change push to chunked uploads from monolithic

2026-09-12 14:10:47 -04:00 · 2023-07-22 16:13:28 -07:00
374 changed files with 48246 additions and 75456 deletions
@@ -1,9 +1,7 @@
+build
+llama/build
+.venv
 .vscode
 ollama
 app
-macapp
-dist
-llm/llama.cpp
-.env
-.cache
-test_data
+web
@@ -1 +0,0 @@
-llm/ext_server/* linguist-vendored
@@ -1,60 +0,0 @@
-name: Bug report
-labels: [bug]
-description: Something isn't working right.
-body:
-  - type: textarea
-    id: description
-    attributes:
-      label: What is the issue?
-      description: What happened? What did you expect to happen?
-    validations:
-      required: true
-  - type: dropdown
-    id: os
-    attributes:
-      label: OS
-      description: Which operating system are you using?
-      multiple: true
-      options:
-        - Linux
-        - macOS
-        - Windows
-        - Docker
-        - WSL2
-    validations:
-      required: false
-  - type: dropdown
-    id: gpu
-    attributes:
-      label: GPU
-      description: Which GPU are you using?
-      multiple: true
-      options:
-        - Nvidia
-        - AMD
-        - Intel
-        - Apple
-        - Other
-    validations:
-      required: false
-  - type: dropdown
-    id: cpu
-    attributes:
-      label: CPU
-      description: Which CPU are you using?
-      multiple: true
-      options:
-        - Intel
-        - AMD
-        - Apple
-        - Other
-    validations:
-      required: false
-  - type: input
-    id: version
-    attributes:
-      label: Ollama version
-      description: What version of Ollama are you using? (`ollama --version`)
-      placeholder: e.g., 0.1.32
-    validations:
-      required: false
@@ -1,6 +0,0 @@
---
-name: Feature request
-about: Request a new feature
-labels: feature request
---
-
@@ -1,5 +0,0 @@
---
-name: Model request
-about: Request support for a new model to be added to Ollama
-labels: model request
---
@@ -1,8 +0,0 @@
-blank_issues_enabled: true
-contact_links:
-  - name: Help
-    url: https://discord.com/invite/ollama
-    about: Please join our Discord server for help using Ollama
-  - name: Troubleshooting
-    url: https://github.com/ollama/ollama/blob/main/docs/faq.md#faq
-    about: See the FAQ for common issues and solutions
@@ -1,24 +0,0 @@
-name: latest
-
-on:
-  release:
-    types: [released]
-
-jobs:
-  update-latest:
-    environment: release
-    runs-on: linux
-    steps:
-      - uses: actions/checkout@v4
-      - name: Login to Docker Hub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ vars.DOCKER_USER }}
-          password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
-      - name: Tag images as latest
-        env:
-          PUSH: "1"
-        shell: bash
-        run: |
-          export "VERSION=${GITHUB_REF_NAME#v}"
-          ./scripts/tag_latest.sh
@@ -1,473 +0,0 @@
-name: release
-
-on:
-  push:
-    tags:
-      - 'v*'
-
-jobs:
-  # Full build of the Mac assets
-  build-darwin:
-    runs-on: macos-12
-    environment: release
-    steps:
-      - uses: actions/checkout@v4
-      - name: Set Version
-        shell: bash
-        run: |
-          echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-          echo "RELEASE_VERSION=$(echo ${GITHUB_REF_NAME} | cut -f1 -d-)" >> $GITHUB_ENV
-      - name: key
-        env:
-          MACOS_SIGNING_KEY: ${{ secrets.MACOS_SIGNING_KEY }}
-          MACOS_SIGNING_KEY_PASSWORD: ${{ secrets.MACOS_SIGNING_KEY_PASSWORD }}
-        run: |
-          echo $MACOS_SIGNING_KEY | base64 --decode > certificate.p12
-          security create-keychain -p password build.keychain
-          security default-keychain -s build.keychain
-          security unlock-keychain -p password build.keychain
-          security import certificate.p12 -k build.keychain -P $MACOS_SIGNING_KEY_PASSWORD -T /usr/bin/codesign
-          security set-key-partition-list -S apple-tool:,apple:,codesign: -s -k password build.keychain
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - name: Build Darwin
-        env:
-          APPLE_IDENTITY: ${{ secrets.APPLE_IDENTITY }}
-          APPLE_PASSWORD: ${{ secrets.APPLE_PASSWORD }}
-          APPLE_TEAM_ID: ${{ vars.APPLE_TEAM_ID }}
-          APPLE_ID: ${{ vars.APPLE_ID }}
-          SDKROOT: /Applications/Xcode_13.4.1.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk
-          DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
-        run: |
-          ./scripts/build_darwin.sh
-
-      - uses: actions/upload-artifact@v4
-        with:
-          name: dist-darwin
-          path: |
-            dist/*arwin*
-            !dist/*-cov
-
-  # Windows builds take a long time to both install the dependencies and build, so parallelize
-  # CPU generation step
-  generate-windows-cpu:
-    environment: release
-    runs-on: windows
-    env:
-      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
-    steps:
-      - uses: actions/checkout@v4
-      - name: Set Version
-        shell: bash
-        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-      - uses: 'google-github-actions/auth@v2'
-        with:
-          project_id: 'ollama'
-          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
-      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
-      - name: install Windows SDK 8.1 to get signtool
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading SDK"
-          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
-          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
-          write-host "Win SDK 8.1 installed"
-          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
-      - name: install signing plugin
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading plugin"
-          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
-          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
-          write-host "Installing plugin"
-          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
-          write-host "plugin installed"
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - run: go get ./...
-      - run: |
-          $gopath=(get-command go).source | split-path -parent
-          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
-          cd $env:GITHUB_WORKSPACE
-          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
-          $env:PATH="$gopath;$env:PATH"
-          go generate -x ./...
-        name: go generate
-      - uses: actions/upload-artifact@v4
-        with:
-          name: generate-windows-cpu
-          path: |
-            llm/build/**/bin/*
-            llm/build/**/*.a
-            dist/windows-amd64/**
-
-  # ROCm generation step
-  generate-windows-rocm:
-    environment: release
-    runs-on: windows
-    env:
-      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
-    steps:
-      - uses: actions/checkout@v4
-      - name: Set Version
-        shell: bash
-        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-      - uses: 'google-github-actions/auth@v2'
-        with:
-          project_id: 'ollama'
-          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
-      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
-      - name: install Windows SDK 8.1 to get signtool
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading SDK"
-          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
-          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
-          write-host "Win SDK 8.1 installed"
-          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
-      - name: install signing plugin
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading plugin"
-          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
-          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
-          write-host "Installing plugin"
-          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
-          write-host "plugin installed"
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - name: 'Install ROCm'
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading AMD HIP Installer"
-          Invoke-WebRequest -Uri "https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-23.Q4-WinSvr2022-For-HIP.exe" -OutFile "${env:RUNNER_TEMP}\rocm-install.exe"
-          write-host "Installing AMD HIP"
-          Start-Process "${env:RUNNER_TEMP}\rocm-install.exe" -ArgumentList '-install' -NoNewWindow -Wait
-          write-host "Completed AMD HIP"
-      - name: 'Verify ROCm'
-        run: |
-          & 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' --version
-      - run: go get ./...
-      - run: |
-          $gopath=(get-command go).source | split-path -parent
-          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
-          cd $env:GITHUB_WORKSPACE
-          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
-          $env:PATH="$gopath;$env:PATH"
-          $env:OLLAMA_SKIP_CPU_GENERATE="1"
-          $env:HIP_PATH=$(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | split-path | split-path)
-          go generate -x ./...
-        name: go generate
-      - name: 'gather rocm dependencies'
-        run: |
-          $HIP_PATH=$(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | split-path | split-path)
-          md "dist\deps\bin\rocblas\library"
-          cp "${HIP_PATH}\bin\hipblas.dll" "dist\deps\bin\"
-          cp "${HIP_PATH}\bin\rocblas.dll" "dist\deps\bin\"
-          cp "${HIP_PATH}\bin\rocblas\library\*" "dist\deps\bin\rocblas\library\"
-      - uses: actions/upload-artifact@v4
-        with:
-          name: generate-windows-rocm
-          path: |
-            llm/build/**/bin/*
-            dist/windows-amd64/**
-      - uses: actions/upload-artifact@v4
-        with:
-          name: windows-rocm-deps
-          path: dist/deps/*
-
-  # CUDA generation step
-  generate-windows-cuda:
-    environment: release
-    runs-on: windows
-    env:
-      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
-    steps:
-      - uses: actions/checkout@v4
-      - name: Set Version
-        shell: bash
-        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-      - uses: 'google-github-actions/auth@v2'
-        with:
-          project_id: 'ollama'
-          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
-      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
-      - name: install Windows SDK 8.1 to get signtool
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading SDK"
-          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
-          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
-          write-host "Win SDK 8.1 installed"
-          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
-      - name: install signing plugin
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading plugin"
-          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
-          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
-          write-host "Installing plugin"
-          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
-          write-host "plugin installed"
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - name: 'Install CUDA'
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading CUDA Installer"
-          Invoke-WebRequest -Uri "https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.89_win10.exe" -OutFile "${env:RUNNER_TEMP}\cuda-install.exe"
-          write-host "Installing CUDA"
-          Start-Process "${env:RUNNER_TEMP}\cuda-install.exe" -ArgumentList '-s' -NoNewWindow -Wait
-          write-host "Completed CUDA"
-          $cudaPath=((resolve-path "c:\Program Files\NVIDIA*\CUDA\v*\bin\nvcc.exe")[0].path | split-path | split-path)
-          $cudaVer=($cudaPath | split-path -leaf ) -replace 'v(\d+).(\d+)', '$1_$2' 
-          echo "$cudaPath\bin" >> $env:GITHUB_PATH
-          echo "CUDA_PATH=$cudaPath" >> $env:GITHUB_ENV
-          echo "CUDA_PATH_V${cudaVer}=$cudaPath" >> $env:GITHUB_ENV
-          echo "CUDA_PATH_VX_Y=CUDA_PATH_V${cudaVer}" >> $env:GITHUB_ENV
-      - name: 'Verify CUDA'
-        run: nvcc -V
-      - run: go get ./...
-      - name: go generate
-        run: |
-          $gopath=(get-command go).source | split-path -parent
-          $cudabin=(get-command nvcc).source | split-path
-          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
-          cd $env:GITHUB_WORKSPACE
-          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
-          $env:PATH="$gopath;$cudabin;$env:PATH"
-          $env:OLLAMA_SKIP_CPU_GENERATE="1"
-          go generate -x ./...
-      - name: 'gather cuda dependencies'
-        run: |
-          $NVIDIA_DIR=(resolve-path 'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\*\bin\')[0]
-          md "dist\deps"
-          cp "${NVIDIA_DIR}\cudart64_*.dll" "dist\deps\"
-          cp "${NVIDIA_DIR}\cublas64_*.dll" "dist\deps\"
-          cp "${NVIDIA_DIR}\cublasLt64_*.dll" "dist\deps\"
-      - uses: actions/upload-artifact@v4
-        with:
-          name: generate-windows-cuda
-          path: |
-            llm/build/**/bin/*
-            dist/windows-amd64/**
-      - uses: actions/upload-artifact@v4
-        with:
-          name: windows-cuda-deps
-          path: dist/deps/*
-
-  # Import the prior generation steps and build the final windows assets
-  build-windows:
-    environment: release
-    runs-on: windows
-    needs:
-      - generate-windows-cuda
-      - generate-windows-rocm
-      - generate-windows-cpu
-    env:
-      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          submodules: recursive
-      - name: Set Version
-        shell: bash
-        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-      - uses: 'google-github-actions/auth@v2'
-        with:
-          project_id: 'ollama'
-          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
-      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
-      - name: install Windows SDK 8.1 to get signtool
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading SDK"
-          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
-          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
-          write-host "Win SDK 8.1 installed"
-          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
-      - name: install signing plugin
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading plugin"
-          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
-          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
-          write-host "Installing plugin"
-          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
-          write-host "plugin installed"
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - run: go get
-      - uses: actions/download-artifact@v4
-        with:
-          name: generate-windows-cpu
-      - uses: actions/download-artifact@v4
-        with:
-          name: generate-windows-cuda
-      - uses: actions/download-artifact@v4
-        with:
-          name: windows-cuda-deps
-      - uses: actions/download-artifact@v4
-        with:
-          name: windows-rocm-deps
-      - uses: actions/download-artifact@v4
-        with:
-          name: generate-windows-rocm
-      - run: dir llm/build
-      - run: |
-          $gopath=(get-command go).source | split-path -parent
-          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
-          cd $env:GITHUB_WORKSPACE
-          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
-          $env:PATH="$gopath;$env:PATH"
-          $env:OLLAMA_SKIP_GENERATE="1"
-          & .\scripts\build_windows.ps1
-      - uses: actions/upload-artifact@v4
-        with:
-          name: dist-windows
-          path: |
-            dist/OllamaSetup.exe
-            dist/ollama-windows-*.zip
-
-  # Linux x86 assets built using the container based build
-  build-linux-amd64:
-    environment: release
-    runs-on: linux
-    env:
-      OLLAMA_SKIP_MANIFEST_CREATE: '1'
-      BUILD_ARCH: amd64
-      PUSH: '1'
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          submodules: recursive
-      - name: Set Version
-        shell: bash
-        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-      - name: Login to Docker Hub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ vars.DOCKER_USER }}
-          password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
-      - run: |
-          ./scripts/build_linux.sh
-          ./scripts/build_docker.sh
-          mv dist/deps/* dist/
-      - uses: actions/upload-artifact@v4
-        with:
-          name: dist-linux-amd64
-          path: |
-            dist/*linux*
-            !dist/*-cov
-
-  # Linux ARM assets built using the container based build
-  # (at present, docker isn't pre-installed on arm ubunutu images)
-  build-linux-arm64:
-    environment: release
-    runs-on: linux-arm64
-    env:
-      OLLAMA_SKIP_MANIFEST_CREATE: '1'
-      BUILD_ARCH: arm64
-      PUSH: '1'
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          submodules: recursive
-      - name: Set Version
-        shell: bash
-        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-      - name: 'Install Docker'
-        run: |
-          # Add Docker's official GPG key:
-          env
-          uname -a
-          sudo apt-get update
-          sudo apt-get install -y ca-certificates curl
-          sudo install -m 0755 -d /etc/apt/keyrings
-          sudo curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyrings/docker.asc
-          sudo chmod a+r /etc/apt/keyrings/docker.asc
-
-          # Add the repository to Apt sources:
-          echo \
-            "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \
-            $(. /etc/os-release && echo "$VERSION_CODENAME") stable" | \
-            sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
-          sudo apt-get update
-          sudo apt-get install -y docker-ce docker-ce-cli containerd.io
-          sudo usermod -aG docker $USER
-          sudo apt-get install acl
-          sudo setfacl --modify user:$USER:rw /var/run/docker.sock
-      - name: Login to Docker Hub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ vars.DOCKER_USER }}
-          password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
-      - run: |
-          ./scripts/build_linux.sh
-          ./scripts/build_docker.sh
-      - uses: actions/upload-artifact@v4
-        with:
-          name: dist-linux-arm64
-          path: |
-            dist/*linux*
-            !dist/*-cov
-
-  # Aggregate all the assets and ship a release
-  release:
-    needs:
-      - build-darwin
-      - build-windows
-      - build-linux-amd64
-      - build-linux-arm64
-    runs-on: linux
-    environment: release
-    permissions:
-      contents: write
-    env:
-      OLLAMA_SKIP_IMAGE_BUILD: '1'
-      PUSH: '1'
-    steps:
-      - uses: actions/checkout@v4
-      - name: Set Version
-        shell: bash
-        run: |
-          echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
-          echo "RELEASE_VERSION=$(echo ${GITHUB_REF_NAME} | cut -f1 -d-)" >> $GITHUB_ENV
-      - name: Login to Docker Hub
-        uses: docker/login-action@v3
-        with:
-          username: ${{ vars.DOCKER_USER }}
-          password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
-      - run: ./scripts/build_docker.sh
-      - name: Retrieve built artifact
-        uses: actions/download-artifact@v4
-        with:
-          path: dist
-          pattern: dist-*
-          merge-multiple: true
-      - run: |
-          ls -lh dist/
-          (cd dist; sha256sum * > sha256sum.txt)
-          cat dist/sha256sum.txt
-      - uses: ncipollo/release-action@v1
-        with:
-          name: ${{ env.RELEASE_VERSION }}
-          allowUpdates: true
-          artifacts: 'dist/*'
-          draft: true
-          prerelease: true
-          omitBodyDuringUpdate: true
-          generateReleaseNotes: true
-          omitDraftDuringUpdate: true
-          omitPrereleaseDuringUpdate: true
-          replacesArtifacts: true
@@ -1,319 +0,0 @@
-name: test
-
-concurrency:
-  # For PRs, later CI runs preempt previous ones. e.g. a force push on a PR
-  # cancels running CI jobs and starts all new ones.
-  #
-  # For non-PR pushes, concurrency.group needs to be unique for every distinct
-  # CI run we want to have happen. Use run_id, which in practice means all
-  # non-PR CI runs will be allowed to run without preempting each other.
-  group: ${{ github.workflow }}-$${{ github.pull_request.number || github.run_id }}
-  cancel-in-progress: true
-
-on:
-  pull_request:
-    paths:
-      - '**/*'
-      - '!docs/**'
-      - '!README.md'
-
-jobs:
-  changes:
-    runs-on: ubuntu-latest
-    outputs:
-      GENERATE: ${{ steps.changes.outputs.GENERATE }}
-      GENERATE_CUDA: ${{ steps.changes.outputs.GENERATE_CUDA }}
-      GENERATE_ROCM: ${{ steps.changes.outputs.GENERATE_ROCM }}
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          fetch-depth: 0
-      - id: changes
-        run: |
-          changed() {
-            git diff-tree -r --no-commit-id --name-only \
-              $(git merge-base ${{ github.event.pull_request.base.sha }} ${{ github.event.pull_request.head.sha }}) \
-              ${{ github.event.pull_request.head.sha }} \
-              | xargs python3 -c "import sys; print(any([x.startswith('$1') for x in sys.argv[1:]]))"
-          }
-
-          {
-            echo GENERATE=$(changed llm/)
-            echo GENERATE_CUDA=$(changed llm/)
-            echo GENERATE_ROCM=$(changed llm/)
-          } >>$GITHUB_OUTPUT
-
-  generate:
-    needs: [changes]
-    if: ${{ needs.changes.outputs.GENERATE == 'True' }}
-    strategy:
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-2019]
-        arch: [amd64, arm64]
-        exclude:
-          - os: ubuntu-latest
-            arch: arm64
-          - os: windows-2019
-            arch: arm64
-    runs-on: ${{ matrix.os }}
-    env:
-      GOARCH: ${{ matrix.arch }}
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - run: go get ./...
-      - run: |
-          $gopath=(get-command go).source | split-path -parent
-          $gccpath=(get-command gcc).source | split-path -parent
-          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
-          cd $env:GITHUB_WORKSPACE
-          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
-          $env:PATH="$gopath;$gccpath;$env:PATH"
-          echo $env:PATH
-          go generate -x ./...
-        if: ${{ startsWith(matrix.os, 'windows-') }}
-        name: 'Windows Go Generate'
-      - run: go generate -x ./...
-        if: ${{ ! startsWith(matrix.os, 'windows-') }}
-        name: 'Unix Go Generate'
-      - uses: actions/upload-artifact@v4
-        with:
-          name: ${{ matrix.os }}-${{ matrix.arch }}-libraries
-          path: |
-            llm/build/**/bin/*
-            llm/build/**/*.a
-  generate-cuda:
-    needs: [changes]
-    if: ${{ needs.changes.outputs.GENERATE_CUDA == 'True' }}
-    strategy:
-      matrix:
-        cuda-version:
-          - '11.8.0'
-    runs-on: linux
-    container: nvidia/cuda:${{ matrix.cuda-version }}-devel-ubuntu20.04
-    steps:
-      - run: |
-          apt-get update && apt-get install -y git build-essential curl
-          curl -fsSL https://github.com/Kitware/CMake/releases/download/v3.28.1/cmake-3.28.1-linux-x86_64.tar.gz \
-            | tar -zx -C /usr --strip-components 1
-        env:
-          DEBIAN_FRONTEND: noninteractive
-      - uses: actions/checkout@v4
-      - uses: actions/setup-go@v4
-        with:
-          go-version-file: go.mod
-          cache: true
-      - run: go get ./...
-      - run: |
-          git config --global --add safe.directory /__w/ollama/ollama
-          go generate -x ./...
-        env:
-          OLLAMA_SKIP_CPU_GENERATE: '1'
-      - uses: actions/upload-artifact@v4
-        with:
-          name: cuda-${{ matrix.cuda-version }}-libraries
-          path: |
-            llm/build/**/bin/*
-            dist/windows-amd64/**
-  generate-rocm:
-    needs: [changes]
-    if: ${{ needs.changes.outputs.GENERATE_ROCM == 'True' }}
-    strategy:
-      matrix:
-        rocm-version:
-          - '6.0.2'
-    runs-on: linux
-    container: rocm/dev-ubuntu-20.04:${{ matrix.rocm-version }}
-    steps:
-      - run: |
-          apt-get update && apt-get install -y git build-essential curl rocm-libs
-          curl -fsSL https://github.com/Kitware/CMake/releases/download/v3.28.1/cmake-3.28.1-linux-x86_64.tar.gz \
-            | tar -zx -C /usr --strip-components 1
-        env:
-          DEBIAN_FRONTEND: noninteractive
-      - uses: actions/checkout@v4
-      - uses: actions/setup-go@v4
-        with:
-          go-version-file: go.mod
-          cache: true
-      - run: go get ./...
-      - run: |
-          git config --global --add safe.directory /__w/ollama/ollama
-          go generate -x ./...
-        env:
-          OLLAMA_SKIP_CPU_GENERATE: '1'
-      - uses: actions/upload-artifact@v4
-        with:
-          name: rocm-${{ matrix.rocm-version }}-libraries
-          path: |
-            llm/build/**/bin/*
-            dist/windows-amd64/**
-
-  # ROCm generation step
-  generate-windows-rocm:
-    needs: [changes]
-    if: ${{ needs.changes.outputs.GENERATE_ROCM == 'True' }}
-    runs-on: windows
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - name: 'Install ROCm'
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading AMD HIP Installer"
-          Invoke-WebRequest -Uri "https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-23.Q4-WinSvr2022-For-HIP.exe" -OutFile "${env:RUNNER_TEMP}\rocm-install.exe"
-          write-host "Installing AMD HIP"
-          Start-Process "${env:RUNNER_TEMP}\rocm-install.exe" -ArgumentList '-install' -NoNewWindow -Wait
-          write-host "Completed AMD HIP"
-      - name: 'Verify ROCm'
-        run: |
-          & 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' --version
-      - run: go get ./...
-      - run: |
-          $gopath=(get-command go).source | split-path -parent
-          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
-          cd $env:GITHUB_WORKSPACE
-          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
-          $env:PATH="$gopath;$env:PATH"
-          $env:OLLAMA_SKIP_CPU_GENERATE="1"
-          $env:HIP_PATH=$(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | split-path | split-path)
-          go generate -x ./...
-        name: go generate
-        env:
-          OLLAMA_SKIP_CPU_GENERATE: '1'
-      # TODO - do we need any artifacts?
-
-  # CUDA generation step
-  generate-windows-cuda:
-    needs: [changes]
-    if: ${{ needs.changes.outputs.GENERATE_CUDA == 'True' }}
-    runs-on: windows
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - name: 'Install CUDA'
-        run: |
-          $ErrorActionPreference = "Stop"
-          write-host "downloading CUDA Installer"
-          Invoke-WebRequest -Uri "https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.89_win10.exe" -OutFile "${env:RUNNER_TEMP}\cuda-install.exe"
-          write-host "Installing CUDA"
-          Start-Process "${env:RUNNER_TEMP}\cuda-install.exe" -ArgumentList '-s' -NoNewWindow -Wait
-          write-host "Completed CUDA"
-          $cudaPath=((resolve-path "c:\Program Files\NVIDIA*\CUDA\v*\bin\nvcc.exe")[0].path | split-path | split-path)
-          $cudaVer=($cudaPath | split-path -leaf ) -replace 'v(\d+).(\d+)', '$1_$2' 
-          echo "$cudaPath\bin" >> $env:GITHUB_PATH
-          echo "CUDA_PATH=$cudaPath" >> $env:GITHUB_ENV
-          echo "CUDA_PATH_V${cudaVer}=$cudaPath" >> $env:GITHUB_ENV
-          echo "CUDA_PATH_VX_Y=CUDA_PATH_V${cudaVer}" >> $env:GITHUB_ENV
-      - name: 'Verify CUDA'
-        run: nvcc -V
-      - run: go get ./...
-      - name: go generate
-        run: |
-          $gopath=(get-command go).source | split-path -parent
-          $cudabin=(get-command nvcc).source | split-path
-          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
-          cd $env:GITHUB_WORKSPACE
-          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
-          $env:PATH="$gopath;$cudabin;$env:PATH"
-          $env:OLLAMA_SKIP_CPU_GENERATE="1"
-          go generate -x ./...
-        env:
-          OLLAMA_SKIP_CPU_GENERATE: '1'
-      # TODO - do we need any artifacts?
-
-  lint:
-    strategy:
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-2019]
-        arch: [amd64, arm64]
-        exclude:
-          - os: ubuntu-latest
-            arch: arm64
-          - os: windows-2019
-            arch: arm64
-          - os: macos-latest
-            arch: amd64
-    runs-on: ${{ matrix.os }}
-    env:
-      GOARCH: ${{ matrix.arch }}
-      CGO_ENABLED: '1'
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          submodules: recursive
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: false
-      - run: |
-          case ${{ matrix.arch }} in
-            amd64) echo ARCH=x86_64 ;;
-            arm64) echo ARCH=arm64 ;;
-          esac >>$GITHUB_ENV
-        shell: bash
-      - run: |
-          mkdir -p llm/build/linux/$ARCH/stub/bin
-          touch llm/build/linux/$ARCH/stub/bin/ollama_llama_server
-        if: ${{ startsWith(matrix.os, 'ubuntu-') }}
-      - run: |
-          mkdir -p llm/build/darwin/$ARCH/stub/bin
-          touch llm/build/darwin/$ARCH/stub/bin/ollama_llama_server
-        if: ${{ startsWith(matrix.os, 'macos-') }}
-      - uses: golangci/golangci-lint-action@v4
-        with:
-          args: --timeout 8m0s -v
-  test:
-    strategy:
-      matrix:
-        os: [ubuntu-latest, macos-latest, windows-2019]
-        arch: [amd64]
-        exclude:
-          - os: ubuntu-latest
-            arch: arm64
-          - os: windows-2019
-            arch: arm64
-    runs-on: ${{ matrix.os }}
-    env:
-      GOARCH: ${{ matrix.arch }}
-      CGO_ENABLED: '1'
-      OLLAMA_CPU_TARGET: 'static'
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          submodules: recursive
-      - uses: actions/setup-go@v5
-        with:
-          go-version-file: go.mod
-          cache: true
-      - run: |
-          case ${{ matrix.arch }} in
-            amd64) echo ARCH=x86_64 ;;
-            arm64) echo ARCH=arm64 ;;
-          esac >>$GITHUB_ENV
-        shell: bash
-      - run: |
-          mkdir -p llm/build/linux/$ARCH/stub/bin
-          touch llm/build/linux/$ARCH/stub/bin/ollama_llama_server
-        if: ${{ startsWith(matrix.os, 'ubuntu-') }}
-      - run: |
-          mkdir -p llm/build/darwin/$ARCH/stub/bin
-          touch llm/build/darwin/$ARCH/stub/bin/ollama_llama_server
-        if: ${{ startsWith(matrix.os, 'macos-') }}
-        shell: bash
-      - run: go generate ./...
-      - run: go build
-      - run: go test -v ./...
-      - uses: actions/upload-artifact@v4
-        with:
-          name: ${{ matrix.os }}-binaries
-          path: ollama
@@ -5,10 +5,3 @@
 .swp
 dist
 ollama
-ggml-metal.metal
-.cache
-*.exe
-.idea
-test_data
-*.crt
-llm/build
@@ -1,4 +0,0 @@
-[submodule "llama.cpp"]
-	path = llm/llama.cpp
-	url = https://github.com/ggerganov/llama.cpp.git
-	shallow = true
@@ -1,17 +0,0 @@
-run:
-  timeout: 5m
-linters:
-  enable:
-    - asasalint
-    - bidichk
-    - bodyclose
-    - containedctx
-    - contextcheck
-    - exportloopref
-    - gocheckcompilerdirectives
-    # FIXME: for some reason this errors on windows
-    # - gofmt
-    # - goimports
-    - misspell
-    - nilerr
-    - unused
@@ -1,144 +1,15 @@
-ARG GOLANG_VERSION=1.22.1
-ARG CMAKE_VERSION=3.22.1
-# this CUDA_VERSION corresponds with the one specified in docs/gpu.md
-ARG CUDA_VERSION=11.3.1
-ARG ROCM_VERSION=6.0.2
-
-# Copy the minimal context we need to run the generate scripts
-FROM scratch AS llm-code
-COPY .git .git
-COPY .gitmodules .gitmodules
-COPY llm llm
-
-FROM --platform=linux/amd64 nvidia/cuda:$CUDA_VERSION-devel-centos7 AS cuda-build-amd64
-ARG CMAKE_VERSION
-COPY ./scripts/rh_linux_deps.sh /
-RUN CMAKE_VERSION=${CMAKE_VERSION} sh /rh_linux_deps.sh
-ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
-COPY --from=llm-code / /go/src/github.com/ollama/ollama/
-WORKDIR /go/src/github.com/ollama/ollama/llm/generate
-ARG CGO_CFLAGS
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
-
-FROM --platform=linux/arm64 nvidia/cuda:$CUDA_VERSION-devel-rockylinux8 AS cuda-build-arm64
-ARG CMAKE_VERSION
-COPY ./scripts/rh_linux_deps.sh /
-RUN CMAKE_VERSION=${CMAKE_VERSION} sh /rh_linux_deps.sh
-ENV PATH /opt/rh/gcc-toolset-10/root/usr/bin:$PATH
-COPY --from=llm-code / /go/src/github.com/ollama/ollama/
-WORKDIR /go/src/github.com/ollama/ollama/llm/generate
-ARG CGO_CFLAGS
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
-
-FROM --platform=linux/amd64 rocm/dev-centos-7:${ROCM_VERSION}-complete AS rocm-build-amd64
-ARG CMAKE_VERSION
-COPY ./scripts/rh_linux_deps.sh /
-RUN CMAKE_VERSION=${CMAKE_VERSION} sh /rh_linux_deps.sh
-ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
-ENV LIBRARY_PATH /opt/amdgpu/lib64
-COPY --from=llm-code / /go/src/github.com/ollama/ollama/
-WORKDIR /go/src/github.com/ollama/ollama/llm/generate
-ARG CGO_CFLAGS
-ARG AMDGPU_TARGETS
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
-RUN mkdir /tmp/scratch && \
-    for dep in $(zcat /go/src/github.com/ollama/ollama/llm/build/linux/x86_64/rocm*/bin/deps.txt.gz) ; do \
-        cp ${dep} /tmp/scratch/ || exit 1 ; \
-    done && \
-    (cd /opt/rocm/lib && tar cf - rocblas/library) | (cd /tmp/scratch/ && tar xf - ) && \
-    mkdir -p /go/src/github.com/ollama/ollama/dist/deps/ && \
-    (cd /tmp/scratch/ && tar czvf /go/src/github.com/ollama/ollama/dist/deps/ollama-linux-amd64-rocm.tgz . )
-
-
-FROM --platform=linux/amd64 centos:7 AS cpu-builder-amd64
-ARG CMAKE_VERSION
-ARG GOLANG_VERSION
-COPY ./scripts/rh_linux_deps.sh /
-RUN CMAKE_VERSION=${CMAKE_VERSION} GOLANG_VERSION=${GOLANG_VERSION} sh /rh_linux_deps.sh
-ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
-COPY --from=llm-code / /go/src/github.com/ollama/ollama/
-ARG OLLAMA_CUSTOM_CPU_DEFS
-ARG CGO_CFLAGS
-WORKDIR /go/src/github.com/ollama/ollama/llm/generate
-
-FROM --platform=linux/amd64 cpu-builder-amd64 AS static-build-amd64
-RUN OLLAMA_CPU_TARGET="static" sh gen_linux.sh
-FROM --platform=linux/amd64 cpu-builder-amd64 AS cpu-build-amd64
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu" sh gen_linux.sh
-FROM --platform=linux/amd64 cpu-builder-amd64 AS cpu_avx-build-amd64
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu_avx" sh gen_linux.sh
-FROM --platform=linux/amd64 cpu-builder-amd64 AS cpu_avx2-build-amd64
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu_avx2" sh gen_linux.sh
-
-FROM --platform=linux/arm64 centos:7 AS cpu-builder-arm64
-ARG CMAKE_VERSION
-ARG GOLANG_VERSION
-COPY ./scripts/rh_linux_deps.sh /
-RUN CMAKE_VERSION=${CMAKE_VERSION} GOLANG_VERSION=${GOLANG_VERSION} sh /rh_linux_deps.sh
-ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
-COPY --from=llm-code / /go/src/github.com/ollama/ollama/
-ARG OLLAMA_CUSTOM_CPU_DEFS
-ARG CGO_CFLAGS
-WORKDIR /go/src/github.com/ollama/ollama/llm/generate
-
-FROM --platform=linux/arm64 cpu-builder-arm64 AS static-build-arm64
-RUN OLLAMA_CPU_TARGET="static" sh gen_linux.sh
-FROM --platform=linux/arm64 cpu-builder-arm64 AS cpu-build-arm64
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu" sh gen_linux.sh
-
-
-# Intermediate stage used for ./scripts/build_linux.sh
-FROM --platform=linux/amd64 cpu-build-amd64 AS build-amd64
-ENV CGO_ENABLED 1
-WORKDIR /go/src/github.com/ollama/ollama
+FROM golang:1.20
+WORKDIR /go/src/github.com/jmorganca/ollama
 COPY . .
-COPY --from=static-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=cpu_avx-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=cpu_avx2-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=cuda-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=rocm-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=rocm-build-amd64 /go/src/github.com/ollama/ollama/dist/deps/ ./dist/deps/
-ARG GOFLAGS
-ARG CGO_CFLAGS
-RUN go build -trimpath .
+RUN CGO_ENABLED=1 go build -ldflags '-linkmode external -extldflags "-static"' .

