security : add collaborator guidance

* graph : reuse hybrid graphs

* graph : reuse recurrent graphs

* graph : fix reuse check for recurrent inputs

* memory : move the recurrent state into the memory context

* Revert "memory : move the recurrent state into the memory context"

This reverts commit 00f115fe810815d4a22a6dee0acc346131e970e1.

* cont : fix build

.github/workflows/server-webui.yml

@@ -0,0 +1,295 @@
+# Server WebUI build and tests
+name: Server WebUI
+
+on:
+  workflow_dispatch: # allows manual triggering
+    inputs:
+      sha:
+        description: 'Commit SHA1 to build'
+        required: false
+        type: string
+      slow_tests:
+        description: 'Run slow tests'
+        required: true
+        type: boolean
+  push:
+    branches:
+      - master
+    paths: ['.github/workflows/server-webui.yml', 'tools/server/webui/**.*', 'tools/server/tests/**.*', 'tools/server/public/**']
+  pull_request:
+    types: [opened, synchronize, reopened]
+    paths: ['.github/workflows/server-webui.yml', 'tools/server/webui/**.*', 'tools/server/tests/**.*', 'tools/server/public/**']
+
+env:
+  LLAMA_LOG_COLORS: 1
+  LLAMA_LOG_PREFIX: 1
+  LLAMA_LOG_TIMESTAMPS: 1
+  LLAMA_LOG_VERBOSITY: 10
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}-${{ github.head_ref || github.run_id }}
+  cancel-in-progress: true
+
+jobs:
+  webui-setup:
+    name: WebUI Setup
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "22"
+          cache: "npm"
+          cache-dependency-path: "tools/server/webui/package-lock.json"
+
+      - name: Cache node_modules
+        uses: actions/cache@v4
+        id: cache-node-modules
+        with:
+          path: tools/server/webui/node_modules
+          key: ${{ runner.os }}-node-modules-${{ hashFiles('tools/server/webui/package-lock.json') }}
+          restore-keys: |
+            ${{ runner.os }}-node-modules-
+
+      - name: Install dependencies
+        if: steps.cache-node-modules.outputs.cache-hit != 'true'
+        run: npm ci
+        working-directory: tools/server/webui
+
+  webui-check:
+    needs: webui-setup
+    name: WebUI Check
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "22"
+
+      - name: Restore node_modules cache
+        uses: actions/cache@v4
+        with:
+          path: tools/server/webui/node_modules
+          key: ${{ runner.os }}-node-modules-${{ hashFiles('tools/server/webui/package-lock.json') }}
+          restore-keys: |
+            ${{ runner.os }}-node-modules-
+
+      - name: Run type checking
+        run: npm run check
+        working-directory: tools/server/webui
+
+      - name: Run linting
+        run: npm run lint
+        working-directory: tools/server/webui
+
+  webui-build:
+    needs: webui-check
+    name: WebUI Build
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "22"
+
+      - name: Restore node_modules cache
+        uses: actions/cache@v4
+        with:
+          path: tools/server/webui/node_modules
+          key: ${{ runner.os }}-node-modules-${{ hashFiles('tools/server/webui/package-lock.json') }}
+          restore-keys: |
+            ${{ runner.os }}-node-modules-
+
+      - name: Build application
+        run: npm run build
+        working-directory: tools/server/webui
+
+  webui-tests:
+    needs: webui-build
+    name: Run WebUI tests
+    permissions:
+      contents: read
+
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "22"
+
+      - name: Restore node_modules cache
+        uses: actions/cache@v4
+        with:
+          path: tools/server/webui/node_modules
+          key: ${{ runner.os }}-node-modules-${{ hashFiles('tools/server/webui/package-lock.json') }}
+          restore-keys: |
+            ${{ runner.os }}-node-modules-
+
+      - name: Install Playwright browsers
+        run: npx playwright install --with-deps
+        working-directory: tools/server/webui
+
+      - name: Build Storybook
+        run: npm run build-storybook
+        working-directory: tools/server/webui
+
+      - name: Run Client tests
+        run: npm run test:client
+        working-directory: tools/server/webui
+
+      - name: Run Server tests
+        run: npm run test:server
+        working-directory: tools/server/webui
+
+      - name: Run UI tests
+        run: npm run test:ui -- --testTimeout=60000
+        working-directory: tools/server/webui
+
+      - name: Run E2E tests
+        run: npm run test:e2e
+        working-directory: tools/server/webui
+
+  server-build:
+    needs: [webui-tests]
+    runs-on: ubuntu-latest
+
+    strategy:
+      matrix:
+        sanitizer: [ADDRESS, UNDEFINED] # THREAD is broken
+        build_type: [RelWithDebInfo]
+        include:
+          - build_type: Release
+            sanitizer: ""
+      fail-fast: false # While -DLLAMA_SANITIZE_THREAD=ON is broken
+
+    steps:
+      - name: Dependencies
+        id: depends
+        run: |
+          sudo apt-get update
+          sudo apt-get -y install \
+            build-essential \
+            xxd \
+            git \
+            cmake \
+            curl \
+            wget \
+            language-pack-en \
+            libssl-dev
+
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
+
+      - name: Python setup
+        id: setup_python
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.11'
+
+      - name: Tests dependencies
+        id: test_dependencies
+        run: |
+          pip install -r tools/server/tests/requirements.txt
+
+      - name: Setup Node.js for WebUI
+        uses: actions/setup-node@v4
+        with:
+          node-version: "22"
+          cache: "npm"
+          cache-dependency-path: "tools/server/webui/package-lock.json"
+
+      - name: Install WebUI dependencies
+        run: npm ci
+        working-directory: tools/server/webui
+
+      - name: Build WebUI
+        run: npm run build
+        working-directory: tools/server/webui
+
+      - name: Build (no OpenMP)
+        id: cmake_build_no_openmp
+        if: ${{ matrix.sanitizer == 'THREAD' }}
+        run: |
+          cmake -B build \
+              -DGGML_NATIVE=OFF \
+              -DLLAMA_CURL=OFF \
+              -DLLAMA_OPENSSL=ON \
+              -DLLAMA_BUILD_SERVER=ON \
+              -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
+              -DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON \
+              -DGGML_OPENMP=OFF ;
+          cmake --build build --config ${{ matrix.build_type }} -j $(nproc) --target llama-server
+
+      - name: Build (sanitizers)
+        id: cmake_build_sanitizers
+        if: ${{ matrix.sanitizer != '' && matrix.sanitizer != 'THREAD' }}
+        run: |
+          cmake -B build \
+              -DGGML_NATIVE=OFF \
+              -DLLAMA_CURL=OFF \
+              -DLLAMA_OPENSSL=ON \
+              -DLLAMA_BUILD_SERVER=ON \
+              -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
+              -DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON ;
+          cmake --build build --config ${{ matrix.build_type }} -j $(nproc) --target llama-server
+
+      - name: Build (sanitizers)
+        id: cmake_build
+        if: ${{ matrix.sanitizer == '' }}
+        run: |
+          cmake -B build \
+              -DGGML_NATIVE=OFF \
+              -DLLAMA_CURL=OFF \
+              -DLLAMA_OPENSSL=ON \
+              -DLLAMA_BUILD_SERVER=ON \
+              -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} ;
+          cmake --build build --config ${{ matrix.build_type }} -j $(nproc) --target llama-server
+
+      - name: Tests
+        id: server_integration_tests
+        if: ${{ matrix.sanitizer == '' }}
+        env:
+          GITHUB_ACTIONS: "true"
+        run: |
+          cd tools/server/tests
+          ./tests.sh
+
+      - name: Tests (sanitizers)
+        id: server_integration_tests_sanitizers
+        if: ${{ matrix.sanitizer != '' }}
+        run: |
+          cd tools/server/tests
+          LLAMA_SANITIZE=1 ./tests.sh
+
+      - name: Slow tests
+        id: server_integration_tests_slow
+        if: ${{ (github.event.schedule || github.event.inputs.slow_tests == 'true') && matrix.build_type == 'Release' }}
+        run: |
+          cd tools/server/tests
+          SLOW_TESTS=1 ./tests.sh

.github/workflows/server.yml

@@ -76,270 +76,6 @@ jobs:
         run: |
           pip install -r tools/server/tests/requirements.txt
 
-  webui-setup:
-    name: WebUI Setup
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
-
-      - name: Setup Node.js
-        uses: actions/setup-node@v4
-        with:
-          node-version: "22"
-          cache: "npm"
-          cache-dependency-path: "tools/server/webui/package-lock.json"
-
-      - name: Cache node_modules
-        uses: actions/cache@v4
-        id: cache-node-modules
-        with:
-          path: tools/server/webui/node_modules
-          key: ${{ runner.os }}-node-modules-${{ hashFiles('tools/server/webui/package-lock.json') }}
-          restore-keys: |
-            ${{ runner.os }}-node-modules-
-
-      - name: Install dependencies
-        if: steps.cache-node-modules.outputs.cache-hit != 'true'
-        run: npm ci
-        working-directory: tools/server/webui
-
-  webui-check:
-    needs: webui-setup
-    name: WebUI Check
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
-
-      - name: Setup Node.js
-        uses: actions/setup-node@v4
-        with:
-          node-version: "22"
-
-      - name: Restore node_modules cache
-        uses: actions/cache@v4
-        with:
-          path: tools/server/webui/node_modules
-          key: ${{ runner.os }}-node-modules-${{ hashFiles('tools/server/webui/package-lock.json') }}
-          restore-keys: |
-            ${{ runner.os }}-node-modules-
-
-      - name: Run type checking
-        run: npm run check
-        working-directory: tools/server/webui
-
-      - name: Run linting
-        run: npm run lint
-        working-directory: tools/server/webui
-
-  webui-build:
-    needs: webui-check
-    name: WebUI Build
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
-
-      - name: Setup Node.js
-        uses: actions/setup-node@v4
-        with:
-          node-version: "22"
-
-      - name: Restore node_modules cache
-        uses: actions/cache@v4
-        with:
-          path: tools/server/webui/node_modules
-          key: ${{ runner.os }}-node-modules-${{ hashFiles('tools/server/webui/package-lock.json') }}
-          restore-keys: |
-            ${{ runner.os }}-node-modules-
-
-      - name: Build application
-        run: npm run build
-        working-directory: tools/server/webui
-
-  webui-tests:
-    needs: webui-build
-    name: Run WebUI tests
-    permissions:
-      contents: read
-
-    runs-on: ubuntu-latest
-
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v4
-
-      - name: Setup Node.js
-        uses: actions/setup-node@v4
-        with:
-          node-version: "22"
-
-      - name: Restore node_modules cache
-        uses: actions/cache@v4
-        with:
-          path: tools/server/webui/node_modules
-          key: ${{ runner.os }}-node-modules-${{ hashFiles('tools/server/webui/package-lock.json') }}
-          restore-keys: |
-            ${{ runner.os }}-node-modules-
-
-      - name: Install Playwright browsers
-        run: npx playwright install --with-deps
-        working-directory: tools/server/webui
-
-      - name: Build Storybook
-        run: npm run build-storybook
-        working-directory: tools/server/webui
-
-      - name: Run Client tests
-        run: npm run test:client
-        working-directory: tools/server/webui
-
-      - name: Run Server tests
-        run: npm run test:server
-        working-directory: tools/server/webui
-
-      - name: Run UI tests
-        run: npm run test:ui -- --testTimeout=60000
-        working-directory: tools/server/webui
-
-      - name: Run E2E tests
-        run: npm run test:e2e
-        working-directory: tools/server/webui
-
-  server-build:
-    needs: [webui-tests]
-    runs-on: ubuntu-latest
-
-    strategy:
-      matrix:
-        sanitizer: [ADDRESS, UNDEFINED] # THREAD is broken
-        build_type: [RelWithDebInfo]
-        include:
-          - build_type: Release
-            sanitizer: ""
-      fail-fast: false # While -DLLAMA_SANITIZE_THREAD=ON is broken
-
-    steps:
-      - name: Dependencies
-        id: depends
-        run: |
-          sudo apt-get update
-          sudo apt-get -y install \
-            build-essential \
-            xxd \
-            git \
-            cmake \
-            curl \
-            wget \
-            language-pack-en \
-            libssl-dev
-
-      - name: Clone
-        id: checkout
-        uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-          ref: ${{ github.event.inputs.sha || github.event.pull_request.head.sha || github.sha || github.head_ref || github.ref_name }}
-
-      - name: Python setup
-        id: setup_python
-        uses: actions/setup-python@v5
-        with:
-          python-version: '3.11'
-
-      - name: Tests dependencies
-        id: test_dependencies
-        run: |
-          pip install -r tools/server/tests/requirements.txt
-
-      - name: Setup Node.js for WebUI
-        uses: actions/setup-node@v4
-        with:
-          node-version: "22"
-          cache: "npm"
-          cache-dependency-path: "tools/server/webui/package-lock.json"
-
-      - name: Install WebUI dependencies
-        run: npm ci
-        working-directory: tools/server/webui
-
-      - name: Build WebUI
-        run: npm run build
-        working-directory: tools/server/webui
-
-      - name: Build (no OpenMP)
-        id: cmake_build_no_openmp
-        if: ${{ matrix.sanitizer == 'THREAD' }}
-        run: |
-          cmake -B build \
-              -DGGML_NATIVE=OFF \
-              -DLLAMA_CURL=OFF \
-              -DLLAMA_OPENSSL=ON \
-              -DLLAMA_BUILD_SERVER=ON \
-              -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
-              -DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON \
-              -DGGML_OPENMP=OFF ;
-          cmake --build build --config ${{ matrix.build_type }} -j $(nproc) --target llama-server
-
-      - name: Build (sanitizers)
-        id: cmake_build_sanitizers
-        if: ${{ matrix.sanitizer != '' && matrix.sanitizer != 'THREAD' }}
-        run: |
-          cmake -B build \
-              -DGGML_NATIVE=OFF \
-              -DLLAMA_CURL=OFF \
-              -DLLAMA_OPENSSL=ON \
-              -DLLAMA_BUILD_SERVER=ON \
-              -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
-              -DLLAMA_SANITIZE_${{ matrix.sanitizer }}=ON ;
-          cmake --build build --config ${{ matrix.build_type }} -j $(nproc) --target llama-server
-
-      - name: Build (sanitizers)
-        id: cmake_build
-        if: ${{ matrix.sanitizer == '' }}
-        run: |
-          cmake -B build \
-              -DGGML_NATIVE=OFF \
-              -DLLAMA_CURL=OFF \
-              -DLLAMA_OPENSSL=ON \
-              -DLLAMA_BUILD_SERVER=ON \
-              -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} ;
-          cmake --build build --config ${{ matrix.build_type }} -j $(nproc) --target llama-server
-
-      - name: Tests
-        id: server_integration_tests
-        if: ${{ matrix.sanitizer == '' }}
-        env:
-          GITHUB_ACTIONS: "true"
-        run: |
-          cd tools/server/tests
-          ./tests.sh
-
-      - name: Tests (sanitizers)
-        id: server_integration_tests_sanitizers
-        if: ${{ matrix.sanitizer != '' }}
-        run: |
-          cd tools/server/tests
-          LLAMA_SANITIZE=1 ./tests.sh
-
-      - name: Slow tests
-        id: server_integration_tests_slow
-        if: ${{ (github.event.schedule || github.event.inputs.slow_tests == 'true') && matrix.build_type == 'Release' }}
-        run: |
-          cd tools/server/tests
-          SLOW_TESTS=1 ./tests.sh
-
-
   server-windows:
     runs-on: windows-2022
 

SECURITY.md

@@ -68,3 +68,6 @@ Please disclose it as a private [security advisory](https://github.com/ggml-org/
 Please note that using AI to identify vulnerabilities and generate reports is permitted. However, you must (1) explicitly disclose how AI was used and (2) conduct a thorough manual review before submitting the report.
 
