From 5d3575e615d87f43fd23306d93625eb06fe26039 Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 25 Jul 2025 17:08:48 -0400
Subject: [PATCH 01/11] Format changes first

---
 .github/workflows/test.yml | 110 ++++++++++++++++++-------------------
 1 file changed, 54 insertions(+), 56 deletions(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 6e8ea34..d7550c7 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -19,7 +19,7 @@ on:
       # Skip running the source code checks when only documentation has been updated
       - "!**.md"
       - "!**.rst"
-      - "!**.txt"  # Any requirements file changes will also involve changing other files
+      - "!**.txt" # Any requirements file changes will also involve changing other files
   push:
     branches:
       - main
@@ -37,7 +37,7 @@ jobs:
   tests:
     runs-on: ${{ matrix.os }}
     strategy:
-      fail-fast: false  # Always report results for all targets
+      fail-fast: false # Always report results for all targets
       max-parallel: 8
       matrix:
         python-version: ["3.10", "3.11", "3.12", "3.13"]
@@ -51,61 +51,59 @@ jobs:
     # for latest versions if the standard actions start emitting warnings
 
     steps:
-    - uses: actions/checkout@v4
-      with:
-        persist-credentials: false
-
-    - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v5
-      with:
-        python-version: ${{ matrix.python-version }}
-
-    - name: Get pip cache dir
-      id: pip-cache
-      run: |
-        echo "dir=$(python -m pip cache dir)" >> $GITHUB_OUTPUT
-
-    - name: Cache bootstrapping dependencies
-      uses: actions/cache@v4
-      with:
-        path: ${{ steps.pip-cache.outputs.dir }}
-        key:
-          pip-${{ matrix.os }}-${{ matrix.python-version }}-v1-${{ hashFiles('pdm.lock') }}
-        restore-keys: |
-          pip-${{ matrix.os }}-${{ matrix.python-version }}-v1-
-
-    - name: Install PDM
-      run: |
-        # Ensure `pdm` uses the same version as specified in `pdm.lock`
-        # while avoiding the error raised by https://github.com/pypa/pip/issues/12889
-        python -m pip install --upgrade -r ci-bootstrap-requirements.txt
-
-    - name: Create development virtual environment
-      run: |
-        python -m pdm sync --no-self --dev
-        # Handle Windows vs non-Windows differences in .venv layout
-        VIRTUAL_ENV_BIN_DIR="$PWD/.venv/bin"
-        test -e "$VIRTUAL_ENV_BIN_DIR" || VIRTUAL_ENV_BIN_DIR="$PWD/.venv/Scripts"
-        echo "VIRTUAL_ENV_BIN_DIR=$VIRTUAL_ENV_BIN_DIR" >> "$GITHUB_ENV"
-
-    - name: Static checks
-      run: |
-        source "$VIRTUAL_ENV_BIN_DIR/activate"
-        python -m tox -v -m static
-
-    - name: CI-compatible tests
-      run: |
-        source "$VIRTUAL_ENV_BIN_DIR/activate"
-        python -m tox -v -- -m 'not lmstudio'
-
-    - name: Upload coverage data
-      uses: actions/upload-artifact@v4
-      with:
-        name: coverage-data-${{ matrix.os }}-py${{ matrix.python-version }}
-        path: .coverage.*
-        include-hidden-files: true
-        if-no-files-found: ignore
+      - uses: actions/checkout@v4
+        with:
+          persist-credentials: false
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
 
