 @@ var _ = Describe("Hardware-driven config defaults", func() {
 		Entry("unknown 0", uint64(0), 1),
 	)
 
+	Describe("ParallelSlotsForContext (per-device VRAM headroom)", func() {
+		It("keeps the VRAM-scaled slot count when the context fits the device", func() {
+			// 16 GiB card, small context: plenty of room for concurrency.
+			Expect(ParallelSlotsForContext(GPU{VRAM: 16 * gib}, 8192)).To(Equal(4))
+		})
+		It("drops to a single slot when a large context already fills the device", func() {
+			// Regression guard for issue #10485: 16 GiB consumer Blackwell, ~200k
+			// context. Even with unified KV, the per-slot compute/checkpoint
+			// scratch from 4 slots is the straw that overflows the tighter device.
+			Expect(ParallelSlotsForContext(GPU{VRAM: 16 * gib}, 204800)).To(Equal(1))
+		})
+		It("keeps concurrency on a large unified-memory device (GB10)", func() {
+			// GB10 reports system RAM (~119 GiB): a 200k context leaves headroom.
+			Expect(ParallelSlotsForContext(GPU{VRAM: 119 * gib}, 204800)).To(Equal(8))
+		})
+		It("keeps concurrency on a big datacenter card with a large context", func() {
+			// 80 GiB A100: 200k context is a small fraction, concurrency stays.
+			Expect(ParallelSlotsForContext(GPU{VRAM: 80 * gib}, 204800)).To(Equal(8))
+		})
+		It("stays a single slot on small/unknown VRAM regardless of context", func() {
+			Expect(ParallelSlotsForContext(GPU{VRAM: 2 * gib}, 8192)).To(Equal(1))
+			Expect(ParallelSlotsForContext(GPU{}, 8192)).To(Equal(1))
+		})
+	})
+
 	Describe("ApplyHardwareDefaults parallel slots", func() {
 		It("adds a VRAM-scaled parallel option on a capable GPU", func() {
 			cfg := &ModelConfig{}
 			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "12.1", VRAM: 119 * gib})
 			Expect(cfg.Options).To(ContainElement("parallel:8"))
 		})
+		It("adds no parallel option when a large context already fills one device", func() {
+			// Regression guard for issue #10485: 16 GiB card + ~200k context. The
+			// model barely fits; defaulting concurrency tips the tighter GPU into
+			// CUDA OOM during the final (MTP draft) KV allocation.
+			ctx := 204800
+			cfg := &ModelConfig{LLMConfig: LLMConfig{ContextSize: &ctx}}
+			ApplyHardwareDefaults(cfg, GPU{ComputeCapability: "12.0", VRAM: 16 * gib})
+			Expect(cfg.Options).ToNot(ContainElement(ContainSubstring("parallel")))
+		})
 		It("scales the slot count down with VRAM", func() {
 			cfg := &ModelConfig{}
 			ApplyHardwareDefaults(cfg, GPU{VRAM: 24 * gib})

core/config/model_config.go

@@ -1204,11 +1204,6 @@ func (cfg *ModelConfig) SetDefaults(opts ...ConfigLoaderOption) {
 	// This ensures gallery-installed and runtime-loaded models get optimal parameters.
 	ApplyInferenceDefaults(cfg, cfg.Name, cfg.Model)
 
-	// Apply hardware-driven defaults (e.g. a larger physical batch on Blackwell).
-	// Uses the local GPU here; in distributed mode the router re-applies the same
-	// heuristics for the selected node's GPU before loading. Explicit config wins.
-	ApplyHardwareDefaults(cfg, localGPU())
-
 	// Apply serving-policy defaults (device-independent): cross-request prefix
 	// caching. Propagates to distributed nodes via the model options.
 	ApplyServingDefaults(cfg)
@@ -1247,6 +1242,16 @@ func (cfg *ModelConfig) SetDefaults(opts ...ConfigLoaderOption) {
 		cfg.ContextSize = &ctx
 	}
 	runBackendHooks(cfg, lo.modelPath)
+
+	// Apply hardware-driven defaults (e.g. a larger physical batch on Blackwell)
+	// LAST, after the context size is fully resolved (explicit config, LoadOptions,
+	// then the GGUF guess inside runBackendHooks): the Blackwell batch guard sizes
+	// the per-device compute buffer against this model's context, so it must see
+	// the final value, not a pre-guess nil. Uses the local GPU here; in distributed
+	// mode the router re-applies the same heuristics for the selected node's GPU
+	// before loading. Explicit config always wins.
+	ApplyHardwareDefaults(cfg, localGPU())
+
 	cfg.syncKnownUsecasesFromString()
 }
 

