improve scratch buffer estimates

2024-01-19 13:24:24 -05:00
286 changed files with 6271 additions and 60195 deletions
--- a/.dockerignore
+++ b/.dockerignore
@@ -1,9 +1,8 @@
 .vscode
 ollama
 app
-macapp
 dist
 llm/llama.cpp
 .env
 .cache
-test_data
+test_data
--- a/.gitattributes
+++ b/.gitattributes
@@ -1 +0,0 @@
-llm/ext_server/* linguist-vendored
--- a/.github/ISSUE_TEMPLATE/10_bug_report.yml
+++ b/.github/ISSUE_TEMPLATE/10_bug_report.yml
@@ -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
--- a/.github/ISSUE_TEMPLATE/20_feature_request.md
+++ b/.github/ISSUE_TEMPLATE/20_feature_request.md
@@ -1,6 +0,0 @@
---
-name: Feature request
-about: Request a new feature
-labels: feature request
---
-
--- a/.github/ISSUE_TEMPLATE/30_model_request.md
+++ b/.github/ISSUE_TEMPLATE/30_model_request.md
@@ -1,5 +0,0 @@
---
-name: Model request
-about: Request support for a new model to be added to Ollama
-labels: model request
---
--- a/.github/ISSUE_TEMPLATE/config.yml
+++ b/.github/ISSUE_TEMPLATE/config.yml
@@ -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
--- a/.github/workflows/latest.yaml
+++ b/.github/workflows/latest.yaml
@@ -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
--- a/.github/workflows/release.yaml
+++ b/.github/workflows/release.yaml
@@ -1,474 +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
-          security set-keychain-settings -lut 3600 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
--- a/.github/workflows/test.yaml
+++ b/.github/workflows/test.yaml
@@ -1,59 +1,18 @@
 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]
+        os: [ubuntu-latest, macos-latest, windows-latest]
        arch: [amd64, arm64]
        exclude:
          - os: ubuntu-latest
            arch: arm64
-          - os: windows-2019
+          - os: windows-latest
            arch: arm64
    runs-on: ${{ matrix.os }}
    env:
@@ -62,258 +21,86 @@ jobs:
      - uses: actions/checkout@v4
      - uses: actions/setup-go@v5
        with:
-          go-version-file: go.mod
+          go-version: '1.21'
          cache: true
+      - if: ${{ startsWith(matrix.os, 'windows-') }}
+        shell: pwsh
+        run: |
+          $path = vswhere -latest -products * -requires Microsoft.VisualStudio.Component.VC.Tools.x86.x64 -property installationPath
+          if ($path) {
+              $path = join-path $path 'Common7\Tools\vsdevcmd.bat'
+              if (test-path $path) {
+                  cmd /s /c """$path"" $args && set" | where { $_ -match '(\w+)=(.*)' } | foreach {
+                      echo "$($Matches[1])=$($Matches[2])" | Out-File -FilePath $Env:GITHUB_ENV -Encoding utf8 -Append
+                  }
+              }
+          }
+
+          echo "C:\Program Files\Git\usr\bin" | Out-File -FilePath $Env:GITHUB_PATH -Encoding utf8 -Append
      - 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?
-
+            llm/llama.cpp/build/**/lib/*
  lint:
+    needs: generate
    strategy:
      matrix:
-        os: [ubuntu-latest, macos-latest, windows-2019]
+        os: [ubuntu-latest, macos-latest, windows-latest]
        arch: [amd64, arm64]
        exclude:
          - os: ubuntu-latest
            arch: arm64
-          - os: windows-2019
+          - os: windows-latest
            arch: arm64
          - os: macos-latest
            arch: amd64
    runs-on: ${{ matrix.os }}
    env:
      GOARCH: ${{ matrix.arch }}
-      CGO_ENABLED: '1'
+      CGO_ENABLED: "1"
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-go@v5
        with:
-          go-version-file: go.mod
+          go-version: '1.21'
          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
+      - uses: actions/download-artifact@v4
        with:
-          args: --timeout 8m0s -v
+          name: ${{ matrix.os }}-${{ matrix.arch }}-libraries
+          path: llm/llama.cpp/build
+      - uses: golangci/golangci-lint-action@v3
  test:
+    needs: generate
    strategy:
      matrix:
-        os: [ubuntu-latest, macos-latest, windows-2019]
+        os: [ubuntu-latest, macos-latest, windows-latest]
        arch: [amd64]
        exclude:
          - os: ubuntu-latest
            arch: arm64
-          - os: windows-2019
+          - os: windows-latest
            arch: arm64
    runs-on: ${{ matrix.os }}
    env:
      GOARCH: ${{ matrix.arch }}
-      CGO_ENABLED: '1'
-      OLLAMA_CPU_TARGET: 'static'
+      CGO_ENABLED: "1"
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-go@v5
        with:
-          go-version-file: go.mod
+          go-version: '1.21'
          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 get
+      - uses: actions/download-artifact@v4
+        with:
+          name: ${{ matrix.os }}-${{ matrix.arch }}-libraries
+          path: llm/llama.cpp/build
      - run: go build
      - run: go test -v ./...
-      - uses: actions/upload-artifact@v4
-        with:
-          name: ${{ matrix.os }}-binaries
-          path: ollama
--- a/.gitignore
+++ b/.gitignore
@@ -9,7 +9,4 @@ ggml-metal.metal
 .cache
 *.exe
 .idea
-test_data
-*.crt
-llm/build
-__debug_bin*
+test_data
--- a/.golangci.yaml
+++ b/.golangci.yaml
@@ -15,3 +15,13 @@ linters:
    - misspell
    - nilerr
    - unused
+linters-settings:
+  errcheck:
+    # exclude the following functions since we don't generally
+    # need to be concerned with the returned errors
+    exclude-functions:
+      - encoding/binary.Read
+      - (*os.File).Seek
+      - (*bufio.Writer).WriteString
+      - (*github.com/spf13/pflag.FlagSet).Set
+      - (*github.com/jmorganca/ollama/llm.readSeekOffset).Seek
--- a/145
+++ b/145
@@ -1,144 +1,29 @@
-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
+FROM nvidia/cuda:11.8.0-devel-ubuntu22.04

-# 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
+ARG TARGETARCH
+ARG GOFLAGS="'-ldflags=-w -s'"

-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:$PATH
-COPY --from=llm-code / /go/src/github.com/ollama/ollama/
-WORKDIR /go/src/github.com/ollama/ollama/llm/generate
-ARG CGO_CFLAGS
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+WORKDIR /go/src/github.com/jmorganca/ollama
+RUN apt-get update && apt-get install -y git build-essential cmake
+ADD https://dl.google.com/go/go1.21.3.linux-$TARGETARCH.tar.gz /tmp/go1.21.3.tar.gz
+RUN mkdir -p /usr/local && tar xz -C /usr/local </tmp/go1.21.3.tar.gz

-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:$PATH
-COPY --from=llm-code / /go/src/github.com/ollama/ollama/
-ARG OLLAMA_CUSTOM_CPU_DEFS
-ARG CGO_CFLAGS
-WORKDIR /go/src/github.com/ollama/ollama/llm/generate
-
-FROM --platform=linux/arm64 cpu-builder-arm64 AS static-build-arm64
-RUN OLLAMA_CPU_TARGET="static" sh gen_linux.sh
-FROM --platform=linux/arm64 cpu-builder-arm64 AS cpu-build-arm64
-RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu" sh gen_linux.sh
-
-
-# Intermediate stage used for ./scripts/build_linux.sh
-FROM --platform=linux/amd64 cpu-build-amd64 AS build-amd64
-ENV CGO_ENABLED 1
-WORKDIR /go/src/github.com/ollama/ollama
 COPY . .
-COPY --from=static-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=cpu_avx-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=cpu_avx2-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=cuda-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=rocm-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
-COPY --from=rocm-build-amd64 /go/src/github.com/ollama/ollama/dist/deps/ ./dist/deps/
-ARG GOFLAGS
-ARG CGO_CFLAGS
-RUN go build -trimpath .
+ENV GOARCH=$TARGETARCH
+ENV GOFLAGS=$GOFLAGS
+RUN /usr/local/go/bin/go generate ./... \
+    && /usr/local/go/bin/go build .

-# 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
+FROM ubuntu:22.04
 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
+COPY --from=0 /go/src/github.com/jmorganca/ollama/ollama /bin/ollama
 EXPOSE 11434
 ENV OLLAMA_HOST 0.0.0.0

-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
+# set some environment variable for better NVIDIA compatibility
+ENV PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:/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"]
--- a/Dockerfile.build
+++ b/Dockerfile.build
@@ -0,0 +1,99 @@
+ARG GOLANG_VERSION=1.21.3
+ARG CMAKE_VERSION=3.22.1
+ARG CUDA_VERSION=11.3.1
+
+# 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:$CUDA_VERSION-devel-centos7 AS cuda-build-amd64
+ARG CMAKE_VERSION
+ARG CGO_CFLAGS
+COPY ./scripts/rh_linux_deps.sh /
+RUN CMAKE_VERSION=${CMAKE_VERSION} sh /rh_linux_deps.sh
+ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
+COPY --from=llm-code / /go/src/github.com/jmorganca/ollama/
+WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+RUN OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+
+FROM --platform=linux/arm64 nvidia/cuda:$CUDA_VERSION-devel-rockylinux8 AS cuda-build-arm64
+ARG CMAKE_VERSION
+ARG CGO_CFLAGS
+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/jmorganca/ollama/
+WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+RUN OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+
+FROM --platform=linux/amd64 rocm/dev-centos-7:5.7.1-complete AS rocm-5-build-amd64
+ARG CMAKE_VERSION
+ARG CGO_CFLAGS
+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/jmorganca/ollama/
+WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+RUN OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+
+FROM --platform=linux/amd64 rocm/dev-centos-7:6.0-complete AS rocm-6-build-amd64
+ARG CMAKE_VERSION
+ARG CGO_CFLAGS
+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/jmorganca/ollama/
+WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+RUN OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+
+FROM --platform=linux/amd64 centos:7 AS cpu-build-amd64
+ARG CMAKE_VERSION
+ARG GOLANG_VERSION
+ARG OLLAMA_CUSTOM_CPU_DEFS
+ARG CGO_CFLAGS
+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:$PATH
+COPY --from=llm-code / /go/src/github.com/jmorganca/ollama/
+WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+RUN sh gen_linux.sh
+
+FROM --platform=linux/arm64 centos:7 AS cpu-build-arm64
+ARG CMAKE_VERSION
+ARG GOLANG_VERSION
+ARG OLLAMA_CUSTOM_CPU_DEFS
+ARG CGO_CFLAGS
+COPY ./scripts/rh_linux_deps.sh /
+RUN CMAKE_VERSION=${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/jmorganca/ollama/
+WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+RUN sh gen_linux.sh
+
+
+FROM --platform=linux/amd64 cpu-build-amd64 AS build-amd64
+ENV CGO_ENABLED 1
+ARG GOFLAGS
+ARG CGO_CFLAGS
+WORKDIR /go/src/github.com/jmorganca/ollama
+COPY . .
+COPY --from=cuda-build-amd64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
+COPY --from=rocm-5-build-amd64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
+COPY --from=rocm-6-build-amd64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
+RUN go build .
+
+FROM --platform=linux/arm64 cpu-build-arm64 AS build-arm64
+ENV CGO_ENABLED 1
+ARG GOLANG_VERSION
+ARG GOFLAGS
+ARG CGO_CFLAGS
+WORKDIR /go/src/github.com/jmorganca/ollama
+COPY . .
+COPY --from=cuda-build-arm64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
+RUN go build .
+
+FROM build-$TARGETARCH
--- a/README.md
+++ b/README.md
@@ -1,5 +1,8 @@
 <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
@@ -10,59 +13,53 @@ Get up and running with large language models locally.

 ### macOS

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

-### Windows preview
+### Windows

-[Download](https://ollama.com/download/OllamaSetup.exe)
+Coming soon! For now, you can install Ollama on Windows via WSL2.

