achievements: community log, privacy filter and a dry-run preview

Phase 0 of the community plan: the car works out what it would share and writes it to files. Nothing is uploaded. - community.py keeps community.db next to state.json: 6 km map tiles (H3 resolution 6) with the day first seen, the tile where each drive starts and ends, engaged distance and time per day, and engaged distance per region per day. Positions only ever become tiles. - The privacy filter follows the build plan: a tile where drives start or end on 3+ days is a frequent stop and it and its 6 neighbors never leave the car; other start/end tiles wait 14 days; hidden tiles never leave; tile batches are sorted with no dates; tiles that become private after being sent turn into delete requests; home region is where you drove most on most of the last 90 days, with detected changes sent at most every 30 days. - preview.py writes the daily reports, the tile batch, delete requests and a map page (your 6 km map, the 45 km blocks others see, what stays on the car) to achievements/preview/ after every drive. - h3lite.py ports the H3 functions we need to pure Python, since sunnypilot doesn't add packages. Checked against h3-py on 2.8M points: identical cells, centers, parents and rings, including pentagons, poles and the antimeridian. - regions.json.gz (207 KB) maps tiles to ISO 3166-2 codes for US, CA, MX, BR, AU, DE and country codes elsewhere, from Natural Earth (public domain). tools/build_regions.py rebuilds it. - replay.py runs everything over qlogs already on the car, into its own folder. - achievementsd feeds the log each sample; any community error turns it off for the drive without touching achievements. 92 tests pass (engine, store, H3 port, privacy filter, replay feeder). The daemon itself was only compiled and linted here; cereal and msgq are not built in this environment. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ATDgLQKGkv2ToqFZNWSF8j
achievements: count driving stats and unlock achievements on the device
2026-11-10 22:33:43 +08:00 · 2026-11-10 13:58:14 +00:00 · 2026-11-10 13:29:45 +00:00 · 2026-11-09 21:00:33 -08:00 · 2026-11-09 01:13:29 -08:00 · 2026-11-09 00:38:20 -08:00
51 changed files with 4167 additions and 59 deletions
@@ -14,3 +14,8 @@ openpilot/sunnypilot/modeld_v2/models/*.pkl filter=lfs diff=lfs merge=lfs -text

 openpilot/selfdrive/car/tests/test_models_segs.txt filter=lfs diff=lfs merge=lfs -text
 openpilot/common/hardware/comma/updater filter=lfs diff=lfs merge=lfs -text
+
+# Achievements icons are kept out of LFS on this dev branch (no push access to the LFS store).
+# Move them into LFS before this goes upstream.
+openpilot/sunnypilot/selfdrive/assets/icons_mici/achievements.png -filter binary
+openpilot/sunnypilot/selfdrive/assets/offroad/icon_achievements.png -filter binary
@@ -34,6 +34,11 @@ on:
        required: false
        type: string
        default: 'qcom'
+      on_device:
+        description: 'Compile on the tici / chestnut runner instead of on a hosted runner with tinygrad''s qcom / amd mock'
+        required: false
+        type: boolean
+        default: false
      hf_repo:
        description: 'Hugging Face dataset repository (e.g. sunnypilot/sunnypilot_models_v1)'
        required: false
@@ -107,6 +112,11 @@ on:
        options:
          - qcom
          - chestnut
+      on_device:
+        description: 'Compile on the tici / chestnut runner instead of on a hosted runner with tinygrad''s qcom / amd mock'
+        required: false
+        type: boolean
+        default: false
      hf_repo:
        description: 'Hugging Face dataset repository'
        required: false
@@ -130,6 +140,7 @@ jobs:
      is_20hz: ${{ inputs.is_20hz }}
      artifact_suffix: ${{ inputs.artifact_suffix }}
      target_hardware: ${{ inputs.target_hardware }}
+      on_device: ${{ inputs.on_device || false }}
    secrets: inherit