-# Intermediate stage used for ./scripts/build_linux.sh
-FROM --platform=linux/arm64 cpu-build-arm64 AS build-arm64
-ENV CGO_ENABLED 1
-ARG GOLANG_VERSION
-WORKDIR /go/src/github.com/ollama/ollama
-COPY . .
-COPY --from=static-build-arm64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=cuda-build-arm64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-ARG GOFLAGS
-ARG CGO_CFLAGS
-RUN go build -trimpath .
-
-# Runtime stages
-FROM --platform=linux/amd64 ubuntu:22.04 as runtime-amd64
-RUN apt-get update && apt-get install -y ca-certificates
-COPY --from=build-amd64 /go/src/github.com/ollama/ollama/ollama /bin/ollama
-FROM --platform=linux/arm64 ubuntu:22.04 as runtime-arm64
-RUN apt-get update && apt-get install -y ca-certificates
-COPY --from=build-arm64 /go/src/github.com/ollama/ollama/ollama /bin/ollama
-
-# Radeon images are much larger so we keep it distinct from the CPU/CUDA image
-FROM --platform=linux/amd64 rocm/dev-centos-7:${ROCM_VERSION}-complete as runtime-rocm
-RUN update-pciids
-COPY --from=build-amd64 /go/src/github.com/ollama/ollama/ollama /bin/ollama
+FROM alpine
+COPY --from=0 /go/src/github.com/jmorganca/ollama/ollama /bin/ollama
 EXPOSE 11434
-ENV OLLAMA_HOST 0.0.0.0
-
+ARG USER=ollama
+ARG GROUP=ollama
+RUN addgroup -g 1000 $GROUP && adduser -u 1000 -DG $GROUP $USER
+USER $USER:$GROUP
 ENTRYPOINT ["/bin/ollama"]
-CMD ["serve"]
-
-FROM runtime-$TARGETARCH
-EXPOSE 11434
 ENV OLLAMA_HOST 0.0.0.0
-ENV PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
-ENV LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64
-ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
-ENV NVIDIA_VISIBLE_DEVICES=all
-
-ENTRYPOINT ["/bin/ollama"]
 CMD ["serve"]
@@ -1,114 +1,74 @@
 <div align="center">
-  <img alt="ollama" height="200px" src="https://github.com/ollama/ollama/assets/3325447/0d0b44e2-8f4a-4e99-9b52-a5c1c741c8f7">
+  <picture>
+    <source media="(prefers-color-scheme: dark)" height="200px" srcset="https://github.com/jmorganca/ollama/assets/3325447/56ea1849-1284-4645-8970-956de6e51c3c">
+    <img alt="logo" height="200px" src="https://github.com/jmorganca/ollama/assets/3325447/0d0b44e2-8f4a-4e99-9b52-a5c1c741c8f7">
+  </picture>
 </div>

 # Ollama

 [![Discord](https://dcbadge.vercel.app/api/server/ollama?style=flat&compact=true)](https://discord.gg/ollama)

-Get up and running with large language models locally.
+> Note: Ollama is in early preview. Please report any issues you find.

-### macOS
+Run, create, and share large language models (LLMs).

-[Download](https://ollama.com/download/Ollama-darwin.zip)
+## Download

-### Windows preview
-
-[Download](https://ollama.com/download/OllamaSetup.exe)
-
-### Linux
-
-```
-curl -fsSL https://ollama.com/install.sh | sh
-```
-
-[Manual install instructions](https://github.com/ollama/ollama/blob/main/docs/linux.md)
-
-### Docker
-
-The official [Ollama Docker image](https://hub.docker.com/r/ollama/ollama) `ollama/ollama` is available on Docker Hub.
-
-### Libraries
-
- [ollama-python](https://github.com/ollama/ollama-python)
- [ollama-js](https://github.com/ollama/ollama-js)
+- [Download](https://ollama.ai/download) for macOS on Apple Silicon (Intel coming soon)
+- Download for Windows and Linux (coming soon)
+- Build [from source](#building)

 ## Quickstart

-To run and chat with [Llama 3](https://ollama.com/library/llama3):
+To run and chat with [Llama 2](https://ai.meta.com/llama), the new model by Meta:

 ```
-ollama run llama3
+ollama run llama2
 ```

 ## Model library

-Ollama supports a list of models available on [ollama.com/library](https://ollama.com/library 'ollama model library')
+`ollama` includes a library of open-source models:

-Here are some example models that can be downloaded:
+| Model                    | Parameters | Size  | Download                    |
+| ------------------------ | ---------- | ----- | --------------------------- |
+| Llama2                   | 7B         | 3.8GB | `ollama pull llama2`        |
+| Llama2 13B               | 13B        | 7.3GB | `ollama pull llama2:13b`    |
+| Orca Mini                | 3B         | 1.9GB | `ollama pull orca`          |
+| Vicuna                   | 7B         | 3.8GB | `ollama pull vicuna`        |
+| Nous-Hermes              | 13B        | 7.3GB | `ollama pull nous-hermes`   |
+| Wizard Vicuna Uncensored | 13B        | 7.3GB | `ollama pull wizard-vicuna` |

-| Model              | Parameters | Size  | Download                       |
-| ------------------ | ---------- | ----- | ------------------------------ |
-| Llama 3            | 8B         | 4.7GB | `ollama run llama3`            |
-| Llama 3            | 70B        | 40GB  | `ollama run llama3:70b`        |
-| Phi-3              | 3.8B       | 2.3GB | `ollama run phi3`              |
-| Mistral            | 7B         | 4.1GB | `ollama run mistral`           |
-| Neural Chat        | 7B         | 4.1GB | `ollama run neural-chat`       |
-| Starling           | 7B         | 4.1GB | `ollama run starling-lm`       |
-| Code Llama         | 7B         | 3.8GB | `ollama run codellama`         |
-| Llama 2 Uncensored | 7B         | 3.8GB | `ollama run llama2-uncensored` |
-| LLaVA              | 7B         | 4.5GB | `ollama run llava`             |
-| Gemma              | 2B         | 1.4GB | `ollama run gemma:2b`          |
-| Gemma              | 7B         | 4.8GB | `ollama run gemma:7b`          |
-| Solar              | 10.7B      | 6.1GB | `ollama run solar`             |
+> Note: You should have at least 8 GB of RAM to run the 3B models, 16 GB to run the 7B models, and 32 GB to run the 13B models.

-> Note: You should have at least 8 GB of RAM available to run the 7B models, 16 GB to run the 13B models, and 32 GB to run the 33B models.
+## Examples

-## Customize a model
-
-### Import from GGUF
-
-Ollama supports importing GGUF models in the Modelfile:
-
-1. Create a file named `Modelfile`, with a `FROM` instruction with the local filepath to the model you want to import.
-
-   ```
-   FROM ./vicuna-33b.Q4_0.gguf
-   ```
-
-2. Create the model in Ollama
-
-   ```
-   ollama create example -f Modelfile
-   ```
-
-3. Run the model
-
-   ```
-   ollama run example
-   ```
-
-### Import from PyTorch or Safetensors
-
-See the [guide](docs/import.md) on importing models for more information.
-
-### Customize a prompt
-
-Models from the Ollama library can be customized with a prompt. For example, to customize the `llama3` model:
+### Run a model

 ```
-ollama pull llama3
+ollama run llama2
+>>> hi
+Hello! How can I help you today?
+```
+
+### Create a custom model
+
+Pull a base model:
+
+```
+ollama pull llama2
 ```

 Create a `Modelfile`:

 ```
-FROM llama3
+FROM llama2

 # set the temperature to 1 [higher is more creative, lower is more coherent]
 PARAMETER temperature 1

-# set the system message
+# set the system prompt
 SYSTEM """
 You are Mario from Super Mario Bros. Answer as Mario, the assistant, only.
 """
@@ -123,265 +83,55 @@ ollama run mario
 Hello! It's your friend Mario.
 ```

-For more examples, see the [examples](examples) directory. For more information on working with a Modelfile, see the [Modelfile](docs/modelfile.md) documentation.
+For more examples, see the [examples](./examples) directory.

-## CLI Reference
-
-### Create a model
-
-`ollama create` is used to create a model from a Modelfile.
+### Pull a model from the registry

 ```
-ollama create mymodel -f ./Modelfile
+ollama pull orca
 ```

-### Pull a model
-
-```
-ollama pull llama3
-```
-
-> This command can also be used to update a local model. Only the diff will be pulled.
-
-### Remove a model
-
-```
-ollama rm llama3
-```
-
-### Copy a model
-
-```
-ollama cp llama3 my-model
-```
-
-### Multiline input
-
-For multiline input, you can wrap text with `"""`:
-
-```
->>> """Hello,
-... world!
-... """
-I'm a basic program that prints the famous "Hello, world!" message to the console.
-```
-
-### Multimodal models
-
-```
->>> What's in this image? /Users/jmorgan/Desktop/smile.png
-The image features a yellow smiley face, which is likely the central focus of the picture.
-```
-
-### Pass in prompt as arguments
-
-```
-$ ollama run llama3 "Summarize this file: $(cat README.md)"
- Ollama is a lightweight, extensible framework for building and running language models on the local machine. It provides a simple API for creating, running, and managing models, as well as a library of pre-built models that can be easily used in a variety of applications.
-```
-
-### List models on your computer
+### Listing local models

 ```
 ollama list
 ```

-### Start Ollama
+## Model packages

-`ollama serve` is used when you want to start ollama without running the desktop application.
+### Overview
+
+Ollama bundles model weights, configuration, and data into a single package, defined by a [Modelfile](./docs/modelfile.md).
+
+<picture>
+  <source media="(prefers-color-scheme: dark)" height="480" srcset="https://github.com/jmorganca/ollama/assets/251292/2fd96b5f-191b-45c1-9668-941cfad4eb70">
+  <img alt="logo" height="480" src="https://github.com/jmorganca/ollama/assets/251292/2fd96b5f-191b-45c1-9668-941cfad4eb70">
+</picture>

 ## Building

-Install `cmake` and `go`:
-
-```
-brew install cmake go
-```
-
-Then generate dependencies:
-
-```
-go generate ./...
-```
-
-Then build the binary:
-
 ```
 go build .
 ```

-More detailed instructions can be found in the [developer guide](https://github.com/ollama/ollama/blob/main/docs/development.md)
-
-### Running local builds
-
-Next, start the server:
+To run it start the server:

 ```
-./ollama serve
+./ollama serve &
 ```

-Finally, in a separate shell, run a model:
+Finally, run a model!

 ```
-./ollama run llama3
+./ollama run llama2
 ```

 ## REST API

-Ollama has a REST API for running and managing models.
+### `POST /api/generate`

-### Generate a response
+Generate text from a model.

 ```
-curl http://localhost:11434/api/generate -d '{
-  "model": "llama3",
-  "prompt":"Why is the sky blue?"
-}'
+curl -X POST http://localhost:11434/api/generate -d '{"model": "llama2", "prompt":"Why is the sky blue?"}'
 ```
-
-### Chat with a model
-
-```
-curl http://localhost:11434/api/chat -d '{
-  "model": "llama3",
-  "messages": [
-    { "role": "user", "content": "why is the sky blue?" }
-  ]
-}'
-```
-
-See the [API documentation](./docs/api.md) for all endpoints.
-
-## Community Integrations
-
-### Web & Desktop
-
- [Open WebUI](https://github.com/open-webui/open-webui)
- [Enchanted (macOS native)](https://github.com/AugustDev/enchanted)
- [Lollms-Webui](https://github.com/ParisNeo/lollms-webui)
- [LibreChat](https://github.com/danny-avila/LibreChat)
- [Bionic GPT](https://github.com/bionic-gpt/bionic-gpt)
- [HTML UI](https://github.com/rtcfirefly/ollama-ui)
- [Saddle](https://github.com/jikkuatwork/saddle)
- [Chatbot UI](https://github.com/ivanfioravanti/chatbot-ollama)
- [Chatbot UI v2](https://github.com/mckaywrigley/chatbot-ui)
- [Typescript UI](https://github.com/ollama-interface/Ollama-Gui?tab=readme-ov-file)
- [Minimalistic React UI for Ollama Models](https://github.com/richawo/minimal-llm-ui)
- [Ollamac](https://github.com/kevinhermawan/Ollamac)
- [big-AGI](https://github.com/enricoros/big-AGI/blob/main/docs/config-local-ollama.md)
- [Cheshire Cat assistant framework](https://github.com/cheshire-cat-ai/core)
- [Amica](https://github.com/semperai/amica)
- [chatd](https://github.com/BruceMacD/chatd)
- [Ollama-SwiftUI](https://github.com/kghandour/Ollama-SwiftUI)
- [Dify.AI](https://github.com/langgenius/dify)
- [MindMac](https://mindmac.app)
- [NextJS Web Interface for Ollama](https://github.com/jakobhoeg/nextjs-ollama-llm-ui)
- [Msty](https://msty.app)
- [Chatbox](https://github.com/Bin-Huang/Chatbox)
- [WinForm Ollama Copilot](https://github.com/tgraupmann/WinForm_Ollama_Copilot)
- [NextChat](https://github.com/ChatGPTNextWeb/ChatGPT-Next-Web) with [Get Started Doc](https://docs.nextchat.dev/models/ollama)
- [Alpaca WebUI](https://github.com/mmo80/alpaca-webui)
- [OllamaGUI](https://github.com/enoch1118/ollamaGUI)
- [OpenAOE](https://github.com/InternLM/OpenAOE)
- [Odin Runes](https://github.com/leonid20000/OdinRunes)
- [LLM-X: Progressive Web App](https://github.com/mrdjohnson/llm-x)
- [AnythingLLM (Docker + MacOs/Windows/Linux native app)](https://github.com/Mintplex-Labs/anything-llm)
- [Ollama Basic Chat: Uses HyperDiv Reactive UI](https://github.com/rapidarchitect/ollama_basic_chat)
- [Ollama-chats RPG](https://github.com/drazdra/ollama-chats)
- [QA-Pilot: Chat with Code Repository](https://github.com/reid41/QA-Pilot)
- [ChatOllama: Open Source Chatbot based on Ollama with Knowledge Bases](https://github.com/sugarforever/chat-ollama)
- [CRAG Ollama Chat: Simple Web Search with Corrective RAG](https://github.com/Nagi-ovo/CRAG-Ollama-Chat)
- [RAGFlow: Open-source Retrieval-Augmented Generation engine based on deep document understanding](https://github.com/infiniflow/ragflow)
- [chat: chat web app for teams](https://github.com/swuecho/chat)
- [Lobe Chat](https://github.com/lobehub/lobe-chat) with [Integrating Doc](https://lobehub.com/docs/self-hosting/examples/ollama)
- [Ollama RAG Chatbot: Local Chat with multiples PDFs using Ollama and RAG.](https://github.com/datvodinh/rag-chatbot.git)
-
-### Terminal
-
- [oterm](https://github.com/ggozad/oterm)
- [Ellama Emacs client](https://github.com/s-kostyaev/ellama)
- [Emacs client](https://github.com/zweifisch/ollama)
- [gen.nvim](https://github.com/David-Kunz/gen.nvim)
- [ollama.nvim](https://github.com/nomnivore/ollama.nvim)
- [ollero.nvim](https://github.com/marco-souza/ollero.nvim)
- [ollama-chat.nvim](https://github.com/gerazov/ollama-chat.nvim)
- [ogpt.nvim](https://github.com/huynle/ogpt.nvim)
- [gptel Emacs client](https://github.com/karthink/gptel)
- [Oatmeal](https://github.com/dustinblackman/oatmeal)
- [cmdh](https://github.com/pgibler/cmdh)
- [ooo](https://github.com/npahlfer/ooo)
- [shell-pilot](https://github.com/reid41/shell-pilot)
- [tenere](https://github.com/pythops/tenere)
- [llm-ollama](https://github.com/taketwo/llm-ollama) for [Datasette's LLM CLI](https://llm.datasette.io/en/stable/).
- [typechat-cli](https://github.com/anaisbetts/typechat-cli)
- [ShellOracle](https://github.com/djcopley/ShellOracle)
- [tlm](https://github.com/yusufcanb/tlm)
- [podman-ollama](https://github.com/ericcurtin/podman-ollama)
-
-### Database
-
- [MindsDB](https://github.com/mindsdb/mindsdb/blob/staging/mindsdb/integrations/handlers/ollama_handler/README.md) (Connects Ollama models with nearly 200 data platforms and apps)
- [chromem-go](https://github.com/philippgille/chromem-go/blob/v0.5.0/embed_ollama.go) with [example](https://github.com/philippgille/chromem-go/tree/v0.5.0/examples/rag-wikipedia-ollama)
-
-### Package managers
-
- [Pacman](https://archlinux.org/packages/extra/x86_64/ollama/)
- [Helm Chart](https://artifacthub.io/packages/helm/ollama-helm/ollama)
-
-### Libraries
-
- [LangChain](https://python.langchain.com/docs/integrations/llms/ollama) and [LangChain.js](https://js.langchain.com/docs/modules/model_io/models/llms/integrations/ollama) with [example](https://js.langchain.com/docs/use_cases/question_answering/local_retrieval_qa)
- [LangChainGo](https://github.com/tmc/langchaingo/) with [example](https://github.com/tmc/langchaingo/tree/main/examples/ollama-completion-example)
- [LangChain4j](https://github.com/langchain4j/langchain4j) with [example](https://github.com/langchain4j/langchain4j-examples/tree/main/ollama-examples/src/main/java)
- [LlamaIndex](https://gpt-index.readthedocs.io/en/stable/examples/llm/ollama.html)
- [LiteLLM](https://github.com/BerriAI/litellm)
- [OllamaSharp for .NET](https://github.com/awaescher/OllamaSharp)
- [Ollama for Ruby](https://github.com/gbaptista/ollama-ai)
- [Ollama-rs for Rust](https://github.com/pepperoni21/ollama-rs)
- [Ollama4j for Java](https://github.com/amithkoujalgi/ollama4j)
- [ModelFusion Typescript Library](https://modelfusion.dev/integration/model-provider/ollama)
- [OllamaKit for Swift](https://github.com/kevinhermawan/OllamaKit)
- [Ollama for Dart](https://github.com/breitburg/dart-ollama)
- [Ollama for Laravel](https://github.com/cloudstudio/ollama-laravel)
- [LangChainDart](https://github.com/davidmigloz/langchain_dart)
- [Semantic Kernel - Python](https://github.com/microsoft/semantic-kernel/tree/main/python/semantic_kernel/connectors/ai/ollama)
- [Haystack](https://github.com/deepset-ai/haystack-integrations/blob/main/integrations/ollama.md)
- [Elixir LangChain](https://github.com/brainlid/langchain)
- [Ollama for R - rollama](https://github.com/JBGruber/rollama)
- [Ollama-ex for Elixir](https://github.com/lebrunel/ollama-ex)
- [Ollama Connector for SAP ABAP](https://github.com/b-tocs/abap_btocs_ollama)
- [Testcontainers](https://testcontainers.com/modules/ollama/)
-
-### Mobile
-
- [Enchanted](https://github.com/AugustDev/enchanted)
- [Maid](https://github.com/Mobile-Artificial-Intelligence/maid)
-
-### Extensions & Plugins
-
- [Raycast extension](https://github.com/MassimilianoPasquini97/raycast_ollama)
- [Discollama](https://github.com/mxyng/discollama) (Discord bot inside the Ollama discord channel)
- [Continue](https://github.com/continuedev/continue)
- [Obsidian Ollama plugin](https://github.com/hinterdupfinger/obsidian-ollama)
- [Logseq Ollama plugin](https://github.com/omagdy7/ollama-logseq)
- [NotesOllama](https://github.com/andersrex/notesollama) (Apple Notes Ollama plugin)
- [Dagger Chatbot](https://github.com/samalba/dagger-chatbot)
- [Discord AI Bot](https://github.com/mekb-turtle/discord-ai-bot)
- [Ollama Telegram Bot](https://github.com/ruecat/ollama-telegram)
- [Hass Ollama Conversation](https://github.com/ej52/hass-ollama-conversation)
- [Rivet plugin](https://github.com/abrenneke/rivet-plugin-ollama)
- [Llama Coder](https://github.com/ex3ndr/llama-coder) (Copilot alternative using Ollama)
- [Obsidian BMO Chatbot plugin](https://github.com/longy2k/obsidian-bmo-chatbot)
- [Cliobot](https://github.com/herval/cliobot) (Telegram bot with Ollama support)
- [Copilot for Obsidian plugin](https://github.com/logancyang/obsidian-copilot)
- [Obsidian Local GPT plugin](https://github.com/pfrankov/obsidian-local-gpt)
- [Open Interpreter](https://docs.openinterpreter.com/language-model-setup/local-models/ollama)
- [twinny](https://github.com/rjmacarthy/twinny) (Copilot and Copilot chat alternative using Ollama)
- [Wingman-AI](https://github.com/RussellCanfield/wingman-ai) (Copilot code and chat alternative using Ollama and HuggingFace)
- [Page Assist](https://github.com/n4ze3m/page-assist) (Chrome Extension)
- [AI Telegram Bot](https://github.com/tusharhero/aitelegrambot) (Telegram bot using Ollama in backend)
- [AI ST Completion](https://github.com/yaroslavyaroslav/OpenAI-sublime-text) (Sublime Text 4 AI assistant plugin with Ollama support)
- [Discord-Ollama Chat Bot](https://github.com/kevinthedang/discord-ollama) (Generalized TypeScript Discord Bot w/ Tuning Documentation)
-
-### Supported backends 
- [llama.cpp](https://github.com/ggerganov/llama.cpp) project founded by Georgi Gerganov. 
@@ -1,9 +1,3 @@
-// Package api implements the client-side API for code wishing to interact
-// with the ollama service. The methods of the [Client] type correspond to
-// the ollama REST API as described in https://github.com/ollama/ollama/blob/main/docs/api.md
-//
-// The ollama command-line client itself uses this package to interact with
-// the backend service.
 package api

 import (
@@ -13,27 +7,18 @@ import (
 	"encoding/json"
 	"fmt"
 	"io"
-	"net"
 	"net/http"
 	"net/url"
-	"os"
-	"runtime"
-	"strconv"
-	"strings"
-
-	"github.com/ollama/ollama/format"
-	"github.com/ollama/ollama/version"
 )

-// Client encapsulates client state for interacting with the ollama
-// service. Use [ClientFromEnvironment] to create new Clients.
 type Client struct {
-	base *url.URL
-	http *http.Client
+	base    url.URL
+	HTTP    http.Client
+	Headers http.Header
 }

 func checkError(resp *http.Response, body []byte) error {
-	if resp.StatusCode < http.StatusBadRequest {
+	if resp.StatusCode >= 200 && resp.StatusCode < 400 {
 		return nil
 	}

@@ -48,80 +33,15 @@ func checkError(resp *http.Response, body []byte) error {
 	return apiError
 }

-// ClientFromEnvironment creates a new [Client] using configuration from the
-// environment variable OLLAMA_HOST, which points to the network host and
-// port on which the ollama service is listenting. The format of this variable
-// is:
-//
-//	<scheme>://<host>:<port>
-//
-// If the variable is not specified, a default ollama host and port will be
-// used.
-func ClientFromEnvironment() (*Client, error) {
-	ollamaHost, err := GetOllamaHost()
-	if err != nil {
-		return nil, err
+func NewClient(hosts ...string) *Client {
+	host := "127.0.0.1:11434"
+	if len(hosts) > 0 {
+		host = hosts[0]
 	}

 	return &Client{
-		base: &url.URL{
-			Scheme: ollamaHost.Scheme,
-			Host:   net.JoinHostPort(ollamaHost.Host, ollamaHost.Port),
-		},
-		http: http.DefaultClient,
-	}, nil
-}
-
-type OllamaHost struct {
-	Scheme string
-	Host   string
-	Port   string
-}
-
-func GetOllamaHost() (OllamaHost, error) {
-	defaultPort := "11434"
-
-	hostVar := os.Getenv("OLLAMA_HOST")
-	hostVar = strings.TrimSpace(strings.Trim(strings.TrimSpace(hostVar), "\"'"))
-
-	scheme, hostport, ok := strings.Cut(hostVar, "://")
-	switch {
-	case !ok:
-		scheme, hostport = "http", hostVar
-	case scheme == "http":
-		defaultPort = "80"
-	case scheme == "https":
-		defaultPort = "443"
-	}
-
-	// trim trailing slashes
-	hostport = strings.TrimRight(hostport, "/")
-
-	host, port, err := net.SplitHostPort(hostport)
-	if err != nil {
-		host, port = "127.0.0.1", defaultPort
-		if ip := net.ParseIP(strings.Trim(hostport, "[]")); ip != nil {
-			host = ip.String()
-		} else if hostport != "" {
-			host = hostport
-		}
-	}
-
-	if portNum, err := strconv.ParseInt(port, 10, 32); err != nil || portNum > 65535 || portNum < 0 {
-		return OllamaHost{}, ErrInvalidHostPort
-	}
-
-	return OllamaHost{
-		Scheme: scheme,
-		Host:   host,
-		Port:   port,
-	}, nil
-}
-
-func NewClient(base *url.URL, http *http.Client) *Client {
-	return &Client{
-		base: base,
-		http: http,
+		base: url.URL{Scheme: "http", Host: host},
+		HTTP: http.Client{},
 	}
 }

@@ -129,33 +49,29 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData
 	var reqBody io.Reader
 	var data []byte
 	var err error
-
-	switch reqData := reqData.(type) {
-	case io.Reader:
-		// reqData is already an io.Reader
-		reqBody = reqData
-	case nil:
-		// noop
-	default:
+	if reqData != nil {
 		data, err = json.Marshal(reqData)
 		if err != nil {
 			return err
 		}
-
 		reqBody = bytes.NewReader(data)
 	}

-	requestURL := c.base.JoinPath(path)
-	request, err := http.NewRequestWithContext(ctx, method, requestURL.String(), reqBody)
+	url := c.base.JoinPath(path).String()
+
+	req, err := http.NewRequestWithContext(ctx, method, url, reqBody)
 	if err != nil {
 		return err
 	}

-	request.Header.Set("Content-Type", "application/json")
-	request.Header.Set("Accept", "application/json")
-	request.Header.Set("User-Agent", fmt.Sprintf("ollama/%s (%s %s) Go/%s", version.Version, runtime.GOARCH, runtime.GOOS, runtime.Version()))
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("Accept", "application/json")

-	respObj, err := c.http.Do(request)
+	for k, v := range c.Headers {
+		req.Header[k] = v
+	}
+
+	respObj, err := c.HTTP.Do(req)
 	if err != nil {
 		return err
 	}
@@ -178,8 +94,6 @@ func (c *Client) do(ctx context.Context, method, path string, reqData, respData
 	return nil
 }

-const maxBufferSize = 512 * format.KiloByte
-
 func (c *Client) stream(ctx context.Context, method, path string, data any, fn func([]byte) error) error {
 	var buf *bytes.Buffer
 	if data != nil {
@@ -191,26 +105,21 @@ func (c *Client) stream(ctx context.Context, method, path string, data any, fn f
 		buf = bytes.NewBuffer(bts)
 	}

-	requestURL := c.base.JoinPath(path)
-	request, err := http.NewRequestWithContext(ctx, method, requestURL.String(), buf)
+	request, err := http.NewRequestWithContext(ctx, method, c.base.JoinPath(path).String(), buf)
 	if err != nil {
 		return err
 	}

 	request.Header.Set("Content-Type", "application/json")
-	request.Header.Set("Accept", "application/x-ndjson")
-	request.Header.Set("User-Agent", fmt.Sprintf("ollama/%s (%s %s) Go/%s", version.Version, runtime.GOARCH, runtime.GOOS, runtime.Version()))
+	request.Header.Set("Accept", "application/json")

-	response, err := c.http.Do(request)
+	response, err := http.DefaultClient.Do(request)
 	if err != nil {
 		return err
 	}
 	defer response.Body.Close()

 	scanner := bufio.NewScanner(response.Body)
-	// increase the buffer size to avoid running out of space
-	scanBuf := make([]byte, 0, maxBufferSize)
-	scanner.Buffer(scanBuf, maxBufferSize)
 	for scanner.Scan() {
 		var errorResponse struct {
 			Error string `json:"error,omitempty"`
@@ -222,10 +131,10 @@ func (c *Client) stream(ctx context.Context, method, path string, data any, fn f
 		}

 		if errorResponse.Error != "" {
-			return fmt.Errorf(errorResponse.Error)
+			return fmt.Errorf("stream: %s", errorResponse.Error)
 		}

-		if response.StatusCode >= http.StatusBadRequest {
+		if response.StatusCode >= 400 {
 			return StatusError{
 				StatusCode:   response.StatusCode,
 				Status:       response.Status,
@@ -241,14 +150,8 @@ func (c *Client) stream(ctx context.Context, method, path string, data any, fn f
 	return nil
 }

-// GenerateResponseFunc is a function that [Client.Generate] invokes every time
-// a response is received from the service. If this function returns an error,
-// [Client.Generate] will stop generating and return this error.
 type GenerateResponseFunc func(GenerateResponse) error

-// Generate generates a response for a given prompt. The req parameter should
-// be populated with prompt details. fn is called for each response (there may
-// be multiple responses, e.g. in case streaming is enabled).
 func (c *Client) Generate(ctx context.Context, req *GenerateRequest, fn GenerateResponseFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/generate", req, func(bts []byte) error {
 		var resp GenerateResponse
@@ -260,34 +163,8 @@ func (c *Client) Generate(ctx context.Context, req *GenerateRequest, fn Generate
 	})
 }

-// ChatResponseFunc is a function that [Client.Chat] invokes every time
-// a response is received from the service. If this function returns an error,
-// [Client.Chat] will stop generating and return this error.
-type ChatResponseFunc func(ChatResponse) error
-
-// Chat generates the next message in a chat. [ChatRequest] may contain a
-// sequence of messages which can be used to maintain chat history with a model.
-// fn is called for each response (there may be multiple responses, e.g. if case
-// streaming is enabled).
-func (c *Client) Chat(ctx context.Context, req *ChatRequest, fn ChatResponseFunc) error {
-	return c.stream(ctx, http.MethodPost, "/api/chat", req, func(bts []byte) error {
-		var resp ChatResponse
-		if err := json.Unmarshal(bts, &resp); err != nil {
-			return err
-		}
-
-		return fn(resp)
-	})
-}
-
-// PullProgressFunc is a function that [Client.Pull] invokes every time there
-// is progress with a "pull" request sent to the service. If this function
-// returns an error, [Client.Pull] will stop the process and return this error.
 type PullProgressFunc func(ProgressResponse) error

-// Pull downloads a model from the ollama library. fn is called each time
-// progress is made on the request and can be used to display a progress bar,
-// etc.
 func (c *Client) Pull(ctx context.Context, req *PullRequest, fn PullProgressFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/pull", req, func(bts []byte) error {
 		var resp ProgressResponse
@@ -312,11 +189,11 @@ func (c *Client) Push(ctx context.Context, req *PushRequest, fn PushProgressFunc
 	})
 }

-type CreateProgressFunc func(ProgressResponse) error
+type CreateProgressFunc func(CreateProgress) error

 func (c *Client) Create(ctx context.Context, req *CreateRequest, fn CreateProgressFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/create", req, func(bts []byte) error {
-		var resp ProgressResponse
+		var resp CreateProgress
 		if err := json.Unmarshal(bts, &resp); err != nil {
 			return err
 		}
@@ -333,54 +210,9 @@ func (c *Client) List(ctx context.Context) (*ListResponse, error) {
 	return &lr, nil
 }

-func (c *Client) Copy(ctx context.Context, req *CopyRequest) error {
-	if err := c.do(ctx, http.MethodPost, "/api/copy", req, nil); err != nil {
-		return err
-	}
-	return nil
-}
-
 func (c *Client) Delete(ctx context.Context, req *DeleteRequest) error {
 	if err := c.do(ctx, http.MethodDelete, "/api/delete", req, nil); err != nil {
 		return err
 	}
 	return nil
 }
-
-func (c *Client) Show(ctx context.Context, req *ShowRequest) (*ShowResponse, error) {
-	var resp ShowResponse
-	if err := c.do(ctx, http.MethodPost, "/api/show", req, &resp); err != nil {
-		return nil, err
-	}
-	return &resp, nil
-}
-
-func (c *Client) Heartbeat(ctx context.Context) error {
-	if err := c.do(ctx, http.MethodHead, "/", nil, nil); err != nil {
-		return err
-	}
-	return nil
-}
-func (c *Client) Embeddings(ctx context.Context, req *EmbeddingRequest) (*EmbeddingResponse, error) {
-	var resp EmbeddingResponse
-	if err := c.do(ctx, http.MethodPost, "/api/embeddings", req, &resp); err != nil {
-		return nil, err
-	}
-	return &resp, nil
-}
-
-func (c *Client) CreateBlob(ctx context.Context, digest string, r io.Reader) error {
-	return c.do(ctx, http.MethodPost, fmt.Sprintf("/api/blobs/%s", digest), r, nil)
-}
-
-func (c *Client) Version(ctx context.Context) (string, error) {
-	var version struct {
-		Version string `json:"version"`
-	}
-
-	if err := c.do(ctx, http.MethodGet, "/api/version", nil, &version); err != nil {
-		return "", err
-	}
-
-	return version.Version, nil
-}
@@ -1,85 +0,0 @@
-package api
-
-import (
-	"fmt"
-	"net"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-)
-
-func TestClientFromEnvironment(t *testing.T) {
-	type testCase struct {
-		value  string
-		expect string
-		err    error
-	}
-
-	testCases := map[string]*testCase{
-		"empty":                      {value: "", expect: "http://127.0.0.1:11434"},
-		"only address":               {value: "1.2.3.4", expect: "http://1.2.3.4:11434"},
-		"only port":                  {value: ":1234", expect: "http://:1234"},
-		"address and port":           {value: "1.2.3.4:1234", expect: "http://1.2.3.4:1234"},
-		"scheme http and address":    {value: "http://1.2.3.4", expect: "http://1.2.3.4:80"},
-		"scheme https and address":   {value: "https://1.2.3.4", expect: "https://1.2.3.4:443"},
-		"scheme, address, and port":  {value: "https://1.2.3.4:1234", expect: "https://1.2.3.4:1234"},
-		"hostname":                   {value: "example.com", expect: "http://example.com:11434"},
-		"hostname and port":          {value: "example.com:1234", expect: "http://example.com:1234"},
-		"scheme http and hostname":   {value: "http://example.com", expect: "http://example.com:80"},
-		"scheme https and hostname":  {value: "https://example.com", expect: "https://example.com:443"},
-		"scheme, hostname, and port": {value: "https://example.com:1234", expect: "https://example.com:1234"},
-		"trailing slash":             {value: "example.com/", expect: "http://example.com:11434"},
-		"trailing slash port":        {value: "example.com:1234/", expect: "http://example.com:1234"},
-	}
-
-	for k, v := range testCases {
-		t.Run(k, func(t *testing.T) {
-			t.Setenv("OLLAMA_HOST", v.value)
-
-			client, err := ClientFromEnvironment()
-			if err != v.err {
-				t.Fatalf("expected %s, got %s", v.err, err)
-			}
-
-			if client.base.String() != v.expect {
-				t.Fatalf("expected %s, got %s", v.expect, client.base.String())
-			}
-		})
-	}
-
-	hostTestCases := map[string]*testCase{
-		"empty":               {value: "", expect: "127.0.0.1:11434"},
-		"only address":        {value: "1.2.3.4", expect: "1.2.3.4:11434"},
-		"only port":           {value: ":1234", expect: ":1234"},
-		"address and port":    {value: "1.2.3.4:1234", expect: "1.2.3.4:1234"},
-		"hostname":            {value: "example.com", expect: "example.com:11434"},
-		"hostname and port":   {value: "example.com:1234", expect: "example.com:1234"},
-		"zero port":           {value: ":0", expect: ":0"},
-		"too large port":      {value: ":66000", err: ErrInvalidHostPort},
-		"too small port":      {value: ":-1", err: ErrInvalidHostPort},
-		"ipv6 localhost":      {value: "[::1]", expect: "[::1]:11434"},
-		"ipv6 world open":     {value: "[::]", expect: "[::]:11434"},
-		"ipv6 no brackets":    {value: "::1", expect: "[::1]:11434"},
-		"ipv6 + port":         {value: "[::1]:1337", expect: "[::1]:1337"},
-		"extra space":         {value: " 1.2.3.4 ", expect: "1.2.3.4:11434"},
-		"extra quotes":        {value: "\"1.2.3.4\"", expect: "1.2.3.4:11434"},
-		"extra space+quotes":  {value: " \" 1.2.3.4 \" ", expect: "1.2.3.4:11434"},
-		"extra single quotes": {value: "'1.2.3.4'", expect: "1.2.3.4:11434"},
-	}
-
-	for k, v := range hostTestCases {
-		t.Run(k, func(t *testing.T) {
-			t.Setenv("OLLAMA_HOST", v.value)
-
-			oh, err := GetOllamaHost()
-			if err != v.err {
-				t.Fatalf("expected %s, got %s", v.err, err)
-			}
-
-			if err == nil {
-				host := net.JoinHostPort(oh.Host, oh.Port)
-				assert.Equal(t, v.expect, host, fmt.Sprintf("%s: expected %s, got %s", k, v.expect, host))
-			}
-		})
-	}
-}
@@ -1,14 +1,9 @@
 package api

 import (
-	"encoding/json"
-	"errors"
 	"fmt"
-	"math"
 	"os"
-	"reflect"
-	"strconv"
-	"strings"
+	"runtime"
 	"time"
 )