 A team of volunteers on a reasonable-effort basis maintains this project. As such, please give us at least 90 days to work on a fix before public exposure.
+
+> [!IMPORTANT]
+> For collaborators: if you are interested in helping out with reviewing privting security disclosures, please see: https://github.com/ggml-org/llama.cpp/discussions/18080

convert_hf_to_gguf.py

@@ -136,19 +136,11 @@ class ModelBase:
         self.remote_hf_model_id = remote_hf_model_id
         self.sentence_transformers_dense_modules = sentence_transformers_dense_modules
         self.hparams = ModelBase.load_hparams(self.dir_model, self.is_mistral_format) if hparams is None else hparams
-        self.rope_parameters = self.hparams.get("rope_parameters", self.hparams.get("rope_scaling")) or {}
         self.model_tensors = self.index_tensors(remote_hf_model_id=remote_hf_model_id)
         self.metadata_override = metadata_override
         self.model_name = model_name
         self.dir_model_card = dir_model  # overridden in convert_lora_to_gguf.py
 
-        # Ensure "rope_theta" and "rope_type" is mirrored in rope_parameters
-        if "full_attention" not in self.rope_parameters and "sliding_attention" not in self.rope_parameters:
-            if "rope_theta" not in self.rope_parameters and (rope_theta := self.find_hparam(["rope_theta", "global_rope_theta", "rotary_emb_base"], optional=True)) is not None:
-                self.rope_parameters["rope_theta"] = rope_theta
-            if "rope_type" not in self.rope_parameters and (rope_type := self.rope_parameters.get("type")) is not None:
-                self.rope_parameters["rope_type"] = rope_type
-
         # Apply heuristics to figure out typical tensor encoding based on first layer tensor encoding type
         if self.ftype == gguf.LlamaFileType.GUESSED:
             # NOTE: can't use field "torch_dtype" in config.json, because some finetunes lie.
@@ -765,6 +757,15 @@ class TextModel(ModelBase):
         self.block_count = self.find_hparam(["n_layers", "num_hidden_layers", "n_layer", "num_layers"])
         self.tensor_map = gguf.get_tensor_name_map(self.model_arch, self.block_count)
 
+        self.rope_parameters = self.hparams.get("rope_parameters", self.hparams.get("rope_scaling")) or {}
+
+        # Ensure "rope_theta" and "rope_type" is mirrored in rope_parameters
+        if "full_attention" not in self.rope_parameters and "sliding_attention" not in self.rope_parameters:
+            if "rope_theta" not in self.rope_parameters and (rope_theta := self.find_hparam(["rope_theta", "global_rope_theta", "rotary_emb_base"], optional=True)) is not None:
+                self.rope_parameters["rope_theta"] = rope_theta
+            if "rope_type" not in self.rope_parameters and (rope_type := self.rope_parameters.get("type")) is not None:
+                self.rope_parameters["rope_type"] = rope_type
+
     @classmethod
     def __init_subclass__(cls):
         # can't use an abstract property, because overriding it without type errors
@@ -8489,8 +8490,18 @@ class GraniteHybridModel(Mamba2Model, GraniteMoeModel):
 class NemotronHModel(GraniteHybridModel):
     """Hybrid mamba2/attention model from NVIDIA"""
     model_arch = gguf.MODEL_ARCH.NEMOTRON_H
+    is_moe: bool = False
 
     def __init__(self, *args, **kwargs):
+        # We have to determine the correct model architecture (MoE vs non-MoE) before
+        # calling the parent __init__. This is because the parent constructor
+        # uses self.model_arch to build the tensor name map, and all MoE-specific
+        # mappings would be missed if it were called with the default non-MoE arch.
+        hparams = ModelBase.load_hparams(args[0], self.is_mistral_format)
+        if "num_experts_per_tok" in hparams:
+            self.model_arch = gguf.MODEL_ARCH.NEMOTRON_H_MOE
+            self.is_moe = True
+
         super().__init__(*args, **kwargs)
 
         # Save the top-level head_dim for later
@@ -8502,9 +8513,11 @@ class NemotronHModel(GraniteHybridModel):
 
         # Update the ssm / attn / mlp layers
         # M: Mamba2, *: Attention, -: MLP
+        # MoE:
+        # M: Mamba2, *: Attention, E: Expert
         hybrid_override_pattern = self.hparams["hybrid_override_pattern"]
         self._ssm_layers = [i for i, val in enumerate(hybrid_override_pattern) if val == "M"]
-        self._mlp_layers = [i for i, val in enumerate(hybrid_override_pattern) if val == "-"]
+        self._mlp_layers = [i for i, val in enumerate(hybrid_override_pattern) if val == ("E" if self.is_moe else "-")]
 
     def get_attn_layers(self):
         hybrid_override_pattern = self.hparams["hybrid_override_pattern"]
@@ -8520,10 +8533,28 @@ class NemotronHModel(GraniteHybridModel):
         # Set feed_forward_length
         # NOTE: This will trigger an override warning. This is preferrable to
         #   duplicating all the parent logic
-        n_ff = self.find_hparam(["intermediate_size", "n_inner", "hidden_dim"])
-        self.gguf_writer.add_feed_forward_length([
-            n_ff if i in self._mlp_layers else 0 for i in range(self.block_count)
-        ])
+        if not self.is_moe:
+            n_ff = self.find_hparam(["intermediate_size", "n_inner", "hidden_dim"])
+            self.gguf_writer.add_feed_forward_length([
+                n_ff if i in self._mlp_layers else 0 for i in range(self.block_count)
+            ])
+        else:
+            moe_intermediate_size = self.hparams["moe_intermediate_size"]
+            self.gguf_writer.add_feed_forward_length([
+                moe_intermediate_size if i in self._mlp_layers else 0 for i in range(self.block_count)
+            ])
+            self.gguf_writer.add_expert_used_count(self.hparams["num_experts_per_tok"])
+            self.gguf_writer.add_expert_feed_forward_length(self.hparams["moe_intermediate_size"])
+            self.gguf_writer.add_expert_shared_feed_forward_length(self.hparams["moe_shared_expert_intermediate_size"])
+            self.gguf_writer.add_expert_count(self.hparams["n_routed_experts"])
+            self.gguf_writer.add_expert_shared_count(self.hparams["n_shared_experts"])
+            self.gguf_writer.add_expert_weights_norm(self.hparams["norm_topk_prob"])
+            self.gguf_writer.add_expert_weights_scale(self.hparams["routed_scaling_factor"])
+            self.gguf_writer.add_expert_group_count(self.hparams["n_group"])
+
+            # number of experts used per token (top-k)
+            if (n_experts_used := self.hparams.get("num_experts_per_tok")) is not None:
+                self.gguf_writer.add_expert_used_count(n_experts_used)
 
     def set_vocab(self):
         super().set_vocab()
@@ -8531,7 +8562,81 @@ class NemotronHModel(GraniteHybridModel):
         # The tokenizer _does_ add a BOS token (via post_processor type
         # TemplateProcessing) but does not set add_bos_token to true in the
         # config, so we need to explicitly override it here.
-        self.gguf_writer.add_add_bos_token(True)
+        if not self.is_moe:
+            self.gguf_writer.add_add_bos_token(True)
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        if self.is_moe and bid is not None:
+            if name.endswith("mixer.gate.e_score_correction_bias"):
+                new_name = name.replace("e_score_correction_bias", "e_score_correction.bias")
+                mapped_name = self.map_tensor_name(new_name)
+                return [(mapped_name, data_torch)]
+
+            if name.endswith("mixer.dt_bias"):
+                new_name = name.replace("dt_bias", "dt.bias")
+                mapped_name = self.map_tensor_name(new_name)
+                return [(mapped_name, data_torch)]
+
+            if name.endswith("mixer.conv1d.weight"):
+                squeezed_data = data_torch.squeeze()
+                mapped_name = self.map_tensor_name(name)
+                return [(mapped_name, squeezed_data)]
+
+            if name.endswith("mixer.A_log"):
+                transformed_data = -torch.exp(data_torch)
+                reshaped_data = transformed_data.squeeze().reshape(-1, 1)
+                mapped_name = self.map_tensor_name(name)
+                return [(mapped_name, reshaped_data)]
+
+            if name.endswith("mixer.D"):
+                reshaped_data = data_torch.squeeze().reshape(-1, 1)
+                mapped_name = self.map_tensor_name(name)
+                return [(mapped_name, reshaped_data)]
+
+            if name.endswith("mixer.norm.weight"):
+                reshaped_data = data_torch.reshape(8, 512)
+                mapped_name = self.map_tensor_name(name)
+                return [(mapped_name, reshaped_data)]
+
+            if name.find("mixer.experts") != -1:
+                n_experts = self.hparams["n_routed_experts"]
+                assert bid is not None
+
+                if self._experts is None:
+                    self._experts = [{} for _ in range(self.block_count)]
+
+                self._experts[bid][name] = data_torch
+
+                if len(self._experts[bid]) >= n_experts * 2:
+                    # merge the experts into a single tensor
+                    tensors: list[tuple[str, Tensor]] = []
+                    for w_name in ["down_proj", "up_proj"]:
+                        datas: list[Tensor] = []
+
+                        for xid in range(n_experts):
+                            ename = f"backbone.layers.{bid}.mixer.experts.{xid}.{w_name}.weight"
+                            datas.append(self._experts[bid][ename])
+                            del self._experts[bid][ename]
+
+                        data_torch = torch.stack(datas, dim=0)
+                        merged_name = f"model.layers.{bid}.mlp.experts.{w_name}.weight"
+                        new_name = self.map_tensor_name(merged_name)
+                        tensors.append((new_name, data_torch))
+
+                    return tensors
+                else:
+                    return []
+
+        return super().modify_tensors(data_torch, name, bid)
+
+    def prepare_tensors(self):
+        super().prepare_tensors()
+
+        if self._experts is not None:
+            # flatten `list[dict[str, Tensor]]` into `list[str]`
+            experts = [k for d in self._experts for k in d.keys()]
+            if len(experts) > 0:
+                raise ValueError(f"Unprocessed experts: {experts}")
 
 
 @ModelBase.register("BailingMoeForCausalLM")

ggml/src/ggml-hexagon/ggml-hexagon.cpp

@@ -1976,9 +1976,6 @@ static bool ggml_hexagon_supported_mul_mat(const struct ggml_hexagon_session * s
             break;
 
         case GGML_TYPE_F16:
-            if (!opt_experimental) {
-                return false;
-            }
             break;
 
         default:

ggml/src/ggml-hexagon/htp/matmul-ops.c

@@ -903,7 +903,7 @@ static void vec_dot_f16_f32(const int n, float * restrict s, const void * restri
         const float * restrict vy  = (const float * restrict) y;
 
         for (uint32_t i = 0; i < n; i++) {
-            rsum += vx[i] * (__fp16) vy[i];
+            rsum += (float)vx[i] * vy[i];
         }
         *s = rsum;
         return;
@@ -917,7 +917,7 @@ static void vec_dot_f16_f32(const int n, float * restrict s, const void * restri
 
     // for some reason we need volatile here so that the compiler doesn't try anything funky
     volatile HVX_Vector rsum = Q6_V_vsplat_R(0);
-
+    float r_sum_scalar = 0.0f;
     uint32_t i = 0;
 
     for (i = 0; i < nv0; i++) {
@@ -926,31 +926,42 @@ static void vec_dot_f16_f32(const int n, float * restrict s, const void * restri
         HVX_Vector     x  = vx[i];
         HVX_VectorPair xp = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(x), Q6_Vh_vsplat_R(0x3C00));  // mul by 1.0
 
-        HVX_Vector hi = Q6_Vqf32_vmpy_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_V_hi_W(xp)), Q6_V_hi_W(yp));
-        HVX_Vector lo = Q6_Vqf32_vmpy_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_V_lo_W(xp)), Q6_V_lo_W(yp));
+        //NOTE: need volatile here to prevent compiler optimization
+        // Seem compiler cannot guarantee read-after-write??
+        volatile HVX_Vector hi = Q6_Vqf32_vmpy_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_V_hi_W(xp)), Q6_V_hi_W(yp));
+        volatile HVX_Vector lo = Q6_Vqf32_vmpy_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_V_lo_W(xp)), Q6_V_lo_W(yp));
 
         HVX_Vector sum = Q6_Vqf32_vadd_Vqf32Vqf32(hi, lo);
         rsum           = Q6_Vqf32_vadd_Vqf32Vqf32(rsum, sum);
     }
 
     if (nv1) {
-        HVX_VectorPair yp = vy[i];
+        // HVX_VectorPair yp = vy[i];
 
-        HVX_Vector     x  = vx[i];
-        HVX_VectorPair xp = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(x), Q6_Vh_vsplat_R(0x3C00));  // mul by 1.0
+        // HVX_Vector     x  = vx[i];
+        // HVX_VectorPair xp = Q6_Wqf32_vmpy_VhfVhf(Q6_Vh_vshuff_Vh(x), Q6_Vh_vsplat_R(0x3C00));  // mul by 1.0
 
-        if (nv1 >= 32) {
-            HVX_Vector hi = Q6_Vqf32_vmpy_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_V_hi_W(xp)), Q6_V_hi_W(yp));
-            rsum          = Q6_Vqf32_vadd_Vqf32Vqf32(rsum, hi);
-            nv1 -= 32;
-        }
+        // if (nv1 >= 32) {
+        //     volatile HVX_Vector hi = Q6_Vqf32_vmpy_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_V_hi_W(xp)), Q6_V_hi_W(yp));
+        //     rsum          = Q6_Vqf32_vadd_Vqf32Vqf32(rsum, hi);
+        //     nv1 -= 32;
+        // }
 
+        // rsum = hvx_vec_qf32_reduce_sum(rsum);
+
+        // if (nv1) {
+        //     volatile HVX_Vector lo  = Q6_Vqf32_vmpy_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_V_lo_W(xp)), Q6_V_lo_W(yp));
+        //     HVX_Vector sum = hvx_vec_qf32_reduce_sum_n(lo, nv1);
+        //     rsum           = Q6_Vqf32_vadd_Vqf32Vqf32(rsum, sum);
+        // }
+
+        //process the remainder using scalar loop
         rsum = hvx_vec_qf32_reduce_sum(rsum);
+        const __fp16 * restrict sx = (const __fp16 * restrict) x;
+        const float * restrict sy  = (const float * restrict) y;
 