  publish_model:
@@ -0,0 +1,61 @@
+name: Chestnut mock build parity
+description: >-
+  Convert one commaai precompiled chestnut model (LS2) on this runner under tinygrad's amd mock, as a hosted chestnut build does, and require the
+  exact bytes the chestnut runner built (the published artifact). Needs setup-amd-mock, numpy and zstandard; tinygrad_repo checked out.
+runs:
+  using: composite
+  steps:
+    - name: Key this check by everything its result depends on
+      id: key
+      shell: bash
+      env:
+        SOURCES: ${{ hashFiles('tinygrad_repo/tinygrad/**/*.py', 'tinygrad_repo/test/mockgpu/**', 'openpilot/sunnypilot/modeld_v2/*.py', '.github/workflows/chestnut-build-parity/action.yml') }}
+      run: echo "key=chestnut-build-parity-$SOURCES-$(sha256sum "$LLVM_PATH" | cut -c1-16)-$(python3 -c 'import platform; print(platform.python_version())')" >> "$GITHUB_OUTPUT"
+
+    - name: Passed before with all of it identical?
+      id: passed
+      uses: actions/cache/restore@v4
+      with:
+        path: ~/.chestnut_build_parity.passed
+        key: ${{ steps.key.outputs.key }}
+
+    - name: Rebuild LS2 on the mock and require the chestnut runner's artifact byte for byte
+      if: steps.passed.outputs.cache-hit != 'true'
+      shell: bash
+      env:
+        REF: b92fc4828664f8767bce5e92f1630ed319cbe509  # commaai/openpilot: precompiled big_driving_tinygrad.pkl, pins tinygrad ec91b74b5b
+        PKL_OID: b4dead3ee091d66ab002a6b8e716c26f4f79e3c931e4be5f48e017ba92f6e960
+        PKL_SIZE: "767228907"
+        TG_SHA: ec91b74b5b6b0fc924b72d3f27689679ca2438b9
+        WANT: 680a8d629785f2b1d1a2b0e40d0dfbcd0f7291c21e20c51259f1d768a22ec217  # driving_light_speed_2_november_04_2026_tinygrad.pkl, recompiled46
+      run: |
+        set -euo pipefail
+        work="$RUNNER_TEMP/chestnut_parity" && rm -rf "$work" && mkdir -p "$work"
+        href=$(curl -fsSL --retry 3 -X POST -H "Accept: application/vnd.git-lfs+json" -H "Content-Type: application/vnd.git-lfs+json" \
+          https://huggingface.co/commaai/openpilot-lfs.git/info/lfs/objects/batch \
+          -d "{\"operation\":\"download\",\"transfer\":[\"basic\"],\"objects\":[{\"oid\":\"$PKL_OID\",\"size\":$PKL_SIZE}]}" | jq -r '.objects[0].actions.download.href')
+        curl -fsSL --retry 3 -o "$work/big_driving_tinygrad.pkl" "$href"
+        echo "$PKL_OID  $work/big_driving_tinygrad.pkl" | sha256sum -c -
+        tg="$work/tinygrad" && git init -q "$tg" && git -C "$tg" fetch -q --depth 1 https://github.com/tinygrad/tinygrad "$TG_SHA" && git -C "$tg" checkout -q FETCH_HEAD
+        CHECK_OOB=0 GMMU=0 PYTHONPATH="$GITHUB_WORKSPACE/tinygrad_repo" python3 tinygrad_repo/test/mockgpu/usb_server.py \
+          env DEV=USB+AMD PYTHONPATH="$tg:$GITHUB_WORKSPACE" python3 openpilot/sunnypilot/modeld_v2/convert_upstream_pkl.py "$work/big_driving_tinygrad.pkl" "$work/LS2.pkl"
+        got=$(sha256sum "$work/LS2.pkl" | cut -d' ' -f1)
+        if [ "$got" != "$WANT" ]; then
+          echo "::error::the mock's LS2 build ($got) is not the chestnut runner's ($WANT): this runner's LLVM or the mock no longer builds what the chestnut does"
+          exit 1
+        fi
+        echo "LS2 rebuilt on the mock: byte-identical to the chestnut runner's build"
+        touch ~/.chestnut_build_parity.passed
+
+    - name: Remember the pass
+      if: steps.passed.outputs.cache-hit != 'true'
+      uses: actions/cache/save@v4
+      with:
+        path: ~/.chestnut_build_parity.passed
+        key: ${{ steps.key.outputs.key }}
+
+    - name: Already passed
+      if: steps.passed.outputs.cache-hit == 'true'
+      shell: bash
+      run: |
+        echo "LS2 rebuild skipped: it already matched the chestnut runner's build with this exact mock, converter, LLVM and python"
@@ -0,0 +1,60 @@
+name: Set up tinygrad's amd mock
+description: LLVM 20 and ROCm comgr, as tinygrad_repo/.github/actions/setup-tinygrad installs them for the mock gpu
+inputs:
+  emu_cache:
+    description: >-
+      Keep tinygrad's cache db (the emulator's compiled kernels and gate results) between runs under this name, e.g. the model's file name.
+      Empty: no cache.
+    default: ''
+runs:
+  using: composite
+  steps:
+    - name: Install LLVM 20 and ROCm comgr
+      shell: bash
+      run: |
+        . /etc/os-release
+        curl -fsSL https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc > /dev/null
+        echo "deb http://apt.llvm.org/$VERSION_CODENAME/ llvm-toolchain-$VERSION_CODENAME-20 main" | sudo tee /etc/apt/sources.list.d/llvm.list
+        sudo mkdir -p /etc/apt/keyrings
+        curl -fsSL https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor | sudo tee /etc/apt/keyrings/rocm.gpg > /dev/null
+        echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/7.1 $VERSION_CODENAME main" | sudo tee /etc/apt/sources.list.d/rocm.list
+        printf 'Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600\n' | sudo tee /etc/apt/preferences.d/rocm-pin-600
+        sudo apt-get -qq update && sudo apt-get -qq install -y --no-install-recommends libllvm20 clang-20 lld-20 comgr
+        printf '/opt/rocm/lib\n/opt/rocm/lib64\n' | sudo tee /etc/ld.so.conf.d/rocm.conf && sudo ldconfig
+        echo "/usr/lib/llvm-20/bin" >> "$GITHUB_PATH"
+        llvm=$(ls /usr/lib/llvm-20/lib/libLLVM*.so* /usr/lib/*-linux-gnu/libLLVM-20.so* 2>/dev/null | head -1 || true)
+        [ -n "$llvm" ] || { echo "::error::libLLVM 20 not found"; exit 1; }
+        echo "LLVM_PATH=$llvm" >> "$GITHUB_ENV"
+
+    - name: Cache the chestnut's firmware
+      uses: actions/cache@v4
+      with:
+        path: ~/.cache/tinygrad/downloads/fw
+        key: chestnut-fw-${{ hashFiles('tinygrad_repo/tinygrad/runtime/autogen/am/fw.py') }}
+
+    - name: Fetch the chestnut's firmware (gitlab times out now and then; the mock boots from it)
+      shell: bash
+      run: |
+        for i in 1 2 3 4 5; do
+          PYTHONPATH="$GITHUB_WORKSPACE/tinygrad_repo" python3 -c "
+        from tinygrad.helpers import fetch_fw
+        from tinygrad.runtime.autogen.am import fw
+        for n, h in fw.hashes.items():
+          if n.startswith(('psp_14_0_2_', 'smu_14_0_2.', 'sdma_7_0_0.', 'gc_12_0_0_')): fetch_fw('amdgpu', n, h)" && exit 0
+          sleep $((i * 10))
+        done
+        echo "::error::could not fetch the chestnut's firmware from gitlab"; exit 1
+
+    - name: runner cpu
+      if: inputs.emu_cache != ''
+      id: cpu
+      shell: bash
+      run: echo "cpu=$( (grep -m1 '^model name' /proc/cpuinfo; grep -m1 '^flags' /proc/cpuinfo) | sha256sum | cut -c1-16)" >> "$GITHUB_OUTPUT"
+
+    - name: Keep the emulator's compiled kernels between runs
+      if: inputs.emu_cache != ''
+      uses: actions/cache@v4
+      with:
+        path: ~/.cache/tinygrad/cache.db
+        key: emu-${{ steps.cpu.outputs.cpu }}-${{ hashFiles('tinygrad_repo/tinygrad/**/*.py', 'tinygrad_repo/test/mockgpu/**') }}-${{ inputs.emu_cache }}
+        restore-keys: emu-${{ steps.cpu.outputs.cpu }}-${{ hashFiles('tinygrad_repo/tinygrad/**/*.py', 'tinygrad_repo/test/mockgpu/**') }}-
@@ -35,6 +35,11 @@ on:
        required: false
        type: string
        default: 'qcom'
+      on_device:
+        description: 'Compile on the tici / chestnut runner instead of on a hosted runner with tinygrad''s qcom / amd mock'
+        required: false
+        type: boolean
+        default: false
  workflow_dispatch:
    inputs:
      upstream_branch:
@@ -59,6 +64,11 @@ on:
          - qcom
          - chestnut
        default: 'qcom'
+      on_device:
+        description: 'Compile on the tici / chestnut runner instead of on a hosted runner with tinygrad''s qcom / amd mock'
+        required: false
+        type: boolean
+        default: false