@@ -32,494 +27,160 @@ func (e StatusError) Error() string {
 	}
 }

-type ImageData []byte
-
-// GenerateRequest describes a request sent by [Client.Generate]. While you
-// have to specify the Model and Prompt fields, all the other fields have
-// reasonable defaults for basic uses.
 type GenerateRequest struct {
-	// Model is the model name; it should be a name familiar to Ollama from
-	// the library at https://ollama.com/library
-	Model string `json:"model"`
+	Model   string `json:"model"`
+	Prompt  string `json:"prompt"`
+	Context []int  `json:"context,omitempty"`

-	// Prompt is the textual prompt to send to the model.
-	Prompt string `json:"prompt"`
-
-	// System overrides the model's default system message/prompt.
-	System string `json:"system"`
-
-	// Template overrides the model's default prompt template.
-	Template string `json:"template"`
-
-	// Context is the context parameter returned from a previous call to
-	// Generate call. It can be used to keep a short conversational memory.
-	Context []int `json:"context,omitempty"`
-
-	// Stream specifies whether the response is streaming; it is true by default.
-	Stream *bool `json:"stream,omitempty"`
-
-	// Raw set to true means that no formatting will be applied to the prompt.
-	Raw bool `json:"raw,omitempty"`
-
-	// Format specifies the format to return a response in.
-	Format string `json:"format"`
-
-	// KeepAlive controls how long the model will stay loaded in memory following
-	// this request.
-	KeepAlive *Duration `json:"keep_alive,omitempty"`
-
-	// Images is an optional list of base64-encoded images accompanying this
-	// request, for multimodal models.
-	Images []ImageData `json:"images,omitempty"`
-
-	// Options lists model-specific options. For example, temperature can be
-	// set through this field, if the model supports it.
-	Options map[string]interface{} `json:"options"`
+	Options `json:"options"`
 }

-type ChatRequest struct {
-	Model     string    `json:"model"`
-	Messages  []Message `json:"messages"`
-	Stream    *bool     `json:"stream,omitempty"`
-	Format    string    `json:"format"`
-	KeepAlive *Duration `json:"keep_alive,omitempty"`
-
-	Options map[string]interface{} `json:"options"`
+type CreateRequest struct {
+	Name string `json:"name"`
+	Path string `json:"path"`
 }

-type Message struct {
-	Role    string      `json:"role"` // one of ["system", "user", "assistant"]
-	Content string      `json:"content"`
-	Images  []ImageData `json:"images,omitempty"`
+type CreateProgress struct {
+	Status string `json:"status"`
 }

-type ChatResponse struct {
+type DeleteRequest struct {
+	Name string `json:"name"`
+}
+
+type PullRequest struct {
+	Name     string `json:"name"`
+	Insecure bool   `json:"insecure,omitempty"`
+	Username string `json:"username"`
+	Password string `json:"password"`
+}
+
+type ProgressResponse struct {
+	Status    string `json:"status"`
+	Digest    string `json:"digest,omitempty"`
+	Total     int    `json:"total,omitempty"`
+	Completed int    `json:"completed,omitempty"`
+}
+
+type PushRequest struct {
+	Name     string `json:"name"`
+	Insecure bool   `json:"insecure,omitempty"`
+	Username string `json:"username"`
+	Password string `json:"password"`
+}
+
+type ListResponse struct {
+	Models []ListResponseModel `json:"models"`
+}
+
+type ListResponseModel struct {
+	Name       string    `json:"name"`
+	ModifiedAt time.Time `json:"modified_at"`
+	Size       int       `json:"size"`
+}
+
+type GenerateResponse struct {
 	Model     string    `json:"model"`
 	CreatedAt time.Time `json:"created_at"`
-	Message   Message   `json:"message"`
+	Response  string    `json:"response,omitempty"`

-	Done bool `json:"done"`
+	Done    bool  `json:"done"`
+	Context []int `json:"context,omitempty"`

-	Metrics
-}
-
-type Metrics struct {
 	TotalDuration      time.Duration `json:"total_duration,omitempty"`
-	LoadDuration       time.Duration `json:"load_duration,omitempty"`
 	PromptEvalCount    int           `json:"prompt_eval_count,omitempty"`
 	PromptEvalDuration time.Duration `json:"prompt_eval_duration,omitempty"`
 	EvalCount          int           `json:"eval_count,omitempty"`
 	EvalDuration       time.Duration `json:"eval_duration,omitempty"`
 }

-// Options specified in GenerateRequest, if you add a new option here add it to the API docs also
+func (r *GenerateResponse) Summary() {
+	if r.TotalDuration > 0 {
+		fmt.Fprintf(os.Stderr, "total duration:       %v\n", r.TotalDuration)
+	}
+
+	if r.PromptEvalCount > 0 {
+		fmt.Fprintf(os.Stderr, "prompt eval count:    %d token(s)\n", r.PromptEvalCount)
+	}
+
+	if r.PromptEvalDuration > 0 {
+		fmt.Fprintf(os.Stderr, "prompt eval duration: %s\n", r.PromptEvalDuration)
+		fmt.Fprintf(os.Stderr, "prompt eval rate:     %.2f tokens/s\n", float64(r.PromptEvalCount)/r.PromptEvalDuration.Seconds())
+	}
+
+	if r.EvalCount > 0 {
+		fmt.Fprintf(os.Stderr, "eval count:           %d token(s)\n", r.EvalCount)
+	}
+
+	if r.EvalDuration > 0 {
+		fmt.Fprintf(os.Stderr, "eval duration:        %s\n", r.EvalDuration)
+		fmt.Fprintf(os.Stderr, "eval rate:            %.2f tokens/s\n", float64(r.EvalCount)/r.EvalDuration.Seconds())
+	}
+}
+
 type Options struct {
-	Runner
+	Seed int `json:"seed,omitempty"`

-	// Predict options used at runtime
-	NumKeep          int      `json:"num_keep,omitempty"`
-	Seed             int      `json:"seed,omitempty"`
-	NumPredict       int      `json:"num_predict,omitempty"`
-	TopK             int      `json:"top_k,omitempty"`
-	TopP             float32  `json:"top_p,omitempty"`
-	TFSZ             float32  `json:"tfs_z,omitempty"`
-	TypicalP         float32  `json:"typical_p,omitempty"`
-	RepeatLastN      int      `json:"repeat_last_n,omitempty"`
-	Temperature      float32  `json:"temperature,omitempty"`
-	RepeatPenalty    float32  `json:"repeat_penalty,omitempty"`
-	PresencePenalty  float32  `json:"presence_penalty,omitempty"`
-	FrequencyPenalty float32  `json:"frequency_penalty,omitempty"`
-	Mirostat         int      `json:"mirostat,omitempty"`
-	MirostatTau      float32  `json:"mirostat_tau,omitempty"`
-	MirostatEta      float32  `json:"mirostat_eta,omitempty"`
-	PenalizeNewline  bool     `json:"penalize_newline,omitempty"`
-	Stop             []string `json:"stop,omitempty"`
-}
+	// Backend options
+	UseNUMA bool `json:"numa,omitempty"`

-// Runner options which must be set when the model is loaded into memory
-type Runner struct {
-	UseNUMA   bool `json:"numa,omitempty"`
-	NumCtx    int  `json:"num_ctx,omitempty"`
-	NumBatch  int  `json:"num_batch,omitempty"`
-	NumGQA    int  `json:"num_gqa,omitempty"`
-	NumGPU    int  `json:"num_gpu,omitempty"`
-	MainGPU   int  `json:"main_gpu,omitempty"`
-	LowVRAM   bool `json:"low_vram,omitempty"`
-	F16KV     bool `json:"f16_kv,omitempty"`
-	LogitsAll bool `json:"logits_all,omitempty"`
-	VocabOnly bool `json:"vocab_only,omitempty"`
-	UseMMap   bool `json:"use_mmap,omitempty"`
-	UseMLock  bool `json:"use_mlock,omitempty"`
-	NumThread int  `json:"num_thread,omitempty"`
+	// Model options
+	NumCtx        int  `json:"num_ctx,omitempty"`
+	NumBatch      int  `json:"num_batch,omitempty"`
+	NumGPU        int  `json:"num_gpu,omitempty"`
+	MainGPU       int  `json:"main_gpu,omitempty"`
+	LowVRAM       bool `json:"low_vram,omitempty"`
+	F16KV         bool `json:"f16_kv,omitempty"`
+	LogitsAll     bool `json:"logits_all,omitempty"`
+	VocabOnly     bool `json:"vocab_only,omitempty"`
+	UseMMap       bool `json:"use_mmap,omitempty"`
+	UseMLock      bool `json:"use_mlock,omitempty"`
+	EmbeddingOnly bool `json:"embedding_only,omitempty"`

-	// Unused: RopeFrequencyBase is ignored. Instead the value in the model will be used
-	RopeFrequencyBase float32 `json:"rope_frequency_base,omitempty"`
-	// Unused: RopeFrequencyScale is ignored. Instead the value in the model will be used
-	RopeFrequencyScale float32 `json:"rope_frequency_scale,omitempty"`
-}
+	// Predict options
+	RepeatLastN      int     `json:"repeat_last_n,omitempty"`
+	RepeatPenalty    float32 `json:"repeat_penalty,omitempty"`
+	FrequencyPenalty float32 `json:"frequency_penalty,omitempty"`
+	PresencePenalty  float32 `json:"presence_penalty,omitempty"`
+	Temperature      float32 `json:"temperature,omitempty"`
+	TopK             int     `json:"top_k,omitempty"`
+	TopP             float32 `json:"top_p,omitempty"`
+	TFSZ             float32 `json:"tfs_z,omitempty"`
+	TypicalP         float32 `json:"typical_p,omitempty"`
+	Mirostat         int     `json:"mirostat,omitempty"`
+	MirostatTau      float32 `json:"mirostat_tau,omitempty"`
+	MirostatEta      float32 `json:"mirostat_eta,omitempty"`

-type EmbeddingRequest struct {
-	Model     string    `json:"model"`
-	Prompt    string    `json:"prompt"`
-	KeepAlive *Duration `json:"keep_alive,omitempty"`
-
-	Options map[string]interface{} `json:"options"`
-}
-
-type EmbeddingResponse struct {
-	Embedding []float64 `json:"embedding"`
-}
-
-type CreateRequest struct {
-	Model        string `json:"model"`
-	Path         string `json:"path"`
-	Modelfile    string `json:"modelfile"`
-	Stream       *bool  `json:"stream,omitempty"`
-	Quantization string `json:"quantization,omitempty"`
-
-	// Name is deprecated, see Model
-	Name string `json:"name"`
-}
-
-type DeleteRequest struct {
-	Model string `json:"model"`
-
-	// Name is deprecated, see Model
-	Name string `json:"name"`
-}
-
-type ShowRequest struct {
-	Model    string `json:"model"`
-	System   string `json:"system"`
-	Template string `json:"template"`
-
-	Options map[string]interface{} `json:"options"`
-
-	// Name is deprecated, see Model
-	Name string `json:"name"`
-}
-
-type ShowResponse struct {
-	License    string       `json:"license,omitempty"`
-	Modelfile  string       `json:"modelfile,omitempty"`
-	Parameters string       `json:"parameters,omitempty"`
-	Template   string       `json:"template,omitempty"`
-	System     string       `json:"system,omitempty"`
-	Details    ModelDetails `json:"details,omitempty"`
-	Messages   []Message    `json:"messages,omitempty"`
-}
-
-type CopyRequest struct {
-	Source      string `json:"source"`
-	Destination string `json:"destination"`
-}
-
-type PullRequest struct {
-	Model    string `json:"model"`
-	Insecure bool   `json:"insecure,omitempty"`
-	Username string `json:"username"`
-	Password string `json:"password"`
-	Stream   *bool  `json:"stream,omitempty"`
-
-	// Name is deprecated, see Model
-	Name string `json:"name"`
-}
-
-type ProgressResponse struct {
-	Status    string `json:"status"`
-	Digest    string `json:"digest,omitempty"`
-	Total     int64  `json:"total,omitempty"`
-	Completed int64  `json:"completed,omitempty"`
-}
-
-type PushRequest struct {
-	Model    string `json:"model"`
-	Insecure bool   `json:"insecure,omitempty"`
-	Username string `json:"username"`
-	Password string `json:"password"`
-	Stream   *bool  `json:"stream,omitempty"`
-
-	// Name is deprecated, see Model
-	Name string `json:"name"`
-}
-
-type ListResponse struct {
-	Models []ModelResponse `json:"models"`
-}
-
-type ModelResponse struct {
-	Name       string       `json:"name"`
-	Model      string       `json:"model"`
-	ModifiedAt time.Time    `json:"modified_at"`
-	Size       int64        `json:"size"`
-	Digest     string       `json:"digest"`
-	Details    ModelDetails `json:"details,omitempty"`
-}
-
-type TokenResponse struct {
-	Token string `json:"token"`
-}
-
-type GenerateResponse struct {
-	Model     string    `json:"model"`
-	CreatedAt time.Time `json:"created_at"`
-	Response  string    `json:"response"`
-
-	Done    bool  `json:"done"`
-	Context []int `json:"context,omitempty"`
-
-	Metrics
-}
-
-type ModelDetails struct {
-	ParentModel       string   `json:"parent_model"`
-	Format            string   `json:"format"`
-	Family            string   `json:"family"`
-	Families          []string `json:"families"`
-	ParameterSize     string   `json:"parameter_size"`
-	QuantizationLevel string   `json:"quantization_level"`
-}
-
-func (m *Metrics) Summary() {
-	if m.TotalDuration > 0 {
-		fmt.Fprintf(os.Stderr, "total duration:       %v\n", m.TotalDuration)
-	}
-
-	if m.LoadDuration > 0 {
-		fmt.Fprintf(os.Stderr, "load duration:        %v\n", m.LoadDuration)
-	}
-
-	if m.PromptEvalCount > 0 {
-		fmt.Fprintf(os.Stderr, "prompt eval count:    %d token(s)\n", m.PromptEvalCount)
-	}
-
-	if m.PromptEvalDuration > 0 {
-		fmt.Fprintf(os.Stderr, "prompt eval duration: %s\n", m.PromptEvalDuration)
-		fmt.Fprintf(os.Stderr, "prompt eval rate:     %.2f tokens/s\n", float64(m.PromptEvalCount)/m.PromptEvalDuration.Seconds())
-	}
-
-	if m.EvalCount > 0 {
-		fmt.Fprintf(os.Stderr, "eval count:           %d token(s)\n", m.EvalCount)
-	}
-
-	if m.EvalDuration > 0 {
-		fmt.Fprintf(os.Stderr, "eval duration:        %s\n", m.EvalDuration)
-		fmt.Fprintf(os.Stderr, "eval rate:            %.2f tokens/s\n", float64(m.EvalCount)/m.EvalDuration.Seconds())
-	}
-}
-
-var ErrInvalidOpts = errors.New("invalid options")
-var ErrInvalidHostPort = errors.New("invalid port specified in OLLAMA_HOST")
-
-func (opts *Options) FromMap(m map[string]interface{}) error {
-	valueOpts := reflect.ValueOf(opts).Elem() // names of the fields in the options struct
-	typeOpts := reflect.TypeOf(opts).Elem()   // types of the fields in the options struct
-
-	// build map of json struct tags to their types
-	jsonOpts := make(map[string]reflect.StructField)
-	for _, field := range reflect.VisibleFields(typeOpts) {
-		jsonTag := strings.Split(field.Tag.Get("json"), ",")[0]
-		if jsonTag != "" {
-			jsonOpts[jsonTag] = field
-		}
-	}
-
-	invalidOpts := []string{}
-	for key, val := range m {
-		if opt, ok := jsonOpts[key]; ok {
-			field := valueOpts.FieldByName(opt.Name)
-			if field.IsValid() && field.CanSet() {
-				if val == nil {
-					continue
-				}
-
-				switch field.Kind() {
-				case reflect.Int:
-					switch t := val.(type) {
-					case int64:
-						field.SetInt(t)
-					case float64:
-						// when JSON unmarshals numbers, it uses float64, not int
-						field.SetInt(int64(t))
-					default:
-						return fmt.Errorf("option %q must be of type integer", key)
-					}
-				case reflect.Bool:
-					val, ok := val.(bool)
-					if !ok {
-						return fmt.Errorf("option %q must be of type boolean", key)
-					}
-					field.SetBool(val)
-				case reflect.Float32:
-					// JSON unmarshals to float64
-					val, ok := val.(float64)
-					if !ok {
-						return fmt.Errorf("option %q must be of type float32", key)
-					}
-					field.SetFloat(val)
-				case reflect.String:
-					val, ok := val.(string)
-					if !ok {
-						return fmt.Errorf("option %q must be of type string", key)
-					}
-					field.SetString(val)
-				case reflect.Slice:
-					// JSON unmarshals to []interface{}, not []string
-					val, ok := val.([]interface{})
-					if !ok {
-						return fmt.Errorf("option %q must be of type array", key)
-					}
-					// convert []interface{} to []string
-					slice := make([]string, len(val))
-					for i, item := range val {
-						str, ok := item.(string)
-						if !ok {
-							return fmt.Errorf("option %q must be of an array of strings", key)
-						}
-						slice[i] = str
-					}
-					field.Set(reflect.ValueOf(slice))
-				default:
-					return fmt.Errorf("unknown type loading config params: %v", field.Kind())
-				}
-			}
-		} else {
-			invalidOpts = append(invalidOpts, key)
-		}
-	}
-
-	if len(invalidOpts) > 0 {
-		return fmt.Errorf("%w: %v", ErrInvalidOpts, strings.Join(invalidOpts, ", "))
-	}
-	return nil
+	NumThread int `json:"num_thread,omitempty"`
 }

 func DefaultOptions() Options {
 	return Options{
-		// options set on request to runner
-		NumPredict: -1,
+		Seed: -1,

-		// set a minimal num_keep to avoid issues on context shifts
-		NumKeep:          4,
+		UseNUMA: false,
+
+		NumCtx:   2048,
+		NumBatch: 512,
+		NumGPU:   1,
+		LowVRAM:  false,
+		F16KV:    true,
+		UseMMap:  true,
+		UseMLock: false,
+
+		RepeatLastN:      512,
+		RepeatPenalty:    1.1,
+		FrequencyPenalty: 0.0,
+		PresencePenalty:  0.0,
 		Temperature:      0.8,
 		TopK:             40,
 		TopP:             0.9,
 		TFSZ:             1.0,
 		TypicalP:         1.0,
-		RepeatLastN:      64,
-		RepeatPenalty:    1.1,
-		PresencePenalty:  0.0,
-		FrequencyPenalty: 0.0,
 		Mirostat:         0,
 		MirostatTau:      5.0,
 		MirostatEta:      0.1,
-		PenalizeNewline:  true,
-		Seed:             -1,

-		Runner: Runner{
-			// options set when the model is loaded
-			NumCtx:    2048,
-			NumBatch:  512,
-			NumGPU:    -1, // -1 here indicates that NumGPU should be set dynamically
-			NumGQA:    1,
-			NumThread: 0, // let the runtime decide
-			LowVRAM:   false,
-			F16KV:     true,
-			UseMLock:  false,
-			UseMMap:   true,
-			UseNUMA:   false,
-		},
+		NumThread: runtime.NumCPU(),
 	}
 }
-
-type Duration struct {
-	time.Duration
-}
-
-func (d *Duration) UnmarshalJSON(b []byte) (err error) {
-	var v any
-	if err := json.Unmarshal(b, &v); err != nil {
-		return err
-	}
-
-	d.Duration = 5 * time.Minute
-
-	switch t := v.(type) {
-	case float64:
-		if t < 0 {
-			d.Duration = time.Duration(math.MaxInt64)
-		} else {
-			d.Duration = time.Duration(t * float64(time.Second))
-		}
-	case string:
-		d.Duration, err = time.ParseDuration(t)
-		if err != nil {
-			return err
-		}
-		if d.Duration < 0 {
-			d.Duration = time.Duration(math.MaxInt64)
-		}
-	}
-
-	return nil
-}
-
-// FormatParams converts specified parameter options to their correct types
-func FormatParams(params map[string][]string) (map[string]interface{}, error) {
-	opts := Options{}
-	valueOpts := reflect.ValueOf(&opts).Elem() // names of the fields in the options struct
-	typeOpts := reflect.TypeOf(opts)           // types of the fields in the options struct
-
-	// build map of json struct tags to their types
-	jsonOpts := make(map[string]reflect.StructField)
-	for _, field := range reflect.VisibleFields(typeOpts) {
-		jsonTag := strings.Split(field.Tag.Get("json"), ",")[0]
-		if jsonTag != "" {
-			jsonOpts[jsonTag] = field
-		}
-	}
-
-	out := make(map[string]interface{})
-	// iterate params and set values based on json struct tags
-	for key, vals := range params {
-		if opt, ok := jsonOpts[key]; !ok {
-			return nil, fmt.Errorf("unknown parameter '%s'", key)
-		} else {
-			field := valueOpts.FieldByName(opt.Name)
-			if field.IsValid() && field.CanSet() {
-				switch field.Kind() {
-				case reflect.Float32:
-					floatVal, err := strconv.ParseFloat(vals[0], 32)
-					if err != nil {
-						return nil, fmt.Errorf("invalid float value %s", vals)
-					}
-
-					out[key] = float32(floatVal)
-				case reflect.Int:
-					intVal, err := strconv.ParseInt(vals[0], 10, 64)
-					if err != nil {
-						return nil, fmt.Errorf("invalid int value %s", vals)
-					}
-
-					out[key] = intVal
-				case reflect.Bool:
-					boolVal, err := strconv.ParseBool(vals[0])
-					if err != nil {
-						return nil, fmt.Errorf("invalid bool value %s", vals)
-					}
-
-					out[key] = boolVal
-				case reflect.String:
-					out[key] = vals[0]
-				case reflect.Slice:
-					// TODO: only string slices are supported right now
-					out[key] = vals
-				default:
-					return nil, fmt.Errorf("unknown type %s for %s", field.Kind(), key)
-				}
-			}
-		}
-	}
-
-	return out, nil
-}
@@ -1,50 +0,0 @@
-package api
-
-import (
-	"encoding/json"
-	"math"
-	"testing"
-	"time"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestKeepAliveParsingFromJSON(t *testing.T) {
-	tests := []struct {
-		name string
-		req  string
-		exp  *Duration
-	}{
-		{
-			name: "Positive Integer",
-			req:  `{ "keep_alive": 42 }`,
-			exp:  &Duration{42 * time.Second},
-		},
-		{
-			name: "Positive Integer String",
-			req:  `{ "keep_alive": "42m" }`,
-			exp:  &Duration{42 * time.Minute},
-		},
-		{
-			name: "Negative Integer",
-			req:  `{ "keep_alive": -1 }`,
-			exp:  &Duration{math.MaxInt64},
-		},
-		{
-			name: "Negative Integer String",
-			req:  `{ "keep_alive": "-1m" }`,
-			exp:  &Duration{math.MaxInt64},
-		},
-	}
-
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			var dec ChatRequest
-			err := json.Unmarshal([]byte(test.req), &dec)
-			require.NoError(t, err)
-
-			assert.Equal(t, test.exp, dec.KeepAlive)
-		})
-	}
-}
@@ -1 +1,92 @@
-ollama.syso
+# Logs
+logs
+*.log
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+lerna-debug.log*
+
+# Diagnostic reports (https://nodejs.org/api/report.html)
+report.[0-9]*.[0-9]*.[0-9]*.[0-9]*.json
+
+# Runtime data
+pids
+*.pid
+*.seed
+*.pid.lock
+.DS_Store
+
+# Directory for instrumented libs generated by jscoverage/JSCover
+lib-cov
+
+# Coverage directory used by tools like istanbul
+coverage
+*.lcov
+
+# nyc test coverage
+.nyc_output
+
+# node-waf configuration
+.lock-wscript
+
+# Compiled binary addons (https://nodejs.org/api/addons.html)
+build/Release
+
+# Dependency directories
+node_modules/
+jspm_packages/
+
+# TypeScript v1 declaration files
+typings/
+
+# TypeScript cache
+*.tsbuildinfo
+
+# Optional npm cache directory
+.npm
+
+# Optional eslint cache
+.eslintcache
+
+# Optional REPL history
+.node_repl_history
+
+# Output of 'npm pack'
+*.tgz
+
+# Yarn Integrity file
+.yarn-integrity
+
+# dotenv environment variables file
+.env
+.env.test
+
+# parcel-bundler cache (https://parceljs.org/)
+.cache
+
+# next.js build output
+.next
+
+# nuxt.js build output
+.nuxt
+
+# vuepress build output
+.vuepress/dist
+
+# Serverless directories
+.serverless/
+
+# FuseBox cache
+.fusebox/
+
+# DynamoDB Local files
+.dynamodb/
+
+# Webpack
+.webpack/
+
+# Vite
+.vite/
+
+# Electron-Forge
+out/
@@ -1,22 +1,27 @@
-# Ollama App
+# Desktop

-## Linux
+_Note: the Ollama desktop app is a work in progress and is not ready yet for general use._

-TODO
+This app builds upon Ollama to provide a desktop experience for running models.