-### Linux
+### Linux & WSL2

 ```
-curl -fsSL https://ollama.com/install.sh | sh
+curl https://ollama.ai/install.sh | sh
 ```

-[Manual install instructions](https://github.com/ollama/ollama/blob/main/docs/linux.md)
+[Manual install instructions](https://github.com/jmorganca/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)
-
 ## Quickstart

-To run and chat with [Llama 3](https://ollama.com/library/llama3):
+To run and chat with [Llama 2](https://ollama.ai/library/llama2):

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

-Here are some example models that can be downloaded:
+Here are some example open-source models that can be downloaded:

 | Model              | Parameters | Size  | Download                       |
 | ------------------ | ---------- | ----- | ------------------------------ |
-| Llama 3            | 8B         | 4.7GB | `ollama run llama3`            |
-| Llama 3            | 70B        | 40GB  | `ollama run llama3:70b`        |
-| Phi 3 Mini         | 3.8B       | 2.3GB | `ollama run phi3`              |
-| Phi 3 Medium       | 14B        | 7.9GB | `ollama run phi3:medium`       |
-| Gemma              | 2B         | 1.4GB | `ollama run gemma:2b`          |
-| Gemma              | 7B         | 4.8GB | `ollama run gemma:7b`          |
+| Llama 2            | 7B         | 3.8GB | `ollama run llama2`            |
 | Mistral            | 7B         | 4.1GB | `ollama run mistral`           |
-| Moondream 2        | 1.4B       | 829MB | `ollama run moondream`         |
+| Dolphin Phi        | 2.7B       | 1.6GB | `ollama run dolphin-phi`       |
+| Phi-2              | 2.7B       | 1.7GB | `ollama run phi`               |
 | 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` |
+| Llama 2 13B        | 13B        | 7.3GB | `ollama run llama2:13b`        |
+| Llama 2 70B        | 70B        | 39GB  | `ollama run llama2:70b`        |
+| Orca Mini          | 3B         | 1.9GB | `ollama run orca-mini`         |
+| Vicuna             | 7B         | 3.8GB | `ollama run vicuna`            |
 | LLaVA              | 7B         | 4.5GB | `ollama run llava`             |
-| Solar              | 10.7B      | 6.1GB | `ollama run solar`             |

 > 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.

@@ -96,16 +93,16 @@ 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:
+Models from the Ollama library can be customized with a prompt. For example, to customize the `llama2` model:

 ```
-ollama pull llama3
+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
@@ -140,7 +137,7 @@ ollama create mymodel -f ./Modelfile
 ### Pull a model

 ```
-ollama pull llama3
+ollama pull llama2
 ```

 > This command can also be used to update a local model. Only the diff will be pulled.
@@ -148,13 +145,13 @@ ollama pull llama3
 ### Remove a model

 ```
-ollama rm llama3
+ollama rm llama2
 ```

 ### Copy a model

 ```
-ollama cp llama3 my-model
+ollama cp llama2 my-llama2
 ```

 ### Multiline input
@@ -175,10 +172,10 @@ I'm a basic program that prints the famous "Hello, world!" message to the consol
 The image features a yellow smiley face, which is likely the central focus of the picture.
 ```

-### Pass the prompt as an argument
+### Pass in prompt as arguments

 ```
-$ ollama run llama3 "Summarize this file: $(cat README.md)"
+$ ollama run llama2 "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.
 ```

@@ -194,10 +191,25 @@ ollama list

 ## Building

-See the [developer guide](https://github.com/ollama/ollama/blob/main/docs/development.md)
+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/jmorganca/ollama/blob/main/docs/development.md)
+

 ### Running local builds
-
 Next, start the server:

 ```
@@ -207,7 +219,7 @@ Next, start the server:
 Finally, in a separate shell, run a model:

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

 ## REST API
@@ -218,7 +230,7 @@ Ollama has a REST API for running and managing models.

 ```
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama3",
+  "model": "llama2",
  "prompt":"Why is the sky blue?"
 }'
 ```
@@ -227,7 +239,7 @@ curl http://localhost:11434/api/generate -d '{

 ```
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama3",
+  "model": "mistral",
  "messages": [
    { "role": "user", "content": "why is the sky blue?" }
  ]
@@ -236,55 +248,26 @@ curl http://localhost:11434/api/chat -d '{

 See the [API documentation](./docs/api.md) for all endpoints.

+## Integrations
+
+- [ollama-python](https://github.com/jmorganca/ollama-python)
+
 ## Community Integrations

 ### Web & Desktop
-
- [Open WebUI](https://github.com/open-webui/open-webui)
- [Enchanted (macOS native)](https://github.com/AugustDev/enchanted)
- [Hollama](https://github.com/fmaclen/hollama)
- [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)
+- [Web UI](https://github.com/ollama-webui/ollama-webui)
 - [Ollamac](https://github.com/kevinhermawan/Ollamac)
- [big-AGI](https://github.com/enricoros/big-AGI/blob/main/docs/config-local-ollama.md)
+- [big-AGI](https://github.com/enricoros/big-agi/blob/main/docs/config-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](https://github.com/mrdjohnson/llm-x) (Progressive Web App)
- [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](https://github.com/reid41/QA-Pilot) (Chat with Code Repository)
- [ChatOllama](https://github.com/sugarforever/chat-ollama) (Open Source Chatbot based on Ollama with Knowledge Bases)
- [CRAG Ollama Chat](https://github.com/Nagi-ovo/CRAG-Ollama-Chat) (Simple Web Search with Corrective RAG)
- [RAGFlow](https://github.com/infiniflow/ragflow) (Open-source Retrieval-Augmented Generation engine based on deep document understanding)
- [StreamDeploy](https://github.com/StreamDeploy-DevRel/streamdeploy-llm-app-scaffold) (LLM Application Scaffold)
- [chat](https://github.com/swuecho/chat) (chat web app for teams)
- [Lobe Chat](https://github.com/lobehub/lobe-chat) with [Integrating Doc](https://lobehub.com/docs/self-hosting/examples/ollama)
- [Ollama RAG Chatbot](https://github.com/datvodinh/rag-chatbot.git) (Local Chat with multiple PDFs using Ollama and RAG)
- [BrainSoup](https://www.nurgo-software.com/products/brainsoup) (Flexible native client with RAG & multi-agent automation)
- [macai](https://github.com/Renset/macai) (macOS client for Ollama, ChatGPT, and other compatible API back-ends)
- [Olpaka](https://github.com/Otacon/olpaka) (User-friendly Flutter Web App for Ollama)
- [OllamaSpring](https://github.com/CrazyNeil/OllamaSpring) (Ollama Client for macOS)
+

 ### Terminal

@@ -293,37 +276,23 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [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)
+- [MindsDB](https://github.com/mindsdb/mindsdb/blob/staging/mindsdb/integrations/handlers/ollama_handler/README.md)

 ### Package managers

 - [Pacman](https://archlinux.org/packages/extra/x86_64/ollama/)
- [Helm Chart](https://artifacthub.io/packages/helm/ollama-helm/ollama)
- [Guix channel](https://codeberg.org/tusharhero/ollama-guix)

 ### 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)
@@ -337,15 +306,8 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [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 for R - ollama-r](https://github.com/hauselin/ollama-r)
- [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/)
- [Portkey](https://portkey.ai/docs/welcome/integration-guides/ollama)
- [PromptingTools.jl](https://github.com/svilupp/PromptingTools.jl) with an [example](https://svilupp.github.io/PromptingTools.jl/dev/examples/working_with_ollama)
- [LlamaScript](https://github.com/Project-Llama/llamascript)
+
+
 ### Mobile

 - [Enchanted](https://github.com/AugustDev/enchanted)
@@ -358,27 +320,11 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [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)
- [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)
 - [Llama Coder](https://github.com/ex3ndr/llama-coder) (Copilot alternative using Ollama)
- [Ollama Copilot](https://github.com/bernardo-bruning/ollama-copilot) (Proxy that allows you to use ollama as a copilot like Github copilot)
- [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)
- [Discord AI chat/moderation bot](https://github.com/rapmd73/Companion) Chat/moderation bot written in python. Uses Ollama to create personalities.
-
-### Supported backends 
- [llama.cpp](https://github.com/ggerganov/llama.cpp) project founded by Georgi Gerganov.
-
+- [Obsidian BMO Chatbot plugin](https://github.com/longy2k/obsidian-bmo-chatbot)
+- [Open Interpreter](https://docs.openinterpreter.com/language-model-setup/local-models/ollama)
--- a/api/client.go
+++ b/api/client.go
@@ -1,16 +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 [the API documentation].
-// The ollama command-line client itself uses this package to interact with
-// the backend service.
-//
-// # Examples
-//
-// Several examples of using this package are available [in the GitHub
-// repository].
-//
-// [the API documentation]: https://github.com/ollama/ollama/blob/main/docs/api.md
-// [in the GitHub repository]: https://github.com/ollama/ollama/tree/main/examples
 package api

 import (
@@ -18,6 +5,7 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"net"
@@ -25,18 +13,15 @@ import (
 	"net/url"
 	"os"
 	"runtime"
-	"strconv"
 	"strings"

-	"github.com/ollama/ollama/format"
-	"github.com/ollama/ollama/version"
+	"github.com/jmorganca/ollama/format"
+	"github.com/jmorganca/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
+	http http.Client
 }

 func checkError(resp *http.Response, body []byte) error {
@@ -55,46 +40,13 @@ 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
-	}
-
-	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, "://")
+	scheme, hostport, ok := strings.Cut(os.Getenv("OLLAMA_HOST"), "://")
 	switch {
 	case !ok:
-		scheme, hostport = "http", hostVar
+		scheme, hostport = "http", os.Getenv("OLLAMA_HOST")
 	case scheme == "http":
 		defaultPort = "80"
 	case scheme == "https":
@@ -114,22 +66,30 @@ func GetOllamaHost() (OllamaHost, error) {
 		}
 	}

-	if portNum, err := strconv.ParseInt(port, 10, 32); err != nil || portNum > 65535 || portNum < 0 {
-		return OllamaHost{}, ErrInvalidHostPort
+	client := Client{
+		base: &url.URL{
+			Scheme: scheme,
+			Host:   net.JoinHostPort(host, port),
+		},
 	}

-	return OllamaHost{
-		Scheme: scheme,
-		Host:   host,
-		Port:   port,
-	}, nil
-}
-
-func NewClient(base *url.URL, http *http.Client) *Client {
-	return &Client{
-		base: base,
-		http: http,
+	mockRequest, err := http.NewRequest(http.MethodHead, client.base.String(), nil)
+	if err != nil {
+		return nil, err
 	}
+
+	proxyURL, err := http.ProxyFromEnvironment(mockRequest)
+	if err != nil {
+		return nil, err
+	}
+
+	client.http = http.Client{
+		Transport: &http.Transport{
+			Proxy: http.ProxyURL(proxyURL),
+		},
+	}
+
+	return &client, nil
 }

 func (c *Client) do(ctx context.Context, method, path string, reqData, respData any) error {
@@ -248,14 +208,8 @@ 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
@@ -267,15 +221,8 @@ 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
@@ -287,14 +234,8 @@ func (c *Client) Chat(ctx context.Context, req *ChatRequest, fn ChatResponseFunc
 	})
 }

-// 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
@@ -306,14 +247,8 @@ func (c *Client) Pull(ctx context.Context, req *PullRequest, fn PullProgressFunc
 	})
 }

-// PushProgressFunc is a function that [Client.Push] invokes when progress is
-// made.
-// It's similar to other progress function types like [PullProgressFunc].
 type PushProgressFunc func(ProgressResponse) error

-// Push uploads a model to the model library; requires registering for ollama.ai
-// and adding a public key first. fn is called each time progress is made on
-// the request and can be used to display a progress bar, etc.
 func (c *Client) Push(ctx context.Context, req *PushRequest, fn PushProgressFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/push", req, func(bts []byte) error {
 		var resp ProgressResponse
@@ -325,15 +260,8 @@ func (c *Client) Push(ctx context.Context, req *PushRequest, fn PushProgressFunc
 	})
 }

-// CreateProgressFunc is a function that [Client.Create] invokes when progress
-// is made.
-// It's similar to other progress function types like [PullProgressFunc].
 type CreateProgressFunc func(ProgressResponse) error

-// Create creates a model from a [Modelfile]. fn is a progress function that
-// behaves similarly to other methods (see [Client.Pull]).
-//
-// [Modelfile]: https://github.com/ollama/ollama/blob/main/docs/modelfile.md
 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
@@ -345,7 +273,6 @@ func (c *Client) Create(ctx context.Context, req *CreateRequest, fn CreateProgre
 	})
 }

-// List lists models that are available locally.
 func (c *Client) List(ctx context.Context) (*ListResponse, error) {
 	var lr ListResponse
 	if err := c.do(ctx, http.MethodGet, "/api/tags", nil, &lr); err != nil {
@@ -354,17 +281,6 @@ func (c *Client) List(ctx context.Context) (*ListResponse, error) {
 	return &lr, nil
 }

-// List running models.
-func (c *Client) ListRunning(ctx context.Context) (*ListResponse, error) {
-	var lr ListResponse
-	if err := c.do(ctx, http.MethodGet, "/api/ps", nil, &lr); err != nil {
-		return nil, err
-	}
-	return &lr, nil
-}
-
-// Copy copies a model - creating a model with another name from an existing
-// model.
 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
@@ -372,7 +288,6 @@ func (c *Client) Copy(ctx context.Context, req *CopyRequest) error {
 	return nil
 }