 run-name: Build model [${{ inputs.custom_name || inputs.upstream_branch }}] from ref [${{ inputs.upstream_branch }}]
@@ -153,18 +163,67 @@ jobs:
          if-no-files-found: error

  build_model:
-    runs-on: [self-hosted, "${{ inputs.target_hardware == 'chestnut' && 'chestnut' || 'tici' }}"]
+    runs-on: ${{ inputs.on_device && fromJSON(inputs.target_hardware == 'chestnut' && '["self-hosted","chestnut"]' || '["self-hosted","tici"]') || 'ubuntu-24.04' }}
    needs: get_model
    env:
      MODEL_NAME: ${{ inputs.custom_name || inputs.upstream_branch }} (${{ needs.get_model.outputs.model_date }})
      REF: ${{ inputs.upstream_branch }}
+      MOCK_BUILD: ${{ !inputs.on_device }}
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 1
          submodules: recursive

+      - name: Set up the hosted mock build
+        if: env.MOCK_BUILD == 'true'
+        uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+
+      - name: Install LLVM 20 and ROCm comgr for the amd mock
+        if: env.MOCK_BUILD == 'true' && inputs.target_hardware == 'chestnut'
+        uses: ./.github/workflows/setup-amd-mock
+        with:
+          emu_cache: build
+
+      - name: Install the amd mock build dependencies
+        if: env.MOCK_BUILD == 'true' && inputs.target_hardware == 'chestnut'
+        run: |
+          python3 -m pip install -q numpy zstandard
+
+      - name: Install the qcom mock build dependencies
+        if: env.MOCK_BUILD == 'true' && inputs.target_hardware != 'chestnut'
+        id: qcom_deps
+        run: |
+          sudo apt-get -qq update && sudo apt-get -qq install -y --no-install-recommends qemu-user-static aria2
+          python3 -m pip install -q numpy zstandard tinymesa==25.2.7.2
+          echo "qemu=$(dpkg-query -W -f='${Version}' qemu-user-static)" >> "$GITHUB_OUTPUT"
+
+      - name: Compiler check passed before with all of it identical?
+        if: env.MOCK_BUILD == 'true' && inputs.target_hardware != 'chestnut'
+        id: qcom_compiler
+        uses: actions/cache/restore@v4
+        with:
+          path: ~/.qcom_compiler_parity.passed
+          key: qcom-compiler-parity-${{ steps.qcom_deps.outputs.qemu }}-${{ hashFiles('tinygrad_repo/test/device/qcom/compiler_parity.py', 'tinygrad_repo/test/mockgpu/qcom/a630_compiled.pkl.xz', 'tinygrad_repo/tinygrad/runtime/support/compiler_qcom.py') }}
+
+      - name: Check the off-device qcom compiler builds the a630's bytes
+        if: env.MOCK_BUILD == 'true' && inputs.target_hardware != 'chestnut' && steps.qcom_compiler.outputs.cache-hit != 'true'
+        run: |
+          # every kernel an a630 has compiled must compile to the same bytes here before a model is built here
+          PYTHONPATH="$TINYGRAD_PATH" python3 "$TINYGRAD_PATH/test/device/qcom/compiler_parity.py" check "$TINYGRAD_PATH/test/mockgpu/qcom/a630_compiled.pkl.xz"
+          touch ~/.qcom_compiler_parity.passed
+
+      - name: Remember the compiler check passed
+        if: env.MOCK_BUILD == 'true' && inputs.target_hardware != 'chestnut' && steps.qcom_compiler.outputs.cache-hit != 'true'
+        uses: actions/cache/save@v4
+        with:
+          path: ~/.qcom_compiler_parity.passed
+          key: ${{ steps.qcom_compiler.outputs.cache-primary-key }}
+
      - name: Set environment variables
+        if: env.MOCK_BUILD != 'true'
        id: set-env
        run: |
          # Set up common environment
@@ -177,6 +236,7 @@ jobs:
          printenv >> $GITHUB_ENV

      - name: Setup build environment
+        if: env.MOCK_BUILD != 'true'
        run: |
          mkdir -p "${BUILD_DIR}/"
          sudo find $BUILD_DIR/ -mindepth 1 -delete
@@ -198,21 +258,28 @@ jobs:

      - name: Build Model
        run: |
-          source /etc/profile
-          export UV_PROJECT_ENVIRONMENT=${HOME}/venv
-          export VIRTUAL_ENV=$UV_PROJECT_ENVIRONMENT
-          source ${UV_PROJECT_ENVIRONMENT}/bin/activate
+          if [ "$MOCK_BUILD" = "true" ]; then
+            TASKSET=""
+            export CHECK_OOB=0 GMMU=0  # as test/mockgpu runs the amd mock
+          else
+            source /etc/profile
+            export UV_PROJECT_ENVIRONMENT=${HOME}/venv
+            export VIRTUAL_ENV=$UV_PROJECT_ENVIRONMENT
+            source ${UV_PROJECT_ENVIRONMENT}/bin/activate
+            TASKSET="taskset -c 7"
+          fi
          export PYTHONPATH="${PYTHONPATH}:${{ env.TINYGRAD_PATH }}:${{ github.workspace }}"