-## MacOS
+## Developing

-TODO
-
-## Windows
-
-If you want to build the installer, youll need to install
- https://jrsoftware.org/isinfo.php
-
-
-In the top directory of this repo, run the following powershell script
-to build the ollama CLI, ollama app, and ollama installer.
+First, build the `ollama` binary:

 ```
-powershell -ExecutionPolicy Bypass -File .\scripts\build_windows.ps1
+make -C ..
 ```
+
+Then run the desktop app with `npm start`:
+
+```
+npm install
+npm start
+```
+
+## Coming soon
+
+- Browse the latest available models on Hugging Face and other sources
+- Keep track of previous conversations with models
+- Switch quickly between models
+- Connect to remote Ollama servers to run models
@@ -1,17 +0,0 @@
-package assets
-
-import (
-	"embed"
-	"io/fs"
-)
-
-//go:embed *.ico
-var icons embed.FS
-
-func ListIcons() ([]string, error) {
-	return fs.Glob(icons, "*")
-}
-
-func GetIcon(filename string) ([]byte, error) {
-	return icons.ReadFile(filename)
-}
@@ -18,15 +18,12 @@ const config: ForgeConfig = {
    asar: true,
    icon: './assets/icon.icns',
    extraResource: [
-      '../dist/ollama',
-      path.join(__dirname, './assets/iconTemplate.png'),
-      path.join(__dirname, './assets/iconTemplate@2x.png'),
-      path.join(__dirname, './assets/iconUpdateTemplate.png'),
-      path.join(__dirname, './assets/iconUpdateTemplate@2x.png'),
-      path.join(__dirname, './assets/iconDarkTemplate.png'),
-      path.join(__dirname, './assets/iconDarkTemplate@2x.png'),
-      path.join(__dirname, './assets/iconDarkUpdateTemplate.png'),
-      path.join(__dirname, './assets/iconDarkUpdateTemplate@2x.png'),
+      '../ollama',
+      path.join(__dirname, './assets/ollama_icon_16x16Template.png'),
+      path.join(__dirname, './assets/ollama_icon_16x16Template@2x.png'),
+      path.join(__dirname, './assets/ollama_outline_icon_16x16Template.png'),
+      path.join(__dirname, './assets/ollama_outline_icon_16x16Template@2x.png'),
+      ...(process.platform === 'darwin' ? ['../llama/ggml-metal.metal'] : []),
    ],
    ...(process.env.SIGN
      ? {
@@ -41,12 +38,19 @@ const config: ForgeConfig = {
          },
        }
      : {}),
-    osxUniversal: {
-      x64ArchFiles: '**/ollama',
-    },
  },
  rebuildConfig: {},
  makers: [new MakerSquirrel({}), new MakerZIP({}, ['darwin'])],
+  publishers: [
+    new PublisherGithub({
+      repository: {
+        name: 'ollama',
+        owner: 'jmorganca',
+      },
+      draft: false,
+      prerelease: true,
+    }),
+  ],
  hooks: {
    readPackageJson: async (_, packageJson) => {
      return { ...packageJson, version: process.env.VERSION || packageJson.version }
@@ -1,9 +0,0 @@
-//go:build !windows
-
-package lifecycle
-
-import "fmt"
-
-func GetStarted() error {
-	return fmt.Errorf("GetStarted not implemented")
-}
@@ -1,44 +0,0 @@
-package lifecycle
-
-import (
-	"fmt"
-	"log/slog"
-	"os"
-	"os/exec"
-	"path/filepath"
-	"syscall"
-)
-
-func GetStarted() error {
-	const CREATE_NEW_CONSOLE = 0x00000010
-	var err error
-	bannerScript := filepath.Join(AppDir, "ollama_welcome.ps1")
-	args := []string{
-		// TODO once we're signed, the execution policy bypass should be removed
-		"powershell", "-noexit", "-ExecutionPolicy", "Bypass", "-nologo", "-file", bannerScript,
-	}
-	args[0], err = exec.LookPath(args[0])
-	if err != nil {
-		return err
-	}
-
-	// Make sure the script actually exists
-	_, err = os.Stat(bannerScript)
-	if err != nil {
-		return fmt.Errorf("getting started banner script error %s", err)
-	}
-
-	slog.Info(fmt.Sprintf("opening getting started terminal with %v", args))
-	attrs := &os.ProcAttr{
-		Files: []*os.File{os.Stdin, os.Stdout, os.Stderr},
-		Sys:   &syscall.SysProcAttr{CreationFlags: CREATE_NEW_CONSOLE, HideWindow: false},
-	}
-	proc, err := os.StartProcess(args[0], args, attrs)
-
-	if err != nil {
-		return fmt.Errorf("unable to start getting started shell %w", err)
-	}
-
-	slog.Debug(fmt.Sprintf("getting started terminal PID: %d", proc.Pid))
-	return proc.Release()
-}
@@ -1,92 +0,0 @@
-package lifecycle
-
-import (
-	"context"
-	"fmt"
-	"log"
-	"log/slog"
-	"os"
-	"os/signal"
-	"syscall"
-
-	"github.com/ollama/ollama/app/store"
-	"github.com/ollama/ollama/app/tray"
-)
-
-func Run() {
-	InitLogging()
-
-	ctx, cancel := context.WithCancel(context.Background())
-	var done chan int
-
-	t, err := tray.NewTray()
-	if err != nil {
-		log.Fatalf("Failed to start: %s", err)
-	}
-	callbacks := t.GetCallbacks()
-
-	signals := make(chan os.Signal, 1)
-	signal.Notify(signals, syscall.SIGINT, syscall.SIGTERM)
-
-	go func() {
-		slog.Debug("starting callback loop")
-		for {
-			select {
-			case <-callbacks.Quit:
-				slog.Debug("quit called")
-				t.Quit()
-			case <-signals:
-				slog.Debug("shutting down due to signal")
-				t.Quit()
-			case <-callbacks.Update:
-				err := DoUpgrade(cancel, done)
-				if err != nil {
-					slog.Warn(fmt.Sprintf("upgrade attempt failed: %s", err))
-				}
-			case <-callbacks.ShowLogs:
-				ShowLogs()
-			case <-callbacks.DoFirstUse:
-				err := GetStarted()
-				if err != nil {
-					slog.Warn(fmt.Sprintf("Failed to launch getting started shell: %s", err))
-				}
-			}
-		}
-	}()
-
-	// Are we first use?
-	if !store.GetFirstTimeRun() {
-		slog.Debug("First time run")
-		err = t.DisplayFirstUseNotification()
-		if err != nil {
-			slog.Debug(fmt.Sprintf("XXX failed to display first use notification %v", err))
-		}
-		store.SetFirstTimeRun(true)
-	} else {
-		slog.Debug("Not first time, skipping first run notification")
-	}
-
-	if IsServerRunning(ctx) {
-		slog.Info("Detected another instance of ollama running, exiting")
-		os.Exit(1)
-	} else {
-		done, err = SpawnServer(ctx, CLIName)
-		if err != nil {
-			// TODO - should we retry in a backoff loop?
-			// TODO - should we pop up a warning and maybe add a menu item to view application logs?
-			slog.Error(fmt.Sprintf("Failed to spawn ollama server %s", err))
-			done = make(chan int, 1)
-			done <- 1
-		}
-	}
-
-	StartBackgroundUpdaterChecker(ctx, t.UpdateAvailable)
-
-	t.Run()
-	cancel()
-	slog.Info("Waiting for ollama server to shutdown...")
-	if done != nil {
-		<-done
-	}
-	slog.Info("Ollama app exiting")
-}
@@ -1,46 +0,0 @@
-package lifecycle
-
-import (
-	"fmt"
-	"log/slog"
-	"os"
-	"path/filepath"
-)
-
-func InitLogging() {
-	level := slog.LevelInfo
-
-	if debug := os.Getenv("OLLAMA_DEBUG"); debug != "" {
-		level = slog.LevelDebug
-	}
-
-	var logFile *os.File
-	var err error
-	// Detect if we're a GUI app on windows, and if not, send logs to console
-	if os.Stderr.Fd() != 0 {
-		// Console app detected
-		logFile = os.Stderr
-		// TODO - write one-line to the app.log file saying we're running in console mode to help avoid confusion
-	} else {
-		logFile, err = os.OpenFile(AppLogFile, os.O_APPEND|os.O_WRONLY|os.O_CREATE, 0755)
-		if err != nil {
-			slog.Error(fmt.Sprintf("failed to create server log %v", err))
-			return
-		}
-	}
-	handler := slog.NewTextHandler(logFile, &slog.HandlerOptions{
-		Level:     level,
-		AddSource: true,
-		ReplaceAttr: func(_ []string, attr slog.Attr) slog.Attr {
-			if attr.Key == slog.SourceKey {
-				source := attr.Value.Any().(*slog.Source)
-				source.File = filepath.Base(source.File)
-			}
-			return attr
-		},
-	})
-
-	slog.SetDefault(slog.New(handler))
-
-	slog.Info("ollama app started")
-}
@@ -1,9 +0,0 @@
-//go:build !windows
-
-package lifecycle
-
-import "log/slog"
-
-func ShowLogs() {
-	slog.Warn("ShowLogs not yet implemented")
-}
@@ -1,19 +0,0 @@
-package lifecycle
-
-import (
-	"fmt"
-	"log/slog"
-	"os/exec"
-	"syscall"
-)
-
-func ShowLogs() {
-	cmd_path := "c:\\Windows\\system32\\cmd.exe"
-	slog.Debug(fmt.Sprintf("viewing logs with start %s", AppDataDir))
-	cmd := exec.Command(cmd_path, "/c", "start", AppDataDir)
-	cmd.SysProcAttr = &syscall.SysProcAttr{HideWindow: false, CreationFlags: 0x08000000}
-	err := cmd.Start()
-	if err != nil {
-		slog.Error(fmt.Sprintf("Failed to open log dir: %s", err))
-	}
-}
@@ -1,79 +0,0 @@
-package lifecycle
-
-import (
-	"errors"
-	"fmt"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"runtime"
-	"strings"
-)
-
-var (
-	AppName    = "ollama app"
-	CLIName    = "ollama"
-	AppDir     = "/opt/Ollama"
-	AppDataDir = "/opt/Ollama"
-	// TODO - should there be a distinct log dir?
-	UpdateStageDir = "/tmp"
-	AppLogFile     = "/tmp/ollama_app.log"
-	ServerLogFile  = "/tmp/ollama.log"
-	UpgradeLogFile = "/tmp/ollama_update.log"
-	Installer      = "OllamaSetup.exe"
-)
-
-func init() {
-	if runtime.GOOS == "windows" {
-		AppName += ".exe"
-		CLIName += ".exe"
-		// Logs, configs, downloads go to LOCALAPPDATA
-		localAppData := os.Getenv("LOCALAPPDATA")
-		AppDataDir = filepath.Join(localAppData, "Ollama")
-		UpdateStageDir = filepath.Join(AppDataDir, "updates")
-		AppLogFile = filepath.Join(AppDataDir, "app.log")
-		ServerLogFile = filepath.Join(AppDataDir, "server.log")
-		UpgradeLogFile = filepath.Join(AppDataDir, "upgrade.log")
-
-		// Executables are stored in APPDATA
-		AppDir = filepath.Join(localAppData, "Programs", "Ollama")
-
-		// Make sure we have PATH set correctly for any spawned children
-		paths := strings.Split(os.Getenv("PATH"), ";")
-		// Start with whatever we find in the PATH/LD_LIBRARY_PATH
-		found := false
-		for _, path := range paths {
-			d, err := filepath.Abs(path)
-			if err != nil {
-				continue
-			}
-			if strings.EqualFold(AppDir, d) {
-				found = true
-			}
-		}
-		if !found {
-			paths = append(paths, AppDir)
-
-			pathVal := strings.Join(paths, ";")
-			slog.Debug("setting PATH=" + pathVal)
-			err := os.Setenv("PATH", pathVal)
-			if err != nil {
-				slog.Error(fmt.Sprintf("failed to update PATH: %s", err))
-			}
-		}
-
-		// Make sure our logging dir exists
-		_, err := os.Stat(AppDataDir)
-		if errors.Is(err, os.ErrNotExist) {
-			if err := os.MkdirAll(AppDataDir, 0o755); err != nil {
-				slog.Error(fmt.Sprintf("create ollama dir %s: %v", AppDataDir, err))
-			}
-		}
-
-	} else if runtime.GOOS == "darwin" {
-		// TODO
-		AppName += ".app"
-		// } else if runtime.GOOS == "linux" {
-		// TODO
-	}
-}
@@ -1,181 +0,0 @@
-package lifecycle
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"io"
-	"log/slog"
-	"os"
-	"os/exec"
-	"path/filepath"
-	"time"
-
-	"github.com/ollama/ollama/api"
-)
-
-func getCLIFullPath(command string) string {
-	cmdPath := ""
-	appExe, err := os.Executable()
-	if err == nil {
-		cmdPath = filepath.Join(filepath.Dir(appExe), command)
-		_, err := os.Stat(cmdPath)
-		if err == nil {
-			return cmdPath
-		}
-	}
-	cmdPath, err = exec.LookPath(command)
-	if err == nil {
-		_, err := os.Stat(cmdPath)
-		if err == nil {
-			return cmdPath
-		}
-	}
-	pwd, err := os.Getwd()
-	if err == nil {
-		cmdPath = filepath.Join(pwd, command)
-		_, err = os.Stat(cmdPath)
-		if err == nil {
-			return cmdPath
-		}
-	}
-
-	return command
-}
-
-func start(ctx context.Context, command string) (*exec.Cmd, error) {
-	cmd := getCmd(ctx, getCLIFullPath(command))
-	stdout, err := cmd.StdoutPipe()
-	if err != nil {
-		return nil, fmt.Errorf("failed to spawn server stdout pipe: %w", err)
-	}
-	stderr, err := cmd.StderrPipe()
-	if err != nil {
-		return nil, fmt.Errorf("failed to spawn server stderr pipe: %w", err)
-	}
-
-	// TODO - rotation
-	logFile, err := os.OpenFile(ServerLogFile, os.O_APPEND|os.O_WRONLY|os.O_CREATE, 0755)
-	if err != nil {
-		return nil, fmt.Errorf("failed to create server log: %w", err)
-	}
-
-	logDir := filepath.Dir(ServerLogFile)
-	_, err = os.Stat(logDir)
-	if err != nil {
-		if !errors.Is(err, os.ErrNotExist) {
-			return nil, fmt.Errorf("stat ollama server log dir %s: %v", logDir, err)
-
-		}
-
-		if err := os.MkdirAll(logDir, 0o755); err != nil {
-			return nil, fmt.Errorf("create ollama server log dir %s: %v", logDir, err)
-		}
-	}
-
-	go func() {
-		defer logFile.Close()
-		io.Copy(logFile, stdout) //nolint:errcheck
-	}()
-	go func() {
-		defer logFile.Close()
-		io.Copy(logFile, stderr) //nolint:errcheck
-	}()
-
-	// Re-wire context done behavior to attempt a graceful shutdown of the server
-	cmd.Cancel = func() error {
-		if cmd.Process != nil {
-			err := terminate(cmd)
-			if err != nil {
-				slog.Warn("error trying to gracefully terminate server", "err", err)
-				return cmd.Process.Kill()
-			}
-
-			tick := time.NewTicker(10 * time.Millisecond)
-			defer tick.Stop()
-
-			for {
-				select {
-				case <-tick.C:
-					exited, err := isProcessExited(cmd.Process.Pid)
-					if err != nil {
-						return err
-					}
-
-					if exited {
-						return nil
-					}
-				case <-time.After(5 * time.Second):
-					slog.Warn("graceful server shutdown timeout, killing", "pid", cmd.Process.Pid)
-					return cmd.Process.Kill()
-				}
-			}
-		}
-		return nil
-	}
-
-	// run the command and wait for it to finish
-	if err := cmd.Start(); err != nil {
-		return nil, fmt.Errorf("failed to start server %w", err)
-	}
-	if cmd.Process != nil {
-		slog.Info(fmt.Sprintf("started ollama server with pid %d", cmd.Process.Pid))
-	}
-	slog.Info(fmt.Sprintf("ollama server logs %s", ServerLogFile))
-
-	return cmd, nil
-}
-
-func SpawnServer(ctx context.Context, command string) (chan int, error) {
-	done := make(chan int)
-
-	go func() {
-		// Keep the server running unless we're shuttind down the app
-		crashCount := 0
-		for {
-			slog.Info("starting server...")
-			cmd, err := start(ctx, command)
-			if err != nil {
-				crashCount++
-				slog.Error(fmt.Sprintf("failed to start server %s", err))
-				time.Sleep(500 * time.Millisecond * time.Duration(crashCount))
-				continue
-			}
-
-			cmd.Wait() //nolint:errcheck
-			var code int
-			if cmd.ProcessState != nil {
-				code = cmd.ProcessState.ExitCode()
-			}
-
-			select {
-			case <-ctx.Done():
-				slog.Info(fmt.Sprintf("server shutdown with exit code %d", code))
-				done <- code
-				return
-			default:
-				crashCount++
-				slog.Warn(fmt.Sprintf("server crash %d - exit code %d - respawning", crashCount, code))
-				time.Sleep(500 * time.Millisecond * time.Duration(crashCount))
-				break
-			}
-		}
-	}()
-
-	return done, nil
-}
-
-func IsServerRunning(ctx context.Context) bool {
-	client, err := api.ClientFromEnvironment()
-	if err != nil {
-		slog.Info("unable to connect to server")
-		return false
-	}
-	err = client.Heartbeat(ctx)
-	if err != nil {
-		slog.Debug(fmt.Sprintf("heartbeat from server: %s", err))
-		slog.Info("unable to connect to server")
-		return false
-	}
-	return true
-}
@@ -1,38 +0,0 @@
-//go:build !windows
-
-package lifecycle
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"os"
-	"os/exec"
-	"syscall"
-)
-
-func getCmd(ctx context.Context, cmd string) *exec.Cmd {
-	return exec.CommandContext(ctx, cmd, "serve")
-}
-
-func terminate(cmd *exec.Cmd) error {
-	return cmd.Process.Signal(os.Interrupt)
-}
-
-func isProcessExited(pid int) (bool, error) {
-	proc, err := os.FindProcess(pid)
-	if err != nil {
-		return false, fmt.Errorf("failed to find process: %v", err)
-	}
-
-	err = proc.Signal(syscall.Signal(0))
-	if err != nil {
-		if errors.Is(err, os.ErrProcessDone) || errors.Is(err, syscall.ESRCH) {
-			return true, nil
-		}
-
-		return false, fmt.Errorf("error signaling process: %v", err)
-	}
-
-	return false, nil
-}
@@ -1,89 +0,0 @@
-package lifecycle
-
-import (
-	"context"
-	"fmt"
-	"os/exec"
-	"syscall"
-
-	"golang.org/x/sys/windows"
-)
-
-func getCmd(ctx context.Context, exePath string) *exec.Cmd {
-	cmd := exec.CommandContext(ctx, exePath, "serve")
-	cmd.SysProcAttr = &syscall.SysProcAttr{
-		HideWindow:    true,
-		CreationFlags: windows.CREATE_NEW_PROCESS_GROUP,
-	}
-
-	return cmd
-}
-
-func terminate(cmd *exec.Cmd) error {
-	dll, err := windows.LoadDLL("kernel32.dll")
-	if err != nil {
-		return err
-	}
-	defer dll.Release() // nolint: errcheck
-
-	pid := cmd.Process.Pid
-
-	f, err := dll.FindProc("AttachConsole")
-	if err != nil {
-		return err
-	}
-
-	r1, _, err := f.Call(uintptr(pid))
-	if r1 == 0 && err != syscall.ERROR_ACCESS_DENIED {
-		return err
-	}
-
-	f, err = dll.FindProc("SetConsoleCtrlHandler")
-	if err != nil {
-		return err
-	}
-
-	r1, _, err = f.Call(0, 1)
-	if r1 == 0 {
-		return err
-	}
-
-	f, err = dll.FindProc("GenerateConsoleCtrlEvent")
-	if err != nil {
-		return err
-	}
-
-	r1, _, err = f.Call(windows.CTRL_BREAK_EVENT, uintptr(pid))
-	if r1 == 0 {
-		return err
-	}
-
-	r1, _, err = f.Call(windows.CTRL_C_EVENT, uintptr(pid))
-	if r1 == 0 {
-		return err
-	}
-
-	return nil
-}
-
-const STILL_ACTIVE = 259
-
-func isProcessExited(pid int) (bool, error) {
-	hProcess, err := windows.OpenProcess(windows.PROCESS_QUERY_INFORMATION, false, uint32(pid))
-	if err != nil {
-		return false, fmt.Errorf("failed to open process: %v", err)
-	}
-	defer windows.CloseHandle(hProcess) // nolint: errcheck
-
-	var exitCode uint32
-	err = windows.GetExitCodeProcess(hProcess, &exitCode)
-	if err != nil {
-		return false, fmt.Errorf("failed to get exit code: %v", err)
-	}
-
-	if exitCode == STILL_ACTIVE {
-		return false, nil
-	}
-
-	return true, nil
-}
@@ -1,228 +0,0 @@
-package lifecycle
-
-import (
-	"context"
-	"crypto/rand"
-	"encoding/json"
-	"errors"
-	"fmt"
-	"io"
-	"log/slog"
-	"mime"
-	"net/http"
-	"net/url"
-	"os"
-	"path"
-	"path/filepath"
-	"runtime"
-	"strings"
-	"time"
-
-	"github.com/ollama/ollama/auth"
-	"github.com/ollama/ollama/version"
-)
-
-var (
-	UpdateCheckURLBase  = "https://ollama.com/api/update"
-	UpdateDownloaded    = false
-	UpdateCheckInterval = 60 * 60 * time.Second
-)
-
-// TODO - maybe move up to the API package?
-type UpdateResponse struct {
-	UpdateURL     string `json:"url"`
-	UpdateVersion string `json:"version"`
-}
-
-func IsNewReleaseAvailable(ctx context.Context) (bool, UpdateResponse) {
-	var updateResp UpdateResponse
-
-	requestURL, err := url.Parse(UpdateCheckURLBase)
-	if err != nil {
-		return false, updateResp
-	}
-
-	query := requestURL.Query()
-	query.Add("os", runtime.GOOS)
-	query.Add("arch", runtime.GOARCH)
-	query.Add("version", version.Version)
-	query.Add("ts", fmt.Sprintf("%d", time.Now().Unix()))
-
-	nonce, err := auth.NewNonce(rand.Reader, 16)
-	if err != nil {
-		return false, updateResp
-	}
-
-	query.Add("nonce", nonce)
-	requestURL.RawQuery = query.Encode()
-
-	data := []byte(fmt.Sprintf("%s,%s", http.MethodGet, requestURL.RequestURI()))
-	signature, err := auth.Sign(ctx, data)
-	if err != nil {
-		return false, updateResp
-	}
-
-	req, err := http.NewRequestWithContext(ctx, http.MethodGet, requestURL.String(), nil)
-	if err != nil {
-		slog.Warn(fmt.Sprintf("failed to check for update: %s", err))
-		return false, updateResp
-	}
-	req.Header.Set("Authorization", signature)
-	req.Header.Set("User-Agent", fmt.Sprintf("ollama/%s (%s %s) Go/%s", version.Version, runtime.GOARCH, runtime.GOOS, runtime.Version()))
-
-	slog.Debug("checking for available update", "requestURL", requestURL)
-	resp, err := http.DefaultClient.Do(req)
-	if err != nil {
-		slog.Warn(fmt.Sprintf("failed to check for update: %s", err))
-		return false, updateResp
-	}
-	defer resp.Body.Close()
-
-	if resp.StatusCode == 204 {
-		slog.Debug("check update response 204 (current version is up to date)")
-		return false, updateResp
-	}
-	body, err := io.ReadAll(resp.Body)
-	if err != nil {
-		slog.Warn(fmt.Sprintf("failed to read body response: %s", err))
-	}
-
-	if resp.StatusCode != 200 {
-		slog.Info(fmt.Sprintf("check update error %d - %.96s", resp.StatusCode, string(body)))
-		return false, updateResp
-	}
-	err = json.Unmarshal(body, &updateResp)
-	if err != nil {
-		slog.Warn(fmt.Sprintf("malformed response checking for update: %s", err))
-		return false, updateResp
-	}
-	// Extract the version string from the URL in the github release artifact path
-	updateResp.UpdateVersion = path.Base(path.Dir(updateResp.UpdateURL))
-
-	slog.Info("New update available at " + updateResp.UpdateURL)
-	return true, updateResp
-}
-
-func DownloadNewRelease(ctx context.Context, updateResp UpdateResponse) error {
-	// Do a head first to check etag info
-	req, err := http.NewRequestWithContext(ctx, http.MethodHead, updateResp.UpdateURL, nil)
-	if err != nil {
-		return err
-	}
-
-	resp, err := http.DefaultClient.Do(req)
-	if err != nil {
-		return fmt.Errorf("error checking update: %w", err)
-	}
-	if resp.StatusCode != 200 {
-		return fmt.Errorf("unexpected status attempting to download update %d", resp.StatusCode)
-	}
-	resp.Body.Close()
-	etag := strings.Trim(resp.Header.Get("etag"), "\"")
-	if etag == "" {
-		slog.Debug("no etag detected, falling back to filename based dedup")
-		etag = "_"
-	}
-	filename := Installer
-	_, params, err := mime.ParseMediaType(resp.Header.Get("content-disposition"))
-	if err == nil {
-		filename = params["filename"]
-	}
-
-	stageFilename := filepath.Join(UpdateStageDir, etag, filename)
-
-	// Check to see if we already have it downloaded
-	_, err = os.Stat(stageFilename)
-	if err == nil {
-		slog.Info("update already downloaded")
-		return nil
-	}
-
-	cleanupOldDownloads()
-
-	req.Method = http.MethodGet
-	resp, err = http.DefaultClient.Do(req)
-	if err != nil {
-		return fmt.Errorf("error checking update: %w", err)
-	}
-	defer resp.Body.Close()
-	etag = strings.Trim(resp.Header.Get("etag"), "\"")
-	if etag == "" {
-		slog.Debug("no etag detected, falling back to filename based dedup") // TODO probably can get rid of this redundant log
-		etag = "_"
-	}
-
-	stageFilename = filepath.Join(UpdateStageDir, etag, filename)
-
-	_, err = os.Stat(filepath.Dir(stageFilename))
-	if errors.Is(err, os.ErrNotExist) {
-		if err := os.MkdirAll(filepath.Dir(stageFilename), 0o755); err != nil {
-			return fmt.Errorf("create ollama dir %s: %v", filepath.Dir(stageFilename), err)
-		}
-	}
-
-	payload, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return fmt.Errorf("failed to read body response: %w", err)
-	}
-	fp, err := os.OpenFile(stageFilename, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0o755)
-	if err != nil {
-		return fmt.Errorf("write payload %s: %w", stageFilename, err)
-	}
-	defer fp.Close()
-	if n, err := fp.Write(payload); err != nil || n != len(payload) {
-		return fmt.Errorf("write payload %s: %d vs %d -- %w", stageFilename, n, len(payload), err)
-	}
-	slog.Info("new update downloaded " + stageFilename)
-
-	UpdateDownloaded = true
-	return nil
-}
-
-func cleanupOldDownloads() {
-	files, err := os.ReadDir(UpdateStageDir)
-	if err != nil && errors.Is(err, os.ErrNotExist) {
-		// Expected behavior on first run
-		return
-	} else if err != nil {
-		slog.Warn(fmt.Sprintf("failed to list stage dir: %s", err))
-		return
-	}
-	for _, file := range files {
-		fullname := filepath.Join(UpdateStageDir, file.Name())
-		slog.Debug("cleaning up old download: " + fullname)
-		err = os.RemoveAll(fullname)
-		if err != nil {
-			slog.Warn(fmt.Sprintf("failed to cleanup stale update download %s", err))
-		}
-	}
-}
-
-func StartBackgroundUpdaterChecker(ctx context.Context, cb func(string) error) {
-	go func() {
-		// Don't blast an update message immediately after startup
-		// time.Sleep(30 * time.Second)
-		time.Sleep(3 * time.Second)
-
-		for {
-			available, resp := IsNewReleaseAvailable(ctx)
-			if available {
-				err := DownloadNewRelease(ctx, resp)
-				if err != nil {
-					slog.Error(fmt.Sprintf("failed to download new release: %s", err))
-				}
-				err = cb(resp.UpdateVersion)
-				if err != nil {
-					slog.Warn(fmt.Sprintf("failed to register update available with tray: %s", err))
-				}
-			}
-			select {
-			case <-ctx.Done():
-				slog.Debug("stopping background update checker")
-				return
-			default:
-				time.Sleep(UpdateCheckInterval)
-			}
-		}
-	}()
-}
@@ -1,12 +0,0 @@
-//go:build !windows
-
-package lifecycle
-
-import (
-	"context"
-	"fmt"
-)
-
-func DoUpgrade(cancel context.CancelFunc, done chan int) error {
-	return fmt.Errorf("DoUpgrade not yet implemented")
-}
@@ -1,80 +0,0 @@
-package lifecycle
-
-import (
-	"context"
-	"fmt"
-	"log/slog"
-	"os"
-	"os/exec"
-	"path/filepath"
-)
-
-func DoUpgrade(cancel context.CancelFunc, done chan int) error {
-	files, err := filepath.Glob(filepath.Join(UpdateStageDir, "*", "*.exe")) // TODO generalize for multiplatform
-	if err != nil {
-		return fmt.Errorf("failed to lookup downloads: %s", err)
-	}
-	if len(files) == 0 {
-		return fmt.Errorf("no update downloads found")
-	} else if len(files) > 1 {
-		// Shouldn't happen
-		slog.Warn(fmt.Sprintf("multiple downloads found, using first one %v", files))
-	}
-	installerExe := files[0]
-
-	slog.Info("starting upgrade with " + installerExe)
-	slog.Info("upgrade log file " + UpgradeLogFile)
-
-	// When running in debug mode, we'll be "verbose" and let the installer pop up and prompt
-	installArgs := []string{
-		"/CLOSEAPPLICATIONS",                    // Quit the tray app if it's still running
-		"/LOG=" + filepath.Base(UpgradeLogFile), // Only relative seems reliable, so set pwd
-		"/FORCECLOSEAPPLICATIONS",               // Force close the tray app - might be needed
-	}
-	// When we're not in debug mode, make the upgrade as quiet as possible (no GUI, no prompts)
-	// TODO - temporarily disable since we're pinning in debug mode for the preview
-	// if debug := os.Getenv("OLLAMA_DEBUG"); debug == "" {
-	installArgs = append(installArgs,
-		"/SP", // Skip the "This will install... Do you wish to continue" prompt
-		"/SUPPRESSMSGBOXES",
-		"/SILENT",
-		"/VERYSILENT",
-	)
-	// }
-
-	// Safeguard in case we have requests in flight that need to drain...
-	slog.Info("Waiting for server to shutdown")
-	cancel()
-	if done != nil {
-		<-done
-	} else {
-		// Shouldn't happen
-		slog.Warn("done chan was nil, not actually waiting")
-	}
-
-	slog.Debug(fmt.Sprintf("starting installer: %s %v", installerExe, installArgs))
-	os.Chdir(filepath.Dir(UpgradeLogFile)) //nolint:errcheck
-	cmd := exec.Command(installerExe, installArgs...)
-
-	if err := cmd.Start(); err != nil {
-		return fmt.Errorf("unable to start ollama app %w", err)
-	}
-
-	if cmd.Process != nil {
-		err = cmd.Process.Release()
-		if err != nil {
-			slog.Error(fmt.Sprintf("failed to release server process: %s", err))
-		}
-	} else {
-		// TODO - some details about why it didn't start, or is this a pedantic error case?
-		return fmt.Errorf("installer process did not start")
-	}
-
-	// TODO should we linger for a moment and check to make sure it's actually running by checking the pid?
-
-	slog.Info("Installer started in background, exiting")
-
-	os.Exit(0)
-	// Not reached
-	return nil
-}
@@ -1,12 +0,0 @@
-package main
-
-// Compile with the following to get rid of the cmd pop up on windows
-// go build -ldflags="-H windowsgui" .
-
-import (
-	"github.com/ollama/ollama/app/lifecycle"
-)
-
-func main() {
-	lifecycle.Run()
-}
@@ -1,156 +0,0 @@
-; Inno Setup Installer for Ollama
-;
-; To build the installer use the build script invoked from the top of the source tree
-; 
-; powershell -ExecutionPolicy Bypass -File .\scripts\build_windows.ps
-
-
-#define MyAppName "Ollama"
-#if GetEnv("PKG_VERSION") != ""
-  #define MyAppVersion GetEnv("PKG_VERSION")
-#else
-  #define MyAppVersion "0.0.0"
-#endif
-#define MyAppPublisher "Ollama"
-#define MyAppURL "https://ollama.com/"
-#define MyAppExeName "ollama app.exe"
-#define MyIcon ".\assets\app.ico"
-
-[Setup]
-; NOTE: The value of AppId uniquely identifies this application. Do not use the same AppId value in installers for other applications.
-; (To generate a new GUID, click Tools | Generate GUID inside the IDE.)
-AppId={{44E83376-CE68-45EB-8FC1-393500EB558C}
-AppName={#MyAppName}
-AppVersion={#MyAppVersion}
-VersionInfoVersion={#MyAppVersion}
-;AppVerName={#MyAppName} {#MyAppVersion}
-AppPublisher={#MyAppPublisher}
-AppPublisherURL={#MyAppURL}
-AppSupportURL={#MyAppURL}
-AppUpdatesURL={#MyAppURL}
-ArchitecturesAllowed=x64 arm64
-ArchitecturesInstallIn64BitMode=x64 arm64
-DefaultDirName={localappdata}\Programs\{#MyAppName}
-DefaultGroupName={#MyAppName}
-DisableProgramGroupPage=yes
-PrivilegesRequired=lowest
-OutputBaseFilename="OllamaSetup"
-SetupIconFile={#MyIcon}
-UninstallDisplayIcon={uninstallexe}
-Compression=lzma2
-SolidCompression=no
-WizardStyle=modern
-ChangesEnvironment=yes
-OutputDir=..\dist\
-
-; Disable logging once everything's battle tested
-; Filename will be %TEMP%\Setup Log*.txt
-SetupLogging=yes
-CloseApplications=yes
-RestartApplications=no
-
-; https://jrsoftware.org/ishelp/index.php?topic=setup_wizardimagefile
-WizardSmallImageFile=.\assets\setup.bmp
-
-; TODO verifty actual min windows version...
-; OG Win 10
-MinVersion=10.0.10240
-
-; First release that supports WinRT UI Composition for win32 apps
-; MinVersion=10.0.17134
-; First release with XAML Islands - possible UI path forward
-; MinVersion=10.0.18362
-
-; quiet...
-DisableDirPage=yes
-DisableFinishedPage=yes
-DisableReadyMemo=yes
-DisableReadyPage=yes
-DisableStartupPrompt=yes
-DisableWelcomePage=yes
-
-; TODO - percentage can't be set less than 100, so how to make it shorter?
-; WizardSizePercent=100,80
-
-#if GetEnv("KEY_CONTAINER")
-SignTool=MySignTool
-SignedUninstaller=yes
-#endif
-
-SetupMutex=OllamaSetupMutex
-
-[Languages]
-Name: "english"; MessagesFile: "compiler:Default.isl"
-
-[LangOptions]
-DialogFontSize=12
-
-[Files]
-Source: ".\app.exe"; DestDir: "{app}"; DestName: "{#MyAppExeName}" ; Flags: ignoreversion 64bit
-Source: "..\ollama.exe"; DestDir: "{app}"; Flags: ignoreversion 64bit
-Source: "..\dist\windows-{#ARCH}\*.dll"; DestDir: "{app}"; Flags: ignoreversion 64bit
-Source: "..\dist\windows-{#ARCH}\ollama_runners\*"; DestDir: "{app}\ollama_runners"; Flags: ignoreversion 64bit recursesubdirs
-Source: "..\dist\ollama_welcome.ps1"; DestDir: "{app}"; Flags: ignoreversion
-Source: ".\assets\app.ico"; DestDir: "{app}"; Flags: ignoreversion
-#if DirExists("..\dist\windows-amd64\rocm")
-  Source: "..\dist\windows-amd64\rocm\*"; DestDir: "{app}\rocm\"; Flags: ignoreversion recursesubdirs
-#endif
-
-
-[Icons]
-Name: "{group}\{#MyAppName}"; Filename: "{app}\{#MyAppExeName}"; IconFilename: "{app}\app.ico"
-Name: "{userstartup}\{#MyAppName}"; Filename: "{app}\{#MyAppExeName}"; IconFilename: "{app}\app.ico"
-Name: "{userprograms}\{#MyAppName}"; Filename: "{app}\{#MyAppExeName}"; IconFilename: "{app}\app.ico"
-
-[Run]
-Filename: "{cmd}"; Parameters: "/C set PATH={app};%PATH% & ""{app}\{#MyAppExeName}"""; Flags: postinstall nowait runhidden
-
-[UninstallRun]
-; Filename: "{cmd}"; Parameters: "/C ""taskkill /im ''{#MyAppExeName}'' /f /t"; Flags: runhidden
-; Filename: "{cmd}"; Parameters: "/C ""taskkill /im ollama.exe /f /t"; Flags: runhidden
-Filename: "taskkill"; Parameters: "/im ""{#MyAppExeName}"" /f /t"; Flags: runhidden
-Filename: "taskkill"; Parameters: "/im ""ollama.exe"" /f /t"; Flags: runhidden
-; HACK!  need to give the server and app enough time to exit
-; TODO - convert this to a Pascal code script so it waits until they're no longer running, then completes
-Filename: "{cmd}"; Parameters: "/c timeout 5"; Flags: runhidden
-
-[UninstallDelete]
-Type: filesandordirs; Name: "{%TEMP}\ollama*"
-Type: filesandordirs; Name: "{%LOCALAPPDATA}\Ollama"
-Type: filesandordirs; Name: "{%LOCALAPPDATA}\Programs\Ollama"
-Type: filesandordirs; Name: "{%USERPROFILE}\.ollama\models"
-Type: filesandordirs; Name: "{%USERPROFILE}\.ollama\history"
-; NOTE: if the user has a custom OLLAMA_MODELS it will be preserved
-
-[Messages]
-WizardReady=Ollama Windows Preview
-ReadyLabel1=%nLet's get you up and running with your own large language models.
-SetupAppRunningError=Another Ollama installer is running.%n%nPlease cancel or finish the other installer, then click OK to continue with this install, or Cancel to exit.
-
-
-;FinishedHeadingLabel=Run your first model
-;FinishedLabel=%nRun this command in a PowerShell or cmd terminal.%n%n%n    ollama run llama3
-;ClickFinish=%n
-
-[Registry]
-Root: HKCU; Subkey: "Environment"; \
-    ValueType: expandsz; ValueName: "Path"; ValueData: "{olddata};{app}"; \
-    Check: NeedsAddPath('{app}')
-
-[Code]
-
-function NeedsAddPath(Param: string): boolean;
-var
-  OrigPath: string;
-begin
-  if not RegQueryStringValue(HKEY_CURRENT_USER,
-    'Environment',
-    'Path', OrigPath)
-  then begin
-    Result := True;
-    exit;
-  end;
-  { look for the path with leading and trailing semicolon }
-  { Pos() returns 0 if not found }
-  Result := Pos(';' + ExpandConstant(Param) + ';', ';' + OrigPath + ';') = 0;
-end;
@@ -1,29 +0,0 @@
-#include <winver.h>
-
-VS_VERSION_INFO VERSIONINFO
- FILEFLAGSMASK 0x3fL
-#ifdef _DEBUG
- FILEFLAGS 0x1L
-#else
- FILEFLAGS 0x0L
-#endif
- FILEOS 0x40004L
- FILETYPE 0x1L
- FILESUBTYPE 0x0L
-BEGIN
-    BLOCK "StringFileInfo"
-    BEGIN
-        BLOCK "040904b0"
-        BEGIN
-            VALUE "FileDescription", "Ollama"
-            VALUE "InternalName", "Ollama"
-            VALUE "OriginalFilename", "ollama app.exe"
-            VALUE "ProductName", "Ollama"
-        END
-    END
-
-    BLOCK "VarFileInfo"
-    BEGIN
-        VALUE "Translation", 0x409, 1200
-    END
-END
@@ -1,8 +0,0 @@
-# TODO - consider ANSI colors and maybe ASCII art...
-write-host ""
-write-host "Welcome to Ollama!"
-write-host ""
-write-host "Run your first model:"
-write-host ""
-write-host "`tollama run llama2"
-write-host ""
@@ -6,10 +6,10 @@
  "main": ".webpack/main",
  "scripts": {
    "start": "electron-forge start",
-    "package": "electron-forge package --arch universal",
-    "package:sign": "SIGN=1 electron-forge package --arch universal",
-    "make": "electron-forge make --arch universal",
-    "make:sign": "SIGN=1 electron-forge make --arch universal",
+    "package": "electron-forge package",
+    "package:sign": "SIGN=1 electron-forge package",
+    "make": "electron-forge make",
+    "make:sign": "SIGN=1 electron-forge make",
    "publish": "SIGN=1 electron-forge publish",
    "lint": "eslint --ext .ts,.tsx .",
    "format": "prettier --check . --ignore-path .gitignore",
@@ -32,7 +32,6 @@
    "@electron-forge/plugin-auto-unpack-natives": "^6.2.1",
    "@electron-forge/plugin-webpack": "^6.2.1",
    "@electron-forge/publisher-github": "^6.2.1",
-    "@electron/universal": "^1.4.1",
    "@svgr/webpack": "^8.0.1",
    "@types/chmodr": "^1.0.0",
    "@types/node": "^20.4.0",
@@ -46,7 +45,7 @@
    "chmodr": "^1.2.0",
    "copy-webpack-plugin": "^11.0.0",
    "css-loader": "^6.8.1",
-    "electron": "25.9.2",
+    "electron": "25.2.0",
    "eslint": "^8.43.0",
    "eslint-plugin-import": "^2.27.5",
    "fork-ts-checker-webpack-plugin": "^7.3.0",
@@ -2,7 +2,7 @@ import { useState } from 'react'
 import copy from 'copy-to-clipboard'
 import { CheckIcon, DocumentDuplicateIcon } from '@heroicons/react/24/outline'
 import Store from 'electron-store'
-import { getCurrentWindow, app } from '@electron/remote'
+import { getCurrentWindow } from '@electron/remote'

 import { install } from './install'
 import OllamaIcon from './ollama.svg'
@@ -19,7 +19,7 @@ export default function () {
  const [step, setStep] = useState<Step>(Step.WELCOME)
  const [commandCopied, setCommandCopied] = useState<boolean>(false)