-        if (nv1) {
-            HVX_Vector lo  = Q6_Vqf32_vmpy_VsfVsf(Q6_Vsf_equals_Vqf32(Q6_V_lo_W(xp)), Q6_V_lo_W(yp));
-            HVX_Vector sum = hvx_vec_qf32_reduce_sum_n(lo, nv1);
-            rsum           = Q6_Vqf32_vadd_Vqf32Vqf32(rsum, sum);
+        for (uint32_t i = nv0 * 64; i < n; i++) {
+            r_sum_scalar += (float) sx[i] * sy[i];
         }
 
         // hvx_vec_dump_fp16("X", x);
@@ -961,7 +972,7 @@ static void vec_dot_f16_f32(const int n, float * restrict s, const void * restri
         rsum = hvx_vec_qf32_reduce_sum(rsum);
     }
 
-    *s = hvx_vec_get_fp32(Q6_Vsf_equals_Vqf32(rsum));
+    *s = hvx_vec_get_fp32(Q6_Vsf_equals_Vqf32(rsum)) + r_sum_scalar;
 
 #    ifdef HTP_DEBUG
     {
@@ -1498,9 +1509,6 @@ static void matmul_f16_f32(struct htp_tensor * restrict src0,
     uint64_t t1, t2;
     t1 = HAP_perf_get_qtimer_count();
 
-    const size_t src0_row_size = sizeof(__fp16) * ne00;
-    const size_t src1_row_size = sizeof(float) * ne10;
-
     assert(ne12 % ne02 == 0);
     assert(ne13 % ne03 == 0);
 
@@ -1510,8 +1518,6 @@ static void matmul_f16_f32(struct htp_tensor * restrict src0,
     // This is the size of the rest of the dimensions of the result
     const uint32_t nr1 = ne1 * ne2 * ne3;
 
-    uint32_t chunk_size = 64;
-
     // distribute the thread work across the inner or outer loop based on which one is larger
     uint32_t nchunk0 = nr0 > nr1 ? nth : 1;  // parallelize by src0 rows
     uint32_t nchunk1 = nr0 > nr1 ? 1 : nth;  // parallelize by src1 rows
@@ -1544,11 +1550,11 @@ static void matmul_f16_f32(struct htp_tensor * restrict src0,
     const uint32_t blck_0 = 64;
     const uint32_t blck_1 = 64;
 
-    float tmp[32];
+    __attribute__((aligned(128))) float tmp[64];
 
     for (uint32_t iir1 = ir1_start; iir1 < ir1_end; iir1 += blck_1) {
         for (uint32_t iir0 = ir0_start; iir0 < ir0_end; iir0 += blck_0) {
-            for (uint32_t ir1 = iir1; ir1 < iir1 + blck_1 && ir1 < ir1_end; ir1++) {
+            for (uint32_t ir1 = iir1; ir1 < MIN(iir1 + blck_1, ir1_end); ir1++) {
                 const uint32_t i13 = (ir1 / (ne12 * ne1));
                 const uint32_t i12 = (ir1 - i13 * ne12 * ne1) / ne1;
                 const uint32_t i11 = (ir1 - i13 * ne12 * ne1 - i12 * ne1);
@@ -1561,13 +1567,16 @@ static void matmul_f16_f32(struct htp_tensor * restrict src0,
                 const uint32_t i2 = i12;
                 const uint32_t i3 = i13;
 
-                const uint8_t * restrict src0_row = (const uint8_t *) src0->data + (0 + i02 * nb02 + i03 * nb03);
+                const uint8_t * restrict src0_base = (const uint8_t *) src0->data + (0 + i02 * nb02 + i03 * nb03);
                 const uint8_t * restrict src1_col =
-                    (const uint8_t *) src1->data + (i11 + i12 * ne11 + i13 * ne12 * ne11) * src1_row_size;
+                    (const uint8_t *) src1->data + (i11 * nb11 + i12 * nb12 + i13 * nb13);
                 float * dst_col = (float *) ((uint8_t * restrict) dst->data + (i1 * nb1 + i2 * nb2 + i3 * nb3));
 
-                for (uint32_t ir0 = iir0; ir0 < iir0 + blck_0 && ir0 < ir0_end; ir0++) {
-                    vec_dot_f16_f32(ne00, &tmp[ir0 - iir0], src0_row + ir0 * src0_row_size, src1_col);
+                const uint32_t ir0_block_end = MIN(iir0 + blck_0, ir0_end);
+                for (uint32_t ir0 = iir0; ir0 < ir0_block_end; ir0++) {
+                    // Use nb01 stride for non-contiguous src0 support
+                    const uint8_t * restrict src0_row = src0_base + ir0 * nb01;
+                    vec_dot_f16_f32(ne00, &tmp[ir0 - iir0], src0_row, src1_col);
                 }
 
                 hvx_copy_fp32_ua((uint8_t *) &dst_col[iir0], (uint8_t *) tmp, MIN(iir0 + blck_0, ir0_end) - iir0);

gguf-py/gguf/constants.py

@@ -413,6 +413,7 @@ class MODEL_ARCH(IntEnum):
     JAIS             = auto()
     NEMOTRON         = auto()
     NEMOTRON_H       = auto()
+    NEMOTRON_H_MOE   = auto()
     EXAONE           = auto()
     EXAONE4          = auto()
     GRANITE          = auto()
@@ -786,6 +787,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
     MODEL_ARCH.JAIS:             "jais",
     MODEL_ARCH.NEMOTRON:         "nemotron",
     MODEL_ARCH.NEMOTRON_H:       "nemotron_h",
+    MODEL_ARCH.NEMOTRON_H_MOE:   "nemotron_h_moe",
     MODEL_ARCH.EXAONE:           "exaone",
     MODEL_ARCH.EXAONE4:          "exaone4",
     MODEL_ARCH.GRANITE:          "granite",
@@ -2529,6 +2531,33 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.FFN_UP,
     ],
+    MODEL_ARCH.NEMOTRON_H_MOE: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.SSM_IN,
+        MODEL_TENSOR.SSM_CONV1D,
+        MODEL_TENSOR.SSM_DT,
+        MODEL_TENSOR.SSM_A,
+        MODEL_TENSOR.SSM_D,
+        MODEL_TENSOR.SSM_NORM,
+        MODEL_TENSOR.SSM_OUT,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+        # experts
+        MODEL_TENSOR.FFN_GATE_INP,
+        MODEL_TENSOR.FFN_UP_EXP,
+        MODEL_TENSOR.FFN_DOWN_EXP,
+        # shared expert
+        MODEL_TENSOR.FFN_DOWN_SHEXP,
+        MODEL_TENSOR.FFN_UP_SHEXP,
+        MODEL_TENSOR.FFN_EXP_PROBS_B,
+    ],
     MODEL_ARCH.EXAONE: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,

gguf-py/gguf/tensor_mapping.py

@@ -379,6 +379,7 @@ class TensorNameMap:
             "model.layers.{bid}.feed_forward.gate",               # lfm2moe
             "model.layers.{bid}.mlp.router.gate",               # afmoe
             "layers.{bid}.gate",                                # mistral-large
+            "backbone.layers.{bid}.mixer.gate",                 # nemotron-h-moe
         ),
 
         MODEL_TENSOR.FFN_GATE_INP_SHEXP: (
@@ -392,6 +393,7 @@ class TensorNameMap:
             "model.layers.{bid}.mlp.expert_bias",                           # afmoe
             "model.layers.{bid}.feed_forward.expert_bias",                  # lfm2moe
             "model.layers.{bid}.block_sparse_moe.e_score_correction",       # minimax-m2
+            "backbone.layers.{bid}.mixer.gate.e_score_correction"           # nemotron-h-moe
         ),
 
         # Feed-forward up
@@ -440,7 +442,7 @@ class TensorNameMap:
             "layers.{bid}.feed_forward.experts.w3",                 # mixtral (merged)
             "transformer.decoder_layer.{bid}.moe.linear_v",         # Grok (merged)
             "transformer.blocks.{bid}.ffn.experts.mlp.v1",          # dbrx
-            "model.layers.{bid}.mlp.experts.up_proj",               # qwen2moe olmoe (merged) ernie4.5-moe
+            "model.layers.{bid}.mlp.experts.up_proj",               # qwen2moe olmoe (merged) ernie4.5-moe, nemotron-h-moe (merged)
             "model.layers.{bid}.block_sparse_moe.experts.w3",       # phimoe (merged)
             "model.layers.{bid}.feed_forward.experts.up_proj",      # llama4
             "encoder.layers.{bid}.mlp.experts.mlp.w1",              # nomic-bert-moe
@@ -454,6 +456,7 @@ class TensorNameMap:
             "model.layers.{bid}.feed_forward.down_proj",
             "model.layers.{bid}.mlp.shared_mlp.up_proj",             # hunyuan
             "layers.{bid}.shared_experts.w3",                        # mistral-large
+            "backbone.layers.{bid}.mixer.shared_experts.up_proj",    # nemotron-h-moe
         ),
 
         MODEL_TENSOR.FFN_UP_CHEXP: (
@@ -548,7 +551,7 @@ class TensorNameMap:
             "layers.{bid}.feed_forward.experts.w2",                 # mixtral (merged)
             "transformer.decoder_layer.{bid}.moe.linear_1",         # Grok (merged)
             "transformer.blocks.{bid}.ffn.experts.mlp.w2",          # dbrx
-            "model.layers.{bid}.mlp.experts.down_proj",             # qwen2moe olmoe (merged) ernie4.5-moe
+            "model.layers.{bid}.mlp.experts.down_proj",             # qwen2moe olmoe (merged) ernie4.5-moe nemotron-h-moe (merged)
             "model.layers.{bid}.block_sparse_moe.output_linear",    # granitemoe
             "model.layers.{bid}.block_sparse_moe.experts.w2",       # phimoe (merged)
             "model.layers.{bid}.feed_forward.experts.down_proj",    # llama4
@@ -563,6 +566,7 @@ class TensorNameMap:
             "model.layers.{bid}.shared_mlp.output_linear",             # granitemoe
             "model.layers.{bid}.mlp.shared_mlp.down_proj",             # hunyuan
             "layers.{bid}.shared_experts.w2",                          # mistral-large
+            "backbone.layers.{bid}.mixer.shared_experts.down_proj",    # nemotron-h-moe
         ),
 
         MODEL_TENSOR.FFN_DOWN_CHEXP: (
@@ -706,6 +710,7 @@ class TensorNameMap:
             "model.layers.{bid}.mamba.dt_proj",         # jamba falcon-h1 granite-hybrid
             "model.layers.layers.{bid}.mixer.dt_proj",  # plamo2
             "model.layers.{bid}.linear_attn.dt_proj",   # qwen3next
+            "backbone.layers.{bid}.mixer.dt",           # nemotron-h-moe
         ),
 
         MODEL_TENSOR.SSM_DT_NORM: (

scripts/snapdragon/adb/run-mtmd.sh

@@ -0,0 +1,65 @@
+#!/bin/sh
+#
+
+# Basedir on device
+basedir=/data/local/tmp/llama.cpp
+
+cli_opts=
+
+branch=.
+[ "$B" != "" ] && branch=$B
+
+adbserial=
+[ "$S" != "" ] && adbserial="-s $S"
+
+model="gemma-3-4b-it-Q4_0.gguf"
+[ "$M" != "" ] && model="$M"
+
+mmproj="mmproj-F16.gguf"
+[ "$MMPROJ" != "" ] && mmproj="$MMPROJ"
+
+image=
+[ "$IMG" != "" ] && image="$IMG"
+
+device="HTP0"
+[ "$D" != "" ] && device="$D"
+
+verbose=
+[ "$V" != "" ] && verbose="GGML_HEXAGON_VERBOSE=$V"
+
+experimental="GGML_HEXAGON_EXPERIMENTAL=1"
+[ "$E" != "" ] && experimental="GGML_HEXAGON_EXPERIMENTAL=$E"
+
+sched=
+[ "$SCHED" != "" ] && sched="GGML_SCHED_DEBUG=2" cli_opts="$cli_opts -v"
+
+profile=
+[ "$PROF" != "" ] && profile="GGML_HEXAGON_PROFILE=$PROF GGML_HEXAGON_OPSYNC=1"
+
+opmask=
+[ "$OPMASK" != "" ] && opmask="GGML_HEXAGON_OPMASK=$OPMASK"
+
+nhvx=
+[ "$NHVX" != "" ] && nhvx="GGML_HEXAGON_NHVX=$NHVX"
+
+ndev=
+[ "$NDEV" != "" ] && ndev="GGML_HEXAGON_NDEV=$NDEV"
+
+# MTMD backend device for vision model (defaults to CPU if not set)
+mtmd_backend=
+[ "$MTMD_DEVICE" != "" ] && mtmd_backend="MTMD_BACKEND_DEVICE=$MTMD_DEVICE"
+
+set -x
+
+adb $adbserial shell " \
+  cd $basedir; ulimit -c unlimited;        \
+    LD_LIBRARY_PATH=$basedir/$branch/lib   \
+    ADSP_LIBRARY_PATH=$basedir/$branch/lib \
+    $verbose $experimental $sched $opmask $profile $nhvx $ndev $mtmd_backend       \
+      ./$branch/bin/llama-mtmd-cli --no-mmap -m $basedir/../gguf/$model   \
+         --mmproj $basedir/../gguf/$mmproj \
+         --image $basedir/../gguf/$image \
+         --poll 1000 -t 6 --cpu-mask 0xfc --cpu-strict 1             \
+         --ctx-size 8192 --batch-size 128 -ctk q8_0 -ctv q8_0 -fa on \
+         -ngl 99 --device $device -v $cli_opts $@ \
+"