+      - name: Get pip cache dir
+        id: pip-cache
+        run: |
+          echo "dir=$(python -m pip cache dir)" >> $GITHUB_OUTPUT
+
+      - name: Cache bootstrapping dependencies
+        uses: actions/cache@v4
+        with:
+          path: ${{ steps.pip-cache.outputs.dir }}
+          key: pip-${{ matrix.os }}-${{ matrix.python-version }}-v1-${{ hashFiles('pdm.lock') }}
+          restore-keys: |
+            pip-${{ matrix.os }}-${{ matrix.python-version }}-v1-
+
+      - name: Install PDM
+        run: |
+          # Ensure `pdm` uses the same version as specified in `pdm.lock`
+          # while avoiding the error raised by https://github.com/pypa/pip/issues/12889
+          python -m pip install --upgrade -r ci-bootstrap-requirements.txt
+
+      - name: Create development virtual environment
+        run: |
+          python -m pdm sync --no-self --dev
+          # Handle Windows vs non-Windows differences in .venv layout
+          VIRTUAL_ENV_BIN_DIR="$PWD/.venv/bin"
+          test -e "$VIRTUAL_ENV_BIN_DIR" || VIRTUAL_ENV_BIN_DIR="$PWD/.venv/Scripts"
+          echo "VIRTUAL_ENV_BIN_DIR=$VIRTUAL_ENV_BIN_DIR" >> "$GITHUB_ENV"
+
+      - name: Static checks
+        run: |
+          source "$VIRTUAL_ENV_BIN_DIR/activate"
+          python -m tox -v -m static
+
+      - name: CI-compatible tests
+        run: |
+          source "$VIRTUAL_ENV_BIN_DIR/activate"
+          python -m tox -v -- -m 'not lmstudio'
+
+      - name: Upload coverage data
+        uses: actions/upload-artifact@v4
+        with:
+          name: coverage-data-${{ matrix.os }}-py${{ matrix.python-version }}
+          path: .coverage.*
+          include-hidden-files: true
+          if-no-files-found: ignore
 
   # Coverage check based on https://hynek.me/articles/ditch-codecov-python/
   coverage:

From 1be7bff06f034152cf96e4d96162e86c194221aa Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 15 Aug 2025 11:35:37 -0400
Subject: [PATCH 02/11] Use docker image to run CI tests

---
 .github/actions/docker-daemon-run/action.yml | 85 ++++++++++++++++++++
 .github/workflows/test.yml                   | 69 ++++++++++++++--
 2 files changed, 147 insertions(+), 7 deletions(-)
 create mode 100644 .github/actions/docker-daemon-run/action.yml

diff --git a/.github/actions/docker-daemon-run/action.yml b/.github/actions/docker-daemon-run/action.yml
new file mode 100644
index 0000000..ba01968
--- /dev/null
+++ b/.github/actions/docker-daemon-run/action.yml
@@ -0,0 +1,85 @@
+name: Run Daemon Docker Container
+description: Pulls and runs LM Studio Daemon Docker image for testing
+
+inputs:
+  docker-image:
+    description: "Full Docker image name"
+    required: true
+  container-name:
+    description: "Name for the container"
+    required: false
+    default: "llmster-test"
+  port:
+    description: "Port to expose (host:container)"
+    required: false
+    default: "1234:1234"
+
+outputs:
+  container-id:
+    description: "The ID of the running container"
+    value: ${{ steps.run-container.outputs.container-id }}
+  container-name:
+    description: "The name of the running container"
+    value: ${{ inputs.container-name }}
+
+runs:
+  using: "composite"
+  steps:
+    - name: Pull Docker image
+      shell: bash
+      run: |
+        echo "Pulling image: ${{ inputs.docker-image }}"
+        docker pull ${{ inputs.docker-image }}
+
+    - name: Run container
+      id: run-container
+      shell: bash
+      run: |
+        echo "Starting container: ${{ inputs.container-name }}"
+        if [ "${{ inputs.use-local-image }}" = "true" ]; then
+          echo "Using local image: ${{ inputs.docker-image }}"
+        else
+          echo "Using registry image: ${{ inputs.docker-image }}"
+        fi
+        CONTAINER_ID=$(docker run -d --name ${{ inputs.container-name }} -p ${{ inputs.port }} ${{ inputs.docker-image }})
+        echo "Container ID: $CONTAINER_ID"
+        echo "container-id=$CONTAINER_ID" >> $GITHUB_OUTPUT
+
+        # Wait for container to become healthy
+        TIMEOUT=120  # timeout in seconds (increased to account for start-period)
+        START_TIME=$(date +%s)
+        END_TIME=$((START_TIME + TIMEOUT))
+
+        # Start with 1 second delay, then exponentially increase
+        DELAY=1
+        MAX_DELAY=16  # Cap maximum delay at 16 seconds
+
+        while [ $(date +%s) -lt $END_TIME ]; do
+          HEALTH_STATUS=$(docker inspect --format='{{.State.Health.Status}}' ${{ inputs.container-name }} 2>/dev/null || echo "unknown")
+
+          if [ "$HEALTH_STATUS" = "healthy" ]; then
+            echo "Container is running!"
+            break
+          elif [ "$HEALTH_STATUS" = "unhealthy" ]; then
+            echo "Container is unhealthy - exiting"
+            docker logs ${{ inputs.container-name }}
+            exit 1
+          fi
+
+          ELAPSED=$(($(date +%s) - START_TIME))
+
+          sleep $DELAY
+          DELAY=$((DELAY * 2))
+          if [ $DELAY -gt $MAX_DELAY ]; then
+            DELAY=$MAX_DELAY
+          fi
+        done
+
+        # Final check after waiting for the maximum timeout
+        # Print logs and the health status
+        if [ $(date +%s) -ge $END_TIME ]; then
+          echo "Container health check timed out after ${TIMEOUT} seconds"
+          echo "Final health status: $(docker inspect --format='{{.State.Health.Status}}' ${{ inputs.container-name }})"
+          docker logs ${{ inputs.container-name }}
+          exit 1
+        fi
diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index d7550c7..d2d0b7f 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -73,6 +73,50 @@ jobs:
           restore-keys: |
             pip-${{ matrix.os }}-${{ matrix.python-version }}-v1-
 