-  const command = 'ollama run llama3'
+  const command = 'ollama run llama2'

  return (
    <div className='drag'>
@@ -51,15 +51,10 @@ export default function () {
              <div className='mx-auto'>
                <button
                  onClick={async () => {
-                    try {
-                      await install()
-                      setStep(Step.FINISH)
-                    } catch (e) {
-                      console.error('could not install: ', e)
-                    } finally {
-                      getCurrentWindow().show()
-                      getCurrentWindow().focus()
-                    }
+                    await install()
+                    getCurrentWindow().show()
+                    getCurrentWindow().focus()
+                    setStep(Step.FINISH)
                  }}
                  className='no-drag rounded-dm mx-auto w-[60%] rounded-md bg-black px-4 py-2 text-sm text-white hover:brightness-110'
                >
@@ -1,21 +1,17 @@
-import { spawn, ChildProcess } from 'child_process'
-import { app, autoUpdater, dialog, Tray, Menu, BrowserWindow, MenuItemConstructorOptions, nativeTheme } from 'electron'
+import { spawn } from 'child_process'
+import { app, autoUpdater, dialog, Tray, Menu, BrowserWindow, nativeTheme } from 'electron'
 import Store from 'electron-store'
 import winston from 'winston'
 import 'winston-daily-rotate-file'
 import * as path from 'path'

-import { v4 as uuidv4 } from 'uuid'
+import { analytics, id } from './telemetry'
 import { installed } from './install'

 require('@electron/remote/main').initialize()

-if (require('electron-squirrel-startup')) {
-  app.quit()
-}
-
 const store = new Store()
-
+let tray: Tray | null = null
 let welcomeWindow: BrowserWindow | null = null

 declare const MAIN_WINDOW_WEBPACK_ENTRY: string
@@ -32,30 +28,10 @@ const logger = winston.createLogger({
  format: winston.format.printf(info => info.message),
 })

-app.on('ready', () => {
-  const gotTheLock = app.requestSingleInstanceLock()
-  if (!gotTheLock) {
-    app.exit(0)
-    return
-  }
-
-  app.on('second-instance', () => {
-    if (app.hasSingleInstanceLock()) {
-      app.releaseSingleInstanceLock()
-    }
-
-    if (proc) {
-      proc.off('exit', restart)
-      proc.kill()
-    }
-
-    app.exit(0)
-  })
-
-  app.focus({ steal: true })
-
-  init()
-})
+const SingleInstanceLock = app.requestSingleInstanceLock()
+if (!SingleInstanceLock) {
+  app.quit()
+}

 function firstRunWindow() {
  // Create the browser window.
@@ -71,74 +47,65 @@ function firstRunWindow() {
      nodeIntegration: true,
      contextIsolation: false,
    },
+    alwaysOnTop: true,
  })

  require('@electron/remote/main').enable(welcomeWindow.webContents)

+  // and load the index.html of the app.
  welcomeWindow.loadURL(MAIN_WINDOW_WEBPACK_ENTRY)
+
  welcomeWindow.on('ready-to-show', () => welcomeWindow.show())
-  welcomeWindow.on('closed', () => {
-    if (process.platform === 'darwin') {
-      app.dock.hide()
+
+  // for debugging
+  // welcomeWindow.webContents.openDevTools()
+
+  if (process.platform === 'darwin') {
+    app.dock.hide()
+  }
+}
+
+function createSystemtray() {
+  let iconPath = nativeTheme.shouldUseDarkColors
+    ? path.join(__dirname, '..', '..', 'assets', 'ollama_icon_16x16Template.png')
+    : path.join(__dirname, '..', '..', 'assets', 'ollama_outline_icon_16x16Template.png')
+
+  if (app.isPackaged) {
+    iconPath = nativeTheme.shouldUseDarkColors
+      ? path.join(process.resourcesPath, 'ollama_icon_16x16Template.png')
+      : path.join(process.resourcesPath, 'ollama_outline_icon_16x16Template.png')
+  }
+
+  tray = new Tray(iconPath)
+
+  nativeTheme.on('updated', function theThemeHasChanged() {
+    if (nativeTheme.shouldUseDarkColors) {
+      app.isPackaged
+        ? tray.setImage(path.join(process.resourcesPath, 'ollama_icon_16x16Template.png'))
+        : tray.setImage(path.join(__dirname, '..', '..', 'assets', 'ollama_icon_16x16Template.png'))
+    } else {
+      app.isPackaged
+        ? tray.setImage(path.join(process.resourcesPath, 'ollama_outline_icon_16x16Template.png'))
+        : tray.setImage(path.join(__dirname, '..', '..', 'assets', 'ollama_outline_icon_16x16Template.png'))
    }
  })
+
+  const contextMenu = Menu.buildFromTemplate([{ role: 'quit', label: 'Quit Ollama', accelerator: 'Command+Q' }])
+
+  tray.setContextMenu(contextMenu)
+  tray.setToolTip('Ollama')
 }

-let tray: Tray | null = null
-let updateAvailable = false
-const assetPath = app.isPackaged ? process.resourcesPath : path.join(__dirname, '..', '..', 'assets')
-
-function trayIconPath() {
-  return nativeTheme.shouldUseDarkColors
-    ? updateAvailable
-      ? path.join(assetPath, 'iconDarkUpdateTemplate.png')
-      : path.join(assetPath, 'iconDarkTemplate.png')
-    : updateAvailable
-    ? path.join(assetPath, 'iconUpdateTemplate.png')
-    : path.join(assetPath, 'iconTemplate.png')
+if (require('electron-squirrel-startup')) {
+  app.quit()
 }

-function updateTrayIcon() {
-  if (tray) {
-    tray.setImage(trayIconPath())
-  }
-}
-
-function updateTray() {
-  const updateItems: MenuItemConstructorOptions[] = [
-    { label: 'An update is available', enabled: false },
-    {
-      label: 'Restart to update',
-      click: () => autoUpdater.quitAndInstall(),
-    },
-    { type: 'separator' },
-  ]
-
-  const menu = Menu.buildFromTemplate([
-    ...(updateAvailable ? updateItems : []),
-    { role: 'quit', label: 'Quit Ollama', accelerator: 'Command+Q' },
-  ])
-
-  if (!tray) {
-    tray = new Tray(trayIconPath())
-  }
-
-  tray.setToolTip(updateAvailable ? 'An update is available' : 'Ollama')
-  tray.setContextMenu(menu)
-  tray.setImage(trayIconPath())
-
-  nativeTheme.off('updated', updateTrayIcon)
-  nativeTheme.on('updated', updateTrayIcon)
-}
-
-let proc: ChildProcess = null
-
 function server() {
  const binary = app.isPackaged
    ? path.join(process.resourcesPath, 'ollama')
    : path.resolve(process.cwd(), '..', 'ollama')

-  proc = spawn(binary, ['serve'])
+  const proc = spawn(binary, ['serve'])

  proc.stdout.on('data', data => {
    logger.info(data.toString().trim())
@@ -148,75 +115,23 @@ function server() {
    logger.error(data.toString().trim())
  })

+  function restart() {
+    setTimeout(server, 3000)
+  }
+
  proc.on('exit', restart)
-}

-function restart() {
-  setTimeout(server, 1000)
-}
-
-app.on('before-quit', () => {
-  if (proc) {
+  app.on('before-quit', () => {
    proc.off('exit', restart)
-    proc.kill('SIGINT') // send SIGINT signal to the server, which also stops any loaded llms
-  }
-})
-
-const updateURL = `https://ollama.ai/api/update?os=${process.platform}&arch=${
-  process.arch
-}&version=${app.getVersion()}&id=${id()}`
-
-let latest = ''
-async function isNewReleaseAvailable() {
-  try {
-    const response = await fetch(updateURL)
-
-    if (!response.ok) {
-      return false
-    }
-
-    if (response.status === 204) {
-      return false
-    }
-
-    const data = await response.json()
-
-    const url = data?.url
-    if (!url) {
-      return false
-    }
-
-    if (latest === url) {
-      return false
-    }
-
-    latest = url
-
-    return true
-  } catch (error) {
-    logger.error(`update check failed - ${error}`)
-    return false
-  }
+    proc.kill()
+  })
 }

-async function checkUpdate() {
-  const available = await isNewReleaseAvailable()
-  if (available) {
-    logger.info('checking for update')
-    autoUpdater.checkForUpdates()
-  }
+if (process.platform === 'darwin') {
+  app.dock.hide()
 }

-function init() {
-  if (app.isPackaged) {
-    checkUpdate()
-    setInterval(() => {
-      checkUpdate()
-    }, 60 * 60 * 1000)
-  }
-
-  updateTray()
-
+app.on('ready', () => {
  if (process.platform === 'darwin') {
    if (app.isPackaged) {
      if (!app.isInApplicationsFolder()) {
@@ -252,13 +167,10 @@ function init() {
    }
  }

+  createSystemtray()
  server()

  if (store.get('first-time-run') && installed()) {
-    if (process.platform === 'darwin') {
-      app.dock.hide()
-    }
-
    app.setLoginItemSettings({ openAtLogin: app.getLoginItemSettings().openAtLogin })
    return
  }
@@ -266,7 +178,7 @@ function init() {
  // This is the first run or the CLI is no longer installed
  app.setLoginItemSettings({ openAtLogin: true })
  firstRunWindow()
-}
+})

 // Quit when all windows are closed, except on macOS. There, it's common
 // for applications and their menu bar to stay active until the user quits
@@ -277,26 +189,45 @@ app.on('window-all-closed', () => {
  }
 })

-function id(): string {
-  const id = store.get('id') as string
+// In this file you can include the rest of your app's specific main process
+// code. You can also put them in separate files and import them here.
+autoUpdater.setFeedURL({
+  url: `https://ollama.ai/api/update?os=${process.platform}&arch=${process.arch}&version=${app.getVersion()}`,
+})

-  if (id) {
-    return id
-  }
-
-  const uuid = uuidv4()
-  store.set('id', uuid)
-  return uuid
+async function heartbeat() {
+  analytics.track({
+    anonymousId: id(),
+    event: 'heartbeat',
+    properties: {
+      version: app.getVersion(),
+    },
+  })
 }

-autoUpdater.setFeedURL({ url: updateURL })
+if (app.isPackaged) {
+  heartbeat()
+  autoUpdater.checkForUpdates()
+  setInterval(() => {
+    heartbeat()
+    autoUpdater.checkForUpdates()
+  }, 60 * 60 * 1000)
+}

 autoUpdater.on('error', e => {
  logger.error(`update check failed - ${e.message}`)
-  console.error(`update check failed - ${e.message}`)
 })

-autoUpdater.on('update-downloaded', () => {
-  updateAvailable = true
-  updateTray()
+autoUpdater.on('update-downloaded', (event, releaseNotes, releaseName) => {
+  dialog
+    .showMessageBox({
+      type: 'info',
+      buttons: ['Restart Now', 'Later'],
+      title: 'New update available',
+      message: process.platform === 'win32' ? releaseNotes : releaseName,
+      detail: 'A new version of Ollama is available. Restart to apply the update.',
+    })
+    .then(returnValue => {
+      if (returnValue.response === 0) autoUpdater.quitAndInstall()
+    })
 })
@@ -15,7 +15,12 @@ export function installed() {
 export async function install() {
  const command = `do shell script "mkdir -p ${path.dirname(
    symlinkPath
-  )} && ln -F -s \\"${ollama}\\" \\"${symlinkPath}\\"" with administrator privileges`
+  )} && ln -F -s ${ollama} ${symlinkPath}" with administrator privileges`