-// Delete deletes a model and its data.
 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
@@ -380,7 +295,6 @@ func (c *Client) Delete(ctx context.Context, req *DeleteRequest) error {
 	return nil
 }

-// Show obtains model information, including details, modelfile, license etc.
 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 {
@@ -389,16 +303,12 @@ func (c *Client) Show(ctx context.Context, req *ShowRequest) (*ShowResponse, err
 	return &resp, nil
 }

-// Hearbeat checks if the server has started and is responsive; if yes, it
-// returns nil, otherwise an error.
 func (c *Client) Heartbeat(ctx context.Context) error {
 	if err := c.do(ctx, http.MethodHead, "/", nil, nil); err != nil {
 		return err
 	}
 	return nil
 }
-
-// Embeddings generates embeddings from a model.
 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 {
@@ -407,13 +317,21 @@ func (c *Client) Embeddings(ctx context.Context, req *EmbeddingRequest) (*Embedd
 	return &resp, nil
 }

-// CreateBlob creates a blob from a file on the server. digest is the
-// expected SHA256 digest of the file, and r represents the file.
 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)
+	if err := c.do(ctx, http.MethodHead, fmt.Sprintf("/api/blobs/%s", digest), nil, nil); err != nil {
+		var statusError StatusError
+		if !errors.As(err, &statusError) || statusError.StatusCode != http.StatusNotFound {
+			return err
+		}
+
+		if err := c.do(ctx, http.MethodPost, fmt.Sprintf("/api/blobs/%s", digest), r, nil); err != nil {
+			return err
+		}
+	}
+
+	return nil
 }

-// Version returns the Ollama server version as a string.
 func (c *Client) Version(ctx context.Context) (string, error) {
 	var version struct {
 		Version string `json:"version"`
--- a/api/client_test.go
+++ b/api/client_test.go
@@ -1,12 +1,6 @@
 package api

-import (
-	"fmt"
-	"net"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-)
+import "testing"

 func TestClientFromEnvironment(t *testing.T) {
 	type testCase struct {
@@ -46,40 +40,4 @@ func TestClientFromEnvironment(t *testing.T) {
 			}
 		})
 	}
-
-	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))
-			}
-		})
-	}
 }
--- a/api/types.go
+++ b/api/types.go
@@ -2,9 +2,7 @@ package api

 import (
 	"encoding/json"
-	"errors"
 	"fmt"
-	"log/slog"
 	"math"
 	"os"
 	"reflect"
@@ -13,7 +11,6 @@ import (
 	"time"
 )

-// StatusError is an error with and HTTP status code.
 type StatusError struct {
 	StatusCode   int
 	Status       string
@@ -34,90 +31,41 @@ func (e StatusError) Error() string {
 	}
 }

-// ImageData represents the raw binary data of an image file.
 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"`
+	System   string      `json:"system"`
+	Template string      `json:"template"`
+	Context  []int       `json:"context,omitempty"`
+	Stream   *bool       `json:"stream,omitempty"`
+	Raw      bool        `json:"raw,omitempty"`
+	Format   string      `json:"format"`
+	Images   []ImageData `json:"images,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"`
 }

-// ChatRequest describes a request sent by [Client.Chat].
 type ChatRequest struct {
-	// Model is the model name, as in [GenerateRequest].
-	Model string `json:"model"`
-
-	// Messages is the messages of the chat - can be used to keep a chat memory.
+	Model    string    `json:"model"`
 	Messages []Message `json:"messages"`
+	Stream   *bool     `json:"stream,omitempty"`
+	Format   string    `json:"format"`

-	// Stream enable streaming of returned response; true by default.
-	Stream *bool `json:"stream,omitempty"`
-
-	// Format is the format to return the response in (e.g. "json").
-	Format string `json:"format"`
-
-	// KeepAlive controls how long the model will stay loaded into memory
-	// followin the request.
-	KeepAlive *Duration `json:"keep_alive,omitempty"`
-
-	// Options lists model-specific options.
 	Options map[string]interface{} `json:"options"`
 }

-// Message is a single message in a chat sequence. The message contains the
-// role ("system", "user", or "assistant"), the content and an optional list
-// of images.
 type Message struct {
-	Role    string      `json:"role"`
+	Role    string      `json:"role"` // one of ["system", "user", "assistant"]
 	Content string      `json:"content"`
 	Images  []ImageData `json:"images,omitempty"`
 }

-// ChatResponse is the response returned by [Client.Chat]. Its fields are
-// similar to [GenerateResponse].
 type ChatResponse struct {
-	Model      string    `json:"model"`
-	CreatedAt  time.Time `json:"created_at"`
-	Message    Message   `json:"message"`
-	DoneReason string    `json:"done_reason,omitempty"`
+	Model     string    `json:"model"`
+	CreatedAt time.Time `json:"created_at"`
+	Message   Message   `json:"message"`

 	Done bool `json:"done"`

@@ -133,8 +81,7 @@ type Metrics struct {
 	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.
+// Options specfied in GenerateRequest, if you add a new option here add it to the API docs also
 type Options struct {
 	Runner

@@ -160,57 +107,45 @@ type Options struct {

 // 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"`
-	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"`
+	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"`
+	EmbeddingOnly      bool    `json:"embedding_only,omitempty"`
+	RopeFrequencyBase  float32 `json:"rope_frequency_base,omitempty"`
+	RopeFrequencyScale float32 `json:"rope_frequency_scale,omitempty"`
+	NumThread          int     `json:"num_thread,omitempty"`
 }

-// EmbeddingRequest is the request passed to [Client.Embeddings].
 type EmbeddingRequest struct {
-	// Model is the model name.
-	Model string `json:"model"`
-
-	// Prompt is the textual prompt to embed.
+	Model  string `json:"model"`
 	Prompt string `json:"prompt"`

-	// KeepAlive controls how long the model will stay loaded in memory following
-	// this request.
-	KeepAlive *Duration `json:"keep_alive,omitempty"`
-
-	// Options lists model-specific options.
 	Options map[string]interface{} `json:"options"`
 }

-// EmbeddingResponse is the response from [Client.Embeddings].
 type EmbeddingResponse struct {
 	Embedding []float64 `json:"embedding"`
 }

-// CreateRequest is the request passed to [Client.Create].
 type CreateRequest struct {
 	Model     string `json:"model"`
 	Path      string `json:"path"`
 	Modelfile string `json:"modelfile"`
 	Stream    *bool  `json:"stream,omitempty"`
-	Quantize  string `json:"quantize,omitempty"`

 	// Name is deprecated, see Model
 	Name string `json:"name"`
-
-	// Quantization is deprecated, see Quantize
-	Quantization string `json:"quantization,omitempty"`
 }

-// DeleteRequest is the request passed to [Client.Delete].
 type DeleteRequest struct {
 	Model string `json:"model"`

@@ -218,7 +153,6 @@ type DeleteRequest struct {
 	Name string `json:"name"`
 }

-// ShowRequest is the request passed to [Client.Show].
 type ShowRequest struct {
 	Model    string `json:"model"`
 	System   string `json:"system"`
@@ -230,7 +164,6 @@ type ShowRequest struct {
 	Name string `json:"name"`
 }

-// ShowResponse is the response returned from [Client.Show].
 type ShowResponse struct {
 	License    string       `json:"license,omitempty"`
 	Modelfile  string       `json:"modelfile,omitempty"`
@@ -238,16 +171,13 @@ type ShowResponse struct {
 	Template   string       `json:"template,omitempty"`
 	System     string       `json:"system,omitempty"`
 	Details    ModelDetails `json:"details,omitempty"`
-	Messages   []Message    `json:"messages,omitempty"`
 }

-// CopyRequest is the request passed to [Client.Copy].
 type CopyRequest struct {
 	Source      string `json:"source"`
 	Destination string `json:"destination"`
 }

-// PullRequest is the request passed to [Client.Pull].
 type PullRequest struct {
 	Model    string `json:"model"`
 	Insecure bool   `json:"insecure,omitempty"`
@@ -259,8 +189,6 @@ type PullRequest struct {
 	Name string `json:"name"`
 }

-// ProgressResponse is the response passed to progress functions like
-// [PullProgressFunc] and [PushProgressFunc].
 type ProgressResponse struct {
 	Status    string `json:"status"`
 	Digest    string `json:"digest,omitempty"`
@@ -268,7 +196,6 @@ type ProgressResponse struct {
 	Completed int64  `json:"completed,omitempty"`
 }

-// PushRequest is the request passed to [Client.Push].
 type PushRequest struct {
 	Model    string `json:"model"`
 	Insecure bool   `json:"insecure,omitempty"`
@@ -280,54 +207,35 @@ type PushRequest struct {
 	Name string `json:"name"`
 }

-// ListResponse is the response from [Client.List].
 type ListResponse struct {
 	Models []ModelResponse `json:"models"`
 }

-// ModelResponse is a single model description in [ListResponse].
 type ModelResponse struct {
 	Name       string       `json:"name"`
 	Model      string       `json:"model"`
-	ModifiedAt time.Time    `json:"modified_at,omitempty"`
+	ModifiedAt time.Time    `json:"modified_at"`
 	Size       int64        `json:"size"`
 	Digest     string       `json:"digest"`
 	Details    ModelDetails `json:"details,omitempty"`
-	ExpiresAt  time.Time    `json:"expires_at,omitempty"`
-	SizeVRAM   int64        `json:"size_vram,omitempty"`
 }

 type TokenResponse struct {
 	Token string `json:"token"`
 }

-// GenerateResponse is the response passed into [GenerateResponseFunc].
 type GenerateResponse struct {
-	// Model is the model name that generated the response.
-	Model string `json:"model"`
-
-	//CreatedAt is the timestamp of the response.
+	Model     string    `json:"model"`
 	CreatedAt time.Time `json:"created_at"`
+	Response  string    `json:"response"`

-	// Response is the textual response itself.
-	Response string `json:"response"`
-
-	// Done specifies if the response is complete.
-	Done bool `json:"done"`
-
-	// DoneReason is the reason the model stopped generating text.
-	DoneReason string `json:"done_reason,omitempty"`
-
-	// Context is an encoding of the conversation used in this response; this
-	// can be sent in the next request to keep a conversational memory.
+	Done    bool  `json:"done"`
 	Context []int `json:"context,omitempty"`

 	Metrics
 }

-// ModelDetails provides details about a model.
 type ModelDetails struct {
-	ParentModel       string   `json:"parent_model"`
 	Format            string   `json:"format"`
 	Family            string   `json:"family"`
 	Families          []string `json:"families"`
@@ -363,7 +271,7 @@ func (m *Metrics) Summary() {
 	}
 }

-var ErrInvalidHostPort = errors.New("invalid port specified in OLLAMA_HOST")
+var ErrInvalidOpts = fmt.Errorf("invalid options")

 func (opts *Options) FromMap(m map[string]interface{}) error {
 	valueOpts := reflect.ValueOf(opts).Elem() // names of the fields in the options struct
@@ -378,83 +286,81 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 		}
 	}

+	invalidOpts := []string{}
 	for key, val := range m {
-		opt, ok := jsonOpts[key]
-		if !ok {
-			slog.Warn("invalid option provided", "option", opt.Name)
-			continue
-		}
+		if opt, ok := jsonOpts[key]; ok {
+			field := valueOpts.FieldByName(opt.Name)
+			if field.IsValid() && field.CanSet() {
+				if val == nil {
+					continue
+				}

-		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)
+				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)
 					}
-					slice[i] = str
+				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())
 				}
-				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
 }

-// DefaultOptions is the default set of options for [GenerateRequest]; these
-// values are used unless the user specifies other values explicitly.
 func DefaultOptions() Options {
 	return Options{
 		// options set on request to runner
-		NumPredict: -1,
-
-		// set a minimal num_keep to avoid issues on context shifts
-		NumKeep:          4,
+		NumPredict:       -1,
+		NumKeep:          0,
 		Temperature:      0.8,
 		TopK:             40,
 		TopP:             0.9,
@@ -472,15 +378,19 @@ func DefaultOptions() Options {

 		Runner: Runner{
 			// options set when the model is loaded
-			NumCtx:    2048,
-			NumBatch:  512,
-			NumGPU:    -1, // -1 here indicates that NumGPU should be set dynamically
-			NumThread: 0,  // let the runtime decide
-			LowVRAM:   false,
-			F16KV:     true,
-			UseMLock:  false,
-			UseMMap:   true,
-			UseNUMA:   false,
+			NumCtx:             2048,
+			RopeFrequencyBase:  10000.0,
+			RopeFrequencyScale: 1.0,
+			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,
+			EmbeddingOnly:      true,
 		},
 	}
 }
@@ -489,13 +399,6 @@ type Duration struct {
 	time.Duration
 }