+      - name: Run the built image
+        id: run
+        uses: ./.github/actions/docker-daemon-run
+        with:
+          docker-image: lmstudio/llmster-preview:cpu
+          # Use the same port as the always on API server
+          port: "41343:1234"
+
+      - name: Download models for tests (Ubuntu only)
+        if: matrix.os == 'ubuntu-22.04'
+        run: |
+          echo "Downloading required models..."
+
+          # Download text LLMs
+          docker exec llmster lms get https://huggingface.co/hugging-quants/Llama-3.2-1B-Instruct-Q4_K_M-GGUF -y
+          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF -y
+          docker exec llmster lms get https://huggingface.co/ZiangWu/MobileVLM_V2-1.7B-GGUF -y
+
+          # Download vision LLM
+          docker exec llmster lms get ZiangWu/MobileVLM_V2-1.7B-GGUF -y
+
+          # Download additional model for speculative decoding examples
+          docker exec llmster lms get qwen2.5-0.5b-instruct -y
+
+          echo "Model downloads complete"
+
+      - name: Load models into LM Studio (Ubuntu only)
+        if: matrix.os == 'ubuntu-22.04'
+        run: |
+          echo "Loading models..."
+
+          # Load embedding model
+          docker exec llmster lms load nomic-embed-text-v1.5 --identifier text-embedding-nomic-embed-text-v1.5 -y
+
+          # Load text LLMs
+          docker exec llmster lms load llama-3.2-1b-instruct --identifier llama-3.2-1b-instruct -y
+          docker exec llmster lms load qwen2.5-7b-instruct --identifier qwen2.5-7b-instruct-1m -y
+          docker exec llmster lms load smollm2-135m-instruct --identifier smollm2-135m-instruct -y
+
+          # Load vision LLM
+          docker exec llmster lms load ZiangWu/MobileVLM_V2-1.7B-GGUF --identifier mobilevlm_v2-1.7b
+
+          echo "Model loading complete"
+
       - name: Install PDM
         run: |
           # Ensure `pdm` uses the same version as specified in `pdm.lock`
@@ -92,11 +136,18 @@ jobs:
           source "$VIRTUAL_ENV_BIN_DIR/activate"
           python -m tox -v -m static
 
-      - name: CI-compatible tests
+      - name: CI-compatible tests (Windows)
+        if: matrix.os == 'windows-2022'
         run: |
           source "$VIRTUAL_ENV_BIN_DIR/activate"
           python -m tox -v -- -m 'not lmstudio'
 
+      - name: All tests including LM Studio (Ubuntu)
+        if: matrix.os == 'ubuntu-22.04'
+        run: |
+          source "$VIRTUAL_ENV_BIN_DIR/activate"
+          python -m tox -v
+
       - name: Upload coverage data
         uses: actions/upload-artifact@v4
         with:
@@ -105,6 +156,12 @@ jobs:
           include-hidden-files: true
           if-no-files-found: ignore
 