          COMPILE_MODELD="${{ github.workspace }}/openpilot/sunnypilot/modeld_v2/compile_modeld.py"
          MODEL_SIZE=$(python3 -c "from openpilot.common.transformations.model import MEDMODEL_INPUT_SIZE as s; print(f'{s[0]}x{s[1]}')")
          CAMERA_RES=$(python3 -c "from openpilot.common.transformations.camera import _ar_ox_fisheye as a, _os_fisheye as o; print(f'{a.width}x{a.height} {o.width}x{o.height}')")

-          TG_FLAGS_QCOM="DEV=QCOM IMAGE=1 FLOAT16=1 NOLOCALS=1 OPENPILOT_HACKS=1"
+          MOCK=$([ "$MOCK_BUILD" = "true" ] && echo MOCK || true)
+          TG_FLAGS_QCOM="DEV=${MOCK:+MOCK+}QCOM IMAGE=1 FLOAT16=1 NOLOCALS=1 OPENPILOT_HACKS=1"
          if [ "${{ inputs.target_hardware }}" == "chestnut" ]; then
            echo "CHESTNUT build"
            export CHESTNUT=1
-            TG_FLAGS="DEBUG=1 DEV=USB+AMD:LLVM FRAME_DEV=CPU FLOAT16=1 GMMU=0 TC_OPT=2 TC_MIN_GLOBALS=32"
+            TG_FLAGS="DEBUG=1 DEV=${MOCK}USB+AMD:LLVM FRAME_DEV=CPU FLOAT16=1 GMMU=0 TC_OPT=2 TC_MIN_GLOBALS=32"
            OUTPUT_PKL="${{ env.MODELS_DIR }}/big_driving_tinygrad.pkl"
          else
            echo "QCOM build"
@@ -227,14 +294,16 @@ jobs:
            echo "Converting commaai's precompiled model with tinygrad $TG_SHA"
            UPSTREAM_TG="${{ runner.temp }}/upstream_tinygrad"
            rm -rf "$UPSTREAM_TG" && git init -q "$UPSTREAM_TG" && git -C "$UPSTREAM_TG" fetch -q --depth 1 "$TG_URL" "$TG_SHA" && git -C "$UPSTREAM_TG" checkout -q FETCH_HEAD
+            ON_MOCK=""
+            if [ "$MOCK_BUILD" = "true" ]; then ON_MOCK="env PYTHONPATH=$TINYGRAD_PATH python3 $TINYGRAD_PATH/test/mockgpu/usb_server.py"; fi
            CONVERTED="${{ runner.temp }}/big_driving_tinygrad_plain.pkl"
-            env DEV=USB+AMD PYTHONPATH="$UPSTREAM_TG:${{ github.workspace }}" taskset -c 7 python3 openpilot/sunnypilot/modeld_v2/convert_upstream_pkl.py \
+            $ON_MOCK env DEV=USB+AMD PYTHONPATH="$UPSTREAM_TG:${{ github.workspace }}" $TASKSET python3 openpilot/sunnypilot/modeld_v2/convert_upstream_pkl.py \
              "${{ env.MODELS_DIR }}/big_driving_tinygrad.pkl" "$CONVERTED"
            mv "$CONVERTED" "$OUTPUT_PKL"
          elif NEW_ARCH_ONNX=$(find "${{ env.MODELS_DIR }}" -maxdepth 1 -name "${{ inputs.target_hardware == 'chestnut' && 'big_' || '' }}*driving*.onnx" -print -quit) &&
               [ -n "$NEW_ARCH_ONNX" ] && grep -qa new_img "$NEW_ARCH_ONNX"; then # the new architecture: one ONNX taking the warped frames as new_img
            echo "Compiling new architecture ONNX: $NEW_ARCH_ONNX"
-            env ${TG_FLAGS} taskset -c 7 python3 openpilot/sunnypilot/modeld_v2/compile_modeld_v2.py \
+            env ${TG_FLAGS} $TASKSET python3 openpilot/sunnypilot/modeld_v2/compile_modeld_v2.py \
              "$NEW_ARCH_ONNX" "$OUTPUT_PKL" --device-input "*" --benchmark-runs 1
          else
            echo "Detected onnx architecture. Routing to native compile_modeld.py."
@@ -284,7 +353,7 @@ jobs:

            if [ -n "$MODEL_TYPE" ]; then
              echo "Detected: $MODEL_TYPE -> $OUTPUT_PKL"
-              env ${TG_FLAGS} taskset -c 7 python3 "$COMPILE_MODELD" \
+              env ${TG_FLAGS} $TASKSET python3 "$COMPILE_MODELD" \
                --model-type $MODEL_TYPE \
                --model-size $MODEL_SIZE \
                --camera-resolutions $CAMERA_RES \
@@ -293,6 +362,13 @@ jobs:
            fi
          fi

+      - name: Replay its check frame on the mock and require the result bit for bit
+        if: env.MOCK_BUILD == 'true'
+        env:
+          DEV: ${{ inputs.target_hardware == 'chestnut' && 'MOCKUSB+AMD' || 'MOCK+QCOM' }}
+          PYTHONPATH: ${{ env.TINYGRAD_PATH }}:${{ github.workspace }}
+        run: python3 -m openpilot.sunnypilot.modeld_v2.mock_parity check "${{ env.MODELS_DIR }}/${{ inputs.target_hardware == 'chestnut' && 'big_' || '' }}driving_tinygrad.pkl"
+
      - name: Prepare Output
        run: |
          sudo rm -rf ${{ env.OUTPUT_DIR }}
@@ -304,7 +380,7 @@ jobs:
            --include='/${{ inputs.target_hardware == 'chestnut' && 'big_' || '' }}driving_tinygrad.pkl' \
            --exclude='*' \
            --delete-excluded \
-            --chown=comma:comma \
+            $([ "$MOCK_BUILD" = "true" ] || echo --chown=comma:comma) \
            ${{ env.MODELS_DIR }}/ ${{ env.OUTPUT_DIR }}/

          python3 "${{ github.workspace }}/release/ci/model_generator.py" \
@@ -333,7 +409,7 @@ jobs:
          path: ${{ env.OUTPUT_DIR }}/artifact_name.txt