src/llama-arch.cpp

@@ -75,6 +75,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
     { LLM_ARCH_JAIS,             "jais"             },
     { LLM_ARCH_NEMOTRON,         "nemotron"         },
     { LLM_ARCH_NEMOTRON_H,       "nemotron_h"       },
+    { LLM_ARCH_NEMOTRON_H_MOE,   "nemotron_h_moe"   },
     { LLM_ARCH_EXAONE,           "exaone"           },
     { LLM_ARCH_EXAONE4,          "exaone4"          },
     { LLM_ARCH_RWKV6,            "rwkv6"            },
@@ -1763,6 +1764,39 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
             { LLM_TENSOR_FFN_UP,         "blk.%d.ffn_up" },
         },
     },
+    {
+        LLM_ARCH_NEMOTRON_H_MOE,
+        {
+            { LLM_TENSOR_TOKEN_EMBD,     "token_embd" },
+            { LLM_TENSOR_OUTPUT_NORM,    "output_norm" },
+            { LLM_TENSOR_OUTPUT,         "output" },
+            { LLM_TENSOR_ATTN_NORM,      "blk.%d.attn_norm" },
+            // mamba(2) ssm layers
+            { LLM_TENSOR_SSM_IN,         "blk.%d.ssm_in" },
+            { LLM_TENSOR_SSM_CONV1D,     "blk.%d.ssm_conv1d" },
+            { LLM_TENSOR_SSM_DT,         "blk.%d.ssm_dt" },
+            { LLM_TENSOR_SSM_A,          "blk.%d.ssm_a" },
+            { LLM_TENSOR_SSM_D,          "blk.%d.ssm_d" },
+            { LLM_TENSOR_SSM_NORM,       "blk.%d.ssm_norm" },
+            { LLM_TENSOR_SSM_OUT,        "blk.%d.ssm_out" },
+            // attention layers
+            { LLM_TENSOR_ATTN_Q,         "blk.%d.attn_q" },
+            { LLM_TENSOR_ATTN_K,         "blk.%d.attn_k" },
+            { LLM_TENSOR_ATTN_V,         "blk.%d.attn_v" },
+            { LLM_TENSOR_ATTN_OUT,       "blk.%d.attn_output" },
+            // dense FFN
+            { LLM_TENSOR_FFN_DOWN,       "blk.%d.ffn_down" },
+            { LLM_TENSOR_FFN_UP,         "blk.%d.ffn_up" },
+            // MoE FFN (for MoE layers)
+            { LLM_TENSOR_FFN_GATE_INP,   "blk.%d.ffn_gate_inp" },
+            { LLM_TENSOR_FFN_UP_EXPS,    "blk.%d.ffn_up_exps" },
+            { LLM_TENSOR_FFN_DOWN_EXPS,  "blk.%d.ffn_down_exps" },
+            { LLM_TENSOR_FFN_EXP_PROBS_B,"blk.%d.exp_probs_b" },
+            // MoE shared expert layer
+            { LLM_TENSOR_FFN_DOWN_SHEXP,  "blk.%d.ffn_down_shexp" },
+            { LLM_TENSOR_FFN_UP_SHEXP,    "blk.%d.ffn_up_shexp" },
+        },
+    },
     {
         LLM_ARCH_EXAONE,
         {
@@ -2817,6 +2851,7 @@ bool llm_arch_is_hybrid(const llm_arch & arch) {
         case LLM_ARCH_LFM2:
         case LLM_ARCH_LFM2MOE:
         case LLM_ARCH_NEMOTRON_H:
+        case LLM_ARCH_NEMOTRON_H_MOE:
         case LLM_ARCH_QWEN3NEXT:
             return true;
         default:

src/llama-arch.h

@@ -79,6 +79,7 @@ enum llm_arch {
     LLM_ARCH_JAIS,
     LLM_ARCH_NEMOTRON,
     LLM_ARCH_NEMOTRON_H,
+    LLM_ARCH_NEMOTRON_H_MOE,
     LLM_ARCH_EXAONE,
     LLM_ARCH_EXAONE4,
     LLM_ARCH_RWKV6,

src/llama-graph.cpp

@@ -254,6 +254,24 @@ void llm_graph_input_rs::set_input(const llama_ubatch * ubatch) {
     }
 }
 
+bool llm_graph_input_rs::can_reuse(const llm_graph_params & params) {
+    const auto * mctx = static_cast<const llama_memory_recurrent_context *>(params.mctx);
+
+    this->mctx = mctx;
+
+    bool res = true;
+
+    res &= s_copy->ne[0] == mctx->get_n_rs();
+
+    res &= s_copy_main->ne[0]  == params.ubatch.n_seqs;
+    res &= s_copy_extra->ne[0] == mctx->get_n_rs() - params.ubatch.n_seqs;
+
+    res &= head == mctx->get_head();
+    res &= rs_z == mctx->get_rs_z();
+
+    return res;
+}
+
 void llm_graph_input_cross_embd::set_input(const llama_ubatch * ubatch) {
     GGML_UNUSED(ubatch);
 
@@ -461,8 +479,46 @@ void llm_graph_input_attn_cross::set_input(const llama_ubatch * ubatch) {
 }
 
 void llm_graph_input_mem_hybrid::set_input(const llama_ubatch * ubatch) {
-    inp_attn->set_input(ubatch);
-    inp_rs->set_input(ubatch);
+    mctx->get_attn()->set_input_k_idxs(inp_attn->self_k_idxs, ubatch);
+    mctx->get_attn()->set_input_v_idxs(inp_attn->self_v_idxs, ubatch);
+
+    mctx->get_attn()->set_input_kq_mask(inp_attn->self_kq_mask, ubatch, cparams.causal_attn);
+
+    const int64_t n_rs = mctx->get_recr()->get_n_rs();
+
+    if (inp_rs->s_copy) {
+        GGML_ASSERT(ggml_backend_buffer_is_host(inp_rs->s_copy->buffer));
+        int32_t * data = (int32_t *) inp_rs->s_copy->data;
+
+        // assuming copy destinations ALWAYS happen ONLY on the cells between head and head+n
+        for (uint32_t i = 0; i < n_rs; ++i) {
+            data[i] = mctx->get_recr()->s_copy(i);
+        }
+    }
+}
+
+bool llm_graph_input_mem_hybrid::can_reuse(const llm_graph_params & params) {
+    const auto * mctx = static_cast<const llama_memory_hybrid_context *>(params.mctx);
+
+    this->mctx = mctx;
+
+    bool res = true;
+
+    res &= inp_attn->self_k_idxs->ne[0] == params.ubatch.n_tokens;
+  //res &= inp_attn->self_v_idxs->ne[0] == params.ubatch.n_tokens; // TODO: need to move this to the unified cache and check there
+
+    res &= inp_attn->self_kq_mask->ne[0] == mctx->get_attn()->get_n_kv();
+    res &= inp_attn->self_kq_mask->ne[1] == params.ubatch.n_tokens;
+
+    res &= inp_rs->s_copy->ne[0] == mctx->get_recr()->get_n_rs();
+
+    res &= inp_rs->s_copy_main->ne[0]  == params.ubatch.n_seqs;
+    res &= inp_rs->s_copy_extra->ne[0] == mctx->get_recr()->get_n_rs() - params.ubatch.n_seqs;
+
+    res &= inp_rs->head == mctx->get_recr()->get_head();
+    res &= inp_rs->rs_z == mctx->get_recr()->get_rs_z();
+
+    return res;
 }
 
 //
@@ -1089,6 +1145,15 @@ ggml_tensor * llm_graph_context::build_moe_ffn(
                 cur = ggml_relu(ctx0, cur);
                 cb(cur, "ffn_moe_relu", il);
             } break;
+        case LLM_FFN_RELU_SQR:
+            if (gate_exps) {
+                // TODO: add support for gated squared relu
+                GGML_ABORT("fatal error: gated squared relu not implemented");
+            } else {
+                cur = ggml_relu(ctx0, cur);
+                cur = ggml_sqr(ctx0, cur);
+                cb(cur, "ffn_moe_relu_sqr", il);
+            } break;
         default:
             GGML_ABORT("fatal error");
     }
@@ -1841,6 +1906,9 @@ static std::unique_ptr<llm_graph_input_rs> build_rs_inp_impl(
     inp->s_copy_main  = ggml_view_1d(ctx0, inp->s_copy, n_seqs, 0);
     inp->s_copy_extra = ggml_view_1d(ctx0, inp->s_copy, n_rs - n_seqs, n_seqs * inp->s_copy->nb[0]);
 
+    inp->head = mctx_cur->get_head();
+    inp->rs_z = mctx_cur->get_rs_z();
+
     return inp;
 }
 
@@ -1909,10 +1977,10 @@ ggml_tensor * llm_graph_context::build_rwkv_token_shift_store(
 llm_graph_input_mem_hybrid * llm_graph_context::build_inp_mem_hybrid() const {
     const auto * mctx_cur = static_cast<const llama_memory_hybrid_context *>(mctx);
 
-    auto inp_rs   = build_rs_inp_impl(ctx0, ubatch, mctx_cur->get_recr());
+    auto inp_rs   = build_rs_inp_impl     (ctx0, ubatch, mctx_cur->get_recr());
     auto inp_attn = build_attn_inp_kv_impl(ctx0, ubatch, hparams, cparams, mctx_cur->get_attn());
 
-    auto inp = std::make_unique<llm_graph_input_mem_hybrid>(std::move(inp_attn), std::move(inp_rs), mctx_cur);
+    auto inp = std::make_unique<llm_graph_input_mem_hybrid>(cparams, std::move(inp_attn), std::move(inp_rs), mctx_cur);
 
     return (llm_graph_input_mem_hybrid *) res->add_input(std::move(inp));
 }

src/llama-graph.h

@@ -225,6 +225,8 @@ public:
 
     void set_input(const llama_ubatch * ubatch) override;
 
+    bool can_reuse(const llm_graph_params & params) override;
+
     ggml_tensor * s_copy;  // I32 [n_rs]
 
     // views of s_copy, computed once per graph
@@ -233,6 +235,10 @@ public:
     ggml_tensor * s_copy_extra;  // I32 [n_rs - n_seqs]
 
     const llama_memory_recurrent_context * mctx;
+
+    // used in view offsets, need to match for valid graph reuse
+    uint32_t head;
+    int32_t rs_z;
 };
 
 class llm_graph_input_cross_embd : public llm_graph_input_i {
@@ -365,22 +371,28 @@ public:
 class llm_graph_input_mem_hybrid : public llm_graph_input_i {
 public:
     llm_graph_input_mem_hybrid(
+            const llama_cparams & cparams,
             std::unique_ptr<llm_graph_input_attn_kv> inp_attn,
-            std::unique_ptr<llm_graph_input_rs>              inp_rs,
-            const llama_memory_hybrid_context *              mctx) :
+            std::unique_ptr<llm_graph_input_rs>      inp_rs,
+            const llama_memory_hybrid_context *      mctx) :
         inp_attn(std::move(inp_attn)),
         inp_rs(std::move(inp_rs)),
+        cparams(cparams),
         mctx(mctx) { }
     virtual ~llm_graph_input_mem_hybrid() = default;
 
     void set_input(const llama_ubatch * ubatch) override;
 
+    bool can_reuse(const llm_graph_params & params) override;
+
     std::unique_ptr<llm_graph_input_attn_kv> inp_attn;
     std::unique_ptr<llm_graph_input_rs>      inp_rs;
 
     llm_graph_input_attn_kv * get_attn() const { return inp_attn.get(); }
     llm_graph_input_rs      * get_recr() const { return inp_rs.get(); }
 
+    const llama_cparams cparams;
+
     const llama_memory_hybrid_context * mctx;
 };
 

src/llama-hparams.cpp

@@ -2,6 +2,7 @@
 
 #include "ggml.h"
 
+#include <algorithm>
 #include <cassert>
 
 void llama_hparams::set_swa_pattern(uint32_t n_pattern, bool dense_first) {

src/llama-memory-hybrid.cpp

@@ -222,7 +222,7 @@ llama_memory_hybrid_context::llama_memory_hybrid_context(
     ubatches(std::move(ubatches)),
     // note: here we copy the ubatches. not sure if this is ideal
     ctx_attn(new llama_kv_cache_context(mem->get_mem_attn(), std::move(sinfos_attn), this->ubatches)),
-    ctx_recr(new llama_memory_recurrent_context(mem->get_mem_recr(),                        this->ubatches)),
+    ctx_recr(new llama_memory_recurrent_context(mem->get_mem_recr(), this->ubatches)),
     status(llama_memory_status_combine(ctx_attn->get_status(), ctx_recr->get_status())) {
 }
 

src/llama-model.cpp

@@ -120,6 +120,7 @@ const char * llm_type_name(llm_type type) {
         case LLM_TYPE_16B_A1B:       return "16B.A1B";
         case LLM_TYPE_21B_A3B:       return "21B.A3B";
         case LLM_TYPE_30B_A3B:       return "30B.A3B";
+        case LLM_TYPE_31B_A3_5B:     return "31B.A3.5B";
         case LLM_TYPE_80B_A3B:       return "80B.A3B";
         case LLM_TYPE_100B_A6B:      return "100B.A6B";
         case LLM_TYPE_106B_A12B:     return "106B.A12B";
@@ -1797,6 +1798,7 @@ void llama_model::load_hparams(llama_model_loader & ml) {
                 }
             } break;
         case LLM_ARCH_NEMOTRON_H:
+        case LLM_ARCH_NEMOTRON_H_MOE:
             {
                 ml.get_key(LLM_KV_SSM_CONV_KERNEL,    hparams.ssm_d_conv);
                 ml.get_key(LLM_KV_SSM_INNER_SIZE,     hparams.ssm_d_inner);
@@ -1812,7 +1814,14 @@ void llama_model::load_hparams(llama_model_loader & ml) {
 
                 ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
 
+                ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,        hparams.n_ff_exp,        false);
+                ml.get_key(LLM_KV_EXPERT_SHARED_FEED_FORWARD_LENGTH, hparams.n_ff_shexp,      false);
+                ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,               hparams.n_expert_shared, false);
+                ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,               hparams.expert_weights_norm, false);
+                ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,              hparams.expert_weights_scale, false);
+
                 switch (hparams.n_layer) {
+                    case 52: type = LLM_TYPE_31B_A3_5B; break; // Nemotron-H_MOE 31B
                     case 56: type = LLM_TYPE_9B; break;
                     default: type = LLM_TYPE_UNKNOWN;
                 }
@@ -5159,6 +5168,7 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                     }
                 } break;
             case LLM_ARCH_NEMOTRON_H:
+            case LLM_ARCH_NEMOTRON_H_MOE:
                 {
                     // mamba2 Mixer SSM params
                     // NOTE: int64_t for tensor dimensions
@@ -5169,6 +5179,9 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                     const int64_t n_group    = hparams.ssm_n_group;
                     const int64_t d_in_proj  = 2*d_inner + 2*n_group*d_state + n_ssm_head;
 
+                    const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used;
+                    const int64_t n_ff_shexp = hparams.n_ff_shexp;
+
                     // embeddings
                     tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
 