+      - name: Stop LM Studio Docker container (Ubuntu only)
+        if: matrix.os == 'ubuntu-22.04' && always()
+        run: |
+          docker stop llmster || true
+          docker rm llmster || true
+
   # Coverage check based on https://hynek.me/articles/ditch-codecov-python/
   coverage:
     name: Combine & check coverage
@@ -143,12 +200,10 @@ jobs:
           # Report again and fail if under 50%.
           # Highest historical coverage: 65%
           # Last noted local test coverage level: 94%
-          # CI coverage percentage is low because many of the tests
-          # aren't CI compatible (they need a local LM Studio instance).
-          # It's only as high as it is because the generated data model
-          # classes make up such a large portion of the total SDK code.
-          # Accept anything over 50% until CI is set up to run LM Studio
-          # in headless mode, and hence is able to run end-to-end tests.
+          # Ubuntu runners now run LM Studio in Docker, so they achieve higher
+          # coverage than Windows runners (which skip LM Studio tests).
+          # The generated data model classes make up a large portion of the SDK code.
+          # Accept anything over 50% as Windows runners still skip LM Studio tests.
           coverage report --fail-under=50
 
       - name: Upload HTML report if check failed

From dc398156294be3bce44eab644aea50093a70fbc3 Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 15 Aug 2025 11:38:12 -0400
Subject: [PATCH 03/11] Remove loading of non-required model

---
 .github/workflows/test.yml | 1 -
 1 file changed, 1 deletion(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index d2d0b7f..e73e0e7 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -110,7 +110,6 @@ jobs:
           # Load text LLMs
           docker exec llmster lms load llama-3.2-1b-instruct --identifier llama-3.2-1b-instruct -y
           docker exec llmster lms load qwen2.5-7b-instruct --identifier qwen2.5-7b-instruct-1m -y
-          docker exec llmster lms load smollm2-135m-instruct --identifier smollm2-135m-instruct -y
 
           # Load vision LLM
           docker exec llmster lms load ZiangWu/MobileVLM_V2-1.7B-GGUF --identifier mobilevlm_v2-1.7b

From 286dff92d4cb07c940e803131e2401d382a6664b Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 15 Aug 2025 12:08:50 -0400
Subject: [PATCH 04/11] Run docker only on ubuntu runners

---
 .github/workflows/test.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index e73e0e7..ccb926c 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -74,6 +74,7 @@ jobs:
             pip-${{ matrix.os }}-${{ matrix.python-version }}-v1-
 
       - name: Run the built image
+        if: matrix.os == 'ubuntu-22.04'
         id: run
         uses: ./.github/actions/docker-daemon-run
         with:

From 86efcc5b643475fe7b06605d5a9ef8897d9df873 Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 15 Aug 2025 12:10:59 -0400
Subject: [PATCH 05/11] Correct container name

---
 .github/workflows/test.yml | 1 +
 1 file changed, 1 insertion(+)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index ccb926c..62acfc2 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -81,6 +81,7 @@ jobs:
           docker-image: lmstudio/llmster-preview:cpu
           # Use the same port as the always on API server
           port: "41343:1234"
+          container-name: llmster
 
       - name: Download models for tests (Ubuntu only)
         if: matrix.os == 'ubuntu-22.04'

From 3164f430ccd3882aae0e784c1a5e1f9aa25e7d8d Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 15 Aug 2025 12:16:38 -0400
Subject: [PATCH 06/11] Fix the redundant lms get

---
 .github/workflows/test.yml | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 62acfc2..8b34a50 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -93,9 +93,6 @@ jobs:
           docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF -y
           docker exec llmster lms get https://huggingface.co/ZiangWu/MobileVLM_V2-1.7B-GGUF -y
 
-          # Download vision LLM
-          docker exec llmster lms get ZiangWu/MobileVLM_V2-1.7B-GGUF -y
-
           # Download additional model for speculative decoding examples
           docker exec llmster lms get qwen2.5-0.5b-instruct -y
 

From ef31ddae46585bf6fd818eb5c564cd8df32213e2 Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 15 Aug 2025 12:19:02 -0400
Subject: [PATCH 07/11] Suppress download logs

---
 .github/workflows/test.yml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 8b34a50..ce5cb41 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -89,12 +89,12 @@ jobs:
           echo "Downloading required models..."
 