      - name: Re-enable powersave
-        if: always()
+        if: always() && env.MOCK_BUILD != 'true'
        run: |
          source ${UV_PROJECT_ENVIRONMENT}/bin/activate
          PYTHONPATH=$PYTHONPATH:${{ github.workspace }}/ ${{ github.workspace }}/scripts/manage-powersave.py --enable
@@ -28,18 +28,7 @@ jobs:
        env:
          DEV: CPU
        run: tools/op.sh test -v openpilot/sunnypilot/modeld_v2/tests/test_plain_artifact.py
-      - name: Install LLVM 20 and ROCm comgr for the amd mock, as tinygrad_repo/.github/actions/setup-tinygrad does
-        run: |
-          . /etc/os-release
-          curl -fsSL https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc > /dev/null
-          echo "deb http://apt.llvm.org/$VERSION_CODENAME/ llvm-toolchain-$VERSION_CODENAME-20 main" | sudo tee /etc/apt/sources.list.d/llvm.list
-          sudo mkdir -p /etc/apt/keyrings
-          curl -fsSL https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor | sudo tee /etc/apt/keyrings/rocm.gpg > /dev/null
-          echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/7.1 $VERSION_CODENAME main" | sudo tee /etc/apt/sources.list.d/rocm.list
-          printf 'Package: *\nPin: release o=repo.radeon.com\nPin-Priority: 600\n' | sudo tee /etc/apt/preferences.d/rocm-pin-600
-          sudo apt-get -qq update && sudo apt-get -qq install -y --no-install-recommends libllvm20 clang-20 lld-20 comgr
-          printf '/opt/rocm/lib\n/opt/rocm/lib64\n' | sudo tee /etc/ld.so.conf.d/rocm.conf && sudo ldconfig
-          echo "/usr/lib/llvm-20/bin" >> "$GITHUB_PATH"
+      - uses: ./.github/workflows/setup-amd-mock
      - name: Convert a pickled USB+AMD jit and replay it under tinygrad's amd mock
        env:
          DEV: MOCKUSB+AMD
@@ -80,6 +69,8 @@ jobs:
    runs-on: ubuntu-latest
    outputs:
      models: ${{ steps.set-matrix.outputs.models }}
+      chestnut: ${{ steps.set-matrix.outputs.chestnut }}
+      qcom: ${{ steps.set-matrix.outputs.qcom }}
    steps:
      - uses: actions/checkout@v4
      - name: Fetch and Parse json
@@ -91,16 +82,22 @@ jobs:
          with open("openpilot/sunnypilot/models/fetcher.py", "r") as f:
            urls = re.findall(r"MODEL_URL(?:_CHESTNUT)?\s*=\s*[\"'"'"']([^\"'"'"']+)[\"'"'"']", f.read())

-          artifacts = []
+          artifacts, chestnut, qcom = [], [], []
          for url in urls:
            data = json.loads(urllib.request.urlopen(url).read())
            for bundle in data.get("bundles", []):
              for model in bundle.get("models", []):
                if "artifact" in model:
                  artifacts.append(model["artifact"])
+                  if "chestnut" in url:
+                    chestnut.append(model["artifact"])
+                  elif "tinygrad" in model["artifact"]["file_name"]:
+                    qcom.append(model["artifact"])

          with open(os.environ["GITHUB_OUTPUT"], "a") as f:
            f.write(f"models={json.dumps(artifacts)}\n")
+            f.write(f"chestnut={json.dumps(chestnut)}\n")
+            f.write(f"qcom={json.dumps(qcom)}\n")
          '