-func (d Duration) MarshalJSON() ([]byte, error) {
-	if d.Duration < 0 {
-		return []byte("-1"), nil
-	}
-	return []byte("\"" + d.Duration.String() + "\""), nil
-}
-
 func (d *Duration) UnmarshalJSON(b []byte) (err error) {
 	var v any
 	if err := json.Unmarshal(b, &v); err != nil {
@@ -507,20 +410,15 @@ func (d *Duration) UnmarshalJSON(b []byte) (err error) {
 	switch t := v.(type) {
 	case float64:
 		if t < 0 {
-			d.Duration = time.Duration(math.MaxInt64)
-		} else {
-			d.Duration = time.Duration(int(t) * int(time.Second))
+			t = math.MaxFloat64
 		}
+
+		d.Duration = time.Duration(t)
 	case string:
 		d.Duration, err = time.ParseDuration(t)
 		if err != nil {
 			return err
 		}
-		if d.Duration < 0 {
-			d.Duration = time.Duration(math.MaxInt64)
-		}
-	default:
-		return fmt.Errorf("Unsupported type: '%s'", reflect.TypeOf(v))
 	}

 	return nil
--- a/api/types_test.go
+++ b/api/types_test.go
@@ -1,107 +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 Float",
-			req:  `{ "keep_alive": 42.5 }`,
-			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 Float",
-			req:  `{ "keep_alive": -3.14 }`,
-			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)
-		})
-	}
-}
-
-func TestDurationMarshalUnmarshal(t *testing.T) {
-	tests := []struct {
-		name     string
-		input    time.Duration
-		expected time.Duration
-	}{
-		{
-			"negative duration",
-			time.Duration(-1),
-			time.Duration(math.MaxInt64),
-		},
-		{
-			"positive duration",
-			time.Duration(42 * time.Second),
-			time.Duration(42 * time.Second),
-		},
-		{
-			"another positive duration",
-			time.Duration(42 * time.Minute),
-			time.Duration(42 * time.Minute),
-		},
-		{
-			"zero duration",
-			time.Duration(0),
-			time.Duration(0),
-		},
-		{
-			"max duration",
-			time.Duration(math.MaxInt64),
-			time.Duration(math.MaxInt64),
-		},
-	}
-
-	for _, test := range tests {
-		t.Run(test.name, func(t *testing.T) {
-			b, err := json.Marshal(Duration{test.input})
-			require.NoError(t, err)
-
-			var d Duration
-			err = json.Unmarshal(b, &d)
-			require.NoError(t, err)
-
-			assert.Equal(t, test.expected, d.Duration, "input %v, marshalled %v, got %v", test.input, string(b), d.Duration)
-		})
-	}
-}
--- a/macapp/.eslintrc.json
+++ b/macapp/.eslintrc.json
--- a/app/.gitignore
+++ b/app/.gitignore
@@ -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/
--- a/app/README.md
+++ b/app/README.md
@@ -1,22 +1,21 @@
-# Ollama App
+# Desktop