           # Download text LLMs
-          docker exec llmster lms get https://huggingface.co/hugging-quants/Llama-3.2-1B-Instruct-Q4_K_M-GGUF -y
-          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF -y
-          docker exec llmster lms get https://huggingface.co/ZiangWu/MobileVLM_V2-1.7B-GGUF -y
+          docker exec llmster lms get https://huggingface.co/hugging-quants/Llama-3.2-1B-Instruct-Q4_K_M-GGUF -y --quite
+          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF -y --quite
+          docker exec llmster lms get https://huggingface.co/ZiangWu/MobileVLM_V2-1.7B-GGUF -y --quite
 
           # Download additional model for speculative decoding examples
-          docker exec llmster lms get qwen2.5-0.5b-instruct -y
+          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF -y --quite
 
           echo "Model downloads complete"
 

From 081cdc3cb00cacad040f6b7aa725fb1d2735cc56 Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 15 Aug 2025 12:20:15 -0400
Subject: [PATCH 08/11] Fix spelling error

---
 .github/workflows/test.yml | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index ce5cb41..6a9f07b 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -89,12 +89,12 @@ jobs:
           echo "Downloading required models..."
 
           # Download text LLMs
-          docker exec llmster lms get https://huggingface.co/hugging-quants/Llama-3.2-1B-Instruct-Q4_K_M-GGUF -y --quite
-          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF -y --quite
-          docker exec llmster lms get https://huggingface.co/ZiangWu/MobileVLM_V2-1.7B-GGUF -y --quite
+          docker exec llmster lms get https://huggingface.co/hugging-quants/Llama-3.2-1B-Instruct-Q4_K_M-GGUF -y --quiet
+          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF -y --quiet
+          docker exec llmster lms get https://huggingface.co/ZiangWu/MobileVLM_V2-1.7B-GGUF -y --quiet
 
           # Download additional model for speculative decoding examples
-          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF -y --quite
+          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct-GGUF -y --quiet
 
           echo "Model downloads complete"
 

From 58f950a270e37a5362519ed980f050667fea0555 Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Fri, 15 Aug 2025 13:09:19 -0400
Subject: [PATCH 09/11] Seperate jobs

---
 .github/workflows/test.yml | 166 +++++++++++++++++++++++++------------
 1 file changed, 113 insertions(+), 53 deletions(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 6a9f07b..272fc76 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -34,21 +34,64 @@ defaults:
     shell: bash
 
 jobs:
-  tests:
-    runs-on: ${{ matrix.os }}
+  tests-windows:
+    runs-on: windows-2022
     strategy:
-      fail-fast: false # Always report results for all targets
-      max-parallel: 8
+      fail-fast: false
       matrix:
         python-version: ["3.10", "3.11", "3.12", "3.13"]
-        # While the main SDK is platform independent, the subprocess execution
-        # in the plugin runner and tests requires some Windows-specific code
-        # Note: a green tick in CI is currently misleading due to
-        # https://github.com/lmstudio-ai/lmstudio-python/issues/140
-        os: [ubuntu-22.04, windows-2022]
 
-    # Check https://github.com/actions/action-versions/tree/main/config/actions
-    # for latest versions if the standard actions start emitting warnings
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          persist-credentials: false
+
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Get pip cache dir
+        id: pip-cache
+        run: |
+          echo "dir=$(python -m pip cache dir)" >> $GITHUB_OUTPUT
+
+      - name: Cache bootstrapping dependencies
+        uses: actions/cache@v4
+        with:
+          path: ${{ steps.pip-cache.outputs.dir }}
+          key: pip-windows-2022-${{ matrix.python-version }}-v1-${{ hashFiles('pdm.lock') }}
+          restore-keys: |
+            pip-windows-2022-${{ matrix.python-version }}-v1-
+
+      - name: Install PDM
+        run: |
+          python -m pip install --upgrade -r ci-bootstrap-requirements.txt
+
+      - name: Create development virtual environment
+        run: |
+          python -m pdm sync --no-self --dev
+          echo "VIRTUAL_ENV_BIN_DIR=$PWD/.venv/Scripts" >> "$GITHUB_ENV"
+
+      - name: CI-compatible tests (Windows)
+        run: |
+          source "$VIRTUAL_ENV_BIN_DIR/activate"
+          python -m tox -v -- -m 'not lmstudio'
+
+      - name: Upload coverage data
+        uses: actions/upload-artifact@v4
+        with:
+          name: coverage-data-windows-py${{ matrix.python-version }}
+          path: .coverage.*
+          include-hidden-files: true
+          if-no-files-found: ignore
+
+  tests-docker:
+    runs-on: ubuntu-22.04
+    strategy:
+      fail-fast: false
+      matrix:
+        python-version: ["3.10", "3.11", "3.12", "3.13"]
 