  test-model:
@@ -137,3 +134,81 @@ jobs:
          export PYTHONPATH="$GITHUB_WORKSPACE/tinygrad_repo:$GITHUB_WORKSPACE:$PYTHONPATH"
          python3 -m pip uninstall -y tinygrad || uv pip uninstall -y tinygrad --system || true
          python3 -m pytest openpilot/sunnypilot/modeld_v2/tests/test_models.py
+
+  chestnut-mock-parity:
+    name: Mock parity ${{ matrix.artifact.file_name }}
+    needs: generate-matrix
+    runs-on: ubuntu-24.04
+    strategy:
+      fail-fast: false
+      matrix:
+        artifact: ${{ fromJson(needs.generate-matrix.outputs.chestnut) }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: true
+      - uses: ./.github/workflows/setup-amd-mock
+        with:
+          emu_cache: ${{ matrix.artifact.file_name }}
+      - run: python3 -m pip install -q numpy
+      - name: Download model
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+          MODEL_URL: ${{ matrix.artifact.download_uri.url }}
+          MODEL_SHA: ${{ matrix.artifact.download_uri.sha256 }}
+        shell: bash
+        run: |
+          AUTH=()
+          if [ -n "$HF_TOKEN" ]; then AUTH=(-H "Authorization: Bearer $HF_TOKEN"); fi
+          curl --retry 3 --retry-all-errors -s -S -f -L "${AUTH[@]}" -o "$RUNNER_TEMP/model.pkl" "$MODEL_URL"
+          if [ -n "$MODEL_SHA" ]; then echo "$MODEL_SHA  $RUNNER_TEMP/model.pkl" | sha256sum -c -; fi
+      - name: Replay its check frame on the amd mock and require the chestnut's result bit for bit
+        env:
+          PYTHONPATH: ${{ github.workspace }}/tinygrad_repo:${{ github.workspace }}
+        run: python3 -m openpilot.sunnypilot.modeld_v2.mock_parity check "$RUNNER_TEMP/model.pkl"
+
+  qcom-mock-parity:
+    name: QCOM mock parity ${{ matrix.artifact.file_name }}
+    needs: generate-matrix
+    runs-on: ubuntu-24.04
+    strategy:
+      fail-fast: false
+      matrix:
+        artifact: ${{ fromJson(needs.generate-matrix.outputs.qcom) }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: true
+      - uses: actions/setup-python@v5
+        with:
+          python-version: '3.12'
+      - run: python3 -m pip install -q numpy tinymesa==25.2.7.2
+      - name: Download model
+        env:
+          HF_TOKEN: ${{ secrets.HF_TOKEN }}
+          MODEL_URL: ${{ matrix.artifact.download_uri.url }}
+          MODEL_SHA: ${{ matrix.artifact.download_uri.sha256 }}
+        shell: bash
+        run: |
+          AUTH=()
+          if [ -n "$HF_TOKEN" ]; then AUTH=(-H "Authorization: Bearer $HF_TOKEN"); fi
+          curl --retry 3 --retry-all-errors -s -S -f -L "${AUTH[@]}" -o "$RUNNER_TEMP/model.pkl" "$MODEL_URL"
+          if [ -n "$MODEL_SHA" ]; then echo "$MODEL_SHA  $RUNNER_TEMP/model.pkl" | sha256sum -c -; fi
+      - name: Replay its check frame on the qcom mock and require the comma device's result bit for bit
+        env:
+          DEV: MOCK+QCOM
+          PYTHONPATH: ${{ github.workspace }}/tinygrad_repo:${{ github.workspace }}
+        run: python3 -m openpilot.sunnypilot.modeld_v2.mock_parity check "$RUNNER_TEMP/model.pkl"
+
+  chestnut-mock-build-parity:
+    name: Chestnut build on the amd mock = chestnut runner build
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: true
+      - uses: ./.github/workflows/setup-amd-mock
+        with:
+          emu_cache: chestnut-build-parity
+      - run: python3 -m pip install -q numpy zstandard
+      - uses: ./.github/workflows/chestnut-build-parity
@@ -0,0 +1,95 @@
+# Achievements: phase 1, on the device
+
+## Status: built, tested off-device, not yet run on a car
+
+Phase 1 of the sunnypilot community plan. The device counts driving stats and unlocks achievements, and shows them on
+an Achievements screen. Nothing leaves the device yet; sharing, the community site, leaderboards, groups, the map and
+contests come in later phases on a separate backend.
+
+## Where it fits
+
+```mermaid
+flowchart LR
+  subgraph device[On the device: this phase]
+    cereal[cereal: carState, selfdriveState,<br/>selfdriveStateSP, modelV2,<br/>driverMonitoringState, GPS] --> d[achievementsd<br/>10 Hz, read-only]
+    d --> f[(state.json<br/>/data/media/0/achievements)]
+    f --> ui[Achievements screen<br/>comma four and comma 3X]
+    ui --> seen[(seen.json)]
+  end
+  f -. later phase: opt-in .-> backend[Achievements backend] --> site[Community site] & forum[Forum badges]
+```
+
+## How a drive is counted
+
+```mermaid
+flowchart TD