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

-TODO
+## Developing

-## MacOS
-
-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
+cd ..
+go build .
 ```
+
+Then run the desktop app with `npm start`:
+
+```
+cd app
+npm install
+npm start
+```
+
--- a/app/assets/app.ico
+++ b/app/assets/app.ico
--- a/app/assets/assets.go
+++ b/app/assets/assets.go
@@ -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)
-}
--- a/macapp/assets/icon.icns
+++ b/macapp/assets/icon.icns
--- a/macapp/assets/iconDarkTemplate.png
+++ b/macapp/assets/iconDarkTemplate.png
--- a/macapp/assets/iconDarkTemplate@2x.png
+++ b/macapp/assets/iconDarkTemplate@2x.png
--- a/macapp/assets/iconDarkUpdateTemplate.png
+++ b/macapp/assets/iconDarkUpdateTemplate.png
--- a/macapp/assets/iconDarkUpdateTemplate@2x.png
+++ b/macapp/assets/iconDarkUpdateTemplate@2x.png
--- a/macapp/assets/iconTemplate.png
+++ b/macapp/assets/iconTemplate.png
--- a/macapp/assets/iconTemplate@2x.png
+++ b/macapp/assets/iconTemplate@2x.png
--- a/macapp/assets/iconUpdateTemplate.png
+++ b/macapp/assets/iconUpdateTemplate.png
--- a/macapp/assets/iconUpdateTemplate@2x.png
+++ b/macapp/assets/iconUpdateTemplate@2x.png
--- a/app/assets/setup.bmp
+++ b/app/assets/setup.bmp
--- a/app/assets/tray.ico
+++ b/app/assets/tray.ico
--- a/app/assets/tray_upgrade.ico
+++ b/app/assets/tray_upgrade.ico
--- a/macapp/forge.config.ts
+++ b/macapp/forge.config.ts
--- a/app/lifecycle/getstarted_nonwindows.go
+++ b/app/lifecycle/getstarted_nonwindows.go
@@ -1,9 +0,0 @@
-//go:build !windows
-
-package lifecycle
-
-import "fmt"
-
-func GetStarted() error {
-	return fmt.Errorf("GetStarted not implemented")
-}
--- a/app/lifecycle/getstarted_windows.go
+++ b/app/lifecycle/getstarted_windows.go
@@ -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()
-}
--- a/app/lifecycle/lifecycle.go
+++ b/app/lifecycle/lifecycle.go
@@ -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")
-}
--- a/app/lifecycle/logging.go
+++ b/app/lifecycle/logging.go
@@ -1,48 +0,0 @@
-package lifecycle
-
-import (
-	"fmt"
-	"log/slog"
-	"os"
-	"path/filepath"
-
-	"github.com/ollama/ollama/envconfig"
-)
-
-func InitLogging() {
-	level := slog.LevelInfo
-
-	if envconfig.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")
-}
--- a/app/lifecycle/logging_nonwindows.go
+++ b/app/lifecycle/logging_nonwindows.go
@@ -1,9 +0,0 @@
-//go:build !windows
-
-package lifecycle
-
-import "log/slog"
-
-func ShowLogs() {
-	slog.Warn("ShowLogs not yet implemented")
-}
--- a/app/lifecycle/logging_windows.go
+++ b/app/lifecycle/logging_windows.go
@@ -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))
-	}
-}
--- a/app/lifecycle/paths.go
+++ b/app/lifecycle/paths.go
@@ -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
-	}
-}
--- a/app/lifecycle/server.go
+++ b/app/lifecycle/server.go
@@ -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
-}
--- a/app/lifecycle/server_unix.go
+++ b/app/lifecycle/server_unix.go
@@ -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
-}
--- a/app/lifecycle/server_windows.go
+++ b/app/lifecycle/server_windows.go
@@ -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
-}
--- a/app/lifecycle/updater.go
+++ b/app/lifecycle/updater.go
@@ -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)
-			}
-		}
-	}()
-}
--- a/app/lifecycle/updater_nonwindows.go
+++ b/app/lifecycle/updater_nonwindows.go
@@ -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")
-}
--- a/app/lifecycle/updater_windows.go
+++ b/app/lifecycle/updater_windows.go
@@ -1,77 +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
-	}
-	// make the upgrade as quiet as possible (no GUI, no prompts)
-	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
-}
--- a/app/main.go
+++ b/app/main.go
@@ -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()
-}
--- a/app/ollama.iss
+++ b/app/ollama.iss
@@ -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;
--- a/app/ollama.rc
+++ b/app/ollama.rc
@@ -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
--- a/app/ollama_welcome.ps1
+++ b/app/ollama_welcome.ps1
@@ -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 ""
--- a/macapp/package-lock.json
+++ b/macapp/package-lock.json
--- a/macapp/package.json
+++ b/macapp/package.json
--- a/macapp/postcss.config.js
+++ b/macapp/postcss.config.js
--- a/macapp/src/app.css
+++ b/macapp/src/app.css
--- a/macapp/src/app.tsx
+++ b/macapp/src/app.tsx
@@ -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'>
--- a/macapp/src/declarations.d.ts
+++ b/macapp/src/declarations.d.ts
--- a/macapp/src/index.html
+++ b/macapp/src/index.html
--- a/macapp/src/index.ts
+++ b/macapp/src/index.ts
@@ -162,7 +162,7 @@ app.on('before-quit', () => {
  }
 })

-const updateURL = `https://ollama.com/api/update?os=${process.platform}&arch=${
+const updateURL = `https://ollama.ai/api/update?os=${process.platform}&arch=${
  process.arch
 }&version=${app.getVersion()}&id=${id()}`

--- a/macapp/src/install.ts
+++ b/macapp/src/install.ts
--- a/macapp/src/ollama.svg
+++ b/macapp/src/ollama.svg
--- a/macapp/src/preload.ts
+++ b/macapp/src/preload.ts
--- a/macapp/src/renderer.tsx
+++ b/macapp/src/renderer.tsx
--- a/app/store/store.go
+++ b/app/store/store.go
@@ -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))
-}
--- a/app/store/store_darwin.go
+++ b/app/store/store_darwin.go
@@ -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")
-}
--- a/app/store/store_linux.go
+++ b/app/store/store_linux.go
@@ -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")
-}
--- a/app/store/store_windows.go
+++ b/app/store/store_windows.go
@@ -1,11 +0,0 @@
-package store
-
-import (
-	"os"
-	"path/filepath"
-)
-
-func getStorePath() string {
-	localAppData := os.Getenv("LOCALAPPDATA")
-	return filepath.Join(localAppData, "Ollama", "config.json")
-}
--- a/macapp/tailwind.config.js
+++ b/macapp/tailwind.config.js
--- a/app/tray/commontray/types.go
+++ b/app/tray/commontray/types.go
@@ -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()
-}
--- a/app/tray/tray.go
+++ b/app/tray/tray.go
@@ -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)
-}
--- a/app/tray/tray_nonwindows.go
+++ b/app/tray/tray_nonwindows.go
@@ -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")
-}
--- a/app/tray/tray_windows.go
+++ b/app/tray/tray_windows.go
@@ -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)
-}
--- a/app/tray/wintray/eventloop.go
+++ b/app/tray/wintray/eventloop.go
@@ -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))
-	}
-}
--- a/app/tray/wintray/menus.go
+++ b/app/tray/wintray/menus.go
@@ -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
-}
--- a/app/tray/wintray/messages.go
+++ b/app/tray/wintray/messages.go
@@ -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"
-)
--- a/app/tray/wintray/notifyicon.go
+++ b/app/tray/wintray/notifyicon.go
@@ -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
-}
--- a/app/tray/wintray/tray.go
+++ b/app/tray/wintray/tray.go
@@ -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()
-}
--- a/app/tray/wintray/w32api.go
+++ b/app/tray/wintray/w32api.go
@@ -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
-}
--- a/app/tray/wintray/winclass.go
+++ b/app/tray/wintray/winclass.go
@@ -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
-}
--- a/macapp/tsconfig.json
+++ b/macapp/tsconfig.json
--- a/macapp/webpack.main.config.ts
+++ b/macapp/webpack.main.config.ts
--- a/macapp/webpack.plugins.ts
+++ b/macapp/webpack.plugins.ts
--- a/macapp/webpack.renderer.config.ts
+++ b/macapp/webpack.renderer.config.ts
--- a/macapp/webpack.rules.ts
+++ b/macapp/webpack.rules.ts
--- a/auth/auth.go
+++ b/auth/auth.go
@@ -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
-}
--- a/cmd/cmd.go
+++ b/cmd/cmd.go
@@ -1,7 +1,6 @@
 package cmd

 import (
-	"archive/zip"
 	"bytes"
 	"context"
 	"crypto/ed25519"
@@ -12,36 +11,28 @@ import (
 	"fmt"
 	"io"
 	"log"
-	"math"
 	"net"
 	"net/http"
 	"os"
+	"os/exec"
 	"os/signal"
 	"path/filepath"
-	"regexp"
 	"runtime"
 	"strings"
 	"syscall"
 	"time"

-	"github.com/containerd/console"
-	"github.com/mattn/go-runewidth"
 	"github.com/olekukonko/tablewriter"
 	"github.com/spf13/cobra"
 	"golang.org/x/crypto/ssh"
-	"golang.org/x/exp/slices"
 	"golang.org/x/term"

-	"github.com/ollama/ollama/api"
-	"github.com/ollama/ollama/auth"
-	"github.com/ollama/ollama/envconfig"
-	"github.com/ollama/ollama/format"
-	"github.com/ollama/ollama/parser"
-	"github.com/ollama/ollama/progress"
-	"github.com/ollama/ollama/server"
-	"github.com/ollama/ollama/types/errtypes"
-	"github.com/ollama/ollama/types/model"
-	"github.com/ollama/ollama/version"
+	"github.com/jmorganca/ollama/api"
+	"github.com/jmorganca/ollama/format"
+	"github.com/jmorganca/ollama/parser"
+	"github.com/jmorganca/ollama/progress"
+	"github.com/jmorganca/ollama/server"
+	"github.com/jmorganca/ollama/version"
 )

 func CreateHandler(cmd *cobra.Command, args []string) error {
@@ -59,13 +50,14 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 	p := progress.NewProgress(os.Stderr)
 	defer p.Stop()

-	f, err := os.Open(filename)
+	bars := make(map[string]*progress.Bar)
+
+	modelfile, err := os.ReadFile(filename)
 	if err != nil {
 		return err
 	}
-	defer f.Close()

-	modelfile, err := parser.ParseFile(f)
+	commands, err := parser.Parse(bytes.NewReader(modelfile))
 	if err != nil {
 		return err
 	}
@@ -79,10 +71,10 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 	spinner := progress.NewSpinner(status)
 	p.Add(status, spinner)

-	for i := range modelfile.Commands {
-		switch modelfile.Commands[i].Name {
+	for _, c := range commands {
+		switch c.Name {
 		case "model", "adapter":
-			path := modelfile.Commands[i].Args
+			path := c.Args
 			if path == "~" {
 				path = home
 			} else if strings.HasPrefix(path, "~/") {
@@ -93,35 +85,29 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 				path = filepath.Join(filepath.Dir(filename), path)
 			}

-			fi, err := os.Stat(path)
-			if errors.Is(err, os.ErrNotExist) && modelfile.Commands[i].Name == "model" {
+			bin, err := os.Open(path)
+			if errors.Is(err, os.ErrNotExist) && c.Name == "model" {
 				continue
 			} else if err != nil {
 				return err
 			}
+			defer bin.Close()

-			if fi.IsDir() {
-				// this is likely a safetensors or pytorch directory
-				// TODO make this work w/ adapters
-				tempfile, err := tempZipFiles(path)
-				if err != nil {
-					return err
-				}
-				defer os.RemoveAll(tempfile)
-
-				path = tempfile
+			hash := sha256.New()
+			if _, err := io.Copy(hash, bin); err != nil {
+				return err
 			}
+			bin.Seek(0, io.SeekStart)

-			digest, err := createBlob(cmd, client, path)
-			if err != nil {
+			digest := fmt.Sprintf("sha256:%x", hash.Sum(nil))
+			if err = client.CreateBlob(cmd.Context(), digest, bin); err != nil {
 				return err
 			}

-			modelfile.Commands[i].Args = "@" + digest
+			modelfile = bytes.ReplaceAll(modelfile, []byte(c.Args), []byte("@"+digest))
 		}
 	}

-	bars := make(map[string]*progress.Bar)
 	fn := func(resp api.ProgressResponse) error {
 		if resp.Digest != "" {
 			spinner.Stop()
@@ -145,9 +131,7 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 		return nil
 	}

-	quantize, _ := cmd.Flags().GetString("quantize")
-
-	request := api.CreateRequest{Name: args[0], Modelfile: modelfile.String(), Quantize: quantize}
+	request := api.CreateRequest{Name: args[0], Modelfile: string(modelfile)}
 	if err := client.Create(cmd.Context(), &request, fn); err != nil {
 		return err
 	}
@@ -155,137 +139,6 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 	return nil
 }

-func tempZipFiles(path string) (string, error) {
-	tempfile, err := os.CreateTemp("", "ollama-tf")
-	if err != nil {
-		return "", err
-	}
-	defer tempfile.Close()
-
-	zipfile := zip.NewWriter(tempfile)
-	defer zipfile.Close()
-
-	detectContentType := func(path string) (string, error) {
-		f, err := os.Open(path)
-		if err != nil {
-			return "", err
-		}
-		defer f.Close()
-
-		var b bytes.Buffer
-		b.Grow(512)
-
-		if _, err := io.CopyN(&b, f, 512); err != nil && !errors.Is(err, io.EOF) {
-			return "", err
-		}
-
-		contentType, _, _ := strings.Cut(http.DetectContentType(b.Bytes()), ";")
-		return contentType, nil
-	}
-
-	glob := func(pattern, contentType string) ([]string, error) {
-		matches, err := filepath.Glob(pattern)
-		if err != nil {
-			return nil, err
-		}
-
-		for _, safetensor := range matches {
-			if ct, err := detectContentType(safetensor); err != nil {
-				return nil, err
-			} else if ct != contentType {
-				return nil, fmt.Errorf("invalid content type: expected %s for %s", ct, safetensor)
-			}
-		}
-
-		return matches, nil
-	}
-
-	var files []string
-	if st, _ := glob(filepath.Join(path, "model*.safetensors"), "application/octet-stream"); len(st) > 0 {
-		// safetensors files might be unresolved git lfs references; skip if they are
-		// covers model-x-of-y.safetensors, model.fp32-x-of-y.safetensors, model.safetensors
-		files = append(files, st...)
-	} else if pt, _ := glob(filepath.Join(path, "pytorch_model*.bin"), "application/zip"); len(pt) > 0 {
-		// pytorch files might also be unresolved git lfs references; skip if they are
-		// covers pytorch_model-x-of-y.bin, pytorch_model.fp32-x-of-y.bin, pytorch_model.bin
-		files = append(files, pt...)
-	} else if pt, _ := glob(filepath.Join(path, "consolidated*.pth"), "application/zip"); len(pt) > 0 {