@@ -5218,12 +5231,26 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                             layer.bk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "bias",   i), {n_embd_k_gqa_i}, TENSOR_NOT_REQUIRED);
                             layer.bv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "bias",   i), {n_embd_v_gqa_i}, TENSOR_NOT_REQUIRED);
                             layer.bo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "bias",   i), {n_embd},         TENSOR_NOT_REQUIRED);
-                        } else {
-                            // mlp layers
-                            layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  hparams.n_ff(i), n_embd}, 0);
-                            layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   hparams.n_ff(i)}, 0);
-                            layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
-                            layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias",   i), {hparams.n_ff(i)}, TENSOR_NOT_REQUIRED);
+                        }  else {
+                            if (n_expert != 0) {
+                                layer.ffn_gate_inp    = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP,  "weight", i), { n_embd, n_expert}, 0);
+                                layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert         }, 0);
+
+                                // MoE branch
+                                layer.ffn_down_exps   = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp,   n_embd, n_expert}, 0);
+                                layer.ffn_up_exps     = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {  n_embd, n_ff_exp, n_expert}, 0);
+
+                                // Shared expert branch
+                                layer.ffn_down_shexp  = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, 0);
+                                layer.ffn_up_shexp    = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp}, 0);
+
+                            } else {
+                                // mlp layers
+                                layer.ffn_down   = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {  hparams.n_ff(i), n_embd}, 0);
+                                layer.ffn_up     = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   hparams.n_ff(i)}, 0);
+                                layer.ffn_down_b = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "bias",   i), {n_embd}, TENSOR_NOT_REQUIRED);
+                                layer.ffn_up_b   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "bias",   i), {hparams.n_ff(i)}, TENSOR_NOT_REQUIRED);
+                            }
                         }
                     }
                 } break;
@@ -6850,7 +6877,8 @@ void llama_model::print_info() const {
         arch == LLM_ARCH_PLAMO2 ||
         arch == LLM_ARCH_GRANITE_HYBRID ||
         arch == LLM_ARCH_QWEN3NEXT ||
-        arch == LLM_ARCH_NEMOTRON_H) {
+        arch == LLM_ARCH_NEMOTRON_H ||
+        arch == LLM_ARCH_NEMOTRON_H_MOE) {
         LLAMA_LOG_INFO("%s: ssm_d_conv       = %u\n",     __func__, hparams.ssm_d_conv);
         LLAMA_LOG_INFO("%s: ssm_d_inner      = %u\n",     __func__, hparams.ssm_d_inner);
         LLAMA_LOG_INFO("%s: ssm_d_state      = %u\n",     __func__, hparams.ssm_d_state);
@@ -6905,7 +6933,8 @@ void llama_model::print_info() const {
     if (arch == LLM_ARCH_MINICPM ||
         arch == LLM_ARCH_GRANITE ||
         arch == LLM_ARCH_GRANITE_MOE ||
-        arch == LLM_ARCH_GRANITE_HYBRID) {
+        arch == LLM_ARCH_GRANITE_HYBRID ||
+        arch == LLM_ARCH_NEMOTRON_H_MOE) {
         LLAMA_LOG_INFO("%s: f_embedding_scale = %f\n", __func__, hparams.f_embedding_scale);
         LLAMA_LOG_INFO("%s: f_residual_scale  = %f\n", __func__, hparams.f_residual_scale);
         LLAMA_LOG_INFO("%s: f_attention_scale = %f\n", __func__, hparams.f_attention_scale);
@@ -7086,7 +7115,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
                     if (arch == LLM_ARCH_FALCON_H1) {
                         filter_attn = [&](int32_t) { return true; };
                         filter_recr = [&](int32_t) { return true; };
-                    } else if (arch == LLM_ARCH_NEMOTRON_H) {
+                    } else if (arch == LLM_ARCH_NEMOTRON_H || arch == LLM_ARCH_NEMOTRON_H_MOE) {
                         filter_attn = [&](int32_t il) {
                             return !hparams.is_recurrent(il) && hparams.n_ff(il) == 0;
                         };
@@ -7457,6 +7486,7 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const {
                 llm = std::make_unique<llm_build_nemotron>(*this, params);
             } break;
         case LLM_ARCH_NEMOTRON_H:
+        case LLM_ARCH_NEMOTRON_H_MOE:
             {
                 llm = std::make_unique<llm_build_nemotron_h>(*this, params);
             } break;
@@ -7741,6 +7771,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
         case LLM_ARCH_ARWKV7:
         case LLM_ARCH_WAVTOKENIZER_DEC:
         case LLM_ARCH_NEMOTRON_H:
+        case LLM_ARCH_NEMOTRON_H_MOE:
             return LLAMA_ROPE_TYPE_NONE;
 
         // use what we call a normal RoPE, operating on pairs of consecutive head values

src/llama-model.h

@@ -113,6 +113,7 @@ enum llm_type {
     LLM_TYPE_16B_A1B,
     LLM_TYPE_21B_A3B, // Ernie MoE small
     LLM_TYPE_30B_A3B,
+    LLM_TYPE_31B_A3_5B,
     LLM_TYPE_80B_A3B, // Qwen3 Next
     LLM_TYPE_100B_A6B,
     LLM_TYPE_106B_A12B, // GLM-4.5-Air

src/models/nemotron-h.cpp

@@ -107,12 +107,41 @@ ggml_tensor * llm_build_nemotron_h::build_attention_layer(ggml_tensor *
 }
 
 ggml_tensor * llm_build_nemotron_h::build_ffn_layer(ggml_tensor * cur, const llama_model & model, const int il) {
-    cur = build_ffn(cur,
-            model.layers[il].ffn_up, model.layers[il].ffn_up_b, NULL,
-            NULL, NULL, NULL,
-            model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
-            NULL, LLM_FFN_RELU_SQR, LLM_FFN_PAR, il);
-    cb(cur, "ffn_out", il);
+    if (model.layers[il].ffn_gate_inp == nullptr) {
+        cur = build_ffn(cur,
+                model.layers[il].ffn_up,   model.layers[il].ffn_up_b,   NULL,
+                NULL,                      NULL,                        NULL,
+                model.layers[il].ffn_down, model.layers[il].ffn_down_b, NULL,
+                NULL,
+                LLM_FFN_RELU_SQR, LLM_FFN_PAR, il);
+        cb(cur, "ffn_out", il);
+    } else {
+        ggml_tensor * ffn_inp = cur;
+        ggml_tensor * moe_out =
+            build_moe_ffn(ffn_inp,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    nullptr, // no gate
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_RELU_SQR, hparams.expert_weights_norm,
+                    true, hparams.expert_weights_scale,
+                    LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID,
+                    il);
+        cb(moe_out, "ffn_moe_out", il);
+
+        ggml_tensor * ffn_shexp = build_ffn(ffn_inp,
+                    model.layers[il].ffn_up_shexp,  NULL, NULL,
+                    NULL /* no gate */           ,  NULL, NULL,
+                    model.layers[il].ffn_down_shexp, NULL, NULL,
+                    NULL,
+                    LLM_FFN_RELU_SQR, LLM_FFN_PAR, il);
+        cb(ffn_shexp, "ffn_shexp", il);
+
+        cur = ggml_add(ctx0, moe_out, ffn_shexp);
+        cb(cur, "ffn_out", il);
+    }
 
     cur = build_cvec(cur, il);
     cb(cur, "l_out", il);

tools/server/webui/README.md

@@ -619,11 +619,12 @@ flowchart TB
 
 ### Test Types
 
-| Type          | Tool               | Location                         | Command             |
-| ------------- | ------------------ | -------------------------------- | ------------------- |
-| **E2E**       | Playwright         | `tests/e2e/`                     | `npm run test:e2e`  |
-| **Unit**      | Vitest             | `tests/client/`, `tests/server/` | `npm run test:unit` |
-| **UI/Visual** | Storybook + Vitest | `tests/stories/`                 | `npm run test:ui`   |
+| Type          | Tool               | Location         | Command             |
+| ------------- | ------------------ | ---------------- | ------------------- |
+| **Unit**      | Vitest             | `tests/unit/`    | `npm run test:unit` |
+| **UI/Visual** | Storybook + Vitest | `tests/stories/` | `npm run test:ui`   |
+| **E2E**       | Playwright         | `tests/e2e/`     | `npm run test:e2e`  |
+| **Client**    | Vitest             | `tests/client/`. | `npm run test:unit` |
 
 ### Running Tests
 

tools/server/webui/package.json

@@ -13,12 +13,11 @@
 		"reset": "rm -rf .svelte-kit node_modules",
 		"format": "prettier --write .",
 		"lint": "prettier --check . && eslint .",
-		"test": "npm run test:ui -- --run && npm run test:client -- --run && npm run test:server -- --run && npm run test:e2e",
+		"test": "npm run test:ui -- --run && npm run test:client -- --run && npm run test:unit -- --run && npm run test:e2e",
 		"test:e2e": "playwright test",
 		"test:client": "vitest --project=client",
-		"test:server": "vitest --project=server",
+		"test:unit": "vitest --project=unit",
 		"test:ui": "vitest --project=ui",
-		"test:unit": "vitest",
 		"storybook": "storybook dev -p 6006",
 		"build-storybook": "storybook build",
 		"cleanup": "rm -rf .svelte-kit build node_modules test-results"

tools/server/webui/src/lib/components/app/chat/ChatAttachments/ChatAttachmentPreview.svelte

@@ -241,7 +241,7 @@
 				</div>
 			{/if}
 		{:else if (isText || (isPdf && pdfViewMode === 'text')) && displayTextContent}
-			<SyntaxHighlightedCode code={displayTextContent} {language} maxWidth="69rem" />
+			<SyntaxHighlightedCode code={displayTextContent} {language} maxWidth="calc(69rem - 2rem)" />
 		{:else if isAudio}
 			<div class="flex items-center justify-center p-8">
 				<div class="w-full max-w-md text-center">

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

@@ -24,7 +24,7 @@
 		MimeTypeImage,
 		MimeTypeText
 	} from '$lib/enums';
-	import { isIMEComposing } from '$lib/utils';
+	import { isIMEComposing, parseClipboardContent } from '$lib/utils';
 	import {
 		AudioRecorder,
 		convertToWav,
@@ -191,7 +191,6 @@
 
 			if ((!message.trim() && uploadedFiles.length === 0) || disabled || isLoading) return;
 
-			// Check if model is selected first
 			if (!checkModelSelected()) return;
 
 			const messageToSend = message.trim();
@@ -228,6 +227,31 @@
 
 		const text = event.clipboardData.getData(MimeTypeText.PLAIN);
 
+		if (text.startsWith('"')) {
+			const parsed = parseClipboardContent(text);
+
+			if (parsed.textAttachments.length > 0) {
+				event.preventDefault();
+
+				message = parsed.message;
+
+				const attachmentFiles = parsed.textAttachments.map(
+					(att) =>
+						new File([att.content], att.name, {
+							type: MimeTypeText.PLAIN
+						})
+				);
+
+				onFileUpload?.(attachmentFiles);
+
+				setTimeout(() => {
+					textareaRef?.focus();
+				}, 10);
+
+				return;
+			}
+		}
+
 		if (
 			text.length > 0 &&
 			pasteLongTextToFileLength > 0 &&

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

@@ -35,7 +35,7 @@
 
 <div class="flex items-center gap-1 {className}">
 	<DropdownMenu.Root>
-		<DropdownMenu.Trigger name="Attach files">
+		<DropdownMenu.Trigger name="Attach files" {disabled}>
 			<Tooltip.Root>
 				<Tooltip.Trigger>
 					<Button

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

@@ -173,6 +173,7 @@
 	/>
 
 	<ModelsSelector
+		{disabled}
 		bind:this={selectorModelRef}
 		currentModel={conversationModel}
 		forceForegroundText={true}

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

@@ -1,6 +1,7 @@
 <script lang="ts">
 	import { chatStore } from '$lib/stores/chat.svelte';
-	import { copyToClipboard, isIMEComposing } from '$lib/utils';
+	import { config } from '$lib/stores/settings.svelte';
+	import { copyToClipboard, isIMEComposing, formatMessageForClipboard } from '$lib/utils';
 	import ChatMessageAssistant from './ChatMessageAssistant.svelte';
 	import ChatMessageUser from './ChatMessageUser.svelte';
 	import ChatMessageSystem from './ChatMessageSystem.svelte';
@@ -87,7 +88,9 @@
 	}
 
 	async function handleCopy() {
-		await copyToClipboard(message.content, 'Message copied to clipboard');
+		const asPlainText = Boolean(config().copyTextAttachmentsAsPlainText);
+		const clipboardContent = formatMessageForClipboard(message.content, message.extra, asPlainText);
+		await copyToClipboard(clipboardContent, 'Message copied to clipboard');
 		onCopy?.(message);
 	}
 

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

@@ -57,6 +57,11 @@
 					label: 'Paste long text to file length',
 					type: 'input'
 				},
+				{
+					key: 'copyTextAttachmentsAsPlainText',
+					label: 'Copy text attachments as plain text',
+					type: 'checkbox'
+				},
 				{
 					key: 'enableContinueGeneration',
 					label: 'Enable "Continue" button',
@@ -109,6 +114,16 @@
 					key: 'disableAutoScroll',
 					label: 'Disable automatic scroll',
 					type: 'checkbox'
+				},
+				{
+					key: 'alwaysShowSidebarOnDesktop',
+					label: 'Always show sidebar on desktop',
+					type: 'checkbox'
+				},
+				{
+					key: 'autoShowSidebarOnNewChat',
+					label: 'Auto-show sidebar on new chat',
+					type: 'checkbox'
 				}
 			]
 		},
@@ -404,7 +419,7 @@
 	</div>
 
 	<!-- Mobile Header with Horizontal Scrollable Menu -->
-	<div class="flex flex-col md:hidden">
+	<div class="flex flex-col pt-6 md:hidden">
 		<div class="border-b border-border/30 py-4">
 			<!-- Horizontal Scrollable Category Menu with Navigation -->
 			<div class="relative flex items-center" style="scroll-padding: 1rem;">

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

@@ -72,9 +72,10 @@
 
 <div
 	class="code-preview-wrapper overflow-auto rounded-lg border border-border bg-muted {className}"
-	style="max-height: {maxHeight};"
+	style="max-height: {maxHeight}; max-width: {maxWidth};"
 >
-	<pre class="m-0 overflow-x-auto p-4 max-w-[{maxWidth}]"><code class="hljs text-sm leading-relaxed"
+	<!-- Needs to be formatted as single line for proper rendering -->
+	<pre class="m-0 overflow-x-auto p-4"><code class="hljs text-sm leading-relaxed"
 			>{@html highlightedHtml}</code
 		></pre>
 </div>

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

@@ -179,51 +179,37 @@
 		});
 	});
 
+	// Handle changes to the model selector pop-down or the model dialog, depending on if the server is in
+	// router mode or not.
 	function handleOpenChange(open: boolean) {
 		if (loading || updating) return;
 
-		if (open) {
-			isOpen = true;
-			searchTerm = '';
-			highlightedIndex = -1;
+		if (isRouter) {
+			if (open) {
+				isOpen = true;
+				searchTerm = '';
+				highlightedIndex = -1;
 
-			// Focus search input after popover opens
-			tick().then(() => {
-				requestAnimationFrame(() => searchInputRef?.focus());
-			});
+				// Focus search input after popover opens
+				tick().then(() => {
+					requestAnimationFrame(() => searchInputRef?.focus());
+				});
 
-			if (isRouter) {
 				modelsStore.fetchRouterModels().then(() => {
 					modelsStore.fetchModalitiesForLoadedModels();
 				});
+			} else {
+				isOpen = false;
+				searchTerm = '';
+				highlightedIndex = -1;
 			}
 		} else {
-			isOpen = false;
-			searchTerm = '';
-			highlightedIndex = -1;
+			showModelDialog = open;
 		}
 	}
 