-  await exec(`osascript -e '${command}'`)
+  try {
+    await exec(`osascript -e '${command}'`)
+  } catch (error) {
+    console.error(`cli: failed to install cli: ${error.message}`)
+    return
+  }
 }
@@ -0,0 +1,19 @@
+import { Analytics } from '@segment/analytics-node'
+import { v4 as uuidv4 } from 'uuid'
+import Store from 'electron-store'
+
+const store = new Store()
+
+export const analytics = new Analytics({ writeKey: process.env.TELEMETRY_WRITE_KEY || '<empty>' })
+
+export function id(): string {
+  const id = store.get('id') as string
+
+  if (id) {
+    return id
+  }
+
+  const uuid = uuidv4()
+  store.set('id', uuid)
+  return uuid
+}
@@ -1,98 +0,0 @@
-package store
-
-import (
-	"encoding/json"
-	"errors"
-	"fmt"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"sync"
-
-	"github.com/google/uuid"
-)
-
-type Store struct {
-	ID           string `json:"id"`
-	FirstTimeRun bool   `json:"first-time-run"`
-}
-
-var (
-	lock  sync.Mutex
-	store Store
-)
-
-func GetID() string {
-	lock.Lock()
-	defer lock.Unlock()
-	if store.ID == "" {
-		initStore()
-	}
-	return store.ID
-
-}
-
-func GetFirstTimeRun() bool {
-	lock.Lock()
-	defer lock.Unlock()
-	if store.ID == "" {
-		initStore()
-	}
-	return store.FirstTimeRun
-}
-
-func SetFirstTimeRun(val bool) {
-	lock.Lock()
-	defer lock.Unlock()
-	if store.FirstTimeRun == val {
-		return
-	}
-	store.FirstTimeRun = val
-	writeStore(getStorePath())
-}
-
-// lock must be held
-func initStore() {
-	storeFile, err := os.Open(getStorePath())
-	if err == nil {
-		defer storeFile.Close()
-		err = json.NewDecoder(storeFile).Decode(&store)
-		if err == nil {
-			slog.Debug(fmt.Sprintf("loaded existing store %s - ID: %s", getStorePath(), store.ID))
-			return
-		}
-	} else if !errors.Is(err, os.ErrNotExist) {
-		slog.Debug(fmt.Sprintf("unexpected error searching for store: %s", err))
-	}
-	slog.Debug("initializing new store")
-	store.ID = uuid.New().String()
-	writeStore(getStorePath())
-}
-
-func writeStore(storeFilename string) {
-	ollamaDir := filepath.Dir(storeFilename)
-	_, err := os.Stat(ollamaDir)
-	if errors.Is(err, os.ErrNotExist) {
-		if err := os.MkdirAll(ollamaDir, 0o755); err != nil {
-			slog.Error(fmt.Sprintf("create ollama dir %s: %v", ollamaDir, err))
-			return
-		}
-	}
-	payload, err := json.Marshal(store)
-	if err != nil {
-		slog.Error(fmt.Sprintf("failed to marshal store: %s", err))
-		return
-	}
-	fp, err := os.OpenFile(storeFilename, os.O_WRONLY|os.O_CREATE|os.O_TRUNC, 0o755)
-	if err != nil {
-		slog.Error(fmt.Sprintf("write store payload %s: %v", storeFilename, err))
-		return
-	}
-	defer fp.Close()
-	if n, err := fp.Write(payload); err != nil || n != len(payload) {
-		slog.Error(fmt.Sprintf("write store payload %s: %d vs %d -- %v", storeFilename, n, len(payload), err))
-		return
-	}
-	slog.Debug("Store contents: " + string(payload))
-	slog.Info(fmt.Sprintf("wrote store: %s", storeFilename))
-}
@@ -1,13 +0,0 @@
-package store
-
-import (
-	"os"
-	"path/filepath"
-)
-
-func getStorePath() string {
-	// TODO - system wide location?
-
-	home := os.Getenv("HOME")
-	return filepath.Join(home, "Library", "Application Support", "Ollama", "config.json")
-}
@@ -1,16 +0,0 @@
-package store
-
-import (
-	"os"
-	"path/filepath"
-)
-
-func getStorePath() string {
-	if os.Geteuid() == 0 {
-		// TODO where should we store this on linux for system-wide operation?
-		return "/etc/ollama/config.json"
-	}
-
-	home := os.Getenv("HOME")
-	return filepath.Join(home, ".ollama", "config.json")
-}
@@ -1,11 +0,0 @@
-package store
-
-import (
-	"os"
-	"path/filepath"
-)
-
-func getStorePath() string {
-	localAppData := os.Getenv("LOCALAPPDATA")
-	return filepath.Join(localAppData, "Ollama", "config.json")
-}
@@ -1,24 +0,0 @@
-package commontray
-
-var (
-	Title   = "Ollama"
-	ToolTip = "Ollama"
-
-	UpdateIconName = "tray_upgrade"
-	IconName       = "tray"
-)
-
-type Callbacks struct {
-	Quit       chan struct{}
-	Update     chan struct{}
-	DoFirstUse chan struct{}
-	ShowLogs   chan struct{}
-}
-
-type OllamaTray interface {
-	GetCallbacks() Callbacks
-	Run()
-	UpdateAvailable(ver string) error
-	DisplayFirstUseNotification() error
-	Quit()
-}
@@ -1,28 +0,0 @@
-package tray
-
-import (
-	"fmt"
-	"runtime"
-
-	"github.com/ollama/ollama/app/assets"
-	"github.com/ollama/ollama/app/tray/commontray"
-)
-
-func NewTray() (commontray.OllamaTray, error) {
-	extension := ".png"
-	if runtime.GOOS == "windows" {
-		extension = ".ico"
-	}
-	iconName := commontray.UpdateIconName + extension
-	updateIcon, err := assets.GetIcon(iconName)
-	if err != nil {
-		return nil, fmt.Errorf("failed to load icon %s: %w", iconName, err)
-	}
-	iconName = commontray.IconName + extension
-	icon, err := assets.GetIcon(iconName)
-	if err != nil {
-		return nil, fmt.Errorf("failed to load icon %s: %w", iconName, err)
-	}
-
-	return InitPlatformTray(icon, updateIcon)
-}
@@ -1,13 +0,0 @@
-//go:build !windows
-
-package tray
-
-import (
-	"fmt"
-
-	"github.com/ollama/ollama/app/tray/commontray"
-)
-
-func InitPlatformTray(icon, updateIcon []byte) (commontray.OllamaTray, error) {
-	return nil, fmt.Errorf("NOT IMPLEMENTED YET")
-}
@@ -1,10 +0,0 @@
-package tray
-
-import (
-	"github.com/ollama/ollama/app/tray/commontray"
-	"github.com/ollama/ollama/app/tray/wintray"
-)
-
-func InitPlatformTray(icon, updateIcon []byte) (commontray.OllamaTray, error) {
-	return wintray.InitTray(icon, updateIcon)
-}
@@ -1,184 +0,0 @@
-//go:build windows
-
-package wintray
-
-import (
-	"fmt"
-	"log/slog"
-	"sync"
-	"unsafe"
-
-	"golang.org/x/sys/windows"
-)
-
-var (
-	quitOnce sync.Once
-)
-
-func (t *winTray) Run() {
-	nativeLoop()
-}
-
-func nativeLoop() {
-	// Main message pump.
-	slog.Debug("starting event handling loop")
-	m := &struct {
-		WindowHandle windows.Handle
-		Message      uint32
-		Wparam       uintptr
-		Lparam       uintptr
-		Time         uint32
-		Pt           point
-		LPrivate     uint32
-	}{}
-	for {
-		ret, _, err := pGetMessage.Call(uintptr(unsafe.Pointer(m)), 0, 0, 0)
-
-		// If the function retrieves a message other than WM_QUIT, the return value is nonzero.
-		// If the function retrieves the WM_QUIT message, the return value is zero.
-		// If there is an error, the return value is -1
-		// https://msdn.microsoft.com/en-us/library/windows/desktop/ms644936(v=vs.85).aspx
-		switch int32(ret) {
-		case -1:
-			slog.Error(fmt.Sprintf("get message failure: %v", err))
-			return
-		case 0:
-			return
-		default:
-			pTranslateMessage.Call(uintptr(unsafe.Pointer(m))) //nolint:errcheck
-			pDispatchMessage.Call(uintptr(unsafe.Pointer(m)))  //nolint:errcheck
-
-		}
-	}
-}
-
-// WindowProc callback function that processes messages sent to a window.
-// https://msdn.microsoft.com/en-us/library/windows/desktop/ms633573(v=vs.85).aspx
-func (t *winTray) wndProc(hWnd windows.Handle, message uint32, wParam, lParam uintptr) (lResult uintptr) {
-	const (
-		WM_RBUTTONUP   = 0x0205
-		WM_LBUTTONUP   = 0x0202
-		WM_COMMAND     = 0x0111
-		WM_ENDSESSION  = 0x0016
-		WM_CLOSE       = 0x0010
-		WM_DESTROY     = 0x0002
-		WM_MOUSEMOVE   = 0x0200
-		WM_LBUTTONDOWN = 0x0201
-	)
-	switch message {
-	case WM_COMMAND:
-		menuItemId := int32(wParam)
-		// https://docs.microsoft.com/en-us/windows/win32/menurc/wm-command#menus
-		switch menuItemId {
-		case quitMenuID:
-			select {
-			case t.callbacks.Quit <- struct{}{}:
-			// should not happen but in case not listening
-			default:
-				slog.Error("no listener on Quit")
-			}
-		case updateMenuID:
-			select {
-			case t.callbacks.Update <- struct{}{}:
-			// should not happen but in case not listening
-			default:
-				slog.Error("no listener on Update")
-			}
-		case diagLogsMenuID:
-			select {
-			case t.callbacks.ShowLogs <- struct{}{}:
-			// should not happen but in case not listening
-			default:
-				slog.Error("no listener on ShowLogs")
-			}
-		default:
-			slog.Debug(fmt.Sprintf("Unexpected menu item id: %d", menuItemId))
-		}
-	case WM_CLOSE:
-		boolRet, _, err := pDestroyWindow.Call(uintptr(t.window))
-		if boolRet == 0 {
-			slog.Error(fmt.Sprintf("failed to destroy window: %s", err))
-		}
-		err = t.wcex.unregister()
-		if err != nil {
-			slog.Error(fmt.Sprintf("failed to uregister windo %s", err))
-		}
-	case WM_DESTROY:
-		// same as WM_ENDSESSION, but throws 0 exit code after all
-		defer pPostQuitMessage.Call(uintptr(int32(0))) //nolint:errcheck
-		fallthrough
-	case WM_ENDSESSION:
-		t.muNID.Lock()
-		if t.nid != nil {
-			err := t.nid.delete()
-			if err != nil {
-				slog.Error(fmt.Sprintf("failed to delete nid: %s", err))
-			}
-		}
-		t.muNID.Unlock()
-	case t.wmSystrayMessage:
-		switch lParam {
-		case WM_MOUSEMOVE, WM_LBUTTONDOWN:
-			// Ignore these...
-		case WM_RBUTTONUP, WM_LBUTTONUP:
-			err := t.showMenu()
-			if err != nil {
-				slog.Error(fmt.Sprintf("failed to show menu: %s", err))
-			}
-		case 0x405: // TODO - how is this magic value derived for the notification left click
-			if t.pendingUpdate {
-				select {
-				case t.callbacks.Update <- struct{}{}:
-				// should not happen but in case not listening
-				default:
-					slog.Error("no listener on Update")
-				}
-			} else {
-				select {
-				case t.callbacks.DoFirstUse <- struct{}{}:
-				// should not happen but in case not listening
-				default:
-					slog.Error("no listener on DoFirstUse")
-				}
-			}
-		case 0x404: // Middle click or close notification
-			// slog.Debug("doing nothing on close of first time notification")
-		default:
-			// 0x402 also seems common - what is it?
-			slog.Debug(fmt.Sprintf("unmanaged app message, lParm: 0x%x", lParam))
-		}
-	case t.wmTaskbarCreated: // on explorer.exe restarts
-		t.muNID.Lock()
-		err := t.nid.add()
-		if err != nil {
-			slog.Error(fmt.Sprintf("failed to refresh the taskbar on explorer restart: %s", err))
-		}
-		t.muNID.Unlock()
-	default:
-		// Calls the default window procedure to provide default processing for any window messages that an application does not process.
-		// https://msdn.microsoft.com/en-us/library/windows/desktop/ms633572(v=vs.85).aspx
-		lResult, _, _ = pDefWindowProc.Call(
-			uintptr(hWnd),
-			uintptr(message),
-			uintptr(wParam),
-			uintptr(lParam),
-		)
-	}
-	return
-}
-
-func (t *winTray) Quit() {
-	quitOnce.Do(quit)
-}
-
-func quit() {
-	boolRet, _, err := pPostMessage.Call(
-		uintptr(wt.window),
-		WM_CLOSE,
-		0,
-		0,
-	)
-	if boolRet == 0 {
-		slog.Error(fmt.Sprintf("failed to post close message on shutdown %s", err))
-	}
-}
@@ -1,71 +0,0 @@
-//go:build windows
-
-package wintray
-
-import (
-	"fmt"
-	"log/slog"
-	"unsafe"
-
-	"golang.org/x/sys/windows"
-)
-
-const (
-	updatAvailableMenuID = 1
-	updateMenuID         = updatAvailableMenuID + 1
-	separatorMenuID      = updateMenuID + 1
-	diagLogsMenuID       = separatorMenuID + 1
-	diagSeparatorMenuID  = diagLogsMenuID + 1
-	quitMenuID           = diagSeparatorMenuID + 1
-)
-
-func (t *winTray) initMenus() error {
-	if err := t.addOrUpdateMenuItem(diagLogsMenuID, 0, diagLogsMenuTitle, false); err != nil {
-		return fmt.Errorf("unable to create menu entries %w\n", err)
-	}
-	if err := t.addSeparatorMenuItem(diagSeparatorMenuID, 0); err != nil {
-		return fmt.Errorf("unable to create menu entries %w", err)
-	}
-	if err := t.addOrUpdateMenuItem(quitMenuID, 0, quitMenuTitle, false); err != nil {
-		return fmt.Errorf("unable to create menu entries %w\n", err)
-	}
-	return nil
-}
-
-func (t *winTray) UpdateAvailable(ver string) error {
-	if !t.updateNotified {
-		slog.Debug("updating menu and sending notification for new update")
-		if err := t.addOrUpdateMenuItem(updatAvailableMenuID, 0, updateAvailableMenuTitle, true); err != nil {
-			return fmt.Errorf("unable to create menu entries %w", err)
-		}
-		if err := t.addOrUpdateMenuItem(updateMenuID, 0, updateMenutTitle, false); err != nil {
-			return fmt.Errorf("unable to create menu entries %w", err)
-		}
-		if err := t.addSeparatorMenuItem(separatorMenuID, 0); err != nil {
-			return fmt.Errorf("unable to create menu entries %w", err)
-		}
-		iconFilePath, err := iconBytesToFilePath(wt.updateIcon)
-		if err != nil {
-			return fmt.Errorf("unable to write icon data to temp file: %w", err)
-		}
-		if err := wt.setIcon(iconFilePath); err != nil {
-			return fmt.Errorf("unable to set icon: %w", err)
-		}
-		t.updateNotified = true
-
-		t.pendingUpdate = true
-		// Now pop up the notification
-		t.muNID.Lock()
-		defer t.muNID.Unlock()
-		copy(t.nid.InfoTitle[:], windows.StringToUTF16(updateTitle))
-		copy(t.nid.Info[:], windows.StringToUTF16(fmt.Sprintf(updateMessage, ver)))
-		t.nid.Flags |= NIF_INFO
-		t.nid.Timeout = 10
-		t.nid.Size = uint32(unsafe.Sizeof(*wt.nid))
-		err = t.nid.modify()
-		if err != nil {
-			return err
-		}
-	}
-	return nil
-}
@@ -1,15 +0,0 @@
-//go:build windows
-
-package wintray
-
-const (
-	firstTimeTitle   = "Ollama is running"
-	firstTimeMessage = "Click here to get started"
-	updateTitle      = "Update available"
-	updateMessage    = "Ollama version %s is ready to install"
-
-	quitMenuTitle            = "Quit Ollama"
-	updateAvailableMenuTitle = "An update is available"
-	updateMenutTitle         = "Restart to update"
-	diagLogsMenuTitle        = "View logs"
-)
@@ -1,66 +0,0 @@
-//go:build windows
-
-package wintray
-
-import (
-	"unsafe"
-
-	"golang.org/x/sys/windows"
-)
-
-// Contains information that the system needs to display notifications in the notification area.
-// Used by Shell_NotifyIcon.
-// https://msdn.microsoft.com/en-us/library/windows/desktop/bb773352(v=vs.85).aspx
-// https://msdn.microsoft.com/en-us/library/windows/desktop/bb762159
-type notifyIconData struct {
-	Size                       uint32
-	Wnd                        windows.Handle
-	ID, Flags, CallbackMessage uint32
-	Icon                       windows.Handle
-	Tip                        [128]uint16
-	State, StateMask           uint32
-	Info                       [256]uint16
-	// Timeout, Version           uint32
-	Timeout uint32
-
-	InfoTitle   [64]uint16
-	InfoFlags   uint32
-	GuidItem    windows.GUID
-	BalloonIcon windows.Handle
-}
-
-func (nid *notifyIconData) add() error {
-	const NIM_ADD = 0x00000000
-	res, _, err := pShellNotifyIcon.Call(
-		uintptr(NIM_ADD),
-		uintptr(unsafe.Pointer(nid)),
-	)
-	if res == 0 {
-		return err
-	}
-	return nil
-}
-
-func (nid *notifyIconData) modify() error {
-	const NIM_MODIFY = 0x00000001
-	res, _, err := pShellNotifyIcon.Call(
-		uintptr(NIM_MODIFY),
-		uintptr(unsafe.Pointer(nid)),
-	)
-	if res == 0 {
-		return err
-	}
-	return nil
-}
-
-func (nid *notifyIconData) delete() error {
-	const NIM_DELETE = 0x00000002
-	res, _, err := pShellNotifyIcon.Call(
-		uintptr(NIM_DELETE),
-		uintptr(unsafe.Pointer(nid)),
-	)
-	if res == 0 {
-		return err
-	}
-	return nil
-}
@@ -1,485 +0,0 @@
-//go:build windows
-
-package wintray
-
-import (
-	"crypto/md5"
-	"encoding/hex"
-	"fmt"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"sort"
-	"sync"
-	"unsafe"
-
-	"github.com/ollama/ollama/app/tray/commontray"
-	"golang.org/x/sys/windows"
-)
-
-// Helpful sources: https://github.com/golang/exp/blob/master/shiny/driver/internal/win32
-
-// Contains information about loaded resources
-type winTray struct {
-	instance,
-	icon,
-	cursor,
-	window windows.Handle
-
-	loadedImages   map[string]windows.Handle
-	muLoadedImages sync.RWMutex
-
-	// menus keeps track of the submenus keyed by the menu item ID, plus 0
-	// which corresponds to the main popup menu.
-	menus    map[uint32]windows.Handle
-	muMenus  sync.RWMutex
-	menuOf   map[uint32]windows.Handle
-	muMenuOf sync.RWMutex
-	// menuItemIcons maintains the bitmap of each menu item (if applies). It's
-	// needed to show the icon correctly when showing a previously hidden menu
-	// item again.
-	// menuItemIcons   map[uint32]windows.Handle
-	// muMenuItemIcons sync.RWMutex
-	visibleItems   map[uint32][]uint32
-	muVisibleItems sync.RWMutex
-
-	nid   *notifyIconData
-	muNID sync.RWMutex
-	wcex  *wndClassEx
-
-	wmSystrayMessage,
-	wmTaskbarCreated uint32
-
-	pendingUpdate  bool
-	updateNotified bool // Only pop up the notification once - TODO consider daily nag?
-	// Callbacks
-	callbacks  commontray.Callbacks
-	normalIcon []byte
-	updateIcon []byte
-}
-
-var wt winTray
-
-func (t *winTray) GetCallbacks() commontray.Callbacks {
-	return t.callbacks
-}
-
-func InitTray(icon, updateIcon []byte) (*winTray, error) {
-	wt.callbacks.Quit = make(chan struct{})
-	wt.callbacks.Update = make(chan struct{})
-	wt.callbacks.ShowLogs = make(chan struct{})
-	wt.callbacks.DoFirstUse = make(chan struct{})
-	wt.normalIcon = icon
-	wt.updateIcon = updateIcon
-	if err := wt.initInstance(); err != nil {
-		return nil, fmt.Errorf("Unable to init instance: %w\n", err)
-	}
-
-	if err := wt.createMenu(); err != nil {
-		return nil, fmt.Errorf("Unable to create menu: %w\n", err)
-	}
-
-	iconFilePath, err := iconBytesToFilePath(wt.normalIcon)
-	if err != nil {
-		return nil, fmt.Errorf("Unable to write icon data to temp file: %w", err)
-	}
-	if err := wt.setIcon(iconFilePath); err != nil {
-		return nil, fmt.Errorf("Unable to set icon: %w", err)
-	}
-
-	return &wt, wt.initMenus()
-}
-
-func (t *winTray) initInstance() error {
-	const (
-		className  = "OllamaClass"
-		windowName = ""
-	)
-
-	t.wmSystrayMessage = WM_USER + 1
-	t.visibleItems = make(map[uint32][]uint32)
-	t.menus = make(map[uint32]windows.Handle)
-	t.menuOf = make(map[uint32]windows.Handle)
-
-	t.loadedImages = make(map[string]windows.Handle)
-
-	taskbarEventNamePtr, _ := windows.UTF16PtrFromString("TaskbarCreated")
-	// https://msdn.microsoft.com/en-us/library/windows/desktop/ms644947
-	res, _, err := pRegisterWindowMessage.Call(
-		uintptr(unsafe.Pointer(taskbarEventNamePtr)),
-	)
-	if res == 0 { // success 0xc000-0xfff
-		return fmt.Errorf("failed to register window: %w", err)
-	}
-	t.wmTaskbarCreated = uint32(res)
-
-	instanceHandle, _, err := pGetModuleHandle.Call(0)
-	if instanceHandle == 0 {
-		return err
-	}
-	t.instance = windows.Handle(instanceHandle)
-
-	// https://msdn.microsoft.com/en-us/library/windows/desktop/ms648072(v=vs.85).aspx
-	iconHandle, _, err := pLoadIcon.Call(0, uintptr(IDI_APPLICATION))
-	if iconHandle == 0 {
-		return err
-	}
-	t.icon = windows.Handle(iconHandle)
-
-	// https://msdn.microsoft.com/en-us/library/windows/desktop/ms648391(v=vs.85).aspx
-	cursorHandle, _, err := pLoadCursor.Call(0, uintptr(IDC_ARROW))
-	if cursorHandle == 0 {
-		return err
-	}
-	t.cursor = windows.Handle(cursorHandle)
-
-	classNamePtr, err := windows.UTF16PtrFromString(className)
-	if err != nil {
-		return err
-	}
-
-	windowNamePtr, err := windows.UTF16PtrFromString(windowName)
-	if err != nil {
-		return err
-	}
-
-	t.wcex = &wndClassEx{
-		Style:      CS_HREDRAW | CS_VREDRAW,
-		WndProc:    windows.NewCallback(t.wndProc),
-		Instance:   t.instance,
-		Icon:       t.icon,
-		Cursor:     t.cursor,
-		Background: windows.Handle(6), // (COLOR_WINDOW + 1)
-		ClassName:  classNamePtr,
-		IconSm:     t.icon,
-	}
-	if err := t.wcex.register(); err != nil {
-		return err
-	}
-
-	windowHandle, _, err := pCreateWindowEx.Call(
-		uintptr(0),
-		uintptr(unsafe.Pointer(classNamePtr)),
-		uintptr(unsafe.Pointer(windowNamePtr)),
-		uintptr(WS_OVERLAPPEDWINDOW),
-		uintptr(CW_USEDEFAULT),
-		uintptr(CW_USEDEFAULT),
-		uintptr(CW_USEDEFAULT),
-		uintptr(CW_USEDEFAULT),
-		uintptr(0),
-		uintptr(0),
-		uintptr(t.instance),
-		uintptr(0),
-	)
-	if windowHandle == 0 {
-		return err
-	}
-	t.window = windows.Handle(windowHandle)
-
-	pShowWindow.Call(uintptr(t.window), uintptr(SW_HIDE)) //nolint:errcheck
-
-	boolRet, _, err := pUpdateWindow.Call(uintptr(t.window))
-	if boolRet == 0 {
-		slog.Error(fmt.Sprintf("failed to update window: %s", err))
-	}
-
-	t.muNID.Lock()
-	defer t.muNID.Unlock()
-	t.nid = &notifyIconData{
-		Wnd:             windows.Handle(t.window),
-		ID:              100,
-		Flags:           NIF_MESSAGE,
-		CallbackMessage: t.wmSystrayMessage,
-	}
-	t.nid.Size = uint32(unsafe.Sizeof(*t.nid))
-
-	return t.nid.add()
-}
-
-func (t *winTray) createMenu() error {
-
-	menuHandle, _, err := pCreatePopupMenu.Call()
-	if menuHandle == 0 {
-		return err
-	}
-	t.menus[0] = windows.Handle(menuHandle)
-
-	// https://msdn.microsoft.com/en-us/library/windows/desktop/ms647575(v=vs.85).aspx
-	mi := struct {
-		Size, Mask, Style, Max uint32
-		Background             windows.Handle
-		ContextHelpID          uint32
-		MenuData               uintptr
-	}{
-		Mask: MIM_APPLYTOSUBMENUS,
-	}
-	mi.Size = uint32(unsafe.Sizeof(mi))
-
-	res, _, err := pSetMenuInfo.Call(
-		uintptr(t.menus[0]),
-		uintptr(unsafe.Pointer(&mi)),
-	)
-	if res == 0 {
-		return err
-	}
-	return nil
-}
-
-// Contains information about a menu item.
-// https://msdn.microsoft.com/en-us/library/windows/desktop/ms647578(v=vs.85).aspx
-type menuItemInfo struct {
-	Size, Mask, Type, State     uint32
-	ID                          uint32
-	SubMenu, Checked, Unchecked windows.Handle
-	ItemData                    uintptr
-	TypeData                    *uint16
-	Cch                         uint32
-	BMPItem                     windows.Handle
-}
-
-func (t *winTray) addOrUpdateMenuItem(menuItemId uint32, parentId uint32, title string, disabled bool) error {
-	titlePtr, err := windows.UTF16PtrFromString(title)
-	if err != nil {
-		return err
-	}
-
-	mi := menuItemInfo{
-		Mask:     MIIM_FTYPE | MIIM_STRING | MIIM_ID | MIIM_STATE,
-		Type:     MFT_STRING,
-		ID:       uint32(menuItemId),
-		TypeData: titlePtr,
-		Cch:      uint32(len(title)),
-	}
-	mi.Size = uint32(unsafe.Sizeof(mi))
-	if disabled {
-		mi.State |= MFS_DISABLED
-	}
-
-	var res uintptr
-	t.muMenus.RLock()
-	menu := t.menus[parentId]
-	t.muMenus.RUnlock()
-	if t.getVisibleItemIndex(parentId, menuItemId) != -1 {
-		// We set the menu item info based on the menuID
-		boolRet, _, err := pSetMenuItemInfo.Call(
-			uintptr(menu),
-			uintptr(menuItemId),
-			0,
-			uintptr(unsafe.Pointer(&mi)),
-		)
-		if boolRet == 0 {
-			return fmt.Errorf("failed to set menu item: %w", err)
-		}
-	}
-
-	if res == 0 {
-		// Menu item does not already exist, create it
-		t.muMenus.RLock()
-		submenu, exists := t.menus[menuItemId]
-		t.muMenus.RUnlock()
-		if exists {
-			mi.Mask |= MIIM_SUBMENU
-			mi.SubMenu = submenu
-		}
-		t.addToVisibleItems(parentId, menuItemId)
-		position := t.getVisibleItemIndex(parentId, menuItemId)
-		res, _, err = pInsertMenuItem.Call(
-			uintptr(menu),
-			uintptr(position),
-			1,
-			uintptr(unsafe.Pointer(&mi)),
-		)
-		if res == 0 {
-			t.delFromVisibleItems(parentId, menuItemId)
-			return err
-		}
-		t.muMenuOf.Lock()
-		t.menuOf[menuItemId] = menu
-		t.muMenuOf.Unlock()
-	}
-
-	return nil
-}
-
-func (t *winTray) addSeparatorMenuItem(menuItemId, parentId uint32) error {
-
-	mi := menuItemInfo{
-		Mask: MIIM_FTYPE | MIIM_ID | MIIM_STATE,
-		Type: MFT_SEPARATOR,
-		ID:   uint32(menuItemId),
-	}
-
-	mi.Size = uint32(unsafe.Sizeof(mi))
-
-	t.addToVisibleItems(parentId, menuItemId)
-	position := t.getVisibleItemIndex(parentId, menuItemId)
-	t.muMenus.RLock()
-	menu := uintptr(t.menus[parentId])
-	t.muMenus.RUnlock()
-	res, _, err := pInsertMenuItem.Call(
-		menu,
-		uintptr(position),
-		1,
-		uintptr(unsafe.Pointer(&mi)),
-	)
-	if res == 0 {
-		return err
-	}
-
-	return nil
-}
-
-// func (t *winTray) hideMenuItem(menuItemId, parentId uint32) error {
-// 	const ERROR_SUCCESS syscall.Errno = 0
-
-// 	t.muMenus.RLock()
-// 	menu := uintptr(t.menus[parentId])
-// 	t.muMenus.RUnlock()
-// 	res, _, err := pRemoveMenu.Call(
-// 		menu,
-// 		uintptr(menuItemId),
-// 		MF_BYCOMMAND,
-// 	)
-// 	if res == 0 && err.(syscall.Errno) != ERROR_SUCCESS {
-// 		return err
-// 	}
-// 	t.delFromVisibleItems(parentId, menuItemId)
-
-// 	return nil
-// }
-
-func (t *winTray) showMenu() error {
-	p := point{}
-	boolRet, _, err := pGetCursorPos.Call(uintptr(unsafe.Pointer(&p)))
-	if boolRet == 0 {
-		return err
-	}
-	boolRet, _, err = pSetForegroundWindow.Call(uintptr(t.window))
-	if boolRet == 0 {
-		slog.Warn(fmt.Sprintf("failed to bring menu to foreground: %s", err))
-	}
-
-	boolRet, _, err = pTrackPopupMenu.Call(
-		uintptr(t.menus[0]),
-		TPM_BOTTOMALIGN|TPM_LEFTALIGN,
-		uintptr(p.X),
-		uintptr(p.Y),
-		0,
-		uintptr(t.window),
-		0,
-	)
-	if boolRet == 0 {
-		return err
-	}
-
-	return nil
-}
-
-func (t *winTray) delFromVisibleItems(parent, val uint32) {
-	t.muVisibleItems.Lock()
-	defer t.muVisibleItems.Unlock()
-	visibleItems := t.visibleItems[parent]
-	for i, itemval := range visibleItems {
-		if val == itemval {
-			t.visibleItems[parent] = append(visibleItems[:i], visibleItems[i+1:]...)
-			break
-		}
-	}
-}
-
-func (t *winTray) addToVisibleItems(parent, val uint32) {
-	t.muVisibleItems.Lock()
-	defer t.muVisibleItems.Unlock()
-	if visibleItems, exists := t.visibleItems[parent]; !exists {
-		t.visibleItems[parent] = []uint32{val}
-	} else {
-		newvisible := append(visibleItems, val)
-		sort.Slice(newvisible, func(i, j int) bool { return newvisible[i] < newvisible[j] })
-		t.visibleItems[parent] = newvisible
-	}
-}
-
-func (t *winTray) getVisibleItemIndex(parent, val uint32) int {
-	t.muVisibleItems.RLock()
-	defer t.muVisibleItems.RUnlock()
-	for i, itemval := range t.visibleItems[parent] {
-		if val == itemval {
-			return i
-		}
-	}
-	return -1
-}
-
-func iconBytesToFilePath(iconBytes []byte) (string, error) {
-	bh := md5.Sum(iconBytes)
-	dataHash := hex.EncodeToString(bh[:])
-	iconFilePath := filepath.Join(os.TempDir(), "ollama_temp_icon_"+dataHash)
-
-	if _, err := os.Stat(iconFilePath); os.IsNotExist(err) {
-		if err := os.WriteFile(iconFilePath, iconBytes, 0644); err != nil {
-			return "", err
-		}
-	}
-	return iconFilePath, nil
-}
-
-// Loads an image from file and shows it in tray.
-// Shell_NotifyIcon: https://msdn.microsoft.com/en-us/library/windows/desktop/bb762159(v=vs.85).aspx
-func (t *winTray) setIcon(src string) error {
-
-	h, err := t.loadIconFrom(src)
-	if err != nil {
-		return err
-	}
-
-	t.muNID.Lock()
-	defer t.muNID.Unlock()
-	t.nid.Icon = h
-	t.nid.Flags |= NIF_ICON
-	t.nid.Size = uint32(unsafe.Sizeof(*t.nid))
-
-	return t.nid.modify()
-}
-
-// Loads an image from file to be shown in tray or menu item.
-// LoadImage: https://msdn.microsoft.com/en-us/library/windows/desktop/ms648045(v=vs.85).aspx
-func (t *winTray) loadIconFrom(src string) (windows.Handle, error) {
-
-	// Save and reuse handles of loaded images
-	t.muLoadedImages.RLock()
-	h, ok := t.loadedImages[src]
-	t.muLoadedImages.RUnlock()
-	if !ok {
-		srcPtr, err := windows.UTF16PtrFromString(src)
-		if err != nil {
-			return 0, err
-		}
-		res, _, err := pLoadImage.Call(
-			0,
-			uintptr(unsafe.Pointer(srcPtr)),
-			IMAGE_ICON,
-			0,
-			0,
-			LR_LOADFROMFILE|LR_DEFAULTSIZE,
-		)
-		if res == 0 {
-			return 0, err
-		}
-		h = windows.Handle(res)
-		t.muLoadedImages.Lock()
-		t.loadedImages[src] = h
-		t.muLoadedImages.Unlock()
-	}
-	return h, nil
-}
-
-func (t *winTray) DisplayFirstUseNotification() error {
-	t.muNID.Lock()
-	defer t.muNID.Unlock()
-	copy(t.nid.InfoTitle[:], windows.StringToUTF16(firstTimeTitle))
-	copy(t.nid.Info[:], windows.StringToUTF16(firstTimeMessage))
-	t.nid.Flags |= NIF_INFO
-	t.nid.Size = uint32(unsafe.Sizeof(*wt.nid))
-
-	return t.nid.modify()
-}
@@ -1,89 +0,0 @@
-//go:build windows
-
-package wintray
-
-import (
-	"runtime"
-
-	"golang.org/x/sys/windows"
-)
-
-var (
-	k32 = windows.NewLazySystemDLL("Kernel32.dll")
-	u32 = windows.NewLazySystemDLL("User32.dll")
-	s32 = windows.NewLazySystemDLL("Shell32.dll")
-
-	pCreatePopupMenu       = u32.NewProc("CreatePopupMenu")
-	pCreateWindowEx        = u32.NewProc("CreateWindowExW")
-	pDefWindowProc         = u32.NewProc("DefWindowProcW")
-	pDestroyWindow         = u32.NewProc("DestroyWindow")
-	pDispatchMessage       = u32.NewProc("DispatchMessageW")
-	pGetCursorPos          = u32.NewProc("GetCursorPos")
-	pGetMessage            = u32.NewProc("GetMessageW")
-	pGetModuleHandle       = k32.NewProc("GetModuleHandleW")
-	pInsertMenuItem        = u32.NewProc("InsertMenuItemW")
-	pLoadCursor            = u32.NewProc("LoadCursorW")
-	pLoadIcon              = u32.NewProc("LoadIconW")
-	pLoadImage             = u32.NewProc("LoadImageW")
-	pPostMessage           = u32.NewProc("PostMessageW")
-	pPostQuitMessage       = u32.NewProc("PostQuitMessage")
-	pRegisterClass         = u32.NewProc("RegisterClassExW")
-	pRegisterWindowMessage = u32.NewProc("RegisterWindowMessageW")
-	pSetForegroundWindow   = u32.NewProc("SetForegroundWindow")
-	pSetMenuInfo           = u32.NewProc("SetMenuInfo")
-	pSetMenuItemInfo       = u32.NewProc("SetMenuItemInfoW")
-	pShellNotifyIcon       = s32.NewProc("Shell_NotifyIconW")
-	pShowWindow            = u32.NewProc("ShowWindow")
-	pTrackPopupMenu        = u32.NewProc("TrackPopupMenu")
-	pTranslateMessage      = u32.NewProc("TranslateMessage")
-	pUnregisterClass       = u32.NewProc("UnregisterClassW")
-	pUpdateWindow          = u32.NewProc("UpdateWindow")
-)
-
-const (
-	CS_HREDRAW          = 0x0002
-	CS_VREDRAW          = 0x0001
-	CW_USEDEFAULT       = 0x80000000
-	IDC_ARROW           = 32512 // Standard arrow
-	IDI_APPLICATION     = 32512
-	IMAGE_ICON          = 1          // Loads an icon
-	LR_DEFAULTSIZE      = 0x00000040 // Loads default-size icon for windows(SM_CXICON x SM_CYICON) if cx, cy are set to zero
-	LR_LOADFROMFILE     = 0x00000010 // Loads the stand-alone image from the file
-	MF_BYCOMMAND        = 0x00000000
-	MFS_DISABLED        = 0x00000003
-	MFT_SEPARATOR       = 0x00000800
-	MFT_STRING          = 0x00000000
-	MIIM_BITMAP         = 0x00000080
-	MIIM_FTYPE          = 0x00000100
-	MIIM_ID             = 0x00000002
-	MIIM_STATE          = 0x00000001
-	MIIM_STRING         = 0x00000040
-	MIIM_SUBMENU        = 0x00000004
-	MIM_APPLYTOSUBMENUS = 0x80000000
-	NIF_ICON            = 0x00000002
-	NIF_INFO            = 0x00000010
-	NIF_MESSAGE         = 0x00000001
-	SW_HIDE             = 0
-	TPM_BOTTOMALIGN     = 0x0020
-	TPM_LEFTALIGN       = 0x0000
-	WM_CLOSE            = 0x0010
-	WM_USER             = 0x0400
-	WS_CAPTION          = 0x00C00000
-	WS_MAXIMIZEBOX      = 0x00010000
-	WS_MINIMIZEBOX      = 0x00020000
-	WS_OVERLAPPED       = 0x00000000
-	WS_OVERLAPPEDWINDOW = WS_OVERLAPPED | WS_CAPTION | WS_SYSMENU | WS_THICKFRAME | WS_MINIMIZEBOX | WS_MAXIMIZEBOX
-	WS_SYSMENU          = 0x00080000
-	WS_THICKFRAME       = 0x00040000
-)
-
-// Not sure if this is actually needed on windows
-func init() {
-	runtime.LockOSThread()
-}
-
-// The POINT structure defines the x- and y- coordinates of a point.
-// https://msdn.microsoft.com/en-us/library/windows/desktop/dd162805(v=vs.85).aspx
-type point struct {
-	X, Y int32
-}
@@ -1,45 +0,0 @@
-//go:build windows
-
-package wintray
-
-import (
-	"unsafe"
-
-	"golang.org/x/sys/windows"
-)
-
-// Contains window class information.
-// It is used with the RegisterClassEx and GetClassInfoEx functions.
-// https://msdn.microsoft.com/en-us/library/ms633577.aspx
-type wndClassEx struct {
-	Size, Style                        uint32
-	WndProc                            uintptr
-	ClsExtra, WndExtra                 int32
-	Instance, Icon, Cursor, Background windows.Handle
-	MenuName, ClassName                *uint16
-	IconSm                             windows.Handle
-}
-
-// Registers a window class for subsequent use in calls to the CreateWindow or CreateWindowEx function.
-// https://msdn.microsoft.com/en-us/library/ms633587.aspx
-func (w *wndClassEx) register() error {
-	w.Size = uint32(unsafe.Sizeof(*w))
-	res, _, err := pRegisterClass.Call(uintptr(unsafe.Pointer(w)))
-	if res == 0 {
-		return err
-	}
-	return nil
-}
-
-// Unregisters a window class, freeing the memory required for the class.
-// https://msdn.microsoft.com/en-us/library/ms644899.aspx
-func (w *wndClassEx) unregister() error {
-	res, _, err := pUnregisterClass.Call(
-		uintptr(unsafe.Pointer(w.ClassName)),
-		uintptr(w.Instance),
-	)
-	if res == 0 {
-		return err
-	}
-	return nil
-}
@@ -1,91 +0,0 @@
-package auth
-
-import (
-	"bytes"
-	"context"
-	"crypto/rand"
-	"encoding/base64"
-	"fmt"
-	"io"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"strings"
-
-	"golang.org/x/crypto/ssh"
-)
-
-const defaultPrivateKey = "id_ed25519"
-
-func keyPath() (string, error) {
-	home, err := os.UserHomeDir()
-	if err != nil {
-		return "", err
-	}
-
-	return filepath.Join(home, ".ollama", defaultPrivateKey), nil
-}
-
-func GetPublicKey() (string, error) {
-	keyPath, err := keyPath()
-	if err != nil {
-		return "", err
-	}
-
-	privateKeyFile, err := os.ReadFile(keyPath)
-	if err != nil {
-		slog.Info(fmt.Sprintf("Failed to load private key: %v", err))
-		return "", err
-	}
-
-	privateKey, err := ssh.ParsePrivateKey(privateKeyFile)
-	if err != nil {
-		return "", err
-	}
-
-	publicKey := ssh.MarshalAuthorizedKey(privateKey.PublicKey())
-
-	return strings.TrimSpace(string(publicKey)), nil
-}
-
-func NewNonce(r io.Reader, length int) (string, error) {
-	nonce := make([]byte, length)
-	if _, err := io.ReadFull(r, nonce); err != nil {
-		return "", err
-	}
-
-	return base64.RawURLEncoding.EncodeToString(nonce), nil
-}
-
-func Sign(ctx context.Context, bts []byte) (string, error) {
-	keyPath, err := keyPath()
-	if err != nil {
-		return "", err
-	}
-
-	privateKeyFile, err := os.ReadFile(keyPath)
-	if err != nil {
-		slog.Info(fmt.Sprintf("Failed to load private key: %v", err))
-		return "", err
-	}
-
-	privateKey, err := ssh.ParsePrivateKey(privateKeyFile)
-	if err != nil {
-		return "", err
-	}
-
-	// get the pubkey, but remove the type
-	publicKey := ssh.MarshalAuthorizedKey(privateKey.PublicKey())
-	parts := bytes.Split(publicKey, []byte(" "))
-	if len(parts) < 2 {
-		return "", fmt.Errorf("malformed public key")
-	}
-
-	signedData, err := privateKey.Sign(rand.Reader, bts)
-	if err != nil {
-		return "", err
-	}
-
-	// signature is <pubkey>:<signature>
-	return fmt.Sprintf("%s:%s", bytes.TrimSpace(parts[1]), base64.StdEncoding.EncodeToString(signedData.Blob)), nil
-}
@@ -1,95 +0,0 @@
-package apitype
-
-import (
-	"cmp"
-	"encoding/json"
-	"log/slog"
-	"net/url"
-	"slices"
-)
-
-type Manifest struct {
-	Layers []*Layer `json:"layers"`
-}
-
-type CompletePart struct {
-	URL  string `json:"url"` // contains partNumber and uploadId from server
-	ETag string `json:"etag"`
-}
-
-func queryFromString(s string) url.Values {
-	u, err := url.Parse(s)
-	if err != nil {
-		return nil
-	}
-	return u.Query()
-}
-
-func (cp *CompletePart) Compare(o *CompletePart) int {
-	qa := queryFromString(cp.URL)
-	qb := queryFromString(o.URL)
-	return cmp.Or(
-		cmp.Compare(qa.Get("partNumber"), qb.Get("partNumber")),
-		cmp.Compare(qa.Get("uploadId"), qb.Get("uploadId")),