+  s[Sample every 0.1 s] --> g{gap over 1 s or<br/>speed outside 0-70 m/s?}
+  g -- yes --> skip[count nothing]
+  g -- no --> e{engaged?<br/>selfdriveState.active<br/>or MADS active}
+  e -- no --> lc
+  e -- yes --> add[add distance, time,<br/>day, model, area, MADS-only km]
+  add --> lc{lane change reached<br/>finishing?}
+  lc --> u[check every tier,<br/>record new unlocks]
+```
+
+- **Engaged** means `selfdriveState.active` (including driver override) or MADS lateral active. MADS-only distance
+  is MADS active without `selfdriveState.active`.
+- **Days** (Regular, Streak, Road Trip, Four Seasons) need a valid wall clock. Before the clock is set, distance still
+  counts but no day does. The calendar day comes from GPS longitude (15 degrees per hour), since the device clock runs
+  in UTC; without GPS it falls back to the system zone. Southern-hemisphere fixes flip the seasons.
+- **Explorer areas** are geohash cells of about 5 km, salted with a random per-device salt and SHA-256 hashed before
+  they are stored. No coordinates are written anywhere.
+- **Attentive** is the longest engaged stretch without a red (level three) driver monitoring alert.
+- **Dev Branch Brave** counts on any channel that is not a release channel, plus staging channels.
+
+## Catalog
+
+`openpilot/sunnypilot/achievements/catalog.json`, 14 achievements and 32 tiers. Unlock ids are `odometer.silver`, or
+just `first_light` for single-tier ones, so a later backend can map them one-to-one to badges.
+
+| Achievement | Measures | Tiers |
+| --- | --- | --- |
+| First Light | engaged km | 0.1 |
+| Odometer | engaged km | 100 / 1,000 / 10,000 / 50,000 |
+| Captain's Hours | engaged hours | 10 / 100 / 1,000 |
+| Regular | days engaged | 7 / 30 / 100 / 365 |
+| Streak | days in a row | 7 / 30 |
+| Steady Hands | MADS-only km | 100 / 1,000 |
+| Smooth Merger | assisted lane changes | 10 / 100 / 1,000 |
+| Road Trip | km in one day | 500 |
+| Explorer | 5 km areas | 5 / 25 / 100 |
+| Model Hopper | models with 50+ km | 3 / 5 / 10 |
+| Dev Branch Brave | km on dev or staging | 100 |
+| Attentive | hours without a red DM alert | 10 / 50 / 100 |
+| Four Seasons | seasons driven | 4 |
+| Linked Up | sunnylink on and registered | yes |
+
+Safety rules for anything added here: never reward speed, hard driving, night or fatigue driving, or fewer
+takeovers, and never show anything while driving.
+
+## Why a state file, not Params
+
+This is a prebuilt branch: a new Params key means rebuilding the ARM64 `libparams_c.so` (see
+`openpilot/tools/params/README.md`). Progress lives in `/data/media/0/achievements/state.json` instead, written
+atomically every 30 s and when the drive ends. It survives updates. The UI keeps its own `seen.json`, so the daemon and
+the UI never write the same file. A corrupt state file is moved to `state.corrupt.json` instead of being overwritten.
+
+## Files
+
+| File | What it is |
+| --- | --- |
+| `sunnypilot/achievements/catalog.json`, `catalog.py` | the goals, and display text for both UIs |
+| `sunnypilot/achievements/engine.py` | pure Python counting and unlocks, no cereal or Params |
+| `sunnypilot/achievements/achievementsd.py` | onroad process that feeds the engine from cereal |
+| `system/manager/process_config.py` | registers `achievementsd` as `only_onroad` |
+| `selfdrive/ui/sunnypilot/achievements_store.py` | shared reader for both screens |
+| `selfdrive/ui/sunnypilot/mici/layouts/achievements.py` | comma four screen, plus the settings button with "N new" |
+| `selfdrive/ui/sunnypilot/layouts/settings/achievements.py` | comma 3X settings panel |
+
+## Testing
+
+`pytest openpilot/sunnypilot/achievements/tests openpilot/selfdrive/ui/sunnypilot/tests/test_achievements_store.py`
+(26 tests, no device libraries needed). The screens and achievementsd were syntax- and lint-checked only: the UI and
+cereal libraries on this branch are ARM64 builds, so neither has run yet. First drive to check: the process starts
+onroad, `state.json` appears within 30 s of engaging, and the comma four settings show "achievements" with
+"2 new" (First Light, and Linked Up if sunnylink is on) after the drive.
@@ -28,6 +28,8 @@ class DesireHelper:
    self.alc = AutoLaneChangeController(self)
    self.lane_turn_controller = LaneTurnController(self)
    self.lane_turn_direction = TurnDirection.none
+    self.turn_frames = 0
+    self.turn_repulse = 98