-	function handleTriggerClick() {
-		if (loading || updating) return;
-
-		if (!isRouter) {
-			// Single model mode: show dialog instead of popover
-			showModelDialog = true;
-		}
-		// For router mode, the Popover handles open/close
-	}
-
 	export function open() {
-		if (isRouter) {
-			handleOpenChange(true);
-		} else {
-			showModelDialog = true;
-		}
-	}
-
-	function closeMenu() {
-		handleOpenChange(false);
+		handleOpenChange(true);
 	}
 
 	function handleSearchKeyDown(event: KeyboardEvent) {
@@ -292,7 +278,7 @@
 		}
 
 		if (shouldCloseMenu) {
-			closeMenu();
+			handleOpenChange(false);
 
 			// Focus the chat textarea after model selection
 			requestAnimationFrame(() => {
@@ -360,8 +346,181 @@
 	{:else}
 		{@const selectedOption = getDisplayOption()}
 
-		<Popover.Root bind:open={isOpen} onOpenChange={handleOpenChange}>
-			<Popover.Trigger
+		{#if isRouter}
+			<Popover.Root bind:open={isOpen} onOpenChange={handleOpenChange}>
+				<Popover.Trigger
+					class={cn(
+						`inline-flex cursor-pointer items-center gap-1.5 rounded-sm bg-muted-foreground/10 px-1.5 py-1 text-xs transition hover:text-foreground focus:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-2 disabled:cursor-not-allowed disabled:opacity-60`,
+						!isCurrentModelInCache()
+							? 'bg-red-400/10 !text-red-400 hover:bg-red-400/20 hover:text-red-400'
+							: forceForegroundText
+								? 'text-foreground'
+								: isHighlightedCurrentModelActive
+									? 'text-foreground'
+									: 'text-muted-foreground',
+						isOpen ? 'text-foreground' : ''
+					)}
+					style="max-width: min(calc(100cqw - 6.5rem), 32rem)"
+					disabled={disabled || updating}
+				>
+					<Package class="h-3.5 w-3.5" />
+
+					<span class="truncate font-medium">
+						{selectedOption?.model || 'Select model'}
+					</span>
+
+					{#if updating}
+						<Loader2 class="h-3 w-3.5 animate-spin" />
+					{:else}
+						<ChevronDown class="h-3 w-3.5" />
+					{/if}
+				</Popover.Trigger>
+
+				<Popover.Content
+					class="group/popover-content w-96 max-w-[calc(100vw-2rem)] p-0"
+					align="end"
+					sideOffset={8}
+					collisionPadding={16}
+				>
+					<div class="flex max-h-[50dvh] flex-col overflow-hidden">
+						<div
+							class="order-1 shrink-0 border-b p-4 group-data-[side=top]/popover-content:order-2 group-data-[side=top]/popover-content:border-t group-data-[side=top]/popover-content:border-b-0"
+						>
+							<SearchInput
+								id="model-search"
+								placeholder="Search models..."
+								bind:value={searchTerm}
+								bind:ref={searchInputRef}
+								onClose={() => handleOpenChange(false)}
+								onKeyDown={handleSearchKeyDown}
+							/>
+						</div>
+						<div
+							class="models-list order-2 min-h-0 flex-1 overflow-y-auto group-data-[side=top]/popover-content:order-1"
+						>
+							{#if !isCurrentModelInCache() && currentModel}
+								<!-- Show unavailable model as first option (disabled) -->
+								<button
+									type="button"
+									class="flex w-full cursor-not-allowed items-center bg-red-400/10 px-4 py-2 text-left text-sm text-red-400"
+									role="option"
+									aria-selected="true"
+									aria-disabled="true"
+									disabled
+								>
+									<span class="truncate">{selectedOption?.name || currentModel}</span>
+									<span class="ml-2 text-xs whitespace-nowrap opacity-70">(not available)</span>
+								</button>
+								<div class="my-1 h-px bg-border"></div>
+							{/if}
+							{#if filteredOptions.length === 0}
+								<p class="px-4 py-3 text-sm text-muted-foreground">No models found.</p>
+							{/if}
+							{#each filteredOptions as option, index (option.id)}
+								{@const status = getModelStatus(option.model)}
+								{@const isLoaded = status === ServerModelStatus.LOADED}
+								{@const isLoading = status === ServerModelStatus.LOADING}
+								{@const isSelected = currentModel === option.model || activeId === option.id}
+								{@const isCompatible = isModelCompatible(option)}
+								{@const isHighlighted = index === highlightedIndex}
+								{@const missingModalities = getMissingModalities(option)}
+
+								<div
+									class={cn(
+										'group flex w-full items-center gap-2 px-4 py-2 text-left text-sm transition focus:outline-none',
+										isCompatible
+											? 'cursor-pointer hover:bg-muted focus:bg-muted'
+											: 'cursor-not-allowed opacity-50',
+										isSelected || isHighlighted
+											? 'bg-accent text-accent-foreground'
+											: isCompatible
+												? 'hover:bg-accent hover:text-accent-foreground'
+												: '',
+										isLoaded ? 'text-popover-foreground' : 'text-muted-foreground'
+									)}
+									role="option"
+									aria-selected={isSelected || isHighlighted}
+									aria-disabled={!isCompatible}
+									tabindex={isCompatible ? 0 : -1}
+									onclick={() => isCompatible && handleSelect(option.id)}
+									onmouseenter={() => (highlightedIndex = index)}
+									onkeydown={(e) => {
+										if (isCompatible && (e.key === 'Enter' || e.key === ' ')) {
+											e.preventDefault();
+											handleSelect(option.id);
+										}
+									}}
+								>
+									<span class="min-w-0 flex-1 truncate">{option.model}</span>
+
+									{#if missingModalities}
+										<span class="flex shrink-0 items-center gap-1 text-muted-foreground/70">
+											{#if missingModalities.vision}
+												<Tooltip.Root>
+													<Tooltip.Trigger>
+														<EyeOff class="h-3.5 w-3.5" />
+													</Tooltip.Trigger>
+													<Tooltip.Content class="z-[9999]">
+														<p>No vision support</p>
+													</Tooltip.Content>
+												</Tooltip.Root>
+											{/if}
+											{#if missingModalities.audio}
+												<Tooltip.Root>
+													<Tooltip.Trigger>
+														<MicOff class="h-3.5 w-3.5" />
+													</Tooltip.Trigger>
+													<Tooltip.Content class="z-[9999]">
+														<p>No audio support</p>
+													</Tooltip.Content>
+												</Tooltip.Root>
+											{/if}
+										</span>
+									{/if}
+
+									{#if isLoading}
+										<Tooltip.Root>
+											<Tooltip.Trigger>
+												<Loader2 class="h-4 w-4 shrink-0 animate-spin text-muted-foreground" />
+											</Tooltip.Trigger>
+											<Tooltip.Content class="z-[9999]">
+												<p>Loading model...</p>
+											</Tooltip.Content>
+										</Tooltip.Root>
+									{:else if isLoaded}
+										<Tooltip.Root>
+											<Tooltip.Trigger>
+												<button
+													type="button"
+													class="relative ml-2 flex h-4 w-4 shrink-0 items-center justify-center"
+													onclick={(e) => {
+														e.stopPropagation();
+														modelsStore.unloadModel(option.model);
+													}}
+												>
+													<span
+														class="mr-2 h-2 w-2 rounded-full bg-green-500 transition-opacity group-hover:opacity-0"
+													></span>
+													<Power
+														class="absolute mr-2 h-4 w-4 text-red-500 opacity-0 transition-opacity group-hover:opacity-100 hover:text-red-600"
+													/>
+												</button>
+											</Tooltip.Trigger>
+											<Tooltip.Content class="z-[9999]">
+												<p>Unload model</p>
+											</Tooltip.Content>
+										</Tooltip.Root>
+									{:else}
+										<span class="mx-2 h-2 w-2 rounded-full bg-muted-foreground/50"></span>
+									{/if}
+								</div>
+							{/each}
+						</div>
+					</div>
+				</Popover.Content>
+			</Popover.Root>
+		{:else}
+			<button
 				class={cn(
 					`inline-flex cursor-pointer items-center gap-1.5 rounded-sm bg-muted-foreground/10 px-1.5 py-1 text-xs transition hover:text-foreground focus:outline-none focus-visible:ring-2 focus-visible:ring-ring focus-visible:ring-offset-2 disabled:cursor-not-allowed disabled:opacity-60`,
 					!isCurrentModelInCache()
@@ -374,165 +533,20 @@
 					isOpen ? 'text-foreground' : ''
 				)}
 				style="max-width: min(calc(100cqw - 6.5rem), 32rem)"
-				onclick={handleTriggerClick}
-				disabled={disabled || updating || !isRouter}
+				onclick={() => handleOpenChange(true)}
+				disabled={disabled || updating}
 			>
 				<Package class="h-3.5 w-3.5" />
 
 				<span class="truncate font-medium">
-					{selectedOption?.model || 'Select model'}
+					{selectedOption?.model}
 				</span>
 
 				{#if updating}
 					<Loader2 class="h-3 w-3.5 animate-spin" />
-				{:else if isRouter}
-					<ChevronDown class="h-3 w-3.5" />
 				{/if}
-			</Popover.Trigger>
-
-			<Popover.Content
-				class="group/popover-content w-96 max-w-[calc(100vw-2rem)] p-0"
-				align="end"
-				sideOffset={8}
-				collisionPadding={16}
-			>
-				<div class="flex max-h-[50dvh] flex-col overflow-hidden">
-					<div
-						class="order-1 shrink-0 border-b p-4 group-data-[side=top]/popover-content:order-2 group-data-[side=top]/popover-content:border-t group-data-[side=top]/popover-content:border-b-0"
-					>
-						<SearchInput
-							id="model-search"
-							placeholder="Search models..."
-							bind:value={searchTerm}
-							bind:ref={searchInputRef}
-							onClose={closeMenu}
-							onKeyDown={handleSearchKeyDown}
-						/>
-					</div>
-					<div
-						class="models-list order-2 min-h-0 flex-1 overflow-y-auto group-data-[side=top]/popover-content:order-1"
-					>
-						{#if !isCurrentModelInCache() && currentModel}
-							<!-- Show unavailable model as first option (disabled) -->
-							<button
-								type="button"
-								class="flex w-full cursor-not-allowed items-center bg-red-400/10 px-4 py-2 text-left text-sm text-red-400"
-								role="option"
-								aria-selected="true"
-								aria-disabled="true"
-								disabled
-							>
-								<span class="truncate">{selectedOption?.name || currentModel}</span>
-								<span class="ml-2 text-xs whitespace-nowrap opacity-70">(not available)</span>
-							</button>
-							<div class="my-1 h-px bg-border"></div>
-						{/if}
-						{#if filteredOptions.length === 0}
-							<p class="px-4 py-3 text-sm text-muted-foreground">No models found.</p>
-						{/if}
-						{#each filteredOptions as option, index (option.id)}
-							{@const status = getModelStatus(option.model)}
-							{@const isLoaded = status === ServerModelStatus.LOADED}
-							{@const isLoading = status === ServerModelStatus.LOADING}
-							{@const isSelected = currentModel === option.model || activeId === option.id}
-							{@const isCompatible = isModelCompatible(option)}
-							{@const isHighlighted = index === highlightedIndex}
-							{@const missingModalities = getMissingModalities(option)}
-
-							<div
-								class={cn(
-									'group flex w-full items-center gap-2 px-4 py-2 text-left text-sm transition focus:outline-none',
-									isCompatible
-										? 'cursor-pointer hover:bg-muted focus:bg-muted'
-										: 'cursor-not-allowed opacity-50',
-									isSelected || isHighlighted
-										? 'bg-accent text-accent-foreground'
-										: isCompatible
-											? 'hover:bg-accent hover:text-accent-foreground'
-											: '',
-									isLoaded ? 'text-popover-foreground' : 'text-muted-foreground'
-								)}
-								role="option"
-								aria-selected={isSelected || isHighlighted}
-								aria-disabled={!isCompatible}
-								tabindex={isCompatible ? 0 : -1}
-								onclick={() => isCompatible && handleSelect(option.id)}
-								onmouseenter={() => (highlightedIndex = index)}
-								onkeydown={(e) => {
-									if (isCompatible && (e.key === 'Enter' || e.key === ' ')) {
-										e.preventDefault();
-										handleSelect(option.id);
-									}
-								}}
-							>
-								<span class="min-w-0 flex-1 truncate">{option.model}</span>
-
-								{#if missingModalities}
-									<span class="flex shrink-0 items-center gap-1 text-muted-foreground/70">
-										{#if missingModalities.vision}
-											<Tooltip.Root>
-												<Tooltip.Trigger>
-													<EyeOff class="h-3.5 w-3.5" />
-												</Tooltip.Trigger>
-												<Tooltip.Content class="z-[9999]">
-													<p>No vision support</p>
-												</Tooltip.Content>
-											</Tooltip.Root>
-										{/if}
-										{#if missingModalities.audio}
-											<Tooltip.Root>
-												<Tooltip.Trigger>
-													<MicOff class="h-3.5 w-3.5" />
-												</Tooltip.Trigger>
-												<Tooltip.Content class="z-[9999]">
-													<p>No audio support</p>
-												</Tooltip.Content>
-											</Tooltip.Root>
-										{/if}
-									</span>
-								{/if}
-
-								{#if isLoading}
-									<Tooltip.Root>
-										<Tooltip.Trigger>
-											<Loader2 class="h-4 w-4 shrink-0 animate-spin text-muted-foreground" />
-										</Tooltip.Trigger>
-										<Tooltip.Content class="z-[9999]">
-											<p>Loading model...</p>
-										</Tooltip.Content>
-									</Tooltip.Root>
-								{:else if isLoaded}
-									<Tooltip.Root>
-										<Tooltip.Trigger>
-											<button
-												type="button"
-												class="relative ml-2 flex h-4 w-4 shrink-0 items-center justify-center"
-												onclick={(e) => {
-													e.stopPropagation();
-													modelsStore.unloadModel(option.model);
-												}}
-											>
-												<span
-													class="mr-2 h-2 w-2 rounded-full bg-green-500 transition-opacity group-hover:opacity-0"
-												></span>
-												<Power
-													class="absolute mr-2 h-4 w-4 text-red-500 opacity-0 transition-opacity group-hover:opacity-100 hover:text-red-600"
-												/>
-											</button>
-										</Tooltip.Trigger>
-										<Tooltip.Content class="z-[9999]">
-											<p>Unload model</p>
-										</Tooltip.Content>
-									</Tooltip.Root>
-								{:else}
-									<span class="mx-2 h-2 w-2 rounded-full bg-muted-foreground/50"></span>
-								{/if}
-							</div>
-						{/each}
-					</div>
-				</div>
-			</Popover.Content>
-		</Popover.Root>
+			</button>
+		{/if}
 	{/if}
 </div>
 