-		// pytorch files might also be unresolved git lfs references; skip if they are
-		// covers consolidated.x.pth, consolidated.pth
-		files = append(files, pt...)
-	} else {
-		return "", errors.New("no safetensors or torch files found")
-	}
-
-	// add configuration files, json files are detected as text/plain
-	js, err := glob(filepath.Join(path, "*.json"), "text/plain")
-	if err != nil {
-		return "", err
-	}
-	files = append(files, js...)
-
-	if tks, _ := glob(filepath.Join(path, "tokenizer.model"), "application/octet-stream"); len(tks) > 0 {
-		// add tokenizer.model if it exists, tokenizer.json is automatically picked up by the previous glob
-		// tokenizer.model might be a unresolved git lfs reference; error if it is
-		files = append(files, tks...)
-	} else if tks, _ := glob(filepath.Join(path, "**/tokenizer.model"), "text/plain"); len(tks) > 0 {
-		// some times tokenizer.model is in a subdirectory (e.g. meta-llama/Meta-Llama-3-8B)
-		files = append(files, tks...)
-	}
-
-	for _, file := range files {
-		f, err := os.Open(file)
-		if err != nil {
-			return "", err
-		}
-		defer f.Close()
-
-		fi, err := f.Stat()
-		if err != nil {
-			return "", err
-		}
-
-		zfi, err := zip.FileInfoHeader(fi)
-		if err != nil {
-			return "", err
-		}
-
-		zf, err := zipfile.CreateHeader(zfi)
-		if err != nil {
-			return "", err
-		}
-
-		if _, err := io.Copy(zf, f); err != nil {
-			return "", err
-		}
-	}
-
-	return tempfile.Name(), nil
-}
-
-func createBlob(cmd *cobra.Command, client *api.Client, path string) (string, error) {
-	bin, err := os.Open(path)
-	if err != nil {
-		return "", err
-	}
-	defer bin.Close()
-
-	hash := sha256.New()
-	if _, err := io.Copy(hash, bin); err != nil {
-		return "", err
-	}
-
-	if _, err := bin.Seek(0, io.SeekStart); err != nil {
-		return "", err
-	}
-
-	digest := fmt.Sprintf("sha256:%x", hash.Sum(nil))
-	if err = client.CreateBlob(cmd.Context(), digest, bin); err != nil {
-		return "", err
-	}
-	return digest, nil
-}
-
 func RunHandler(cmd *cobra.Command, args []string) error {
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
@@ -293,121 +146,19 @@ func RunHandler(cmd *cobra.Command, args []string) error {
 	}

 	name := args[0]
-
 	// check if the model exists on the server
-	show, err := client.Show(cmd.Context(), &api.ShowRequest{Name: name})
+	_, err = client.Show(cmd.Context(), &api.ShowRequest{Name: name})
 	var statusError api.StatusError
 	switch {
 	case errors.As(err, &statusError) && statusError.StatusCode == http.StatusNotFound:
 		if err := PullHandler(cmd, []string{name}); err != nil {
 			return err
 		}
-
-		show, err = client.Show(cmd.Context(), &api.ShowRequest{Name: name})
-		if err != nil {
-			return err
-		}
 	case err != nil:
 		return err
 	}

-	interactive := true
-
-	opts := runOptions{
-		Model:       args[0],
-		WordWrap:    os.Getenv("TERM") == "xterm-256color",
-		Options:     map[string]interface{}{},
-		MultiModal:  slices.Contains(show.Details.Families, "clip"),
-		ParentModel: show.Details.ParentModel,
-	}
-
-	format, err := cmd.Flags().GetString("format")
-	if err != nil {
-		return err
-	}
-	opts.Format = format
-
-	keepAlive, err := cmd.Flags().GetString("keepalive")
-	if err != nil {
-		return err
-	}
-	if keepAlive != "" {
-		d, err := time.ParseDuration(keepAlive)
-		if err != nil {
-			return err
-		}
-		opts.KeepAlive = &api.Duration{Duration: d}
-	}
-
-	prompts := args[1:]
-	// prepend stdin to the prompt if provided
-	if !term.IsTerminal(int(os.Stdin.Fd())) {
-		in, err := io.ReadAll(os.Stdin)
-		if err != nil {
-			return err
-		}
-
-		prompts = append([]string{string(in)}, prompts...)
-		opts.WordWrap = false
-		interactive = false
-	}
-	opts.Prompt = strings.Join(prompts, " ")
-	if len(prompts) > 0 {
-		interactive = false
-	}
-
-	nowrap, err := cmd.Flags().GetBool("nowordwrap")
-	if err != nil {
-		return err
-	}
-	opts.WordWrap = !nowrap
-
-	if !interactive {
-		return generate(cmd, opts)
-	}
-
-	return generateInteractive(cmd, opts)
-}
-
-func errFromUnknownKey(unknownKeyErr error) error {
-	// find SSH public key in the error message
-	sshKeyPattern := `ssh-\w+ [^\s"]+`
-	re := regexp.MustCompile(sshKeyPattern)
-	matches := re.FindStringSubmatch(unknownKeyErr.Error())
-
-	if len(matches) > 0 {
-		serverPubKey := matches[0]
-
-		localPubKey, err := auth.GetPublicKey()
-		if err != nil {
-			return unknownKeyErr
-		}
-
-		if runtime.GOOS == "linux" && serverPubKey != localPubKey {
-			// try the ollama service public key
-			svcPubKey, err := os.ReadFile("/usr/share/ollama/.ollama/id_ed25519.pub")
-			if err != nil {
-				return unknownKeyErr
-			}
-			localPubKey = strings.TrimSpace(string(svcPubKey))
-		}
-
-		// check if the returned public key matches the local public key, this prevents adding a remote key to the user's account
-		if serverPubKey != localPubKey {
-			return unknownKeyErr
-		}
-
-		var msg strings.Builder
-		msg.WriteString(unknownKeyErr.Error())
-		msg.WriteString("\n\nYour ollama key is:\n")
-		msg.WriteString(localPubKey)
-		msg.WriteString("\nAdd your key at:\n")
-		msg.WriteString("https://ollama.com/settings/keys")
-
-		return errors.New(msg.String())
-	}
-
-	return unknownKeyErr
+	return RunGenerate(cmd, args)
 }

 func PushHandler(cmd *cobra.Command, args []string) error {
@@ -457,20 +208,6 @@ func PushHandler(cmd *cobra.Command, args []string) error {

 	request := api.PushRequest{Name: args[0], Insecure: insecure}
 	if err := client.Push(cmd.Context(), &request, fn); err != nil {
-		if spinner != nil {
-			spinner.Stop()
-		}
-		if strings.Contains(err.Error(), "access denied") {
-			return errors.New("you are not authorized to push to this namespace, create the model under a namespace you own")
-		}
-		host := model.ParseName(args[0]).Host
-		isOllamaHost := strings.HasSuffix(host, ".ollama.ai") || strings.HasSuffix(host, ".ollama.com")
-		if strings.Contains(err.Error(), errtypes.UnknownOllamaKeyErrMsg) && isOllamaHost {
-			// the user has not added their ollama key to ollama.com
-			// re-throw an error with a more user-friendly message
-			return errFromUnknownKey(err)
-		}
-
 		return err
 	}

@@ -511,52 +248,6 @@ func ListHandler(cmd *cobra.Command, args []string) error {
 	return nil
 }

-func ListRunningHandler(cmd *cobra.Command, args []string) error {
-	client, err := api.ClientFromEnvironment()
-	if err != nil {
-		return err
-	}
-
-	models, err := client.ListRunning(cmd.Context())
-	if err != nil {
-		return err
-	}
-
-	var data [][]string
-
-	for _, m := range models.Models {
-		if len(args) == 0 || strings.HasPrefix(m.Name, args[0]) {
-			var procStr string
-			switch {
-			case m.SizeVRAM == 0:
-				procStr = "100% CPU"
-			case m.SizeVRAM == m.Size:
-				procStr = "100% GPU"
-			case m.SizeVRAM > m.Size || m.Size == 0:
-				procStr = "Unknown"
-			default:
-				sizeCPU := m.Size - m.SizeVRAM
-				cpuPercent := math.Round(float64(sizeCPU) / float64(m.Size) * 100)
-				procStr = fmt.Sprintf("%d%%/%d%% CPU/GPU", int(cpuPercent), int(100-cpuPercent))
-			}
-			data = append(data, []string{m.Name, m.Digest[:12], format.HumanBytes(m.Size), procStr, format.HumanTime(m.ExpiresAt, "Never")})
-		}
-	}
-
-	table := tablewriter.NewWriter(os.Stdout)
-	table.SetHeader([]string{"NAME", "ID", "SIZE", "PROCESSOR", "UNTIL"})
-	table.SetHeaderAlignment(tablewriter.ALIGN_LEFT)
-	table.SetAlignment(tablewriter.ALIGN_LEFT)
-	table.SetHeaderLine(false)
-	table.SetBorder(false)
-	table.SetNoWhiteSpace(true)
-	table.SetTablePadding("\t")
-	table.AppendBulk(data)
-	table.Render()
-
-	return nil
-}
-
 func DeleteHandler(cmd *cobra.Command, args []string) error {
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
@@ -719,21 +410,63 @@ func PullHandler(cmd *cobra.Command, args []string) error {
 	return nil
 }

+func RunGenerate(cmd *cobra.Command, args []string) error {
+	interactive := true
+
+	opts := runOptions{
+		Model:    args[0],
+		WordWrap: os.Getenv("TERM") == "xterm-256color",
+		Options:  map[string]interface{}{},
+	}
+
+	format, err := cmd.Flags().GetString("format")
+	if err != nil {
+		return err
+	}
+	opts.Format = format
+
+	prompts := args[1:]
+	// prepend stdin to the prompt if provided
+	if !term.IsTerminal(int(os.Stdin.Fd())) {
+		in, err := io.ReadAll(os.Stdin)
+		if err != nil {
+			return err
+		}
+
+		prompts = append([]string{string(in)}, prompts...)
+		opts.WordWrap = false
+		interactive = false
+	}
+	opts.Prompt = strings.Join(prompts, " ")
+	if len(prompts) > 0 {
+		interactive = false
+	}
+
+	nowrap, err := cmd.Flags().GetBool("nowordwrap")
+	if err != nil {
+		return err
+	}
+	opts.WordWrap = !nowrap
+
+	if !interactive {
+		return generate(cmd, opts)
+	}
+
+	return generateInteractive(cmd, opts)
+}
+
 type generateContextKey string

 type runOptions struct {
-	Model       string
-	ParentModel string
-	Prompt      string
-	Messages    []api.Message
-	WordWrap    bool
-	Format      string
-	System      string
-	Template    string
-	Images      []api.ImageData
-	Options     map[string]interface{}
-	MultiModal  bool
-	KeepAlive   *api.Duration
+	Model    string
+	Prompt   string
+	Messages []api.Message
+	WordWrap bool
+	Format   string
+	System   string
+	Template string
+	Images   []api.ImageData
+	Options  map[string]interface{}
 }

 type displayResponseState struct {
@@ -746,8 +479,7 @@ func displayResponse(content string, wordWrap bool, state *displayResponseState)
 	if wordWrap && termWidth >= 10 {
 		for _, ch := range content {
 			if state.lineLength+1 > termWidth-5 {
-
-				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
+				if len(state.wordBuffer) > termWidth-10 {
 					fmt.Printf("%s%c", state.wordBuffer, ch)
 					state.wordBuffer = ""
 					state.lineLength = 0
@@ -755,18 +487,12 @@ func displayResponse(content string, wordWrap bool, state *displayResponseState)
 				}

 				// backtrack the length of the last word and clear to the end of the line
-				fmt.Printf("\x1b[%dD\x1b[K\n", runewidth.StringWidth(state.wordBuffer))
+				fmt.Printf("\x1b[%dD\x1b[K\n", len(state.wordBuffer))
 				fmt.Printf("%s%c", state.wordBuffer, ch)
-				chWidth := runewidth.RuneWidth(ch)
-
-				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
+				state.lineLength = len(state.wordBuffer) + 1
 			} else {
 				fmt.Print(string(ch))
-				state.lineLength += runewidth.RuneWidth(ch)
-				if runewidth.RuneWidth(ch) >= 2 {
-					state.wordBuffer = ""
-					continue
-				}
+				state.lineLength += 1

 				switch ch {
 				case ' ':
@@ -835,10 +561,6 @@ func chat(cmd *cobra.Command, opts runOptions) (*api.Message, error) {
 		Options:  opts.Options,
 	}

-	if opts.KeepAlive != nil {
-		req.KeepAlive = opts.KeepAlive
-	}
-
 	if err := client.Chat(cancelCtx, req, fn); err != nil {
 		if errors.Is(err, context.Canceled) {
 			return nil, nil
@@ -906,23 +628,14 @@ func generate(cmd *cobra.Command, opts runOptions) error {
 		return nil
 	}

-	if opts.MultiModal {
-		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
-		if err != nil {
-			return err
-		}
-	}
-
 	request := api.GenerateRequest{
-		Model:     opts.Model,
-		Prompt:    opts.Prompt,
-		Context:   generateContext,
-		Images:    opts.Images,
-		Format:    opts.Format,
-		System:    opts.System,
-		Template:  opts.Template,
-		Options:   opts.Options,
-		KeepAlive: opts.KeepAlive,
+		Model:    opts.Model,
+		Prompt:   opts.Prompt,
+		Context:  generateContext,
+		Format:   opts.Format,
+		System:   opts.System,
+		Template: opts.Template,
+		Options:  opts.Options,
 	}

 	if err := client.Generate(ctx, &request, fn); err != nil {
@@ -957,27 +670,24 @@ func generate(cmd *cobra.Command, opts runOptions) error {
 }

 func RunServer(cmd *cobra.Command, _ []string) error {
-	// retrieve the OLLAMA_HOST environment variable
-	ollamaHost, err := api.GetOllamaHost()
+	host, port, err := net.SplitHostPort(os.Getenv("OLLAMA_HOST"))
 	if err != nil {
-		return err
+		host, port = "127.0.0.1", "11434"
+		if ip := net.ParseIP(strings.Trim(os.Getenv("OLLAMA_HOST"), "[]")); ip != nil {
+			host = ip.String()
+		}
 	}

 	if err := initializeKeypair(); err != nil {
 		return err
 	}

-	ln, err := net.Listen("tcp", net.JoinHostPort(ollamaHost.Host, ollamaHost.Port))
+	ln, err := net.Listen("tcp", net.JoinHostPort(host, port))
 	if err != nil {
 		return err
 	}

-	err = server.Serve(ln)
-	if errors.Is(err, http.ErrServerClosed) {
-		return nil
-	}
-
-	return err
+	return server.Serve(ln)
 }

 func initializeKeypair() error {
@@ -992,42 +702,59 @@ func initializeKeypair() error {
 	_, err = os.Stat(privKeyPath)
 	if os.IsNotExist(err) {
 		fmt.Printf("Couldn't find '%s'. Generating new private key.\n", privKeyPath)
-		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
+		_, privKey, err := ed25519.GenerateKey(rand.Reader)
 		if err != nil {
 			return err
 		}

-		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
+		privKeyBytes, err := format.OpenSSHPrivateKey(privKey, "")
 		if err != nil {
 			return err
 		}

-		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
+		err = os.MkdirAll(filepath.Dir(privKeyPath), 0o755)
+		if err != nil {
 			return fmt.Errorf("could not create directory %w", err)
 		}

-		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
-			return err
-		}
-
-		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