core/http/app.go

@@ -149,6 +149,18 @@ func API(application *application.Application) (*echo.Echo, error) {
 	// Middleware - StripPathPrefix must be registered early as it uses Rewrite which runs before routing
 	e.Pre(httpMiddleware.StripPathPrefix())
 
+	// Stamp the configured external base URL into each request context so
+	// middleware.BaseURL can treat it as authoritative for self-referential
+	// links. Registered as Pre so it runs before routing and handlers.
+	if extBaseURL := application.ApplicationConfig().ExternalBaseURL; extBaseURL != "" {
+		e.Pre(func(next echo.HandlerFunc) echo.HandlerFunc {
+			return func(c echo.Context) error {
+				c.Set("_external_base_url", extBaseURL)
+				return next(c)
+			}
+		})
+	}
+
 	e.Pre(middleware.RemoveTrailingSlash())
 
 	if application.ApplicationConfig().MachineTag != "" {

core/http/auth/db_sqlite.go

@@ -3,10 +3,51 @@
 package auth
 
 import (
+	"net/url"
+	"strings"
+
 	"gorm.io/driver/sqlite"
 	"gorm.io/gorm"
 )
 
 func openSQLiteDialector(path string) (gorm.Dialector, error) {
-	return sqlite.Open(path), nil
+	return sqlite.Open(buildSQLiteDSN(path)), nil
+}
+
+// buildSQLiteDSN augments a SQLite file path with connection pragmas that make
+// the auth DB resilient on slow or contended storage.
+//
+//   - _busy_timeout=5000 makes SQLite retry for up to 5s on SQLITE_BUSY instead
+//     of failing immediately. Network-backed storage (SMB/CIFS/NFS, e.g. Azure
+//     Files) is prone to transient lock contention during migration (see #10506).
+//   - _txlock=immediate takes the write lock at BEGIN, avoiding deadlocks when a
+//     read transaction later upgrades to a write during AutoMigrate.
+//
+// We deliberately do NOT set WAL journal mode: WAL relies on a shared-memory
+// mmap that does not work over SMB/NFS, which is exactly the failing case here.
+//
+// Caller-supplied values for either pragma are preserved.
+func buildSQLiteDSN(path string) string {
+	base := path
+	rawQuery := ""
+	if i := strings.IndexByte(path, '?'); i >= 0 {
+		base = path[:i]
+		rawQuery = path[i+1:]
+	}
+
+	values, err := url.ParseQuery(rawQuery)
+	if err != nil {
+		// An unparseable query string means a hand-crafted DSN we should not
+		// risk corrupting; leave it untouched.
+		return path
+	}
+
+	if values.Get("_busy_timeout") == "" {
+		values.Set("_busy_timeout", "5000")
+	}
+	if values.Get("_txlock") == "" {
+		values.Set("_txlock", "immediate")
+	}
+
+	return base + "?" + values.Encode()
 }