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

@@ -12,9 +12,12 @@ export const SETTING_CONFIG_DEFAULT: Record<string, string | number | boolean> =
 	showMessageStats: true,
 	askForTitleConfirmation: false,
 	pasteLongTextToFileLen: 2500,
+	copyTextAttachmentsAsPlainText: false,
 	pdfAsImage: false,
 	disableAutoScroll: false,
 	renderUserContentAsMarkdown: false,
+	alwaysShowSidebarOnDesktop: false,
+	autoShowSidebarOnNewChat: true,
 	autoMicOnEmpty: false,
 	// make sure these default values are in sync with `common.h`
 	samplers: 'top_k;typ_p;top_p;min_p;temperature',
@@ -50,6 +53,8 @@ export const SETTING_CONFIG_INFO: Record<string, string> = {
 		'Choose the color theme for the interface. You can choose between System (follows your device settings), Light, or Dark.',
 	pasteLongTextToFileLen:
 		'On pasting long text, it will be converted to a file. You can control the file length by setting the value of this parameter. Value 0 means disable.',
+	copyTextAttachmentsAsPlainText:
+		'When copying a message with text attachments, combine them into a single plain text string instead of a special format that can be pasted back as attachments.',
 	samplers:
 		'The order at which samplers are applied, in simplified way. Default is "top_k;typ_p;top_p;min_p;temperature": top_k->typ_p->top_p->min_p->temperature',
 	temperature:
@@ -96,6 +101,10 @@ export const SETTING_CONFIG_INFO: Record<string, string> = {
 	disableAutoScroll:
 		'Disable automatic scrolling while messages stream so you can control the viewport position manually.',
 	renderUserContentAsMarkdown: 'Render user messages using markdown formatting in the chat.',
+	alwaysShowSidebarOnDesktop:
+		'Always keep the sidebar visible on desktop instead of auto-hiding it.',
+	autoShowSidebarOnNewChat:
+		'Automatically show sidebar when starting a new chat. Disable to keep the sidebar hidden until you click on it.',
 	autoMicOnEmpty:
 		'Automatically show microphone button instead of send button when textarea is empty for models with audio modality support.',
 	pyInterpreterEnabled:

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

@@ -0,0 +1,262 @@
+import { toast } from 'svelte-sonner';
+import { AttachmentType } from '$lib/enums';
+import type {
+	DatabaseMessageExtra,
+	DatabaseMessageExtraTextFile,
+	DatabaseMessageExtraLegacyContext
+} from '$lib/types/database';
+
+/**
+ * Copy text to clipboard with toast notification
+ * Uses modern clipboard API when available, falls back to legacy method for non-secure contexts
+ * @param text - Text to copy to clipboard
+ * @param successMessage - Custom success message (optional)
+ * @param errorMessage - Custom error message (optional)
+ * @returns Promise<boolean> - True if successful, false otherwise
+ */
+export async function copyToClipboard(
+	text: string,
+	successMessage = 'Copied to clipboard',
+	errorMessage = 'Failed to copy to clipboard'
+): Promise<boolean> {
+	try {
+		// Try modern clipboard API first (secure contexts only)
+		if (navigator.clipboard && navigator.clipboard.writeText) {
+			await navigator.clipboard.writeText(text);
+			toast.success(successMessage);
+			return true;
+		}
+
+		// Fallback for non-secure contexts
+		const textArea = document.createElement('textarea');
+		textArea.value = text;
+		textArea.style.position = 'fixed';
+		textArea.style.left = '-999999px';
+		textArea.style.top = '-999999px';
+		document.body.appendChild(textArea);
+		textArea.focus();
+		textArea.select();
+
+		const successful = document.execCommand('copy');
+		document.body.removeChild(textArea);
+
+		if (successful) {
+			toast.success(successMessage);
+			return true;
+		} else {
+			throw new Error('execCommand failed');
+		}
+	} catch (error) {
+		console.error('Failed to copy to clipboard:', error);
+		toast.error(errorMessage);
+		return false;
+	}
+}
+
+/**
+ * Copy code with HTML entity decoding and toast notification
+ * @param rawCode - Raw code string that may contain HTML entities
+ * @param successMessage - Custom success message (optional)
+ * @param errorMessage - Custom error message (optional)
+ * @returns Promise<boolean> - True if successful, false otherwise
+ */
+export async function copyCodeToClipboard(
+	rawCode: string,
+	successMessage = 'Code copied to clipboard',
+	errorMessage = 'Failed to copy code'
+): Promise<boolean> {
+	const doc = new DOMParser().parseFromString(rawCode, 'text/html');
+	const decodedCode = doc.body.textContent ?? rawCode;
+
+	return copyToClipboard(decodedCode, successMessage, errorMessage);
+}
+
+/**
+ * Format for text attachments when copied to clipboard
+ */
+export interface ClipboardTextAttachment {
+	type: typeof AttachmentType.TEXT;
+	name: string;
+	content: string;
+}
+
+/**
+ * Parsed result from clipboard content
+ */
+export interface ParsedClipboardContent {
+	message: string;
+	textAttachments: ClipboardTextAttachment[];
+}
+
+/**
+ * Formats a message with text attachments for clipboard copying.
+ *
+ * Default format (asPlainText = false):
+ * ```
+ * "Text message content"
+ * [
+ *   {"type":"TEXT","name":"filename.txt","content":"..."},
+ *   {"type":"TEXT","name":"another.txt","content":"..."}
+ * ]
+ * ```
+ *
+ * Plain text format (asPlainText = true):
+ * ```
+ * Text message content
+ *
+ * file content here
+ *
+ * another file content
+ * ```
+ *
+ * @param content - The message text content
+ * @param extras - Optional array of message attachments
+ * @param asPlainText - If true, format as plain text without JSON structure
+ * @returns Formatted string for clipboard
+ */
+export function formatMessageForClipboard(
+	content: string,
+	extras?: DatabaseMessageExtra[],
+	asPlainText: boolean = false
+): string {
+	// Filter only text attachments (TEXT type and legacy CONTEXT type)
+	const textAttachments =
+		extras?.filter(
+			(extra): extra is DatabaseMessageExtraTextFile | DatabaseMessageExtraLegacyContext =>
+				extra.type === AttachmentType.TEXT || extra.type === AttachmentType.LEGACY_CONTEXT
+		) ?? [];
+
+	if (textAttachments.length === 0) {
+		return content;
+	}
+
+	if (asPlainText) {
+		const parts = [content];
+		for (const att of textAttachments) {
+			parts.push(att.content);
+		}
+		return parts.join('\n\n');
+	}
+
+	const clipboardAttachments: ClipboardTextAttachment[] = textAttachments.map((att) => ({
+		type: AttachmentType.TEXT,
+		name: att.name,
+		content: att.content
+	}));
+
+	return `${JSON.stringify(content)}\n${JSON.stringify(clipboardAttachments, null, 2)}`;
+}
+
+/**
+ * Parses clipboard content to extract message and text attachments.
+ * Supports both plain text and the special format with attachments.
+ *
+ * @param clipboardText - Raw text from clipboard
+ * @returns Parsed content with message and attachments
+ */
+export function parseClipboardContent(clipboardText: string): ParsedClipboardContent {
+	const defaultResult: ParsedClipboardContent = {
+		message: clipboardText,
+		textAttachments: []
+	};
+
+	if (!clipboardText.startsWith('"')) {
+		return defaultResult;
+	}
+
+	try {
+		let stringEndIndex = -1;
+		let escaped = false;
+
+		for (let i = 1; i < clipboardText.length; i++) {
+			const char = clipboardText[i];
+
+			if (escaped) {
+				escaped = false;
+				continue;
+			}
+
+			if (char === '\\') {
+				escaped = true;
+				continue;
+			}
+
+			if (char === '"') {
+				stringEndIndex = i;
+				break;
+			}
+		}
+
+		if (stringEndIndex === -1) {
+			return defaultResult;
+		}
+
+		const jsonStringPart = clipboardText.substring(0, stringEndIndex + 1);
+		const remainingPart = clipboardText.substring(stringEndIndex + 1).trim();
+
+		const message = JSON.parse(jsonStringPart) as string;
+
+		if (!remainingPart || !remainingPart.startsWith('[')) {
+			return {
+				message,
+				textAttachments: []
+			};
+		}
+
+		const attachments = JSON.parse(remainingPart) as unknown[];
+
+		const validAttachments: ClipboardTextAttachment[] = [];
+
+		for (const att of attachments) {
+			if (isValidTextAttachment(att)) {
+				validAttachments.push({
+					type: AttachmentType.TEXT,
+					name: att.name,
+					content: att.content
+				});
+			}
+		}
+
+		return {
+			message,
+			textAttachments: validAttachments
+		};
+	} catch {
+		return defaultResult;
+	}
+}
+
+/**
+ * Type guard to validate a text attachment object
+ * @param obj The object to validate
+ * @returns true if the object is a valid text attachment
+ */
+function isValidTextAttachment(
+	obj: unknown
+): obj is { type: string; name: string; content: string } {
+	if (typeof obj !== 'object' || obj === null) {
+		return false;
+	}
+
+	const record = obj as Record<string, unknown>;
+
+	return (
+		(record.type === AttachmentType.TEXT || record.type === 'TEXT') &&
+		typeof record.name === 'string' &&
+		typeof record.content === 'string'
+	);
+}
+
+/**
+ * Checks if clipboard content contains our special format with attachments
+ * @param clipboardText - Raw text from clipboard
+ * @returns true if the clipboard content contains our special format with attachments
+ */
+export function hasClipboardAttachments(clipboardText: string): boolean {
+	if (!clipboardText.startsWith('"')) {
+		return false;
+	}
+
+	const parsed = parseClipboardContent(clipboardText);
+	return parsed.textAttachments.length > 0;
+}

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

@@ -1,71 +0,0 @@
-import { toast } from 'svelte-sonner';
-
-/**
- * Copy text to clipboard with toast notification
- * Uses modern clipboard API when available, falls back to legacy method for non-secure contexts
- * @param text - Text to copy to clipboard
- * @param successMessage - Custom success message (optional)
- * @param errorMessage - Custom error message (optional)
- * @returns Promise<boolean> - True if successful, false otherwise
- */
-export async function copyToClipboard(
-	text: string,
-	successMessage = 'Copied to clipboard',
-	errorMessage = 'Failed to copy to clipboard'
-): Promise<boolean> {
-	try {
-		// Try modern clipboard API first (secure contexts only)
-		if (navigator.clipboard && navigator.clipboard.writeText) {
-			await navigator.clipboard.writeText(text);
-			toast.success(successMessage);
-			return true;
-		}
-
-		// Fallback for non-secure contexts
-		const textArea = document.createElement('textarea');
-		textArea.value = text;
-		textArea.style.position = 'fixed';
-		textArea.style.left = '-999999px';
-		textArea.style.top = '-999999px';
-		document.body.appendChild(textArea);
-		textArea.focus();
-		textArea.select();
-
-		const successful = document.execCommand('copy');
-		document.body.removeChild(textArea);
-
-		if (successful) {
-			toast.success(successMessage);
-			return true;
-		} else {
-			throw new Error('execCommand failed');
-		}
-	} catch (error) {
-		console.error('Failed to copy to clipboard:', error);
-		toast.error(errorMessage);
-		return false;
-	}
-}
-
-/**
- * Copy code with HTML entity decoding and toast notification
- * @param rawCode - Raw code string that may contain HTML entities
- * @param successMessage - Custom success message (optional)
- * @param errorMessage - Custom error message (optional)
- * @returns Promise<boolean> - True if successful, false otherwise
- */
-export async function copyCodeToClipboard(
-	rawCode: string,
-	successMessage = 'Code copied to clipboard',
-	errorMessage = 'Failed to copy code'
-): Promise<boolean> {
-	// Decode HTML entities
-	const decodedCode = rawCode
-		.replace(/&amp;/g, '&')
-		.replace(/&lt;/g, '<')
-		.replace(/&gt;/g, '>')
-		.replace(/&quot;/g, '"')
-		.replace(/&#39;/g, "'");
-
-	return copyToClipboard(decodedCode, successMessage, errorMessage);
-}

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

@@ -40,7 +40,15 @@ export { setConfigValue, getConfigValue, configToParameterRecord } from './confi
 export { createMessageCountMap, getMessageCount } from './conversation-utils';
 
 // Clipboard utilities
-export { copyToClipboard, copyCodeToClipboard } from './copy';
+export {
+	copyToClipboard,
+	copyCodeToClipboard,
+	formatMessageForClipboard,
+	parseClipboardContent,
+	hasClipboardAttachments,
+	type ClipboardTextAttachment,
+	type ParsedClipboardContent
+} from './clipboard';
 
 // File preview utilities
 export { getFileTypeLabel } from './file-preview';

tools/server/webui/src/routes/+layout.svelte

@@ -14,6 +14,7 @@
 	import { goto } from '$app/navigation';
 	import { modelsStore } from '$lib/stores/models.svelte';
 	import { TOOLTIP_DELAY_DURATION } from '$lib/constants/tooltip-config';
+	import { IsMobile } from '$lib/hooks/is-mobile.svelte';
 
 	let { children } = $props();
 
@@ -21,6 +22,10 @@
 	let isHomeRoute = $derived(page.route.id === '/');
 	let isNewChatMode = $derived(page.url.searchParams.get('new_chat') === 'true');
 	let showSidebarByDefault = $derived(activeMessages().length > 0 || isLoading());
+	let alwaysShowSidebarOnDesktop = $derived(config().alwaysShowSidebarOnDesktop);
+	let autoShowSidebarOnNewChat = $derived(config().autoShowSidebarOnNewChat);
+	let isMobile = new IsMobile();
+	let isDesktop = $derived(!isMobile.current);
 	let sidebarOpen = $state(false);
 	let innerHeight = $state<number | undefined>();
 	let chatSidebar:
@@ -76,6 +81,11 @@
 	}
 
 	$effect(() => {
+		if (alwaysShowSidebarOnDesktop && isDesktop) {
+			sidebarOpen = true;
+			return;
+		}
+
 		if (isHomeRoute && !isNewChatMode) {
 			// Auto-collapse sidebar when navigating to home route (but not in new chat mode)
 			sidebarOpen = false;
@@ -83,8 +93,11 @@
 			// Keep sidebar open in new chat mode
 			sidebarOpen = true;
 		} else if (isChatRoute) {
-			// On chat routes, show sidebar by default
-			sidebarOpen = true;
+			// On chat routes, only auto-show sidebar if setting is enabled
+			if (autoShowSidebarOnNewChat) {
+				sidebarOpen = true;
+			}
+			// If setting is disabled, don't change sidebar state - let user control it manually
 		} else {
 			// Other routes follow default behavior
 			sidebarOpen = showSidebarByDefault;
@@ -190,12 +203,14 @@
 				<ChatSidebar bind:this={chatSidebar} />
 			</Sidebar.Root>
 