+		err = os.WriteFile(privKeyPath, pem.EncodeToMemory(privKeyBytes), 0o600)
 		if err != nil {
 			return err
 		}

-		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
-
-		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
+		sshPrivateKey, err := ssh.NewSignerFromKey(privKey)
+		if err != nil {
 			return err
 		}

-		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
+		pubKeyData := ssh.MarshalAuthorizedKey(sshPrivateKey.PublicKey())
+
+		err = os.WriteFile(pubKeyPath, pubKeyData, 0o644)
+		if err != nil {
+			return err
+		}
+
+		fmt.Printf("Your new public key is: \n\n%s\n", string(pubKeyData))
 	}
 	return nil
 }

-//nolint:unused
-func waitForServer(ctx context.Context, client *api.Client) error {
+func startMacApp(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
+	}
 	// wait for the server to start
 	timeout := time.After(5 * time.Second)
 	tick := time.Tick(500 * time.Millisecond)
@@ -1041,7 +768,6 @@ func waitForServer(ctx context.Context, client *api.Client) error {
 			}
 		}
 	}
-
 }

 func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
@@ -1050,11 +776,15 @@ func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
 		return err
 	}
 	if err := client.Heartbeat(cmd.Context()); err != nil {
-		if !strings.Contains(err.Error(), " refused") {
+		if !strings.Contains(err.Error(), "connection refused") {
 			return err
 		}
-		if err := startApp(cmd.Context(), client); err != nil {
-			return fmt.Errorf("could not connect to ollama app, is it running?")
+		if runtime.GOOS == "darwin" {
+			if err := startMacApp(cmd.Context(), client); err != nil {
+				return fmt.Errorf("could not connect to ollama app, is it running?")
+			}
+		} else {
+			return fmt.Errorf("could not connect to ollama server, run 'ollama serve' to start it")
 		}
 	}
 	return nil
@@ -1080,29 +810,10 @@ func versionHandler(cmd *cobra.Command, _ []string) {
 	}
 }

-func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
-	if len(envs) == 0 {
-		return
-	}
-
-	envUsage := `
-Environment Variables:
-`
-	for _, e := range envs {
-		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
-	}
-
-	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
-}
-
 func NewCLI() *cobra.Command {
 	log.SetFlags(log.LstdFlags | log.Lshortfile)
 	cobra.EnableCommandSorting = false

-	if runtime.GOOS == "windows" {
-		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
-	}
-
 	rootCmd := &cobra.Command{
 		Use:           "ollama",
 		Short:         "Large language model runner",
@@ -1131,8 +842,7 @@ func NewCLI() *cobra.Command {
 		RunE:    CreateHandler,
 	}

-	createCmd.Flags().StringP("file", "f", "Modelfile", "Name of the Modelfile")
-	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
+	createCmd.Flags().StringP("file", "f", "Modelfile", "Name of the Modelfile (default \"Modelfile\")")

 	showCmd := &cobra.Command{
 		Use:     "show MODEL",
@@ -1156,11 +866,11 @@ func NewCLI() *cobra.Command {
 		RunE:    RunHandler,
 	}

-	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
 	runCmd.Flags().Bool("verbose", false, "Show timings for response")
 	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
 	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
 	runCmd.Flags().String("format", "", "Response format (e.g. json)")
+
 	serveCmd := &cobra.Command{
 		Use:     "serve",
 		Aliases: []string{"start"},
@@ -1197,15 +907,8 @@ func NewCLI() *cobra.Command {
 		RunE:    ListHandler,
 	}

-	psCmd := &cobra.Command{
-		Use:     "ps",
-		Short:   "List running models",
-		PreRunE: checkServerHeartbeat,
-		RunE:    ListRunningHandler,
-	}
-
 	copyCmd := &cobra.Command{
-		Use:     "cp SOURCE DESTINATION",
+		Use:     "cp SOURCE TARGET",
 		Short:   "Copy a model",
 		Args:    cobra.ExactArgs(2),
 		PreRunE: checkServerHeartbeat,
@@ -1220,43 +923,6 @@ func NewCLI() *cobra.Command {
 		RunE:    DeleteHandler,
 	}

-	envVars := envconfig.AsMap()
-
-	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
-
-	for _, cmd := range []*cobra.Command{
-		createCmd,
-		showCmd,
-		runCmd,
-		pullCmd,
-		pushCmd,
-		listCmd,
-		psCmd,
-		copyCmd,
-		deleteCmd,
-		serveCmd,
-	} {
-		switch cmd {
-		case runCmd:
-			appendEnvDocs(cmd, []envconfig.EnvVar{envVars["OLLAMA_HOST"], envVars["OLLAMA_NOHISTORY"]})
-		case serveCmd:
-			appendEnvDocs(cmd, []envconfig.EnvVar{
-				envVars["OLLAMA_DEBUG"],
-				envVars["OLLAMA_HOST"],
-				envVars["OLLAMA_KEEP_ALIVE"],
-				envVars["OLLAMA_MAX_LOADED_MODELS"],
-				envVars["OLLAMA_MAX_QUEUE"],
-				envVars["OLLAMA_MODELS"],
-				envVars["OLLAMA_NUM_PARALLEL"],
-				envVars["OLLAMA_NOPRUNE"],
-				envVars["OLLAMA_ORIGINS"],
-				envVars["OLLAMA_TMPDIR"],
-			})
-		default:
-			appendEnvDocs(cmd, envs)
-		}
-	}
-
 	rootCmd.AddCommand(
 		serveCmd,
 		createCmd,
@@ -1265,7 +931,6 @@ func NewCLI() *cobra.Command {
 		pullCmd,
 		pushCmd,
 		listCmd,
-		psCmd,
 		copyCmd,
 		deleteCmd,
 	)
--- a/cmd/interactive.go
+++ b/cmd/interactive.go
@@ -6,19 +6,14 @@ import (
 	"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/envconfig"
-	"github.com/ollama/ollama/progress"
-	"github.com/ollama/ollama/readline"
-	"github.com/ollama/ollama/types/errtypes"
+	"github.com/jmorganca/ollama/api"
+	"github.com/jmorganca/ollama/readline"
 )

 type MultilineState int
@@ -30,88 +25,45 @@ const (
 	MultilineTemplate
 )

-func loadModel(cmd *cobra.Command, opts *runOptions) error {
+func modelIsMultiModal(cmd *cobra.Command, name string) bool {
+	// get model details
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
-		return err
+		fmt.Println("error: couldn't connect to ollama server")
+		return false
 	}

-	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)
+	req := api.ShowRequest{Name: name}
+	resp, err := client.Show(cmd.Context(), &req)
 	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...)
+		return false
 	}

-	chatReq := &api.ChatRequest{
-		Model:    opts.Model,
-		Messages: []api.Message{},
-	}
-
-	if opts.KeepAlive != nil {
-		chatReq.KeepAlive = opts.KeepAlive
-	}
-
-	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
+	return slices.Contains(resp.Details.Families, "clip")
 }

 func generateInteractive(cmd *cobra.Command, opts runOptions) error {
-	opts.Messages = make([]api.Message, 0)
+	multiModal := modelIsMultiModal(cmd, opts.Model)

-	err := loadModel(cmd, &opts)
-	if err != nil {
+	// load the model
+	loadOpts := runOptions{
+		Model:    opts.Model,
+		Prompt:   "",
+		Messages: []api.Message{},
+	}
+	if _, err := chat(cmd, loadOpts); 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, "  /clear          Clear session context")
-		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, "  /set          Set session variables")
+		fmt.Fprintln(os.Stderr, "  /show         Show model information")
+		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, "")
 	}

@@ -139,7 +91,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		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, "  Ctrl + w            Delete the word 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")
@@ -170,7 +121,7 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		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> <string> ...   Set the stop parameters")
+		fmt.Fprintln(os.Stderr, "  /set parameter stop \"<string>\", ...   Set the stop parameters")
 		fmt.Fprintln(os.Stderr, "")
 	}

@@ -184,15 +135,12 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		return err
 	}

-	if envconfig.NoHistory {
-		scanner.HistoryDisable()
-	}
-
 	fmt.Print(readline.StartBracketedPaste)
 	defer fmt.Printf(readline.EndBracketedPaste)

 	var sb strings.Builder
 	var multiline MultilineState
+	opts.Messages = make([]api.Message, 0)

 	for {
 		line, err := scanner.Readline()
@@ -226,7 +174,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 			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:
@@ -246,6 +193,7 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 				fmt.Fprintln(&sb)
 				multiline = MultilinePrompt
 				scanner.Prompt.UseAlt = true
+				break
 			}
 		case scanner.Pasting:
 			fmt.Fprintln(&sb, line)
@@ -255,55 +203,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 			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 {
-				if strings.Contains(err.Error(), errtypes.InvalidModelNameErrMsg) {
-					fmt.Printf("error: The model name '%s' is invalid\n", args[1])
-					continue
-				}
-				return err
-			}
-			fmt.Printf("Created new model '%s'\n", args[1])
-			continue
-		case strings.HasPrefix(line, "/clear"):
-			opts.Messages = []api.Message{}
-			if opts.System != "" {
-				newMessage := api.Message{Role: "system", Content: opts.System}
-				opts.Messages = append(opts.Messages, newMessage)
-			}
-			fmt.Println("Cleared session context")
-			continue
 		case strings.HasPrefix(line, "/set"):
 			args := strings.Fields(line)
 			if len(args) > 1 {
@@ -319,14 +218,10 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 					opts.WordWrap = false
 					fmt.Println("Set 'nowordwrap' mode.")
 				case "verbose":
-					if err := cmd.Flags().Set("verbose", "true"); err != nil {
-						return err
-					}
+					cmd.Flags().Set("verbose", "true")
 					fmt.Println("Set 'verbose' mode.")
 				case "quiet":
-					if err := cmd.Flags().Set("verbose", "false"); err != nil {
-						return err
-					}
+					cmd.Flags().Set("verbose", "false")
 					fmt.Println("Set 'quiet' mode.")
 				case "format":
 					if len(args) < 3 || args[2] != "json" {
@@ -382,21 +277,11 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 					}

 					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)
-						}
+						opts.System = sb.String()
 						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()
@@ -498,13 +383,13 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 			} else {
 				usage()
 			}
-		case strings.HasPrefix(line, "/exit"), strings.HasPrefix(line, "/bye"):
+		case line == "/exit", line == "/bye":
 			return nil
 		case strings.HasPrefix(line, "/"):
 			args := strings.Fields(line)
 			isFile := false

-			if opts.MultiModal {
+			if multiModal {
 				for _, f := range extractFileNames(line) {
 					if strings.HasPrefix(f, args[0]) {
 						isFile = true
@@ -526,23 +411,34 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		if sb.Len() > 0 && multiline == MultilineNone {
 			newMessage := api.Message{Role: "user", Content: sb.String()}

-			if opts.MultiModal {
+			if multiModal {
 				msg, images, err := extractFileData(sb.String())
 				if err != nil {
 					return err
 				}
+				newMessage.Content = msg

-				// clear all previous images for better responses
+				// reset the context if we find another image
 				if len(images) > 0 {
-					for i := range opts.Messages {
-						opts.Messages[i].Images = nil
+					newMessage.Images = append(newMessage.Images, images...)
+					// reset the context for the new image
+					opts.Messages = []api.Message{}
+				} else {
+					if len(opts.Messages) > 1 {
+						newMessage.Images = append(newMessage.Images, opts.Messages[len(opts.Messages)-2].Images...)
 					}
 				}
-
-				newMessage.Content = msg
-				newMessage.Images = images
+				if len(newMessage.Images) == 0 {
+					fmt.Println("This model requires you to add a jpeg, png, or svg image.")
+					fmt.Println()
+					sb.Reset()
+					continue
+				}
 			}

+			if opts.System != "" {
+				opts.Messages = append(opts.Messages, api.Message{Role: "system", Content: opts.System})
+			}
 			opts.Messages = append(opts.Messages, newMessage)

 			assistant, err := chat(cmd, opts)
@@ -558,38 +454,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 	}
 }

-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{
@@ -636,10 +500,10 @@ func extractFileData(input string) (string, []api.ImageData, error) {
 			if os.IsNotExist(err) {
 				continue
 			}
-			fmt.Fprintf(os.Stderr, "Couldn't process image: %q\n", err)
+			fmt.Printf("Couldn't process image: %q\n", err)
 			return "", imgs, err
 		}
-		fmt.Fprintf(os.Stderr, "Added image '%s'\n", nfp)
+		fmt.Printf("Added image '%s'\n", nfp)
 		input = strings.ReplaceAll(input, fp, "")
 		imgs = append(imgs, data)
 	}
@@ -660,7 +524,7 @@ func getImageData(filePath string) ([]byte, error) {
 	}

 	contentType := http.DetectContentType(buf)
-	allowedTypes := []string{"image/jpeg", "image/jpg", "image/png"}
+	allowedTypes := []string{"image/jpeg", "image/jpg", "image/svg+xml", "image/png"}
 	if !slices.Contains(allowedTypes, contentType) {
 		return nil, fmt.Errorf("invalid image type: %s", contentType)
 	}
--- a/cmd/interactive_test.go
+++ b/cmd/interactive_test.go
@@ -1,13 +1,9 @@
 package cmd

 import (
-	"bytes"
 	"testing"
-	"text/template"

 	"github.com/stretchr/testify/assert"
-
-	"github.com/ollama/ollama/api"
 )

 func TestExtractFilenames(t *testing.T) {
@@ -53,64 +49,3 @@ d:\path with\spaces\seven.svg inbetween7 c:\users\jdoe\eight.png inbetween8
 	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)
-}
--- a/cmd/start_darwin.go
+++ b/cmd/start_darwin.go
@@ -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)
-}
--- a/cmd/start_default.go
+++ b/cmd/start_default.go
@@ -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")
-}
--- a/cmd/start_windows.go
+++ b/cmd/start_windows.go
@@ -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)
-}
--- a/convert/convert.go
+++ b/convert/convert.go
@@ -1,200 +0,0 @@
-package convert