     steps:
       - uses: actions/checkout@v4
@@ -69,22 +112,19 @@ jobs:
         uses: actions/cache@v4
         with:
           path: ${{ steps.pip-cache.outputs.dir }}
-          key: pip-${{ matrix.os }}-${{ matrix.python-version }}-v1-${{ hashFiles('pdm.lock') }}
+          key: pip-ubuntu-22.04-${{ matrix.python-version }}-v1-${{ hashFiles('pdm.lock') }}
           restore-keys: |
-            pip-${{ matrix.os }}-${{ matrix.python-version }}-v1-
+            pip-ubuntu-22.04-${{ matrix.python-version }}-v1-
 
       - name: Run the built image
-        if: matrix.os == 'ubuntu-22.04'
         id: run
         uses: ./.github/actions/docker-daemon-run
         with:
           docker-image: lmstudio/llmster-preview:cpu
-          # Use the same port as the always on API server
           port: "41343:1234"
           container-name: llmster
 
-      - name: Download models for tests (Ubuntu only)
-        if: matrix.os == 'ubuntu-22.04'
+      - name: Download models for tests
         run: |
           echo "Downloading required models..."
 
@@ -98,8 +138,7 @@ jobs:
 
           echo "Model downloads complete"
 
-      - name: Load models into LM Studio (Ubuntu only)
-        if: matrix.os == 'ubuntu-22.04'
+      - name: Load models into LM Studio
         run: |
           echo "Loading models..."
 
@@ -117,31 +156,14 @@ jobs:
 
       - name: Install PDM
         run: |
-          # Ensure `pdm` uses the same version as specified in `pdm.lock`
-          # while avoiding the error raised by https://github.com/pypa/pip/issues/12889
           python -m pip install --upgrade -r ci-bootstrap-requirements.txt
 
       - name: Create development virtual environment
         run: |
           python -m pdm sync --no-self --dev
-          # Handle Windows vs non-Windows differences in .venv layout
-          VIRTUAL_ENV_BIN_DIR="$PWD/.venv/bin"
-          test -e "$VIRTUAL_ENV_BIN_DIR" || VIRTUAL_ENV_BIN_DIR="$PWD/.venv/Scripts"
-          echo "VIRTUAL_ENV_BIN_DIR=$VIRTUAL_ENV_BIN_DIR" >> "$GITHUB_ENV"
+          echo "VIRTUAL_ENV_BIN_DIR=$PWD/.venv/bin" >> "$GITHUB_ENV"
 
-      - name: Static checks
-        run: |
-          source "$VIRTUAL_ENV_BIN_DIR/activate"
-          python -m tox -v -m static
-
-      - name: CI-compatible tests (Windows)
-        if: matrix.os == 'windows-2022'
-        run: |
-          source "$VIRTUAL_ENV_BIN_DIR/activate"
-          python -m tox -v -- -m 'not lmstudio'
-
-      - name: All tests including LM Studio (Ubuntu)
-        if: matrix.os == 'ubuntu-22.04'
+      - name: All tests including LM Studio
         run: |
           source "$VIRTUAL_ENV_BIN_DIR/activate"
           python -m tox -v
@@ -149,22 +171,65 @@ jobs:
       - name: Upload coverage data
         uses: actions/upload-artifact@v4
         with:
-          name: coverage-data-${{ matrix.os }}-py${{ matrix.python-version }}
+          name: coverage-data-docker-py${{ matrix.python-version }}
           path: .coverage.*
           include-hidden-files: true
           if-no-files-found: ignore
 
-      - name: Stop LM Studio Docker container (Ubuntu only)
-        if: matrix.os == 'ubuntu-22.04' && always()
+      - name: Stop LM Studio Docker container
+        if: always()
         run: |
           docker stop llmster || true
           docker rm llmster || true
 