-			<Sidebar.Trigger
-				class="transition-left absolute left-0 z-[900] h-8 w-8 duration-200 ease-linear {sidebarOpen
-					? 'md:left-[var(--sidebar-width)]'
-					: ''}"
-				style="translate: 1rem 1rem;"
-			/>
+			{#if !(alwaysShowSidebarOnDesktop && isDesktop)}
+				<Sidebar.Trigger
+					class="transition-left absolute left-0 z-[900] h-8 w-8 duration-200 ease-linear {sidebarOpen
+						? 'md:left-[var(--sidebar-width)]'
+						: ''}"
+					style="translate: 1rem 1rem;"
+				/>
+			{/if}
 
 			<Sidebar.Inset class="flex flex-1 flex-col overflow-hidden">
 				{@render children?.()}

tools/server/webui/tests/server/demo.spec.ts

@@ -1,7 +0,0 @@
-import { describe, it, expect } from 'vitest';
-
-describe('sum test', () => {
-	it('adds 1 + 2 to equal 3', () => {
-		expect(1 + 2).toBe(3);
-	});
-});

tools/server/webui/tests/unit/clipboard.test.ts

@@ -0,0 +1,423 @@
+import { describe, it, expect } from 'vitest';
+import { AttachmentType } from '$lib/enums';
+import {
+	formatMessageForClipboard,
+	parseClipboardContent,
+	hasClipboardAttachments
+} from '$lib/utils/clipboard';
+
+describe('formatMessageForClipboard', () => {
+	it('returns plain content when no extras', () => {
+		const result = formatMessageForClipboard('Hello world', undefined);
+		expect(result).toBe('Hello world');
+	});
+
+	it('returns plain content when extras is empty array', () => {
+		const result = formatMessageForClipboard('Hello world', []);
+		expect(result).toBe('Hello world');
+	});
+
+	it('handles empty string content', () => {
+		const result = formatMessageForClipboard('', undefined);
+		expect(result).toBe('');
+	});
+
+	it('returns plain content when extras has only non-text attachments', () => {
+		const extras = [
+			{
+				type: AttachmentType.IMAGE as const,
+				name: 'image.png',
+				base64Url: 'data:image/png;base64,...'
+			}
+		];
+		const result = formatMessageForClipboard('Hello world', extras);
+		expect(result).toBe('Hello world');
+	});
+
+	it('filters non-text attachments and keeps only text ones', () => {
+		const extras = [
+			{
+				type: AttachmentType.IMAGE as const,
+				name: 'image.png',
+				base64Url: 'data:image/png;base64,...'
+			},
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'file.txt',
+				content: 'Text content'
+			},
+			{
+				type: AttachmentType.PDF as const,
+				name: 'doc.pdf',
+				base64Data: 'data:application/pdf;base64,...',
+				content: 'PDF content',
+				processedAsImages: false
+			}
+		];
+		const result = formatMessageForClipboard('Hello', extras);
+
+		expect(result).toContain('"file.txt"');
+		expect(result).not.toContain('image.png');
+		expect(result).not.toContain('doc.pdf');
+	});
+
+	it('formats message with text attachments', () => {
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'file1.txt',
+				content: 'File 1 content'
+			},
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'file2.txt',
+				content: 'File 2 content'
+			}
+		];
+		const result = formatMessageForClipboard('Hello world', extras);
+
+		expect(result).toContain('"Hello world"');
+		expect(result).toContain('"type": "TEXT"');
+		expect(result).toContain('"name": "file1.txt"');
+		expect(result).toContain('"content": "File 1 content"');
+		expect(result).toContain('"name": "file2.txt"');
+	});
+
+	it('handles content with quotes and special characters', () => {
+		const content = 'Hello "world" with\nnewline';
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'test.txt',
+				content: 'Test content'
+			}
+		];
+		const result = formatMessageForClipboard(content, extras);
+
+		// Should be valid JSON
+		expect(result.startsWith('"')).toBe(true);
+		// The content should be properly escaped
+		const parsed = JSON.parse(result.split('\n')[0]);
+		expect(parsed).toBe(content);
+	});
+
+	it('converts legacy context type to TEXT type', () => {
+		const extras = [
+			{
+				type: AttachmentType.LEGACY_CONTEXT as const,
+				name: 'legacy.txt',
+				content: 'Legacy content'
+			}
+		];
+		const result = formatMessageForClipboard('Hello', extras);
+
+		expect(result).toContain('"type": "TEXT"');
+		expect(result).not.toContain('"context"');
+	});
+
+	it('handles attachment content with special characters', () => {
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'code.js',
+				content: 'const x = "hello\\nworld";\nconst y = `template ${var}`;'
+			}
+		];
+		const formatted = formatMessageForClipboard('Check this code', extras);
+		const parsed = parseClipboardContent(formatted);
+
+		expect(parsed.textAttachments[0].content).toBe(
+			'const x = "hello\\nworld";\nconst y = `template ${var}`;'
+		);
+	});
+
+	it('handles unicode characters in content and attachments', () => {
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'unicode.txt',
+				content: '日本語テスト 🎉 émojis'
+			}
+		];
+		const formatted = formatMessageForClipboard('Привет мир 👋', extras);
+		const parsed = parseClipboardContent(formatted);
+
+		expect(parsed.message).toBe('Привет мир 👋');
+		expect(parsed.textAttachments[0].content).toBe('日本語テスト 🎉 émojis');
+	});
+
+	it('formats as plain text when asPlainText is true', () => {
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'file1.txt',
+				content: 'File 1 content'
+			},
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'file2.txt',
+				content: 'File 2 content'
+			}
+		];
+		const result = formatMessageForClipboard('Hello world', extras, true);
+
+		expect(result).toBe('Hello world\n\nFile 1 content\n\nFile 2 content');
+	});
+
+	it('returns plain content when asPlainText is true but no attachments', () => {
+		const result = formatMessageForClipboard('Hello world', [], true);
+		expect(result).toBe('Hello world');
+	});
+
+	it('plain text mode does not use JSON format', () => {
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'test.txt',
+				content: 'Test content'
+			}
+		];
+		const result = formatMessageForClipboard('Hello', extras, true);
+
+		expect(result).not.toContain('"type"');
+		expect(result).not.toContain('[');
+		expect(result).toBe('Hello\n\nTest content');
+	});
+});
+
+describe('parseClipboardContent', () => {
+	it('returns plain text as message when not in special format', () => {
+		const result = parseClipboardContent('Hello world');
+
+		expect(result.message).toBe('Hello world');
+		expect(result.textAttachments).toHaveLength(0);
+	});
+
+	it('handles empty string input', () => {
+		const result = parseClipboardContent('');
+
+		expect(result.message).toBe('');
+		expect(result.textAttachments).toHaveLength(0);
+	});
+
+	it('handles whitespace-only input', () => {
+		const result = parseClipboardContent('   \n\t  ');
+
+		expect(result.message).toBe('   \n\t  ');
+		expect(result.textAttachments).toHaveLength(0);
+	});
+
+	it('returns plain text as message when starts with quote but invalid format', () => {
+		const result = parseClipboardContent('"Unclosed quote');
+
+		expect(result.message).toBe('"Unclosed quote');
+		expect(result.textAttachments).toHaveLength(0);
+	});
+
+	it('returns original text when JSON array is malformed', () => {
+		const input = '"Hello"\n[invalid json';
+
+		const result = parseClipboardContent(input);
+
+		expect(result.message).toBe('"Hello"\n[invalid json');
+		expect(result.textAttachments).toHaveLength(0);
+	});
+
+	it('parses message with text attachments', () => {
+		const input = `"Hello world"
+[
+  {"type":"TEXT","name":"file1.txt","content":"File 1 content"},
+  {"type":"TEXT","name":"file2.txt","content":"File 2 content"}
+]`;
+
+		const result = parseClipboardContent(input);
+
+		expect(result.message).toBe('Hello world');
+		expect(result.textAttachments).toHaveLength(2);
+		expect(result.textAttachments[0].name).toBe('file1.txt');
+		expect(result.textAttachments[0].content).toBe('File 1 content');
+		expect(result.textAttachments[1].name).toBe('file2.txt');
+		expect(result.textAttachments[1].content).toBe('File 2 content');
+	});
+
+	it('handles escaped quotes in message', () => {
+		const input = `"Hello \\"world\\" with quotes"
+[
+  {"type":"TEXT","name":"file.txt","content":"test"}
+]`;
+
+		const result = parseClipboardContent(input);
+
+		expect(result.message).toBe('Hello "world" with quotes');
+		expect(result.textAttachments).toHaveLength(1);
+	});
+
+	it('handles newlines in message', () => {
+		const input = `"Hello\\nworld"
+[
+  {"type":"TEXT","name":"file.txt","content":"test"}
+]`;
+
+		const result = parseClipboardContent(input);
+
+		expect(result.message).toBe('Hello\nworld');
+		expect(result.textAttachments).toHaveLength(1);
+	});
+
+	it('returns message only when no array follows', () => {
+		const input = '"Just a quoted string"';
+
+		const result = parseClipboardContent(input);
+
+		expect(result.message).toBe('Just a quoted string');
+		expect(result.textAttachments).toHaveLength(0);
+	});
+
+	it('filters out invalid attachment objects', () => {
+		const input = `"Hello"
+[
+  {"type":"TEXT","name":"valid.txt","content":"valid"},
+  {"type":"INVALID","name":"invalid.txt","content":"invalid"},
+  {"name":"missing-type.txt","content":"missing"},
+  {"type":"TEXT","content":"missing name"}
+]`;
+
+		const result = parseClipboardContent(input);
+
+		expect(result.message).toBe('Hello');
+		expect(result.textAttachments).toHaveLength(1);
+		expect(result.textAttachments[0].name).toBe('valid.txt');
+	});
+
+	it('handles empty attachments array', () => {
+		const input = '"Hello"\n[]';
+
+		const result = parseClipboardContent(input);
+
+		expect(result.message).toBe('Hello');
+		expect(result.textAttachments).toHaveLength(0);
+	});
+
+	it('roundtrips correctly with formatMessageForClipboard', () => {
+		const originalContent = 'Hello "world" with\nspecial characters';
+		const originalExtras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'file1.txt',
+				content: 'Content with\nnewlines and "quotes"'
+			},
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'file2.txt',
+				content: 'Another file'
+			}
+		];
+
+		const formatted = formatMessageForClipboard(originalContent, originalExtras);
+		const parsed = parseClipboardContent(formatted);
+
+		expect(parsed.message).toBe(originalContent);
+		expect(parsed.textAttachments).toHaveLength(2);
+		expect(parsed.textAttachments[0].name).toBe('file1.txt');
+		expect(parsed.textAttachments[0].content).toBe('Content with\nnewlines and "quotes"');
+		expect(parsed.textAttachments[1].name).toBe('file2.txt');
+		expect(parsed.textAttachments[1].content).toBe('Another file');
+	});
+});
+
+describe('hasClipboardAttachments', () => {
+	it('returns false for plain text', () => {
+		expect(hasClipboardAttachments('Hello world')).toBe(false);
+	});
+
+	it('returns false for empty string', () => {
+		expect(hasClipboardAttachments('')).toBe(false);
+	});
+
+	it('returns false for quoted string without attachments', () => {
+		expect(hasClipboardAttachments('"Hello world"')).toBe(false);
+	});
+
+	it('returns true for valid format with attachments', () => {
+		const input = `"Hello"
+[{"type":"TEXT","name":"file.txt","content":"test"}]`;
+
+		expect(hasClipboardAttachments(input)).toBe(true);
+	});
+
+	it('returns false for format with empty attachments array', () => {
+		const input = '"Hello"\n[]';
+
+		expect(hasClipboardAttachments(input)).toBe(false);
+	});
+
+	it('returns false for malformed JSON', () => {
+		expect(hasClipboardAttachments('"Hello"\n[broken')).toBe(false);
+	});
+});
+
+describe('roundtrip edge cases', () => {
+	it('preserves empty message with attachments', () => {
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'file.txt',
+				content: 'Content only'
+			}
+		];
+		const formatted = formatMessageForClipboard('', extras);
+		const parsed = parseClipboardContent(formatted);
+
+		expect(parsed.message).toBe('');
+		expect(parsed.textAttachments).toHaveLength(1);
+		expect(parsed.textAttachments[0].content).toBe('Content only');
+	});
+
+	it('preserves attachment with empty content', () => {
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'empty.txt',
+				content: ''
+			}
+		];
+		const formatted = formatMessageForClipboard('Message', extras);
+		const parsed = parseClipboardContent(formatted);
+
+		expect(parsed.message).toBe('Message');
+		expect(parsed.textAttachments).toHaveLength(1);
+		expect(parsed.textAttachments[0].content).toBe('');
+	});
+
+	it('preserves multiple backslashes', () => {
+		const content = 'Path: C:\\\\Users\\\\test\\\\file.txt';
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'path.txt',
+				content: 'D:\\\\Data\\\\file'
+			}
+		];
+		const formatted = formatMessageForClipboard(content, extras);
+		const parsed = parseClipboardContent(formatted);
+
+		expect(parsed.message).toBe(content);
+		expect(parsed.textAttachments[0].content).toBe('D:\\\\Data\\\\file');
+	});
+
+	it('preserves tabs and various whitespace', () => {
+		const content = 'Line1\t\tTabbed\n  Spaced\r\nCRLF';
+		const extras = [
+			{
+				type: AttachmentType.TEXT as const,
+				name: 'whitespace.txt',
+				content: '\t\t\n\n   '
+			}
+		];
+		const formatted = formatMessageForClipboard(content, extras);
+		const parsed = parseClipboardContent(formatted);
+
+		expect(parsed.message).toBe(content);
+		expect(parsed.textAttachments[0].content).toBe('\t\t\n\n   ');
+	});
+});

tools/server/webui/tests/unit/latex-protection.test.ts

@@ -1,6 +1,6 @@
 /* eslint-disable no-irregular-whitespace */
 import { describe, it, expect, test } from 'vitest';
-import { maskInlineLaTeX, preprocessLaTeX } from './latex-protection';
+import { maskInlineLaTeX, preprocessLaTeX } from '$lib/utils/latex-protection';
 
 describe('maskInlineLaTeX', () => {
 	it('should protect LaTeX $x + y$ but not money $3.99', () => {

tools/server/webui/tests/unit/model-names.test.ts

@@ -1,5 +1,5 @@
 import { describe, expect, it } from 'vitest';
-import { isValidModelName, normalizeModelName } from './model-names';
+import { isValidModelName, normalizeModelName } from '$lib/utils/model-names';
 
 describe('normalizeModelName', () => {
 	it('preserves Hugging Face org/model format (single slash)', () => {

tools/server/webui/vite.config.ts

@@ -125,9 +125,9 @@ export default defineConfig({
 			{
 				extends: './vite.config.ts',
 				test: {
-					name: 'server',
+					name: 'unit',
 					environment: 'node',
-					include: ['tests/server/**/*.{test,spec}.{js,ts}']
+					include: ['tests/unit/**/*.{test,spec}.{js,ts}']
 				}
 			},
 			{