-
-import (
-	"cmp"
-	"encoding/binary"
-	"encoding/json"
-	"fmt"
-	"io"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"slices"
-	"strings"
-
-	"google.golang.org/protobuf/proto"
-
-	"github.com/ollama/ollama/convert/sentencepiece"
-	"github.com/ollama/ollama/llm"
-)
-
-const (
-	_ int32 = iota
-	tokenTypeNormal
-	tokenTypeUnknown
-	tokenTypeControl
-	tokenTypeUserDefined
-	tokenTypeUnused
-	tokenTypeByte
-)
-
-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"`
-
-	PreTokenizer string
-
-	ByteOrder
-}
-
-type ByteOrder interface {
-	binary.ByteOrder
-	binary.AppendByteOrder
-}
-
-type ModelArch interface {
-	GetTensors() error
-	LoadVocab() error
-	WriteGGUF(io.WriteSeeker) 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 {
-		if strings.HasSuffix(fn, ".safetensors") {
-			return &SafetensorFormat{}, nil
-		} else if strings.HasSuffix(fn, ".bin") || strings.HasSuffix(fn, ".pth") {
-			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
-	Merges []string
-}
-
-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, tokenTypeUserDefined)
-	}
-	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, tokenTypeUserDefined)
-		}
-	}
-
-	return v, nil
-}
--- a/convert/convert_test.go
+++ b/convert/convert_test.go
@@ -1,103 +0,0 @@
-//go:build slow
-
-package convert
-
-import (
-	"os"
-	"path/filepath"
-	"testing"
-
-	"github.com/ollama/ollama/llm"
-)
-
-func convertFull(t *testing.T, p string) (llm.KV, llm.Tensors) {
-	t.Helper()
-
-	mf, err := GetModelFormat(p)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	params, err := mf.GetParams(p)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	arch, err := mf.GetModelArch("", p, params)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	if err := arch.LoadVocab(); err != nil {
-		t.Fatal(err)
-	}
-
-	if err := arch.GetTensors(); err != nil {
-		t.Fatal(err)
-	}
-
-	f, err := os.CreateTemp(t.TempDir(), "f16")
-	if err != nil {
-		t.Fatal(err)
-	}
-	defer f.Close()
-
-	if err := arch.WriteGGUF(f); err != nil {
-		t.Fatal(err)
-	}
-
-	r, err := os.Open(f.Name())
-	if err != nil {
-		t.Fatal(err)
-	}
-	defer r.Close()
-
-	m, _, err := llm.DecodeGGML(r)
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	return m.KV(), m.Tensors()
-}
-
-func TestConvertFull(t *testing.T) {
-	cases := []struct {
-		path    string
-		arch    string
-		tensors int
-		layers  int
-	}{
-		{"Meta-Llama-3-8B-Instruct", "llama", 291, 35},
-		{"Mistral-7B-Instruct-v0.2", "llama", 291, 35},
-		{"Mixtral-8x7B-Instruct-v0.1", "llama", 291, 35},
-		{"gemma-2b-it", "gemma", 164, 20},
-	}
-
-	for _, tt := range cases {
-		t.Run(tt.path, func(t *testing.T) {
-			p := filepath.Join("testdata", tt.path)
-			if _, err := os.Stat(p); err != nil {
-				t.Skipf("%s not found", p)
-			}
-
-			kv, tensors := convertFull(t, p)
-
-			if kv.Architecture() != tt.arch {
-				t.Fatalf("expected llama, got %s", kv.Architecture())
-			}
-
-			if kv.FileType().String() != "F16" {
-				t.Fatalf("expected F16, got %s", kv.FileType())
-			}
-
-			if len(tensors) != tt.tensors {
-				t.Fatalf("expected %d tensors, got %d", tt.tensors, len(tensors))
-			}
-
-			layers := tensors.Layers()
-			if len(layers) != tt.layers {
-				t.Fatalf("expected %d layers, got %d", tt.layers, len(layers))
-			}
-		})
-	}
-}
--- a/convert/gemma.go
+++ b/convert/gemma.go
@@ -1,103 +0,0 @@
-package convert
-
-import (
-	"fmt"
-	"io"
-	"log/slog"
-	"strings"
-
-	"github.com/pdevine/tensor"
-	"github.com/pdevine/tensor/native"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type GemmaModel struct {
-	ModelData
-}
-
-func addOnes(data []float32, vectorSize int) ([]float32, error) {
-	n := tensor.New(tensor.WithShape(vectorSize), tensor.WithBacking(data))
-	ones := tensor.Ones(tensor.Float32, vectorSize)
-
-	n, err := n.Add(ones)
-	if err != nil {
-		return nil, err
-	}
-
-	ts, err := native.SelectF32(n, 0)
-	if err != nil {
-		return nil, err
-	}
-
-	var f32s []float32
-	for _, t := range ts {
-		f32s = append(f32s, t...)
-	}
-
-
-	return f32s, 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)))
-	for _, l := range t {
-		if strings.HasSuffix(l.Name, "norm.weight") {
-			wt := l.WriterTo.(safetensorWriterTo)
-			wt.repacker = m.Repack
-			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) Repack(_ string, data []float32, shape []uint64) ([]float32, error) {
-	return addOnes(data, int(shape[0]))
-}
-
-func (m *GemmaModel) WriteGGUF(ws io.WriteSeeker) 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,
-	}
-
-	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
-}
--- a/convert/llama.go
+++ b/convert/llama.go
@@ -1,158 +0,0 @@
-package convert
-
-import (
-	"cmp"
-	"errors"
-	"fmt"
-	"io"
-	"os"
-	"path/filepath"
-	"regexp"
-	"strings"
-
-	"github.com/pdevine/tensor"
-	"github.com/pdevine/tensor/native"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type LlamaModel struct {
-	ModelData
-}
-
-func (m *LlamaModel) GetTensors() error {
-	t, err := m.Format.GetTensors(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-
-	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 {
-			switch m.Format.(type) {
-			case *TorchFormat:
-				wt := l.WriterTo.(torchWriterTo)
-				wt.repacker = m.Repack
-				l.WriterTo = wt
-			case *SafetensorFormat:
-				wt := l.WriterTo.(safetensorWriterTo)
-				wt.repacker = m.Repack
-				l.WriterTo = wt
-			}
-		}
-		m.Tensors = append(m.Tensors, l)
-	}
-
-	return nil
-}
-
-func (m *LlamaModel) LoadVocab() (err error) {
-	pre, ts, merges, err := parseTokens(filepath.Join(m.Path, "tokenizer.json"))
-	if errors.Is(err, os.ErrNotExist) {
-		return nil
-	} else if err != nil {
-		return err
-	}
-
-	m.Vocab = &Vocab{}
-	for _, t := range ts {
-		m.Vocab.Tokens = append(m.Vocab.Tokens, t.Content)
-		m.Vocab.Types = append(m.Vocab.Types, t.Type())
-	}
-
-	m.Vocab.Merges = merges
-	m.Params.PreTokenizer = pre
-	return nil
-}
-
-func (m *LlamaModel) WriteGGUF(ws io.WriteSeeker) 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.freq_base":                   float32(m.Params.RopeFrequencyBase),
-		"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":                   "gpt2",
-
-		"tokenizer.ggml.pre":        m.Params.PreTokenizer,
-		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
-		"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),
-	}
-
-	if len(m.Vocab.Merges) > 0 {
-		kv["tokenizer.ggml.merges"] = m.Vocab.Merges
-	} else {
-		kv["tokenizer.ggml.scores"] = m.Vocab.Scores
-	}
-
-	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
-}
-
-func (m *LlamaModel) Repack(name string, data []float32, shape []uint64) ([]float32, error) {
-	return llamaRepack(name, m.Params, data, shape)
-}
-
-func llamaRepack(name string, params *Params, data []float32, shape []uint64) ([]float32, error) {
-	var dims []int
-	for _, dim := range shape {
-		if dim != 0 {
-			dims = append(dims, int(dim))
-		}
-	}
-
-	var heads int
-	if strings.HasSuffix(name, "attn_q.weight") {
-		heads = params.AttentionHeads
-	} else if strings.HasSuffix(name, "attn_k.weight") {
-		heads = cmp.Or(params.KeyValHeads, params.AttentionHeads)
-	} else {
-		return nil, fmt.Errorf("unknown tensor name: %s", name)
-	}
-
-	n := tensor.New(tensor.WithShape(dims...), tensor.WithBacking(data))
-	if err := n.Reshape(append([]int{heads, 2, dims[0] / heads / 2}, dims[1:]...)...); err != nil {
-		return nil, err
-	}
-
-	if err := n.T(0, 2, 1, 3); err != nil {
-		return nil, err
-	}
-
-	if err := n.Reshape(dims...); err != nil {
-		return nil, err
-	}
-
-	if err := n.Transpose(); err != nil {
-		return nil, err
-	}
-
-	ts, err := native.SelectF32(n, 1)
-	if err != nil {
-		return nil, err
-	}
-
-	var f32s []float32
-	for _, t := range ts {
-		f32s = append(f32s, t...)
-	}
-
-	return f32s, nil
-}
--- a/convert/mistral.go
+++ b/convert/mistral.go
@@ -1,79 +0,0 @@
-package convert
-
-import (
-	"io"
-	"regexp"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type MistralModel struct {
-	ModelData
-}
-
-func (m *MistralModel) GetTensors() error {
-	t, err := m.Format.GetTensors(m.Path, m.Params)
-	if err != nil {
-		return err
-	}
-
-	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.repacker = m.Repack
-			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(ws io.WriteSeeker) 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),
-	}
-
-	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
-}
-
-func (m *MistralModel) Repack(name string, data []float32, shape []uint64) ([]float32, error) {
-	return llamaRepack(name, m.Params, data, shape)
-}
--- a/convert/mixtral.go
+++ b/convert/mixtral.go
@@ -1,87 +0,0 @@
-package convert
-
-import (
-	"io"
-	"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
-	}
-
-	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.repacker = m.Repack
-			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(ws io.WriteSeeker) 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,
-	}
-
-	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
-}
-
-func (m *MixtralModel) Repack(name string, data []float32, shape []uint64) ([]float32, error) {
-	return llamaRepack(name, m.Params, data, shape)
-}
--- a/convert/safetensors.go
+++ b/convert/safetensors.go
@@ -1,309 +0,0 @@
-package convert
-
-import (
-	"bytes"
-	"encoding/binary"
-	"encoding/json"
-	"fmt"
-	"io"
-	"os"
-	"path/filepath"
-	"regexp"
-	"slices"
-	"strings"
-
-	"github.com/d4l3k/go-bfloat16"
-	"github.com/x448/float16"
-
-	"github.com/ollama/ollama/llm"
-)
-
-type safetensorWriterTo struct {
-	t *llm.Tensor
-
-	params *Params
-	bo     ByteOrder
-
-	filename string
-	dtype    string
-
-	offset, size int64
-	repacker     func(string, []float32, []uint64) ([]float32, error)
-}
-
-type safetensorMetadata struct {
-	Type    string   `json:"dtype"`
-	Shape   []uint64 `json:"shape"`
-	Offsets []int64  `json:"data_offsets"`
-}
-
-type SafetensorFormat struct{}
-
-func (m *SafetensorFormat) GetTensors(dirpath string, params *Params) ([]llm.Tensor, error) {
-	var tensors []llm.Tensor
-	matches, err := filepath.Glob(filepath.Join(dirpath, "*.safetensors"))
-	if err != nil {
-		return nil, err
-	}
-
-	var offset uint64
-	for _, f := range matches {
-		var t []llm.Tensor
-		var err error
-		t, offset, err = m.readTensors(f, offset, params)
-		if err != nil {
-			return nil, err
-		}
-
-		tensors = append(tensors, t...)
-	}
-	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 n int64
-	if err := binary.Read(f, binary.LittleEndian, &n); err != nil {
-		return nil, 0, err
-	}
-
-	b := bytes.NewBuffer(make([]byte, 0, n))
-	if _, err = io.CopyN(b, f, n); err != nil {
-		return nil, 0, err
-	}
-
-	var headers map[string]safetensorMetadata
-	if err := json.NewDecoder(b).Decode(&headers); err != nil {
-		return nil, 0, err
-	}
-
-	var keys []string
-	for key := range headers {
-		if !strings.HasSuffix(key, "self_attn.rotary_embd.inv_freq") {
-			keys = append(keys, key)
-		}
-	}
-
-	slices.Sort(keys)
-
-	var tensors []llm.Tensor
-	for _, key := range keys {
-		value := headers[key]
-
-		var kind uint32
-		switch len(value.Shape) {
-		case 0:
-			// valuedata
-			continue
-		case 2:
-			kind = 1
-		}
-
-		name, err := m.GetLayerName(key)
-		if err != nil {
-			return nil, 0, err
-		}
-
-		shape := make([]uint64, len(value.Shape))
-		copy(shape, value.Shape)
-
-		pad := func(s int64) int64 {
-			return 8 + n + s
-		}
-
-		t := llm.Tensor{
-			Name:   name,
-			Kind:   kind,
-			Offset: offset,
-			Shape:  shape[:],
-		}
-
-		t.WriterTo = safetensorWriterTo{
-			t:        &t,
-			params:   params,
-			bo:       params.ByteOrder,
-			filename: fn,
-			dtype:    value.Type,
-			offset:   pad(value.Offsets[0]),
-			size:     pad(value.Offsets[1]) - pad(value.Offsets[0]),
-		}
-
-		offset += t.Size()
-		tensors = append(tensors, t)
-	}
-
-	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
-
-	if err := json.NewDecoder(f).Decode(&params); 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(r.offset, io.SeekStart); err != nil {
-		return 0, err
-	}
-
-	var f32s []float32
-	switch r.dtype {
-	case "F32":
-		f32s = make([]float32, r.size/4)
-		if err = binary.Read(f, r.bo, f32s); err != nil {
-			return 0, err
-		}
-	case "F16":
-		u16s := make([]uint16, r.size/2)
-		if err = binary.Read(f, r.bo, u16s); err != nil {
-			return 0, err
-		}
-
-		for _, b := range u16s {
-			f32s = append(f32s, float16.Frombits(b).Float32())
-		}
-
-	case "BF16":
-		u8s := make([]uint8, r.size)
-		if err = binary.Read(f, r.bo, u8s); err != nil {
-			return 0, err
-		}
-
-		f32s = bfloat16.DecodeFloat32(u8s)
-	default:
-		return 0, fmt.Errorf("unknown data type: %s", r.dtype)
-	}
-
-	if r.repacker != nil {
-		f32s, err = r.repacker(r.t.Name, f32s, r.t.Shape)
-		if err != nil {
-			return 0, err
-		}
-	}
-
-	switch r.t.Kind {
-	case 0:
-		return 0, binary.Write(w, r.bo, f32s)
-	case 1:
-		f16s := make([]uint16, len(f32s))
-		for i := range f32s {
-			f16s[i] = float16.Fromfloat32(f32s[i]).Bits()
-		}
-
-		return 0, binary.Write(w, r.bo, f16s)
-	default:
-		return 0, fmt.Errorf("unknown storage type: %d", r.t.Kind)
-	}
-}
-
-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 "LlamaForCausalLM":
-			return &LlamaModel{
-				ModelData{
-					Name:   name,
-					Path:   dirPath,
-					Params: params,
-					Format: m,
-				},
-			}, nil
-		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")
-}
--- a/Show More
+++ b/Show More