-		cmp.Compare(cp.ETag, o.ETag),
-	)
-}
-
-func SortCompleteParts(a []*CompletePart) {
-	slices.SortFunc(a, (*CompletePart).Compare)
-}
-
-type Layer struct {
-	Digest    string `json:"digest"`
-	MediaType string `json:"mediaType"`
-	Size      int64  `json:"size"`
-
-	// If present, URL is a remote location of the layer for fetching.
-	URL string `json:"url,omitempty"`
-}
-
-func (l *Layer) LogValue() slog.Value {
-	return slog.GroupValue(
-		slog.String("digest", l.Digest),
-		slog.String("mediaType", l.MediaType),
-		slog.Int64("size", l.Size),
-		slog.String("url", l.URL),
-	)
-}
-
-type PushRequest struct {
-	Name     string          `json:"ref"`
-	Manifest json.RawMessage `json:"manifest,omitempty"`
-
-	// Parts is a list of upload parts that the client upload in the previous
-	// push.
-	CompleteParts []*CompletePart `json:"part_uploads"`
-}
-
-type Need struct {
-	Digest string `json:"digest"`
-
-	Start int64 `json:"start"`
-	End   int64 `json:"end"`
-
-	// URL is the url to PUT the layer to.
-	//
-	// Clients must include it as the URL, along with the ETag in the
-	// response headers from the PUT request, in the next push request
-	// in the Uploaded field.
-	URL string `json:"url"`
-}
-
-type PushResponse struct {
-	// Needs is a list of digests that the client needs to push before
-	// repushing the manifest.
-	Needs []*Need `json:"requirements,omitempty"`
-}
-
-type PullResponse struct {
-	// Name is the name of the model being pulled.
-	Name string `json:"name"`
-
-	// Manifest is the manifest of the model being pulled.
-	Manifest *Manifest `json:"manifest"`
-}
@@ -1,421 +0,0 @@
-package registry
-
-import (
-	"cmp"
-	"context"
-	"encoding/json"
-	"encoding/xml"
-	"errors"
-	"fmt"
-	"io"
-	"iter"
-	"log/slog"
-	"net/http"
-	"net/url"
-	"os"
-	"sync"
-
-	"github.com/ollama/ollama/client/ollama"
-	"github.com/ollama/ollama/client/registry/apitype"
-	"github.com/ollama/ollama/types/model"
-	"golang.org/x/exp/constraints"
-	"golang.org/x/sync/errgroup"
-)
-
-// Errors
-var (
-	ErrLayerNotFound = errors.New("layer not found")
-)
-
-type Client struct {
-	BaseURL string
-
-	Logger *slog.Logger
-
-	// NameFill is a string that is used to fill in the missing parts of
-	// a name when it is not fully qualified. It is used to make a name
-	// fully qualified before pushing or pulling it. The default is
-	// "registry.ollama.ai/library/_:latest".
-	//
-	// Most users can ignore this field. It is intended for use by
-	// clients that need to push or pull names to registries other than
-	// registry.ollama.ai, and for testing.
-	NameFill string
-}
-
-func (c *Client) log() *slog.Logger {
-	return cmp.Or(c.Logger, slog.Default())
-}
-
-func (c *Client) oclient() *ollama.Client {
-	return &ollama.Client{
-		BaseURL: c.BaseURL,
-	}
-}
-
-type ReadAtSeekCloser interface {
-	io.ReaderAt
-	io.Seeker
-	io.Closer
-}
-
-type Cache interface {
-	// LayerFile returns the absolute file path to the layer file for
-	// the given model digest.
-	//
-	// If the digest is invalid, or the layer does not exist, the empty
-	// string is returned.
-	LayerFile(model.Digest) string
-
-	// OpenLayer opens the layer file for the given model digest and
-	// returns it, or an if any. The caller is responsible for closing
-	// the returned file.
-	OpenLayer(model.Digest) (ReadAtSeekCloser, error)
-
-	// PutLayerFile moves the layer file at fromPath to the cache for
-	// the given model digest. It is a hack intended to short circuit a
-	// file copy operation.
-	//
-	// The file returned is expected to exist for the lifetime of the
-	// cache.
-	//
-	// TODO(bmizerany): remove this; find a better way. Once we move
-	// this into a build package, we should be able to get rid of this.
-	PutLayerFile(_ model.Digest, fromPath string) error
-
-	// SetManifestData sets the provided manifest data for the given
-	// model name. If the manifest data is empty, the manifest is
-	// removed. If the manifeest exists, it is overwritten.
-	//
-	// It is an error to call SetManifestData with a name that is not
-	// complete.
-	SetManifestData(model.Name, []byte) error
-
-	// ManifestData returns the manifest data for the given model name.
-	//
-	// If the name incomplete, or the manifest does not exist, the empty
-	// string is returned.
-	ManifestData(name model.Name) []byte
-}
-
-// Pull pulls the manifest for name, and downloads any of its required
-// layers that are not already in the cache. It returns an error if any part
-// of the process fails, specifically:
-func (c *Client) Pull(ctx context.Context, cache Cache, name string) error {
-	mn := parseNameFill(name, c.NameFill)
-	if !mn.IsFullyQualified() {
-		return fmt.Errorf("ollama: pull: invalid name: %s", name)
-	}
-
-	log := c.log().With("name", name)
-
-	pr, err := ollama.Do[*apitype.PullResponse](ctx, c.oclient(), "GET", "/v1/pull/"+name, nil)
-	if err != nil {
-		return fmt.Errorf("ollama: pull: %w: %s", err, name)
-	}
-
-	if pr.Manifest == nil || len(pr.Manifest.Layers) == 0 {
-		return fmt.Errorf("ollama: pull: invalid manifest: %s: no layers found", name)
-	}
-
-	// download required layers we do not already have
-	for _, l := range pr.Manifest.Layers {
-		d, err := model.ParseDigest(l.Digest)
-		if err != nil {
-			return fmt.Errorf("ollama: reading manifest: %w: %s", err, l.Digest)
-		}
-		if cache.LayerFile(d) != "" {
-			continue
-		}
-		err = func() error {
-			log := log.With("digest", l.Digest, "mediaType", l.MediaType, "size", l.Size)
-			log.Debug("starting download")
-
-			// TODO(bmizerany): stop using temp which might not
-			// be on same device as cache.... instead let cache
-			// give us a place to store parts...
-			tmpFile, err := os.CreateTemp("", "ollama-download-")
-			if err != nil {
-				return err
-			}
-			defer func() {
-				tmpFile.Close()
-				os.Remove(tmpFile.Name()) // in case we fail before committing
-			}()
-
-			g, ctx := errgroup.WithContext(ctx)
-			g.SetLimit(8) // TODO(bmizerany): make this configurable
-
-			// TODO(bmizerany): make chunk size configurable
-			const chunkSize = 50 * 1024 * 1024 // 50MB
-			chunks(l.Size, chunkSize)(func(_ int, rng chunkRange[int64]) bool {
-				g.Go(func() (err error) {
-					defer func() {
-						if err == nil {
-							return
-						}
-						safeURL := redactAmzSignature(l.URL)
-						err = fmt.Errorf("%w: %s %s bytes=%s: %s", err, pr.Name, l.Digest, rng, safeURL)
-					}()
-
-					log.Debug("downloading", "range", rng)
-
-					// TODO(bmizerany): retry
-					// TODO(bmizerany): use real http client
-					// TODO(bmizerany): resumable
-					// TODO(bmizerany): multipart download
-					req, err := http.NewRequestWithContext(ctx, "GET", l.URL, nil)
-					if err != nil {
-						return err
-					}
-					req.Header.Set("Range", "bytes="+rng.String())
-
-					res, err := http.DefaultClient.Do(req)
-					if err != nil {
-						return err
-					}
-					defer res.Body.Close()
-					if res.StatusCode/100 != 2 {
-						log.Debug("unexpected non-2XX status code", "status", res.StatusCode)
-						return fmt.Errorf("unexpected status code fetching layer: %d", res.StatusCode)
-					}
-					if res.ContentLength != rng.Size() {
-						return fmt.Errorf("unexpected content length: %d", res.ContentLength)
-					}
-					w := io.NewOffsetWriter(tmpFile, rng.Start)
-					_, err = io.Copy(w, res.Body)
-					return err
-				})
-				return true
-			})
-			if err := g.Wait(); err != nil {
-				return err
-			}
-
-			tmpFile.Close() // release our hold on the file before moving it
-			return cache.PutLayerFile(d, tmpFile.Name())
-		}()
-		if err != nil {
-			return fmt.Errorf("ollama: pull: %w", err)
-		}
-	}
-
-	// do not store the presigned URLs in the cache
-	for i := range pr.Manifest.Layers {
-		pr.Manifest.Layers[i].URL = ""
-	}
-	data, err := json.Marshal(pr.Manifest)
-	if err != nil {
-		return err
-	}
-
-	// TODO(bmizerany): remove dep on model.Name
-	return cache.SetManifestData(mn, data)
-}
-
-type nopSeeker struct {
-	io.Reader
-}
-
-func (nopSeeker) Seek(int64, int) (int64, error) {
-	return 0, nil
-}
-
-func parseNameFill(name, fill string) model.Name {
-	fill = cmp.Or(fill, "bllamo.com/library/_:latest")
-	f := model.ParseNameBare(fill)
-	if !f.IsFullyQualified() {
-		panic(fmt.Errorf("invalid fill: %q", fill))
-	}
-	return model.Merge(model.ParseNameBare(name), f)
-}
-
-// Push pushes a manifest to the server and responds to the server's
-// requests for layer uploads, if any, and finally commits the manifest for
-// name. It returns an error if any part of the process fails, specifically:
-//
-// If the server requests layers not found in the cache, ErrLayerNotFound is
-// returned.
-func (c *Client) Push(ctx context.Context, cache Cache, name string) error {
-	mn := parseNameFill(name, c.NameFill)
-	if !mn.IsFullyQualified() {
-		return fmt.Errorf("ollama: push: invalid name: %s", name)
-	}
-	manifest := cache.ManifestData(mn)
-	if len(manifest) == 0 {
-		return fmt.Errorf("manifest not found: %s", name)
-	}
-
-	var mu sync.Mutex
-	var completed []*apitype.CompletePart
-	push := func() (*apitype.PushResponse, error) {
-		v, err := ollama.Do[*apitype.PushResponse](ctx, c.oclient(), "POST", "/v1/push", &apitype.PushRequest{
-			Name:          name,
-			Manifest:      manifest,
-			CompleteParts: completed,
-		})
-		if err != nil {
-			return nil, fmt.Errorf("Do: %w", err)
-		}
-		return v, nil
-	}
-
-	pr, err := push()
-	if err != nil {
-		return err
-	}
-
-	var g errgroup.Group
-	for _, need := range pr.Needs {
-		g.Go(func() error {
-			nd, err := model.ParseDigest(need.Digest)
-			if err != nil {
-				return fmt.Errorf("ParseDigest: %w: %s", err, need.Digest)
-			}
-			f, err := cache.OpenLayer(nd)
-			if err != nil {
-				return fmt.Errorf("OpenLayer: %w: %s", err, need.Digest)
-			}
-			defer f.Close()
-
-			c.log().Info("pushing layer", "digest", need.Digest, "start", need.Start, "end", need.End)
-			cp, err := PushLayer(ctx, f, need.URL, need.Start, need.End)
-			if err != nil {
-				return fmt.Errorf("PushLayer: %w: %s", err, need.Digest)
-			}
-			mu.Lock()
-			completed = append(completed, cp)
-			mu.Unlock()
-			return nil
-		})
-	}
-	if err := g.Wait(); err != nil {
-		return fmt.Errorf("Push: Required: %w", err)
-	}
-
-	if len(completed) > 0 {
-		pr, err := push()
-		if err != nil {
-			return err
-		}
-		if len(pr.Needs) > 0 {
-			var errs []error
-			for _, r := range pr.Needs {
-				errs = append(errs, fmt.Errorf("Push: server failed to find part: %q", r.Digest))
-			}
-			return errors.Join(errs...)
-		}
-	}
-
-	return cache.SetManifestData(mn, manifest)
-}
-
-func PushLayer(ctx context.Context, body io.ReaderAt, url string, start, end int64) (*apitype.CompletePart, error) {
-	if start < 0 || end < start {
-		return nil, errors.New("start must satisfy 0 <= start <= end")
-	}
-
-	file := io.NewSectionReader(body, start, end-start+1)
-	req, err := http.NewRequest("PUT", url, file)
-	if err != nil {
-		return nil, err
-	}
-	req.ContentLength = end - start + 1
-
-	// TODO(bmizerany): take content type param
-	req.Header.Set("Content-Type", "text/plain")
-
-	if start != 0 || end != 0 {
-		req.Header.Set("x-amz-copy-source-range", fmt.Sprintf("bytes=%d-%d", start, end))
-	}
-
-	res, err := http.DefaultClient.Do(req)
-	if err != nil {
-		return nil, err
-	}
-	defer res.Body.Close()
-	if res.StatusCode != 200 {
-		e := parseS3Error(res)
-		return nil, fmt.Errorf("unexpected status code: %d; %w", res.StatusCode, e)
-	}
-	cp := &apitype.CompletePart{
-		URL:  url,
-		ETag: res.Header.Get("ETag"),
-		// TODO(bmizerany): checksum
-	}
-	return cp, nil
-}
-
-type s3Error struct {
-	XMLName   xml.Name `xml:"Error"`
-	Code      string   `xml:"Code"`
-	Message   string   `xml:"Message"`
-	Resource  string   `xml:"Resource"`
-	RequestId string   `xml:"RequestId"`
-}
-
-func (e *s3Error) Error() string {
-	return fmt.Sprintf("S3 (%s): %s: %s: %s", e.RequestId, e.Resource, e.Code, e.Message)
-}
-
-// parseS3Error parses an XML error response from S3.
-func parseS3Error(res *http.Response) error {
-	var se *s3Error
-	if err := xml.NewDecoder(res.Body).Decode(&se); err != nil {
-		return err
-	}
-	return se
-}
-
-// TODO: replace below by using upload pkg after we have rangefunc; until
-// then, we need to keep this free of rangefunc for now.
-type chunkRange[I constraints.Integer] struct {
-	// Start is the byte offset of the chunk.
-	Start I
-
-	// End is the byte offset of the last byte in the chunk.
-	End I
-}
-
-func (c chunkRange[I]) Size() I {
-	return c.End - c.Start + 1
-}
-
-func (c chunkRange[I]) String() string {
-	return fmt.Sprintf("%d-%d", c.Start, c.End)
-}
-
-func (c chunkRange[I]) LogValue() slog.Value {
-	return slog.StringValue(c.String())
-}
-
-// Chunks yields a sequence of a part number and a Chunk. The Chunk is the offset
-// and size of the chunk. The last chunk may be smaller than chunkSize if size is
-// not a multiple of chunkSize.
-//
-// The first part number is 1 and increases monotonically.
-func chunks[I constraints.Integer](size, chunkSize I) iter.Seq2[int, chunkRange[I]] {
-	return func(yield func(int, chunkRange[I]) bool) {
-		var n int
-		for off := I(0); off < size; off += chunkSize {
-			n++
-			if !yield(n, chunkRange[I]{
-				Start: off,
-				End:   off + min(chunkSize, size-off) - 1,
-			}) {
-				return
-			}
-		}
-	}
-}
-
-func redactAmzSignature(s string) string {
-	u, err := url.Parse(s)
-	if err != nil {
-		return ""
-	}
-	q := u.Query()
-	q.Set("X-Amz-Signature", "REDACTED")
-	u.RawQuery = q.Encode()
-	return u.String()
-}
@@ -1,667 +0,0 @@
-package cmd
-
-import (
-	"errors"
-	"fmt"
-	"io"
-	"net/http"
-	"os"
-	"path/filepath"
-	"regexp"
-	"sort"
-	"strings"
-
-	"github.com/spf13/cobra"
-	"golang.org/x/exp/slices"
-
-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/progress"
-	"github.com/ollama/ollama/readline"
-)
-
-type MultilineState int
-
-const (
-	MultilineNone MultilineState = iota
-	MultilinePrompt
-	MultilineSystem
-	MultilineTemplate
-)
-
-func loadModel(cmd *cobra.Command, opts *runOptions) error {
-	client, err := api.ClientFromEnvironment()
-	if err != nil {
-		return err
-	}
-
-	p := progress.NewProgress(os.Stderr)
-	defer p.StopAndClear()
-
-	spinner := progress.NewSpinner("")
-	p.Add("", spinner)
-
-	showReq := api.ShowRequest{Name: opts.Model}
-	showResp, err := client.Show(cmd.Context(), &showReq)
-	if err != nil {
-		return err
-	}
-	opts.MultiModal = slices.Contains(showResp.Details.Families, "clip")
-	opts.ParentModel = showResp.Details.ParentModel
-
-	if len(showResp.Messages) > 0 {
-		opts.Messages = append(opts.Messages, showResp.Messages...)
-	}
-
-	chatReq := &api.ChatRequest{
-		Model:    opts.Model,
-		Messages: []api.Message{},
-	}
-	err = client.Chat(cmd.Context(), chatReq, func(resp api.ChatResponse) error {
-		p.StopAndClear()
-		if len(opts.Messages) > 0 {
-			for _, msg := range opts.Messages {
-				switch msg.Role {
-				case "user":
-					fmt.Printf(">>> %s\n", msg.Content)
-				case "assistant":
-					state := &displayResponseState{}
-					displayResponse(msg.Content, opts.WordWrap, state)
-					fmt.Println()
-					fmt.Println()
-				}
-			}
-		}
-		return nil
-	})
-	if err != nil {
-		return err
-	}
-
-	return nil
-}
-
-func generateInteractive(cmd *cobra.Command, opts runOptions) error {
-	opts.Messages = make([]api.Message, 0)
-
-	err := loadModel(cmd, &opts)
-	if err != nil {
-		return err
-	}
-
-	usage := func() {
-		fmt.Fprintln(os.Stderr, "Available Commands:")
-		fmt.Fprintln(os.Stderr, "  /set            Set session variables")
-		fmt.Fprintln(os.Stderr, "  /show           Show model information")
-		fmt.Fprintln(os.Stderr, "  /load <model>   Load a session or model")
-		fmt.Fprintln(os.Stderr, "  /save <model>   Save your current session")
-		fmt.Fprintln(os.Stderr, "  /bye            Exit")
-		fmt.Fprintln(os.Stderr, "  /?, /help       Help for a command")
-		fmt.Fprintln(os.Stderr, "  /? shortcuts    Help for keyboard shortcuts")
-		fmt.Fprintln(os.Stderr, "")
-		fmt.Fprintln(os.Stderr, "Use \"\"\" to begin a multi-line message.")
-
-		if opts.MultiModal {
-			fmt.Fprintf(os.Stderr, "Use %s to include .jpg or .png images.\n", filepath.FromSlash("/path/to/file"))
-		}
-
-		fmt.Fprintln(os.Stderr, "")
-	}
-
-	usageSet := func() {
-		fmt.Fprintln(os.Stderr, "Available Commands:")
-		fmt.Fprintln(os.Stderr, "  /set parameter ...     Set a parameter")
-		fmt.Fprintln(os.Stderr, "  /set system <string>   Set system message")
-		fmt.Fprintln(os.Stderr, "  /set template <string> Set prompt template")
-		fmt.Fprintln(os.Stderr, "  /set history           Enable history")
-		fmt.Fprintln(os.Stderr, "  /set nohistory         Disable history")
-		fmt.Fprintln(os.Stderr, "  /set wordwrap          Enable wordwrap")
-		fmt.Fprintln(os.Stderr, "  /set nowordwrap        Disable wordwrap")
-		fmt.Fprintln(os.Stderr, "  /set format json       Enable JSON mode")
-		fmt.Fprintln(os.Stderr, "  /set noformat          Disable formatting")
-		fmt.Fprintln(os.Stderr, "  /set verbose           Show LLM stats")
-		fmt.Fprintln(os.Stderr, "  /set quiet             Disable LLM stats")
-		fmt.Fprintln(os.Stderr, "")
-	}
-
-	usageShortcuts := func() {
-		fmt.Fprintln(os.Stderr, "Available keyboard shortcuts:")
-		fmt.Fprintln(os.Stderr, "  Ctrl + a            Move to the beginning of the line (Home)")
-		fmt.Fprintln(os.Stderr, "  Ctrl + e            Move to the end of the line (End)")
-		fmt.Fprintln(os.Stderr, "   Alt + b            Move back (left) one word")
-		fmt.Fprintln(os.Stderr, "   Alt + f            Move forward (right) one word")
-		fmt.Fprintln(os.Stderr, "  Ctrl + k            Delete the sentence after the cursor")
-		fmt.Fprintln(os.Stderr, "  Ctrl + u            Delete the sentence before the cursor")
-		fmt.Fprintln(os.Stderr, "")
-		fmt.Fprintln(os.Stderr, "  Ctrl + l            Clear the screen")
-		fmt.Fprintln(os.Stderr, "  Ctrl + c            Stop the model from responding")
-		fmt.Fprintln(os.Stderr, "  Ctrl + d            Exit ollama (/bye)")
-		fmt.Fprintln(os.Stderr, "")
-	}
-
-	usageShow := func() {
-		fmt.Fprintln(os.Stderr, "Available Commands:")
-		fmt.Fprintln(os.Stderr, "  /show info         Show details for this model")
-		fmt.Fprintln(os.Stderr, "  /show license      Show model license")
-		fmt.Fprintln(os.Stderr, "  /show modelfile    Show Modelfile for this model")
-		fmt.Fprintln(os.Stderr, "  /show parameters   Show parameters for this model")
-		fmt.Fprintln(os.Stderr, "  /show system       Show system message")
-		fmt.Fprintln(os.Stderr, "  /show template     Show prompt template")
-		fmt.Fprintln(os.Stderr, "")
-	}
-
-	// only list out the most common parameters
-	usageParameters := func() {
-		fmt.Fprintln(os.Stderr, "Available Parameters:")
-		fmt.Fprintln(os.Stderr, "  /set parameter seed <int>             Random number seed")
-		fmt.Fprintln(os.Stderr, "  /set parameter num_predict <int>      Max number of tokens to predict")
-		fmt.Fprintln(os.Stderr, "  /set parameter top_k <int>            Pick from top k num of tokens")
-		fmt.Fprintln(os.Stderr, "  /set parameter top_p <float>          Pick token based on sum of probabilities")
-		fmt.Fprintln(os.Stderr, "  /set parameter num_ctx <int>          Set the context size")
-		fmt.Fprintln(os.Stderr, "  /set parameter temperature <float>    Set creativity level")
-		fmt.Fprintln(os.Stderr, "  /set parameter repeat_penalty <float> How strongly to penalize repetitions")
-		fmt.Fprintln(os.Stderr, "  /set parameter repeat_last_n <int>    Set how far back to look for repetitions")
-		fmt.Fprintln(os.Stderr, "  /set parameter num_gpu <int>          The number of layers to send to the GPU")
-		fmt.Fprintln(os.Stderr, "  /set parameter stop \"<string>\", ...   Set the stop parameters")
-		fmt.Fprintln(os.Stderr, "")
-	}
-
-	scanner, err := readline.New(readline.Prompt{
-		Prompt:         ">>> ",
-		AltPrompt:      "... ",
-		Placeholder:    "Send a message (/? for help)",
-		AltPlaceholder: `Use """ to end multi-line input`,
-	})
-	if err != nil {
-		return err
-	}
-
-	fmt.Print(readline.StartBracketedPaste)
-	defer fmt.Printf(readline.EndBracketedPaste)
-
-	var sb strings.Builder
-	var multiline MultilineState
-
-	for {
-		line, err := scanner.Readline()
-		switch {
-		case errors.Is(err, io.EOF):
-			fmt.Println()
-			return nil
-		case errors.Is(err, readline.ErrInterrupt):
-			if line == "" {
-				fmt.Println("\nUse Ctrl + d or /bye to exit.")
-			}
-
-			scanner.Prompt.UseAlt = false
-			sb.Reset()
-
-			continue
-		case err != nil:
-			return err
-		}
-
-		switch {
-		case multiline != MultilineNone:
-			// check if there's a multiline terminating string
-			before, ok := strings.CutSuffix(line, `"""`)
-			sb.WriteString(before)
-			if !ok {
-				fmt.Fprintln(&sb)
-				continue
-			}
-
-			switch multiline {
-			case MultilineSystem:
-				opts.System = sb.String()
-				opts.Messages = append(opts.Messages, api.Message{Role: "system", Content: opts.System})
-				fmt.Println("Set system message.")
-				sb.Reset()
-			case MultilineTemplate:
-				opts.Template = sb.String()
-				fmt.Println("Set prompt template.")
-				sb.Reset()
-			}
-
-			multiline = MultilineNone
-			scanner.Prompt.UseAlt = false
-		case strings.HasPrefix(line, `"""`):
-			line := strings.TrimPrefix(line, `"""`)
-			line, ok := strings.CutSuffix(line, `"""`)
-			sb.WriteString(line)
-			if !ok {
-				// no multiline terminating string; need more input
-				fmt.Fprintln(&sb)
-				multiline = MultilinePrompt
-				scanner.Prompt.UseAlt = true
-			}
-		case scanner.Pasting:
-			fmt.Fprintln(&sb, line)
-			continue
-		case strings.HasPrefix(line, "/list"):
-			args := strings.Fields(line)
-			if err := ListHandler(cmd, args[1:]); err != nil {
-				return err
-			}
-		case strings.HasPrefix(line, "/load"):
-			args := strings.Fields(line)
-			if len(args) != 2 {
-				fmt.Println("Usage:\n  /load <modelname>")
-				continue
-			}
-			opts.Model = args[1]
-			opts.Messages = []api.Message{}
-			fmt.Printf("Loading model '%s'\n", opts.Model)
-			if err := loadModel(cmd, &opts); err != nil {
-				return err
-			}
-			continue
-		case strings.HasPrefix(line, "/save"):
-			args := strings.Fields(line)
-			if len(args) != 2 {
-				fmt.Println("Usage:\n  /save <modelname>")
-				continue
-			}
-
-			client, err := api.ClientFromEnvironment()
-			if err != nil {
-				fmt.Println("error: couldn't connect to ollama server")
-				return err
-			}
-
-			req := &api.CreateRequest{
-				Name:      args[1],
-				Modelfile: buildModelfile(opts),
-			}
-			fn := func(resp api.ProgressResponse) error { return nil }
-			err = client.Create(cmd.Context(), req, fn)
-			if err != nil {
-				fmt.Println("error: couldn't save model")
-				return err
-			}
-			fmt.Printf("Created new model '%s'\n", args[1])
-			continue
-		case strings.HasPrefix(line, "/set"):
-			args := strings.Fields(line)
-			if len(args) > 1 {
-				switch args[1] {
-				case "history":
-					scanner.HistoryEnable()
-				case "nohistory":
-					scanner.HistoryDisable()
-				case "wordwrap":
-					opts.WordWrap = true
-					fmt.Println("Set 'wordwrap' mode.")
-				case "nowordwrap":
-					opts.WordWrap = false
-					fmt.Println("Set 'nowordwrap' mode.")
-				case "verbose":
-					if err := cmd.Flags().Set("verbose", "true"); err != nil {
-						return err
-					}
-					fmt.Println("Set 'verbose' mode.")
-				case "quiet":
-					if err := cmd.Flags().Set("verbose", "false"); err != nil {
-						return err
-					}
-					fmt.Println("Set 'quiet' mode.")
-				case "format":
-					if len(args) < 3 || args[2] != "json" {
-						fmt.Println("Invalid or missing format. For 'json' mode use '/set format json'")
-					} else {
-						opts.Format = args[2]
-						fmt.Printf("Set format to '%s' mode.\n", args[2])
-					}
-				case "noformat":
-					opts.Format = ""
-					fmt.Println("Disabled format.")
-				case "parameter":
-					if len(args) < 4 {
-						usageParameters()
-						continue
-					}
-					params := args[3:]
-					fp, err := api.FormatParams(map[string][]string{args[2]: params})
-					if err != nil {
-						fmt.Printf("Couldn't set parameter: %q\n", err)
-						continue
-					}
-					fmt.Printf("Set parameter '%s' to '%s'\n", args[2], strings.Join(params, ", "))
-					opts.Options[args[2]] = fp[args[2]]
-				case "system", "template":
-					if len(args) < 3 {
-						usageSet()
-						continue
-					}
-
-					if args[1] == "system" {
-						multiline = MultilineSystem
-					} else if args[1] == "template" {
-						multiline = MultilineTemplate
-					}
-
-					line := strings.Join(args[2:], " ")
-					line, ok := strings.CutPrefix(line, `"""`)
-					if !ok {
-						multiline = MultilineNone
-					} else {
-						// only cut suffix if the line is multiline
-						line, ok = strings.CutSuffix(line, `"""`)
-						if ok {
-							multiline = MultilineNone
-						}
-					}
-
-					sb.WriteString(line)
-					if multiline != MultilineNone {
-						scanner.Prompt.UseAlt = true
-						continue
-					}
-
-					if args[1] == "system" {
-						opts.System = sb.String() // for display in modelfile
-						newMessage := api.Message{Role: "system", Content: sb.String()}
-						// Check if the slice is not empty and the last message is from 'system'
-						if len(opts.Messages) > 0 && opts.Messages[len(opts.Messages)-1].Role == "system" {
-							// Replace the last message
-							opts.Messages[len(opts.Messages)-1] = newMessage
-						} else {
-							opts.Messages = append(opts.Messages, newMessage)
-						}
-						fmt.Println("Set system message.")
-						sb.Reset()
-					} else if args[1] == "template" {
-						opts.Template = sb.String()
-						fmt.Println("Set prompt template.")
-						sb.Reset()
-					}
-
-					sb.Reset()
-					continue
-				default:
-					fmt.Printf("Unknown command '/set %s'. Type /? for help\n", args[1])
-				}
-			} else {
-				usageSet()
-			}
-		case strings.HasPrefix(line, "/show"):
-			args := strings.Fields(line)
-			if len(args) > 1 {
-				client, err := api.ClientFromEnvironment()
-				if err != nil {
-					fmt.Println("error: couldn't connect to ollama server")
-					return err
-				}
-				req := &api.ShowRequest{
-					Name:     opts.Model,
-					System:   opts.System,
-					Template: opts.Template,
-					Options:  opts.Options,
-				}
-				resp, err := client.Show(cmd.Context(), req)
-				if err != nil {
-					fmt.Println("error: couldn't get model")
-					return err
-				}
-
-				switch args[1] {
-				case "info":
-					fmt.Println("Model details:")
-					if len(resp.Details.Families) > 0 {
-						fmt.Printf("Family              %s\n", strings.Join(resp.Details.Families, ", "))
-					} else if resp.Details.Family != "" {
-						fmt.Printf("Family              %s\n", resp.Details.Family)
-					}
-					fmt.Printf("Parameter Size      %s\n", resp.Details.ParameterSize)
-					fmt.Printf("Quantization Level  %s\n", resp.Details.QuantizationLevel)
-					fmt.Println("")
-				case "license":
-					if resp.License == "" {
-						fmt.Println("No license was specified for this model.")
-					} else {
-						fmt.Println(resp.License)
-					}
-				case "modelfile":
-					fmt.Println(resp.Modelfile)
-				case "parameters":
-					if resp.Parameters == "" {
-						fmt.Println("No parameters were specified for this model.")
-					} else {
-						if len(opts.Options) > 0 {
-							fmt.Println("User defined parameters:")
-							for k, v := range opts.Options {
-								fmt.Printf("%-*s %v\n", 30, k, v)
-							}
-							fmt.Println()
-						}
-						fmt.Println("Model defined parameters:")
-						fmt.Println(resp.Parameters)
-					}
-				case "system":
-					switch {
-					case opts.System != "":
-						fmt.Println(opts.System + "\n")
-					case resp.System != "":
-						fmt.Println(resp.System + "\n")
-					default:
-						fmt.Println("No system message was specified for this model.")
-					}
-				case "template":
-					switch {
-					case opts.Template != "":
-						fmt.Println(opts.Template + "\n")
-					case resp.Template != "":
-						fmt.Println(resp.Template)
-					default:
-						fmt.Println("No prompt template was specified for this model.")
-					}
-				default:
-					fmt.Printf("Unknown command '/show %s'. Type /? for help\n", args[1])
-				}
-			} else {
-				usageShow()
-			}
-		case strings.HasPrefix(line, "/help"), strings.HasPrefix(line, "/?"):
-			args := strings.Fields(line)
-			if len(args) > 1 {
-				switch args[1] {
-				case "set", "/set":
-					usageSet()
-				case "show", "/show":
-					usageShow()
-				case "shortcut", "shortcuts":
-					usageShortcuts()
-				}
-			} else {
-				usage()
-			}
-		case strings.HasPrefix(line, "/exit"), strings.HasPrefix(line, "/bye"):
-			return nil
-		case strings.HasPrefix(line, "/"):
-			args := strings.Fields(line)
-			isFile := false
-
-			if opts.MultiModal {
-				for _, f := range extractFileNames(line) {
-					if strings.HasPrefix(f, args[0]) {
-						isFile = true
-						break
-					}
-				}
-			}
-
-			if !isFile {
-				fmt.Printf("Unknown command '%s'. Type /? for help\n", args[0])
-				continue
-			}
-
-			sb.WriteString(line)
-		default:
-			sb.WriteString(line)
-		}
-
-		if sb.Len() > 0 && multiline == MultilineNone {
-			newMessage := api.Message{Role: "user", Content: sb.String()}
-
-			if opts.MultiModal {
-				msg, images, err := extractFileData(sb.String())
-				if err != nil {
-					return err
-				}
-
-				// clear all previous images for better responses
-				if len(images) > 0 {
-					for i := range opts.Messages {
-						opts.Messages[i].Images = nil
-					}
-				}
-
-				newMessage.Content = msg
-				newMessage.Images = images
-			}
-
-			opts.Messages = append(opts.Messages, newMessage)
-
-			assistant, err := chat(cmd, opts)
-			if err != nil {
-				return err
-			}
-			if assistant != nil {
-				opts.Messages = append(opts.Messages, *assistant)
-			}
-
-			sb.Reset()
-		}
-	}
-}
-
-func buildModelfile(opts runOptions) string {
-	var mf strings.Builder
-	model := opts.ParentModel
-	if model == "" {
-		model = opts.Model
-	}
-	fmt.Fprintf(&mf, "FROM %s\n", model)
-	if opts.System != "" {
-		fmt.Fprintf(&mf, "SYSTEM \"\"\"%s\"\"\"\n", opts.System)
-	}
-
-	if opts.Template != "" {
-		fmt.Fprintf(&mf, "TEMPLATE \"\"\"%s\"\"\"\n", opts.Template)
-	}
-
-	keys := make([]string, 0)
-	for k := range opts.Options {
-		keys = append(keys, k)
-	}
-	sort.Strings(keys)
-	for _, k := range keys {
-		fmt.Fprintf(&mf, "PARAMETER %s %v\n", k, opts.Options[k])
-	}
-	fmt.Fprintln(&mf)
-
-	for _, msg := range opts.Messages {
-		fmt.Fprintf(&mf, "MESSAGE %s \"\"\"%s\"\"\"\n", msg.Role, msg.Content)
-	}
-
-	return mf.String()
-}
-
-func normalizeFilePath(fp string) string {
-	// Define a map of escaped characters and their replacements
-	replacements := map[string]string{
-		"\\ ":  " ",  // Escaped space
-		"\\(":  "(",  // Escaped left parenthesis
-		"\\)":  ")",  // Escaped right parenthesis
-		"\\[":  "[",  // Escaped left square bracket
-		"\\]":  "]",  // Escaped right square bracket
-		"\\{":  "{",  // Escaped left curly brace
-		"\\}":  "}",  // Escaped right curly brace
-		"\\$":  "$",  // Escaped dollar sign
-		"\\&":  "&",  // Escaped ampersand
-		"\\;":  ";",  // Escaped semicolon
-		"\\'":  "'",  // Escaped single quote
-		"\\\\": "\\", // Escaped backslash
-		"\\*":  "*",  // Escaped asterisk
-		"\\?":  "?",  // Escaped question mark
-	}
-
-	for escaped, actual := range replacements {
-		fp = strings.ReplaceAll(fp, escaped, actual)
-	}
-	return fp
-}
-
-func extractFileNames(input string) []string {
-	// Regex to match file paths starting with optional drive letter, / ./ \ or .\ and include escaped or unescaped spaces (\ or %20)
-	// and followed by more characters and a file extension
-	// This will capture non filename strings, but we'll check for file existence to remove mismatches
-	regexPattern := `(?:[a-zA-Z]:)?(?:\./|/|\\)[\S\\ ]+?\.(?i:jpg|jpeg|png|svg)\b`
-	re := regexp.MustCompile(regexPattern)
-
-	return re.FindAllString(input, -1)
-}
-
-func extractFileData(input string) (string, []api.ImageData, error) {
-	filePaths := extractFileNames(input)
-	var imgs []api.ImageData
-
-	for _, fp := range filePaths {
-		nfp := normalizeFilePath(fp)
-		data, err := getImageData(nfp)
-		if err != nil {
-			if os.IsNotExist(err) {
-				continue
-			}
-			fmt.Fprintf(os.Stderr, "Couldn't process image: %q\n", err)
-			return "", imgs, err
-		}
-		fmt.Fprintf(os.Stderr, "Added image '%s'\n", nfp)
-		input = strings.ReplaceAll(input, fp, "")
-		imgs = append(imgs, data)
-	}
-	return input, imgs, nil
-}
-
-func getImageData(filePath string) ([]byte, error) {
-	file, err := os.Open(filePath)
-	if err != nil {
-		return nil, err
-	}
-	defer file.Close()
-
-	buf := make([]byte, 512)
-	_, err = file.Read(buf)
-	if err != nil {
-		return nil, err
-	}
-
-	contentType := http.DetectContentType(buf)
-	allowedTypes := []string{"image/jpeg", "image/jpg", "image/png"}
-	if !slices.Contains(allowedTypes, contentType) {
-		return nil, fmt.Errorf("invalid image type: %s", contentType)
-	}
-
-	info, err := file.Stat()
-	if err != nil {
-		return nil, err
-	}
-
-	// Check if the file size exceeds 100MB
-	var maxSize int64 = 100 * 1024 * 1024 // 100MB in bytes
-	if info.Size() > maxSize {
-		return nil, fmt.Errorf("file size exceeds maximum limit (100MB)")
-	}
-
-	buf = make([]byte, info.Size())
-	_, err = file.Seek(0, 0)
-	if err != nil {
-		return nil, err
-	}
-
-	_, err = io.ReadFull(file, buf)
-	if err != nil {
-		return nil, err
-	}
-
-	return buf, nil
-}
@@ -1,116 +0,0 @@
-package cmd
-
-import (
-	"bytes"
-	"testing"
-	"text/template"
-
-	"github.com/stretchr/testify/assert"
-
-	"github.com/ollama/ollama/api"
-)
-
-func TestExtractFilenames(t *testing.T) {
-	// Unix style paths
-	input := ` some preamble 
- ./relative\ path/one.png inbetween1 ./not a valid two.jpg inbetween2
-/unescaped space /three.jpeg inbetween3 /valid\ path/dir/four.png "./quoted with spaces/five.svg`
-	res := extractFileNames(input)
-	assert.Len(t, res, 5)
-	assert.Contains(t, res[0], "one.png")
-	assert.Contains(t, res[1], "two.jpg")
-	assert.Contains(t, res[2], "three.jpeg")
-	assert.Contains(t, res[3], "four.png")
-	assert.Contains(t, res[4], "five.svg")
-	assert.NotContains(t, res[4], '"')
-	assert.NotContains(t, res, "inbtween")
-
-	// Windows style paths
-	input = ` some preamble
- c:/users/jdoe/one.png inbetween1 c:/program files/someplace/two.jpg inbetween2 
- /absolute/nospace/three.jpeg inbetween3 /absolute/with space/four.png inbetween4
-./relative\ path/five.svg inbetween5 "./relative with/spaces/six.png inbetween6