-  # Coverage check based on https://hynek.me/articles/ditch-codecov-python/
-  coverage:
-    name: Combine & check coverage
+  # Coverage check for Docker tests (higher target since they include LM Studio tests)
+  coverage-docker:
+    name: Docker test coverage
+    if: always()
+    needs: tests-docker
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          persist-credentials: false
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.13"
+
+      - uses: hynek/setup-cached-uv@757bedc3f972eb7227a1aa657651f15a8527c817
+
+      - uses: actions/download-artifact@v4
+        with:
+          pattern: coverage-data-docker-*
+          merge-multiple: true
+
+      - name: Combine coverage & fail if it goes down
+        run: |
+          uv tool install 'coverage[toml]'
+
+          coverage combine
+          coverage html --skip-covered --skip-empty
+
+          # Report and write to summary.
+          coverage report --format=markdown >> $GITHUB_STEP_SUMMARY
+
+          # Higher target for Docker tests since they include LM Studio functionality
+          coverage report --fail-under=75
+
+      - name: Upload HTML report if check failed
+        uses: actions/upload-artifact@v4
+        with:
+          name: html-report-docker
+          path: htmlcov
+        if: ${{ failure() }}
+
+  # Coverage check for Windows tests (lower target since they skip LM Studio tests)
+  coverage-windows:
+    name: Windows test coverage
     if: always()
-    needs: tests
+    needs: tests-windows
     runs-on: ubuntu-latest
 
     steps:
@@ -174,15 +239,13 @@ jobs:
 
       - uses: actions/setup-python@v5
         with:
-          # Use latest Python, so it understands all syntax.
           python-version: "3.13"
 
-      # https://github.com/hynek/setup-cached-uv/releases/tag/v2.3.0
       - uses: hynek/setup-cached-uv@757bedc3f972eb7227a1aa657651f15a8527c817
 
       - uses: actions/download-artifact@v4
         with:
-          pattern: coverage-data-*
+          pattern: coverage-data-windows-*
           merge-multiple: true
 
       - name: Combine coverage & fail if it goes down
@@ -198,15 +261,12 @@ jobs:
           # Report again and fail if under 50%.
           # Highest historical coverage: 65%
           # Last noted local test coverage level: 94%
-          # Ubuntu runners now run LM Studio in Docker, so they achieve higher
-          # coverage than Windows runners (which skip LM Studio tests).
-          # The generated data model classes make up a large portion of the SDK code.
-          # Accept anything over 50% as Windows runners still skip LM Studio tests.
+          # Lower target for Windows tests since they skip LM Studio functionality
           coverage report --fail-under=50
 
       - name: Upload HTML report if check failed
         uses: actions/upload-artifact@v4
         with:
-          name: html-report
+          name: html-report-windows
           path: htmlcov
         if: ${{ failure() }}

From 152aeb785ab1105b0001a80183c3c00fedd96e61 Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Mon, 18 Aug 2025 14:34:28 -0400
Subject: [PATCH 10/11] Split the coverage into two and smaller quantization

---
 .github/workflows/test.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 272fc76..5e4d187 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -130,7 +130,7 @@ jobs:
 
           # Download text LLMs
           docker exec llmster lms get https://huggingface.co/hugging-quants/Llama-3.2-1B-Instruct-Q4_K_M-GGUF -y --quiet
-          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF -y --quiet
+          docker exec llmster lms get https://huggingface.co/Qwen/Qwen2.5-7B-Instruct-GGUF@q3_k_m -y --quiet
           docker exec llmster lms get https://huggingface.co/ZiangWu/MobileVLM_V2-1.7B-GGUF -y --quiet
 
           # Download additional model for speculative decoding examples

From ec632c07767cd434f7ca462e66622640a162f7ec Mon Sep 17 00:00:00 2001
From: Rugved Somwanshi <rugved@lmstudio.ai>
Date: Mon, 18 Aug 2025 14:45:02 -0400
Subject: [PATCH 11/11] Verbose loggin for debug

---
 .github/workflows/test.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml
index 5e4d187..03f6756 100644
--- a/.github/workflows/test.yml
+++ b/.github/workflows/test.yml
@@ -166,7 +166,7 @@ jobs:
       - name: All tests including LM Studio
         run: |
           source "$VIRTUAL_ENV_BIN_DIR/activate"
-          python -m tox -v
+          python -m tox -v -- -v --tb=short
 
       - name: Upload coverage data
         uses: actions/upload-artifact@v4