core/http/auth/db_sqlite_test.go

@@ -0,0 +1,57 @@
+//go:build auth
+
+package auth
+
+import (
+	"net/url"
+	"strings"
+
+	. "github.com/onsi/ginkgo/v2"
+	. "github.com/onsi/gomega"
+)
+
+// parseDSN splits a "base?query" DSN into its base and decoded query values so
+// assertions don't depend on url.Values.Encode()'s key ordering.
+func parseDSN(dsn string) (string, url.Values) {
+	base := dsn
+	rawQuery := ""
+	if i := strings.IndexByte(dsn, '?'); i >= 0 {
+		base = dsn[:i]
+		rawQuery = dsn[i+1:]
+	}
+	values, err := url.ParseQuery(rawQuery)
+	Expect(err).ToNot(HaveOccurred())
+	return base, values
+}
+
+var _ = Describe("buildSQLiteDSN", func() {
+	It("adds busy_timeout and txlock to a plain file path", func() {
+		base, values := parseDSN(buildSQLiteDSN("/data/database.db"))
+		Expect(base).To(Equal("/data/database.db"))
+		Expect(values.Get("_busy_timeout")).To(Equal("5000"))
+		Expect(values.Get("_txlock")).To(Equal("immediate"))
+	})
+
+	It("adds pragmas to an in-memory database", func() {
+		base, values := parseDSN(buildSQLiteDSN(":memory:"))
+		Expect(base).To(Equal(":memory:"))
+		Expect(values.Get("_busy_timeout")).To(Equal("5000"))
+		Expect(values.Get("_txlock")).To(Equal("immediate"))
+	})
+
+	It("preserves an existing query string", func() {
+		base, values := parseDSN(buildSQLiteDSN("/data/database.db?cache=shared"))
+		Expect(base).To(Equal("/data/database.db"))
+		Expect(values.Get("cache")).To(Equal("shared"))
+		Expect(values.Get("_busy_timeout")).To(Equal("5000"))
+		Expect(values.Get("_txlock")).To(Equal("immediate"))
+	})
+
+	It("does not override a caller-supplied busy_timeout or txlock", func() {
+		_, values := parseDSN(buildSQLiteDSN("/data/database.db?_busy_timeout=1000&_txlock=deferred"))
+		Expect(values["_busy_timeout"]).To(HaveLen(1), "_busy_timeout should not be duplicated")
+		Expect(values.Get("_busy_timeout")).To(Equal("1000"))
+		Expect(values["_txlock"]).To(HaveLen(1), "_txlock should not be duplicated")
+		Expect(values.Get("_txlock")).To(Equal("deferred"))
+	})
+})