-d:\path with\spaces\seven.svg inbetween7 c:\users\jdoe\eight.png inbetween8 
- d:\program files\someplace\nine.png inbetween9 "E:\program files\someplace\ten.svg some ending
-`
-	res = extractFileNames(input)
-	assert.Len(t, res, 10)
-	assert.NotContains(t, res, "inbtween")
-	assert.Contains(t, res[0], "one.png")
-	assert.Contains(t, res[0], "c:")
-	assert.Contains(t, res[1], "two.jpg")
-	assert.Contains(t, res[1], "c:")
-	assert.Contains(t, res[2], "three.jpeg")
-	assert.Contains(t, res[3], "four.png")
-	assert.Contains(t, res[4], "five.svg")
-	assert.Contains(t, res[5], "six.png")
-	assert.Contains(t, res[6], "seven.svg")
-	assert.Contains(t, res[6], "d:")
-	assert.Contains(t, res[7], "eight.png")
-	assert.Contains(t, res[7], "c:")
-	assert.Contains(t, res[8], "nine.png")
-	assert.Contains(t, res[8], "d:")
-	assert.Contains(t, res[9], "ten.svg")
-	assert.Contains(t, res[9], "E:")
-}
-
-func TestModelfileBuilder(t *testing.T) {
-	opts := runOptions{
-		Model:    "hork",
-		System:   "You are part horse and part shark, but all hork. Do horklike things",
-		Template: "This is a template.",
-		Messages: []api.Message{
-			{Role: "user", Content: "Hey there hork!"},
-			{Role: "assistant", Content: "Yes it is true, I am half horse, half shark."},
-		},
-		Options: map[string]interface{}{},
-	}
-
-	opts.Options["temperature"] = 0.9
-	opts.Options["seed"] = 42
-	opts.Options["penalize_newline"] = false
-	opts.Options["stop"] = []string{"hi", "there"}
-
-	mf := buildModelfile(opts)
-	expectedModelfile := `FROM {{.Model}}
-SYSTEM """{{.System}}"""
-TEMPLATE """{{.Template}}"""
-PARAMETER penalize_newline false
-PARAMETER seed 42
-PARAMETER stop [hi there]
-PARAMETER temperature 0.9
-
-MESSAGE user """Hey there hork!"""
-MESSAGE assistant """Yes it is true, I am half horse, half shark."""
-`
-
-	tmpl, err := template.New("").Parse(expectedModelfile)
-	assert.Nil(t, err)
-
-	var buf bytes.Buffer
-	err = tmpl.Execute(&buf, opts)
-	assert.Nil(t, err)
-	assert.Equal(t, buf.String(), mf)
-
-	opts.ParentModel = "horseshark"
-	mf = buildModelfile(opts)
-	expectedModelfile = `FROM {{.ParentModel}}
-SYSTEM """{{.System}}"""
-TEMPLATE """{{.Template}}"""
-PARAMETER penalize_newline false
-PARAMETER seed 42
-PARAMETER stop [hi there]
-PARAMETER temperature 0.9
-
-MESSAGE user """Hey there hork!"""
-MESSAGE assistant """Yes it is true, I am half horse, half shark."""
-`
-
-	tmpl, err = template.New("").Parse(expectedModelfile)
-	assert.Nil(t, err)
-
-	var parentBuf bytes.Buffer
-	err = tmpl.Execute(&parentBuf, opts)
-	assert.Nil(t, err)
-	assert.Equal(t, parentBuf.String(), mf)
-}
@@ -0,0 +1,44 @@
+package cmd
+
+import (
+	"fmt"
+	"os"
+	"time"
+
+	"github.com/jmorganca/ollama/progressbar"
+)
+
+type Spinner struct {
+	description string
+	*progressbar.ProgressBar
+}
+
+func NewSpinner(description string) *Spinner {
+	return &Spinner{
+		description: description,
+		ProgressBar: progressbar.NewOptions(-1,
+			progressbar.OptionSetWriter(os.Stderr),
+			progressbar.OptionThrottle(60*time.Millisecond),
+			progressbar.OptionSpinnerType(14),
+			progressbar.OptionSetRenderBlankState(true),
+			progressbar.OptionSetElapsedTime(false),
+			progressbar.OptionClearOnFinish(),
+			progressbar.OptionSetDescription(description),
+		),
+	}
+}
+
+func (s *Spinner) Spin(tick time.Duration) {
+	for range time.Tick(tick) {
+		if s.IsFinished() {
+			break
+		}
+
+		s.Add(1)
+	}
+}
+
+func (s *Spinner) Stop() {
+	s.Finish()
+	fmt.Println(s.description)
+}
@@ -1,30 +0,0 @@
-package cmd
-
-import (
-	"context"
-	"fmt"
-	"os"
-	"os/exec"
-	"strings"
-
-	"github.com/ollama/ollama/api"
-)
-
-func startApp(ctx context.Context, client *api.Client) error {
-	exe, err := os.Executable()
-	if err != nil {
-		return err
-	}
-	link, err := os.Readlink(exe)
-	if err != nil {
-		return err
-	}
-	if !strings.Contains(link, "Ollama.app") {
-		return fmt.Errorf("could not find ollama app")
-	}
-	path := strings.Split(link, "Ollama.app")
-	if err := exec.Command("/usr/bin/open", "-a", path[0]+"Ollama.app").Run(); err != nil {
-		return err
-	}
-	return waitForServer(ctx, client)
-}
@@ -1,14 +0,0 @@
-//go:build !windows && !darwin
-
-package cmd
-
-import (
-	"context"
-	"fmt"
-
-	"github.com/ollama/ollama/api"
-)
-
-func startApp(ctx context.Context, client *api.Client) error {
-	return fmt.Errorf("could not connect to ollama server, run 'ollama serve' to start it")
-}
@@ -1,58 +0,0 @@
-package cmd
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"os"
-	"os/exec"
-	"path/filepath"
-	"strings"
-	"syscall"
-
-	"github.com/ollama/ollama/api"
-)
-
-func startApp(ctx context.Context, client *api.Client) error {
-	// log.Printf("XXX Attempting to find and start ollama app")
-	AppName := "ollama app.exe"
-	exe, err := os.Executable()
-	if err != nil {
-		return err
-	}
-	appExe := filepath.Join(filepath.Dir(exe), AppName)
-	_, err = os.Stat(appExe)
-	if errors.Is(err, os.ErrNotExist) {
-		// Try the standard install location
-		localAppData := os.Getenv("LOCALAPPDATA")
-		appExe = filepath.Join(localAppData, "Ollama", AppName)
-		_, err := os.Stat(appExe)
-		if errors.Is(err, os.ErrNotExist) {
-			// Finally look in the path
-			appExe, err = exec.LookPath(AppName)
-			if err != nil {
-				return fmt.Errorf("could not locate ollama app")
-			}
-		}
-	}
-	// log.Printf("XXX attempting to start app %s", appExe)
-
-	cmd_path := "c:\\Windows\\system32\\cmd.exe"
-	cmd := exec.Command(cmd_path, "/c", appExe)
-	// TODO - these hide flags aren't working - still pops up a command window for some reason
-	cmd.SysProcAttr = &syscall.SysProcAttr{CreationFlags: 0x08000000, HideWindow: true}
-
-	// TODO this didn't help either...
-	cmd.Stdin = strings.NewReader("")
-	cmd.Stdout = os.Stdout
-	cmd.Stderr = os.Stderr
-
-	if err := cmd.Start(); err != nil {
-		return fmt.Errorf("unable to start ollama app %w", err)
-	}
-
-	if cmd.Process != nil {
-		defer cmd.Process.Release() //nolint:errcheck
-	}
-	return waitForServer(ctx, client)
-}
@@ -1,187 +0,0 @@
-package convert
-
-import (
-	"cmp"
-	"encoding/binary"
-	"encoding/json"
-	"fmt"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"slices"
-	"strings"
-
-	"google.golang.org/protobuf/proto"
-
-	"github.com/ollama/ollama/convert/sentencepiece"
-	"github.com/ollama/ollama/llm"
-)
-
-type Params struct {
-	Architectures     []string `json:"architectures"`
-	VocabSize         int      `json:"vocab_size"`
-	HiddenSize        int      `json:"hidden_size"`       // n_embd
-	HiddenLayers      int      `json:"num_hidden_layers"` // n_layer
-	ContextSize       int      `json:"max_position_embeddings"`
-	IntermediateSize  int      `json:"intermediate_size"`
-	AttentionHeads    int      `json:"num_attention_heads"` // n_head
-	KeyValHeads       int      `json:"num_key_value_heads"`
-	NormEPS           float64  `json:"rms_norm_eps"`
-	BoSTokenID        int      `json:"bos_token_id"`
-	EoSTokenID        int      `json:"eos_token_id"`
-	HeadDimension     int      `json:"head_dim"`
-	PaddingTokenID    int      `json:"pad_token_id"`
-	RopeFrequencyBase float64  `json:"rope_theta"`
-
-	Experts     int `json:"num_local_experts"`
-	ExpertsUsed int `json:"num_experts_per_tok"`
-
-	ByteOrder
-}
-
-type ByteOrder interface {
-	binary.ByteOrder
-	binary.AppendByteOrder
-}
-
-type ModelArch interface {
-	GetTensors() error
-	LoadVocab() error
-	WriteGGUF() (string, error)
-}
-
-type ModelFormat interface {
-	GetLayerName(string) (string, error)
-	GetTensors(string, *Params) ([]llm.Tensor, error)
-	GetParams(string) (*Params, error)
-	GetModelArch(string, string, *Params) (ModelArch, error)
-}
-
-type ModelData struct {
-	Path    string
-	Name    string
-	Params  *Params
-	Vocab   *Vocab
-	Tensors []llm.Tensor
-	Format  ModelFormat
-}
-
-func GetModelFormat(dirname string) (ModelFormat, error) {
-	files, err := filepath.Glob(filepath.Join(dirname, "*"))
-	if err != nil {
-		return nil, err
-	}
-
-	for _, fn := range files {
-		slog.Debug(fmt.Sprintf("file = %s", fn))
-		if strings.HasSuffix(fn, ".safetensors") {
-			return &SafetensorFormat{}, nil
-		} else if strings.HasSuffix(fn, ".bin") {
-			slog.Debug("model is torch")
-			return &TorchFormat{}, nil
-		}
-	}
-
-	return nil, fmt.Errorf("couldn't determine model format")
-}
-
-// Details on gguf's tokenizer can be found at:
-// https://github.com/ggerganov/ggml/blob/master/docs/gguf.md#tokenizer
-type Vocab struct {
-	Tokens []string
-	Scores []float32
-	Types  []int32
-}
-
-func LoadSentencePieceTokens(dirpath string, params *Params) (*Vocab, error) {
-	slog.Info(fmt.Sprintf("reading vocab from %s", filepath.Join(dirpath, "tokenizer.model")))
-	in, err := os.ReadFile(filepath.Join(dirpath, "tokenizer.model"))
-	if err != nil {
-		return nil, err
-	}
-
-	// To regenerate sentencepiece from the protobufs use:
-	// protoc -I=./ --go_out=./ sentencepiece_model.proto
-	modelProto := &sentencepiece.ModelProto{}
-	if err := proto.Unmarshal(in, modelProto); err != nil {
-		return nil, err
-	}
-
-	v := &Vocab{
-		Tokens: make([]string, 0),
-		Scores: make([]float32, 0),
-		Types:  make([]int32, 0),
-	}
-
-	pieces := modelProto.GetPieces()
-	for _, p := range pieces {
-		v.Tokens = append(v.Tokens, p.GetPiece())
-		v.Scores = append(v.Scores, p.GetScore())
-		t := p.GetType()
-		switch t {
-		case sentencepiece.ModelProto_SentencePiece_UNKNOWN:
-		case sentencepiece.ModelProto_SentencePiece_CONTROL:
-		case sentencepiece.ModelProto_SentencePiece_UNUSED:
-		case sentencepiece.ModelProto_SentencePiece_BYTE:
-		default:
-			t = sentencepiece.ModelProto_SentencePiece_NORMAL
-		}
-		v.Types = append(v.Types, int32(t))
-	}
-
-	slog.Info(fmt.Sprintf("vocab size: %d", len(v.Tokens)))
-
-	// add any additional tokens
-	addIn, err := os.ReadFile(filepath.Join(dirpath, "added_tokens.json"))
-	if os.IsNotExist(err) {
-		return v, nil
-	} else if err != nil {
-		return nil, err
-	}
-
-	slog.Info("reading user defined tokens")
-
-	var extraTokenData map[string]int
-	if err := json.Unmarshal(addIn, &extraTokenData); err != nil {
-		return nil, err
-	}
-
-	type token struct {
-		key string
-		pos int
-	}
-
-	extraTokens := make([]token, 0)
-	for k, id := range extraTokenData {
-		extraTokens = append(extraTokens, token{k, id})
-	}
-
-	slices.SortFunc(extraTokens, func(a, b token) int {
-		return cmp.Compare(a.pos, b.pos)
-	})
-
-	numToks := len(v.Tokens)
-
-	for cnt, t := range extraTokens {
-		// the token id should match the specific index for the total number of tokens
-		if t.pos != cnt+numToks {
-			return nil, fmt.Errorf("token ID '%d' for '%s' doesn't match total token size", t.pos, t.key)
-		}
-		v.Tokens = append(v.Tokens, t.key)
-		v.Scores = append(v.Scores, -1000.0)
-		v.Types = append(v.Types, int32(llm.GGUFTokenUserDefined))
-	}
-	slog.Info(fmt.Sprintf("vocab size w/ extra tokens: %d", len(v.Tokens)))
-
-	if params.VocabSize > len(v.Tokens) {
-		missingTokens := params.VocabSize - len(v.Tokens)
-		slog.Warn(fmt.Sprintf("vocab is missing %d tokens", missingTokens))
-		for cnt := 0; cnt < missingTokens; cnt++ {
-			v.Tokens = append(v.Tokens, fmt.Sprintf("<dummy%05d>", cnt+1))
-			v.Scores = append(v.Scores, -1)
-			v.Types = append(v.Types, int32(llm.GGUFTokenUserDefined))
-		}
-	}
-
-	return v, nil
-}
@@ -1,137 +0,0 @@
-package convert
-
-import (
-	"encoding/binary"
-	"fmt"
-	"io"
-	"log/slog"
-	"os"
-	"strings"
-
-	"github.com/d4l3k/go-bfloat16"
-	"github.com/pdevine/tensor"
-	"github.com/pdevine/tensor/native"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type GemmaModel struct {
-	ModelData
-}
-
-func gemmaLayerHandler(w io.Writer, r safetensorWriterTo, f *os.File) error {
-	slog.Debug(fmt.Sprintf("converting '%s'", r.t.Name))
-
-	data := make([]byte, r.end-r.start)
-	if err := binary.Read(f, r.bo, data); err != nil {
-		return err
-	}
-
-	tDataF32 := bfloat16.DecodeFloat32(data)
-
-	var err error
-	tDataF32, err = addOnes(tDataF32, int(r.t.Shape[0]))
-	if err != nil {
-		return err
-	}
-
-	if err := binary.Write(w, r.bo, tDataF32); err != nil {
-		return err
-	}
-	return nil
-}
-
-func addOnes(data []float32, vectorSize int) ([]float32, error) {
-	n := tensor.New(tensor.WithShape(vectorSize), tensor.WithBacking(data))
-	ones := tensor.Ones(tensor.Float32, vectorSize)
-
-	var err error
-	n, err = n.Add(ones)
-	if err != nil {
-		return []float32{}, err
-	}
-
-	newN, err := native.SelectF32(n, 0)
-	if err != nil {
-		return []float32{}, err
-	}
-
-	var fullTensor []float32
-	for _, v := range newN {
-		fullTensor = append(fullTensor, v...)
-	}
-
-	return fullTensor, nil
-}
-
-func (m *GemmaModel) GetTensors() error {
-	t, err := m.Format.GetTensors(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-
-	slog.Debug(fmt.Sprintf("Total tensors: %d", len(t)))
-
-	m.Tensors = []llm.Tensor{}
-	for _, l := range t {
-		if strings.HasSuffix(l.Name, "norm.weight") {
-			wt := l.WriterTo.(safetensorWriterTo)
-			wt.handler = gemmaLayerHandler
-			l.WriterTo = wt
-		}
-		m.Tensors = append(m.Tensors, l)
-	}
-
-	return nil
-}
-
-func (m *GemmaModel) LoadVocab() error {
-	v, err := LoadSentencePieceTokens(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-	m.Vocab = v
-	return nil
-}
-
-func (m *GemmaModel) WriteGGUF() (string, error) {
-	kv := llm.KV{
-		"general.architecture":                   "gemma",
-		"general.name":                           m.Name,
-		"gemma.context_length":                   uint32(m.Params.ContextSize),
-		"gemma.embedding_length":                 uint32(m.Params.HiddenSize),
-		"gemma.block_count":                      uint32(m.Params.HiddenLayers),
-		"gemma.feed_forward_length":              uint32(m.Params.IntermediateSize),
-		"gemma.attention.head_count":             uint32(m.Params.AttentionHeads),
-		"gemma.attention.head_count_kv":          uint32(m.Params.KeyValHeads),
-		"gemma.attention.layer_norm_rms_epsilon": float32(m.Params.NormEPS),
-		"gemma.attention.key_length":             uint32(m.Params.HeadDimension),
-		"gemma.attention.value_length":           uint32(m.Params.HeadDimension),
-		"general.file_type":                      uint32(1),
-		"tokenizer.ggml.model":                   "llama",
-
-		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
-		"tokenizer.ggml.scores":     m.Vocab.Scores,
-		"tokenizer.ggml.token_type": m.Vocab.Types,
-
-		"tokenizer.ggml.bos_token_id":     uint32(m.Params.BoSTokenID),
-		"tokenizer.ggml.eos_token_id":     uint32(m.Params.EoSTokenID),
-		"tokenizer.ggml.padding_token_id": uint32(m.Params.PaddingTokenID),
-		"tokenizer.ggml.unknown_token_id": uint32(3),
-		"tokenizer.ggml.add_bos_token":    true,
-		"tokenizer.ggml.add_eos_token":    false,
-	}
-
-	f, err := os.CreateTemp("", "ollama-gguf")
-	if err != nil {
-		return "", err
-	}
-	defer f.Close()
-
-	mod := llm.NewGGUFV3(m.Params.ByteOrder)
-	if err := mod.Encode(f, kv, m.Tensors); err != nil {
-		return "", err
-	}
-
-	return f.Name(), nil
-}
@@ -1,176 +0,0 @@
-package convert
-
-import (
-	"encoding/binary"
-	"fmt"
-	"io"
-	"log/slog"
-	"os"
-	"regexp"
-	"strings"
-
-	"github.com/nlpodyssey/gopickle/pytorch"
-	"github.com/pdevine/tensor"
-	"github.com/pdevine/tensor/native"
-	"github.com/x448/float16"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type LlamaModel struct {
-	ModelData
-}
-
-func llamaLayerHandler(w io.Writer, r torchWriterTo) error {
-	slog.Debug(fmt.Sprintf("repacking layer '%s'", r.t.Name))
-
-	data := r.storage.(*pytorch.HalfStorage).Data
-	tData := make([]uint16, len(data))
-	for cnt, v := range data {
-		tData[cnt] = uint16(float16.Fromfloat32(v))
-	}
-
-	var err error
-	var heads uint32
-	if strings.Contains(r.t.Name, "attn_q") {
-		heads = uint32(r.params.AttentionHeads)
-	} else if strings.Contains(r.t.Name, "attn_k") {
-		heads = uint32(r.params.KeyValHeads)
-		if heads == 0 {
-			heads = uint32(r.params.AttentionHeads)
-		}
-	} else {
-		return fmt.Errorf("unknown layer type")
-	}
-
-	slog.Debug(fmt.Sprintf("heads = %d", heads))
-
-	tData, err = llamaRepack(tData, int(heads), r.t.Shape)
-	if err != nil {
-		return err
-	}
-
-	if err = binary.Write(w, r.bo, tData); err != nil {
-		return err
-	}
-	return nil
-}
-
-func llamaRepack(data []uint16, heads int, shape []uint64) ([]uint16, error) {
-	n := tensor.New(tensor.WithShape(int(shape[0]), int(shape[1])), tensor.WithBacking(data))
-	origShape := n.Shape().Clone()
-
-	// reshape the tensor and swap axes 1 and 2 to unpack the layer for gguf
-	if err := n.Reshape(heads, 2, origShape[0]/heads/2, origShape[1]); err != nil {
-		return nil, err
-	}
-
-	if err := n.T(0, 2, 1, 3); err != nil {
-		return nil, err
-	}
-
-	if err := n.Reshape(origShape...); err != nil {
-		return nil, err
-	}
-
-	if err := n.Transpose(); err != nil {
-		return nil, err
-	}
-	newN, err := native.SelectU16(n, 1)
-	if err != nil {
-		return nil, err
-	}
-
-	var fullTensor []uint16
-	for _, v := range newN {
-		fullTensor = append(fullTensor, v...)
-	}
-	return fullTensor, nil
-}
-
-func (m *LlamaModel) GetTensors() error {
-	t, err := m.Format.GetTensors(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-
-	m.Tensors = []llm.Tensor{}
-
-	pattern := `^blk\.[0-9]+\.attn_(?P<layer>q|k)\.weight$`
-	re, err := regexp.Compile(pattern)
-	if err != nil {
-		return err
-	}
-
-	for _, l := range t {
-		matches := re.FindAllStringSubmatch(l.Name, -1)
-		if len(matches) > 0 {
-			slog.Debug(fmt.Sprintf("setting handler for: %s", l.Name))
-			wt := l.WriterTo.(torchWriterTo)
-			wt.handler = llamaLayerHandler
-			l.WriterTo = wt
-		}
-		m.Tensors = append(m.Tensors, l)
-	}
-
-	return nil
-}
-
-func (m *LlamaModel) LoadVocab() error {
-	var v *Vocab
-	var err error
-
-	slog.Debug("loading vocab")
-	v, err = LoadSentencePieceTokens(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-
-	slog.Debug("vocab loaded")
-
-	m.Vocab = v
-	return nil
-}
-
-func (m *LlamaModel) WriteGGUF() (string, error) {
-	kv := llm.KV{
-		"general.architecture":                   "llama",
-		"general.name":                           m.Name,
-		"llama.vocab_size":                       uint32(len(m.Vocab.Tokens)),
-		"llama.context_length":                   uint32(m.Params.ContextSize),
-		"llama.embedding_length":                 uint32(m.Params.HiddenSize),
-		"llama.block_count":                      uint32(m.Params.HiddenLayers),
-		"llama.feed_forward_length":              uint32(m.Params.IntermediateSize),
-		"llama.rope.dimension_count":             uint32(m.Params.HiddenSize / m.Params.AttentionHeads),
-		"llama.attention.head_count":             uint32(m.Params.AttentionHeads),
-		"llama.attention.head_count_kv":          uint32(m.Params.KeyValHeads),
-		"llama.attention.layer_norm_rms_epsilon": float32(m.Params.NormEPS),
-		"general.file_type":                      uint32(1),
-		"tokenizer.ggml.model":                   "llama",
-
-		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
-		"tokenizer.ggml.scores":     m.Vocab.Scores,
-		"tokenizer.ggml.token_type": m.Vocab.Types,
-
-		"tokenizer.ggml.bos_token_id":     uint32(m.Params.BoSTokenID),
-		"tokenizer.ggml.eos_token_id":     uint32(m.Params.EoSTokenID),
-		"tokenizer.ggml.unknown_token_id": uint32(0),
-		"tokenizer.ggml.add_bos_token":    true,
-		"tokenizer.ggml.add_eos_token":    false,
-	}
-
-	f, err := os.CreateTemp("", "ollama-gguf")
-	if err != nil {
-		return "", err
-	}
-	defer f.Close()
-
-	mod := llm.NewGGUFV3(m.Params.ByteOrder)
-	if err := mod.Encode(f, kv, m.Tensors); err != nil {
-		return "", err
-	}
-
-	slog.Debug(fmt.Sprintf("gguf file = %s", f.Name()))
-
-	return f.Name(), nil
-}
@@ -1,173 +0,0 @@
-package convert
-
-import (
-	"encoding/binary"
-	"fmt"
-	"io"
-	"os"
-	"regexp"
-	"strings"
-
-	"github.com/d4l3k/go-bfloat16"
-	"github.com/pdevine/tensor"
-	"github.com/pdevine/tensor/native"
-	"github.com/x448/float16"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type MistralModel struct {
-	ModelData
-}
-
-func mistralLayerHandler(w io.Writer, r safetensorWriterTo, f *os.File) error {
-	layerSize := r.end - r.start
-
-	var err error
-	tData := make([]uint16, layerSize/2)
-	if err = binary.Read(f, r.bo, tData); err != nil {
-		return err
-	}
-
-	var heads uint32
-	if strings.Contains(r.t.Name, "attn_q") {
-		heads = uint32(r.params.AttentionHeads)
-	} else if strings.Contains(r.t.Name, "attn_k") {
-		heads = uint32(r.params.KeyValHeads)
-		if heads == 0 {
-			heads = uint32(r.params.AttentionHeads)
-		}
-	} else {
-		return fmt.Errorf("unknown layer type")
-	}
-
-	tData, err = repack(tData, int(heads), r.t.Shape)
-	if err != nil {
-		return err
-	}
-
-	var buf []byte
-	for _, n := range tData {
-		buf = r.bo.AppendUint16(buf, n)
-	}
-
-	tempBuf := make([]uint16, len(tData))
-	tDataF32 := bfloat16.DecodeFloat32(buf)
-	for cnt, v := range tDataF32 {
-		tDataF16 := float16.Fromfloat32(v)
-		tempBuf[cnt] = uint16(tDataF16)
-	}
-
-	if err = binary.Write(w, r.bo, tempBuf); err != nil {
-		return err
-	}
-	return nil
-}
-
-func repack(data []uint16, heads int, shape []uint64) ([]uint16, error) {
-	n := tensor.New(tensor.WithShape(int(shape[0]), int(shape[1])), tensor.WithBacking(data))
-	origShape := n.Shape().Clone()
-
-	// reshape the tensor and swap axes 1 and 2 to unpack the layer for gguf
-	if err := n.Reshape(heads, 2, origShape[0]/heads/2, origShape[1]); err != nil {
-		return nil, err
-	}
-
-	if err := n.T(0, 2, 1, 3); err != nil {
-		return nil, err
-	}
-
-	if err := n.Reshape(origShape...); err != nil {
-		return nil, err
-	}
-
-	if err := n.Transpose(); err != nil {
-		return nil, err
-	}
-	newN, err := native.SelectU16(n, 1)
-	if err != nil {
-		return nil, err
-	}
-
-	var fullTensor []uint16
-	for _, v := range newN {
-		fullTensor = append(fullTensor, v...)
-	}
-	return fullTensor, nil
-}
-
-func (m *MistralModel) GetTensors() error {
-	t, err := m.Format.GetTensors(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-
-	m.Tensors = []llm.Tensor{}
-
-	pattern := `^blk\.[0-9]+\.attn_(?P<layer>q|k)\.weight$`
-	re, err := regexp.Compile(pattern)
-	if err != nil {
-		return err
-	}
-
-	for _, l := range t {
-		matches := re.FindAllStringSubmatch(l.Name, -1)
-		if len(matches) > 0 {
-			wt := l.WriterTo.(safetensorWriterTo)
-			wt.handler = mistralLayerHandler
-			l.WriterTo = wt
-		}
-		m.Tensors = append(m.Tensors, l)
-	}
-
-	return nil
-}
-
-func (m *MistralModel) LoadVocab() error {
-	v, err := LoadSentencePieceTokens(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-	m.Vocab = v
-	return nil
-}
-
-func (m *MistralModel) WriteGGUF() (string, error) {
-	kv := llm.KV{
-		"general.architecture":                   "llama",
-		"general.name":                           m.Name,
-		"llama.context_length":                   uint32(m.Params.ContextSize),
-		"llama.embedding_length":                 uint32(m.Params.HiddenSize),
-		"llama.block_count":                      uint32(m.Params.HiddenLayers),
-		"llama.feed_forward_length":              uint32(m.Params.IntermediateSize),
-		"llama.rope.dimension_count":             uint32(m.Params.HiddenSize / m.Params.AttentionHeads),
-		"llama.attention.head_count":             uint32(m.Params.AttentionHeads),
-		"llama.attention.head_count_kv":          uint32(m.Params.KeyValHeads),
-		"llama.attention.layer_norm_rms_epsilon": float32(m.Params.NormEPS),
-		"general.file_type":                      uint32(1),
-		"tokenizer.ggml.model":                   "llama",
-
-		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
-		"tokenizer.ggml.scores":     m.Vocab.Scores,
-		"tokenizer.ggml.token_type": m.Vocab.Types,
-
-		"tokenizer.ggml.bos_token_id":     uint32(m.Params.BoSTokenID),
-		"tokenizer.ggml.eos_token_id":     uint32(m.Params.EoSTokenID),
-		"tokenizer.ggml.add_bos_token":    true,
-		"tokenizer.ggml.add_eos_token":    false,
-		"tokenizer.ggml.unknown_token_id": uint32(0),
-	}
-
-	f, err := os.CreateTemp("", "ollama-gguf")
-	if err != nil {
-		return "", err
-	}
-	defer f.Close()
-
-	mod := llm.NewGGUFV3(m.Params.ByteOrder)
-	if err := mod.Encode(f, kv, m.Tensors); err != nil {
-		return "", err
-	}
-
-	return f.Name(), nil
-}
@@ -1,96 +0,0 @@
-package convert
-
-import (
-	"os"
-	"regexp"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type MixtralModel struct {
-	ModelData
-}
-
-func (m *MixtralModel) GetTensors() error {
-	t, err := m.Format.GetTensors(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-
-	m.Tensors = []llm.Tensor{}
-
-	pattern := `^blk\.[0-9]+\.attn_(?P<layer>q|k)\.weight$`
-	re, err := regexp.Compile(pattern)
-	if err != nil {
-		return err
-	}
-
-	for _, l := range t {
-		matches := re.FindAllStringSubmatch(l.Name, -1)
-		if len(matches) > 0 {
-			wt := l.WriterTo.(safetensorWriterTo)
-			wt.handler = mistralLayerHandler
-			l.WriterTo = wt
-		}
-		m.Tensors = append(m.Tensors, l)
-	}
-
-	return nil
-}
-
-func (m *MixtralModel) LoadVocab() error {
-	v, err := LoadSentencePieceTokens(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-	m.Vocab = v
-	return nil
-}
-
-func (m *MixtralModel) WriteGGUF() (string, error) {
-	kv := llm.KV{
-		"general.architecture":          "llama",
-		"general.name":                  m.Name,
-		"llama.block_count":             uint32(m.Params.HiddenLayers),
-		"llama.context_length":          uint32(m.Params.ContextSize),
-		"llama.embedding_length":        uint32(m.Params.HiddenSize),
-		"llama.feed_forward_length":     uint32(m.Params.IntermediateSize),
-		"llama.attention.head_count":    uint32(m.Params.AttentionHeads),
-		"llama.attention.head_count_kv": uint32(m.Params.KeyValHeads),
-
-		"llama.rope.freq_base":                   float32(m.Params.RopeFrequencyBase),
-		"llama.attention.layer_norm_rms_epsilon": float32(m.Params.NormEPS),
-
-		"llama.expert_count":      uint32(m.Params.Experts),
-		"llama.expert_used_count": uint32(m.Params.ExpertsUsed),
-
-		"llama.vocab_size":           uint32(len(m.Vocab.Tokens)),
-		"llama.rope.dimension_count": uint32(m.Params.HiddenSize / m.Params.AttentionHeads),
-
-		"general.file_type":    uint32(1),
-		"tokenizer.ggml.model": "llama",
-
-		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
-		"tokenizer.ggml.scores":     m.Vocab.Scores,
-		"tokenizer.ggml.token_type": m.Vocab.Types,
-
-		"tokenizer.ggml.bos_token_id":     uint32(m.Params.BoSTokenID),
-		"tokenizer.ggml.eos_token_id":     uint32(m.Params.EoSTokenID),
-		"tokenizer.ggml.unknown_token_id": uint32(0),
-		"tokenizer.ggml.add_bos_token":    true,
-		"tokenizer.ggml.add_eos_token":    false,
-	}
-
-	f, err := os.CreateTemp("", "ollama-gguf")
-	if err != nil {
-		return "", err
-	}
-	defer f.Close()
-
-	mod := llm.NewGGUFV3(m.Params.ByteOrder)
-	if err := mod.Encode(f, kv, m.Tensors); err != nil {
-		return "", err
-	}
-
-	return f.Name(), nil
-}
@@ -1,317 +0,0 @@
-package convert
-
-import (
-	"bytes"
-	"encoding/binary"
-	"encoding/json"
-	"fmt"
-	"io"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"regexp"
-	"slices"
-
-	"github.com/d4l3k/go-bfloat16"
-	"github.com/mitchellh/mapstructure"
-	"github.com/x448/float16"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type safetensorWriterTo struct {
-	t *llm.Tensor
-
-	params *Params
-	bo     ByteOrder
-
-	filename string
-
-	start, end, padding uint64
-	handler             func(w io.Writer, r safetensorWriterTo, f *os.File) error
-}
-
-type tensorMetaData struct {
-	Type    string `mapstructure:"dtype"`
-	Shape   []int  `mapstructure:"shape"`
-	Offsets []int  `mapstructure:"data_offsets"`
-}
-
-type SafetensorFormat struct{}
-
-func (m *SafetensorFormat) GetTensors(dirpath string, params *Params) ([]llm.Tensor, error) {
-	slog.Debug("getting tensor data")
-	var tensors []llm.Tensor
-	files, err := filepath.Glob(filepath.Join(dirpath, "/model-*.safetensors"))
-	if err != nil {
-		return nil, err
-	}
-
-	var offset uint64
-	for _, f := range files {
-		var t []llm.Tensor
-		var err error
-		t, offset, err = m.readTensors(f, offset, params)
-		if err != nil {
-			slog.Error("%v", err)
-			return nil, err
-		}
-		tensors = append(tensors, t...)
-	}
-	slog.Debug(fmt.Sprintf("all tensors = %d", len(tensors)))
-	return tensors, nil
-}
-
-func (m *SafetensorFormat) readTensors(fn string, offset uint64, params *Params) ([]llm.Tensor, uint64, error) {
-	f, err := os.Open(fn)
-	if err != nil {
-		return nil, 0, err
-	}
-	defer f.Close()
-
-	var jsonSize uint64
-	if err := binary.Read(f, binary.LittleEndian, &jsonSize); err != nil {
-		return nil, 0, err
-	}
-
-	buf := make([]byte, jsonSize)
-	_, err = io.ReadFull(f, buf)
-	if err != nil {
-		return nil, 0, err
-	}
-
-	d := json.NewDecoder(bytes.NewBuffer(buf))
-	d.UseNumber()
-	var parsed map[string]interface{}
-	if err = d.Decode(&parsed); err != nil {
-		return nil, 0, err
-	}
-
-	var keys []string
-	for k := range parsed {
-		keys = append(keys, k)
-	}
-
-	slices.Sort(keys)
-	slog.Info("converting layers")
-
-	var tensors []llm.Tensor
-	for _, k := range keys {
-		vals := parsed[k].(map[string]interface{})
-		var data tensorMetaData
-		if err = mapstructure.Decode(vals, &data); err != nil {
-			slog.Error("couldn't decode properly")
-			return nil, 0, err
-		}
-
-		var size uint64
-		var kind uint32
-		switch len(data.Shape) {
-		case 0:
-			// metadata
-			continue
-		case 1:
-			// convert to float32
-			kind = 0
-			size = uint64(data.Shape[0] * 4)
-		case 2:
-			// convert to float16
-			kind = 1
-			size = uint64(data.Shape[0] * data.Shape[1] * 2)
-		}
-
-		ggufName, err := m.GetLayerName(k)
-		if err != nil {
-			slog.Error("%v", err)
-			return nil, 0, err
-		}
-
-		shape := []uint64{0, 0, 0, 0}
-		for i := range data.Shape {
-			shape[i] = uint64(data.Shape[i])
-		}
-
-		t := llm.Tensor{
-			Name:   ggufName,
-			Kind:   kind,
-			Offset: offset,
-			Shape:  shape[:],
-		}
-
-		t.WriterTo = safetensorWriterTo{
-			t:        &t,
-			params:   params,
-			bo:       params.ByteOrder,
-			filename: fn,
-			start:    uint64(data.Offsets[0]),
-			end:      uint64(data.Offsets[1]),
-			padding:  8 + jsonSize,
-		}
-
-		offset += size
-		tensors = append(tensors, t)
-	}
-
-	slog.Debug(fmt.Sprintf("total tensors for file = %d", len(tensors)))
-	slog.Debug(fmt.Sprintf("offset = %d", offset))
-
-	return tensors, offset, nil
-}
-
-func (m *SafetensorFormat) GetParams(dirpath string) (*Params, error) {
-	f, err := os.Open(filepath.Join(dirpath, "config.json"))
-	if err != nil {
-		return nil, err
-	}
-	defer f.Close()
-
-	var params Params
-
-	d := json.NewDecoder(f)
-	err = d.Decode(&params)
-	if err != nil {
-		return nil, err
-	}
-
-	params.ByteOrder = binary.LittleEndian
-	return &params, nil
-}
-
-func (m *SafetensorFormat) GetLayerName(n string) (string, error) {
-	directMap := map[string]string{
-		"model.embed_tokens.weight": "token_embd.weight",
-		"lm_head.weight":            "output.weight",
-		"model.norm.weight":         "output_norm.weight",
-	}
-
-	tMap := map[string]string{
-		"model.layers.(\\d+).input_layernorm.weight":                    "blk.$1.attn_norm.weight",
-		"model.layers.(\\d+).mlp.down_proj.weight":                      "blk.$1.ffn_down.weight",
-		"model.layers.(\\d+).mlp.gate_proj.weight":                      "blk.$1.ffn_gate.weight",
-		"model.layers.(\\d+).mlp.up_proj.weight":                        "blk.$1.ffn_up.weight",
-		"model.layers.(\\d+).post_attention_layernorm.weight":           "blk.$1.ffn_norm.weight",
-		"model.layers.(\\d+).self_attn.k_proj.weight":                   "blk.$1.attn_k.weight",
-		"model.layers.(\\d+).self_attn.o_proj.weight":                   "blk.$1.attn_output.weight",
-		"model.layers.(\\d+).self_attn.q_proj.weight":                   "blk.$1.attn_q.weight",
-		"model.layers.(\\d+).self_attn.v_proj.weight":                   "blk.$1.attn_v.weight",
-		"model.layers.(\\d+).block_sparse_moe.gate.weight":              "blk.$1.ffn_gate_inp.weight",
-		"model.layers.(\\d+).block_sparse_moe.experts.(\\d+).w1.weight": "blk.$1.ffn_gate.$2.weight",
-		"model.layers.(\\d+).block_sparse_moe.experts.(\\d+).w2.weight": "blk.$1.ffn_down.$2.weight",
-		"model.layers.(\\d+).block_sparse_moe.experts.(\\d+).w3.weight": "blk.$1.ffn_up.$2.weight",
-	}
-
-	v, ok := directMap[n]
-	if ok {
-		return v, nil
-	}
-
-	// quick hack to rename the layers to gguf format
-	for k, v := range tMap {
-		re := regexp.MustCompile(k)
-		newName := re.ReplaceAllString(n, v)
-		if newName != n {
-			return newName, nil
-		}
-	}
-
-	return "", fmt.Errorf("couldn't find a layer name for '%s'", n)
-}
-
-func (r safetensorWriterTo) WriteTo(w io.Writer) (n int64, err error) {
-	f, err := os.Open(r.filename)
-	if err != nil {
-		return 0, err
-	}
-	defer f.Close()
-
-	if _, err = f.Seek(int64(r.padding+r.start), 0); err != nil {
-		return 0, err
-	}
-
-	// use the handler if one is present
-	if r.handler != nil {
-		return 0, r.handler(w, r, f)
-	}
-
-	remaining := r.end - r.start
-
-	bufSize := uint64(10240)
-	var finished bool
-	for {
-		data := make([]byte, min(bufSize, remaining))
-
-		b, err := io.ReadFull(f, data)
-		remaining -= uint64(b)
-
-		if err == io.EOF || remaining <= 0 {
-			finished = true
-		} else if err != nil {
-			return 0, err
-		}
-
-		// convert bfloat16 -> ieee float32
-		tDataF32 := bfloat16.DecodeFloat32(data)
-
-		switch r.t.Kind {
-		case 0:
-			if err := binary.Write(w, r.bo, tDataF32); err != nil {
-				return 0, err
-			}
-		case 1:
-			// convert float32 -> float16
-			tempBuf := make([]uint16, len(data)/2)
-			for cnt, v := range tDataF32 {
-				tDataF16 := float16.Fromfloat32(v)
-				tempBuf[cnt] = uint16(tDataF16)
-			}
-			if err := binary.Write(w, r.bo, tempBuf); err != nil {
-				return 0, err
-			}
-		}
-		if finished {
-			break
-		}
-	}
-	return 0, nil
-}
-
-func (m *SafetensorFormat) GetModelArch(name, dirPath string, params *Params) (ModelArch, error) {
-	switch len(params.Architectures) {
-	case 0:
-		return nil, fmt.Errorf("No architecture specified to convert")
-	case 1:
-		switch params.Architectures[0] {
-		case "MistralForCausalLM":
-			return &MistralModel{
-				ModelData{
-					Name:   name,
-					Path:   dirPath,
-					Params: params,
-					Format: m,
-				},
-			}, nil
-		case "MixtralForCausalLM":
-			return &MixtralModel{
-				ModelData{
-					Name:   name,
-					Path:   dirPath,
-					Params: params,
-					Format: m,
-				},
-			}, nil
-		case "GemmaForCausalLM":
-			return &GemmaModel{
-				ModelData{
-					Name:   name,
-					Path:   dirPath,
-					Params: params,
-					Format: m,
-				},
-			}, nil
-		default:
-			return nil, fmt.Errorf("Models based on '%s' are not yet supported", params.Architectures[0])
-		}
-	}
-
-	return nil, fmt.Errorf("Unknown error")
-}