core/http/endpoints/localai/agent_collections.go

@@ -70,7 +70,7 @@ func UploadToCollectionEndpoint(app *application.Application) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		svc := app.AgentPoolService()
 		userID := effectiveUserID(c)
-		name := c.Param("name")
+		name := decodedParam(c, "name")
 		file, err := c.FormFile("file")
 		if err != nil {
 			return c.JSON(http.StatusBadRequest, map[string]string{"error": "file required"})
@@ -116,7 +116,7 @@ func ListCollectionEntriesEndpoint(app *application.Application) echo.HandlerFun
 	return func(c echo.Context) error {
 		svc := app.AgentPoolService()
 		userID := effectiveUserID(c)
-		entries, err := svc.ListCollectionEntriesForUser(userID, c.Param("name"))
+		entries, err := svc.ListCollectionEntriesForUser(userID, decodedParam(c, "name"))
 		if err != nil {
 			if strings.Contains(err.Error(), "not found") {
 				return c.JSON(http.StatusNotFound, map[string]string{"error": err.Error()})
@@ -139,7 +139,7 @@ func GetCollectionEntryContentEndpoint(app *application.Application) echo.Handle
 		if err != nil {
 			entry = entryParam
 		}
-		content, chunkCount, err := svc.GetCollectionEntryContentForUser(userID, c.Param("name"), entry)
+		content, chunkCount, err := svc.GetCollectionEntryContentForUser(userID, decodedParam(c, "name"), entry)
 		if err != nil {
 			if strings.Contains(err.Error(), "not found") {
 				return c.JSON(http.StatusNotFound, map[string]string{"error": err.Error()})
@@ -164,7 +164,7 @@ func SearchCollectionEndpoint(app *application.Application) echo.HandlerFunc {
 		if err := c.Bind(&payload); err != nil {
 			return c.JSON(http.StatusBadRequest, map[string]string{"error": err.Error()})
 		}
-		results, err := svc.SearchCollectionForUser(userID, c.Param("name"), payload.Query, payload.MaxResults)
+		results, err := svc.SearchCollectionForUser(userID, decodedParam(c, "name"), payload.Query, payload.MaxResults)
 		if err != nil {
 			if strings.Contains(err.Error(), "not found") {
 				return c.JSON(http.StatusNotFound, map[string]string{"error": err.Error()})
@@ -182,7 +182,7 @@ func ResetCollectionEndpoint(app *application.Application) echo.HandlerFunc {
 	return func(c echo.Context) error {
 		svc := app.AgentPoolService()
 		userID := effectiveUserID(c)
-		if err := svc.ResetCollectionForUser(userID, c.Param("name")); err != nil {
+		if err := svc.ResetCollectionForUser(userID, decodedParam(c, "name")); err != nil {
 			if strings.Contains(err.Error(), "not found") {
 				return c.JSON(http.StatusNotFound, map[string]string{"error": err.Error()})
 			}
@@ -202,7 +202,7 @@ func DeleteCollectionEntryEndpoint(app *application.Application) echo.HandlerFun
 		if err := c.Bind(&payload); err != nil {
 			return c.JSON(http.StatusBadRequest, map[string]string{"error": err.Error()})
 		}
-		remaining, err := svc.DeleteCollectionEntryForUser(userID, c.Param("name"), payload.Entry)
+		remaining, err := svc.DeleteCollectionEntryForUser(userID, decodedParam(c, "name"), payload.Entry)
 		if err != nil {
 			if strings.Contains(err.Error(), "not found") {
 				return c.JSON(http.StatusNotFound, map[string]string{"error": err.Error()})
@@ -230,7 +230,7 @@ func AddCollectionSourceEndpoint(app *application.Application) echo.HandlerFunc
 		if payload.UpdateInterval < 1 {
 			payload.UpdateInterval = 60
 		}
-		if err := svc.AddCollectionSourceForUser(userID, c.Param("name"), payload.URL, payload.UpdateInterval); err != nil {
+		if err := svc.AddCollectionSourceForUser(userID, decodedParam(c, "name"), payload.URL, payload.UpdateInterval); err != nil {
 			if strings.Contains(err.Error(), "not found") {
 				return c.JSON(http.StatusNotFound, map[string]string{"error": err.Error()})
 			}
@@ -250,7 +250,7 @@ func RemoveCollectionSourceEndpoint(app *application.Application) echo.HandlerFu
 		if err := c.Bind(&payload); err != nil {
 			return c.JSON(http.StatusBadRequest, map[string]string{"error": err.Error()})
 		}
-		if err := svc.RemoveCollectionSourceForUser(userID, c.Param("name"), payload.URL); err != nil {
+		if err := svc.RemoveCollectionSourceForUser(userID, decodedParam(c, "name"), payload.URL); err != nil {
 			return c.JSON(http.StatusInternalServerError, map[string]string{"error": err.Error()})
 		}
 		return c.JSON(http.StatusOK, map[string]string{"status": "ok"})
@@ -267,7 +267,7 @@ func GetCollectionEntryRawFileEndpoint(app *application.Application) echo.Handle
 		if err != nil {
 			entry = entryParam
 		}
-		fpath, err := svc.GetCollectionEntryFilePathForUser(userID, c.Param("name"), entry)
+		fpath, err := svc.GetCollectionEntryFilePathForUser(userID, decodedParam(c, "name"), entry)
 		if err != nil {
 			if strings.Contains(err.Error(), "not found") {
 				return c.JSON(http.StatusNotFound, map[string]string{"error": err.Error()})
@@ -282,7 +282,7 @@ func ListCollectionSourcesEndpoint(app *application.Application) echo.HandlerFun
 	return func(c echo.Context) error {
 		svc := app.AgentPoolService()
 		userID := effectiveUserID(c)
-		sources, err := svc.ListCollectionSourcesForUser(userID, c.Param("name"))
+		sources, err := svc.ListCollectionSourcesForUser(userID, decodedParam(c, "name"))
 		if err != nil {
 			if strings.Contains(err.Error(), "not found") {
 				return c.JSON(http.StatusNotFound, map[string]string{"error": err.Error()})