  @staticmethod
  def get_lane_change_direction(CS):
@@ -95,8 +97,10 @@ class DesireHelper:
    self.prev_one_blinker = one_blinker and lateral_active

    if self.lane_turn_direction != TurnDirection.none:
-      self.desire = TURN_DESIRES[self.lane_turn_direction]
+      self.turn_frames += 1
+      self.desire = TURN_DESIRES[self.lane_turn_direction] if self.turn_frames % self.turn_repulse else log.Desire.none
    else:
+      self.turn_frames = 0
      self.desire = log.Desire.none
      if self.lane_change_state == LaneChangeState.laneChangeStarting:
        if self.lane_change_direction == LaneChangeDirection.left:
@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:65a08adc31d5c456219687d99b7bf5e44d61dae2d49ea67850e76105c7248cce
-size 60918562
+oid sha256:b043d3bbc556e3a872c0edde4081f99b5ba92cbfd98eaea2ee2cbb87e9f277ec
+size 82539558
@@ -0,0 +1,97 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+What both Achievements screens and the settings buttons read: the state file
+achievementsd writes, plus the UI's own list of unlocks already seen.
+No pyray here, so it can be tested without a display.
+"""
+import time
+from pathlib import Path
+
+from openpilot.sunnypilot.achievements.catalog import AchievementStatus, Catalog, format_amount, load_catalog, statuses
+from openpilot.sunnypilot.achievements.engine import State, load_seen, load_state, save_seen, seen_path, state_path
+
+REFRESH_INTERVAL_S = 5.0
+
+
+class AchievementsStore:
+  def __init__(self, state_file: Path | None = None, seen_file: Path | None = None, catalog: Catalog | None = None):
+    self._state_file = state_file
+    self._seen_file = seen_file
+    self.catalog = catalog or load_catalog()
+    self.state = State()
+    self._seen: set[str] = set()
+    self._mtime: float | None = None
+    self._last_check = 0.0
+    self.version = 0  # bumps on every reload or seen change, so several screens can share one store
+
+  def _state_path(self) -> Path:
+    return self._state_file or state_path()
+
+  def _seen_path(self) -> Path:
+    return self._seen_file or seen_path()
+
+  def refresh(self, force: bool = False) -> bool:
+    """Reload if the state file changed. Cheap enough to call every frame. Returns True when reloaded."""
+    now = time.monotonic()
+    if not force and now - self._last_check < REFRESH_INTERVAL_S:
+      return False
+    self._last_check = now
+    try:
+      mtime = self._state_path().stat().st_mtime
+    except OSError:
+      mtime = None
+    if not force and mtime == self._mtime:
+      return False
+    self._mtime = mtime
+    try:
+      self.state = load_state(self._state_path()) if mtime is not None else State()
+    except OSError:
+      return False
+    self._seen = load_seen(self._seen_path())
+    self.version += 1
+    return True
+
+  def statuses(self) -> list[AchievementStatus]:
+    return statuses(self.catalog, self.state.metrics(), self.state.unlocked)
+
+  def status(self, achievement_id: str) -> AchievementStatus | None:
+    return next((s for s in self.statuses() if s.achievement.id == achievement_id), None)
+
+  def unlock_ids(self, achievement_id: str) -> list[str]:
+    a = next((a for a in self.catalog.achievements if a.id == achievement_id), None)
+    return [a.unlock_id(t) for t in a.tiers] if a else []
+
+  def unlocked_count(self) -> tuple[int, int]:
+    all_ids = self.catalog.all_unlock_ids()
+    return sum(1 for uid in all_ids if uid in self.state.unlocked), len(all_ids)
+
+  def new_ids(self) -> set[str]:
+    return {uid for uid in self.state.unlocked if uid not in self._seen and uid in set(self.catalog.all_unlock_ids())}
+
+  def mark_all_seen(self) -> None:
+    new = self.new_ids()
+    if not new:
+      return
+    self._seen |= new
+    self.version += 1
+    try:
+      save_seen(self._seen, self._seen_path())
+    except OSError:
+      pass
+
+  def totals_text(self, is_metric: bool = True) -> str:
+    m = self.state.metrics()
+    unlocked, total = self.unlocked_count()
+    return "\n".join([
+      f"{format_amount(m['engaged_km'], 'km', is_metric)} engaged",
+      f"{format_amount(m['engaged_h'], 'h')} engaged",
+      f"{int(m['active_days'])} days, best streak {int(m['best_streak_days'])}",
+      f"{unlocked} of {total} unlocked",
+    ])
+
+
+achievements_store = AchievementsStore()
@@ -0,0 +1,52 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+comma 3X Achievements panel: a totals row, then one row per achievement.
+Tap a row to read what it is for. Opening the panel marks new unlocks as seen.
+"""
+from openpilot.selfdrive.ui.sunnypilot.achievements_store import AchievementsStore, achievements_store
+from openpilot.selfdrive.ui.ui_state import ui_state
+from openpilot.system.ui.lib.multilang import tr
+from openpilot.system.ui.widgets import Widget
+from openpilot.system.ui.widgets.list_view import text_item
+from openpilot.system.ui.widgets.scroller_tici import Scroller
+
+
+class AchievementsLayout(Widget):
+  def __init__(self, store: AchievementsStore = achievements_store):
+    super().__init__()
+    self._store = store
+    self._new: set[str] = set()
+
+    unlocked = text_item(tr("Unlocked"), self._unlocked_text, description=lambda: self._store.totals_text(ui_state.is_metric))
+    items = [unlocked]
+    for a in self._store.catalog.achievements:
+      items.append(text_item(a.name, lambda aid=a.id: self._row_value(aid), description=a.description))
+    self._scroller = Scroller(items, line_separator=True, spacing=0)
+
+  def _unlocked_text(self) -> str:
+    unlocked, total = self._store.unlocked_count()
+    return f"{unlocked} / {total}"
+
+  def _row_value(self, achievement_id: str) -> str:
+    status = self._store.status(achievement_id)
+    if status is None:
+      return ""
+    is_new = any(uid in self._new for uid in self._store.unlock_ids(achievement_id))
+    return (tr("NEW") + " · " if is_new else "") + status.summary(ui_state.is_metric)
+
+  def show_event(self):
+    self._scroller.show_event()
+    self._store.refresh(force=True)
+    # remembered for this visit, so the NEW labels stay up while the panel is open
+    self._new = self._store.new_ids()
+    self._store.mark_all_seen()
+
+  def _update_state(self):
+    self._store.refresh()
+
+  def _render(self, rect):
+    self._scroller.render(rect)
@@ -11,6 +11,7 @@ import pyray as rl
 from openpilot.selfdrive.ui.layouts.settings import settings as OP
 from openpilot.selfdrive.ui.layouts.settings.firehose import FirehoseLayout
 from openpilot.selfdrive.ui.layouts.settings.toggles import TogglesLayout
+from openpilot.selfdrive.ui.sunnypilot.layouts.settings.achievements import AchievementsLayout
 from openpilot.selfdrive.ui.sunnypilot.layouts.settings.cruise import CruiseLayout
 from openpilot.selfdrive.ui.sunnypilot.layouts.settings.developer import DeveloperLayoutSP
 from openpilot.selfdrive.ui.sunnypilot.layouts.settings.device import DeviceLayoutSP
@@ -50,6 +51,7 @@ OP.PanelType = IntEnum(  # type: ignore[assignment] # ty: ignore[invalid-assignm
    "NAVIGATION",
    "TRIPS",
    "VEHICLE",
+    "ACHIEVEMENTS",
  ],
  start=0,
 )
@@ -122,6 +124,8 @@ class SettingsLayoutSP(OP.SettingsLayout):
      OP.PanelType.OSM: PanelInfo(tr_noop("OSM"), OSMLayout(), icon="../../sunnypilot/selfdrive/assets/offroad/icon_map.png"),
      # OP.PanelType.NAVIGATION: PanelInfo(tr_noop("Navigation"), NavigationLayout(), icon="../../sunnypilot/selfdrive/assets/offroad/icon_map.png"),
      OP.PanelType.TRIPS: PanelInfo(tr_noop("Trips"), TripsLayout(), icon="../../sunnypilot/selfdrive/assets/offroad/icon_trips.png"),
+      OP.PanelType.ACHIEVEMENTS: PanelInfo(tr_noop("Achievements"), AchievementsLayout(),
+                                           icon="../../sunnypilot/selfdrive/assets/offroad/icon_achievements.png"),
      OP.PanelType.VEHICLE: PanelInfo(tr_noop("Vehicle"), VehicleLayout(), icon="../../sunnypilot/selfdrive/assets/offroad/icon_vehicle.png"),
      OP.PanelType.FIREHOSE: PanelInfo(tr_noop("Firehose"), FirehoseLayout(), icon="../../sunnypilot/selfdrive/assets/offroad/icon_firehose.png"),
      OP.PanelType.DEVELOPER: PanelInfo(tr_noop("Developer"), DeveloperLayoutSP(), icon="icons/shell.png"),
@@ -0,0 +1,69 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+comma four Achievements screen: a summary card, then one card per achievement.
+Reads the state file achievementsd writes; opening the screen marks new unlocks as seen.
+"""
+from openpilot.selfdrive.ui.mici.widgets.button import BigButton
+from openpilot.selfdrive.ui.mici.widgets.dialog import BigDialog
+from openpilot.selfdrive.ui.sunnypilot.achievements_store import AchievementsStore, achievements_store
+from openpilot.selfdrive.ui.ui_state import ui_state
+from openpilot.system.ui.lib.application import gui_app
+from openpilot.system.ui.lib.multilang import tr
+from openpilot.system.ui.widgets.scroller import NavScroller
+
+
+class AchievementsLayoutMici(NavScroller):
+  def __init__(self, store: AchievementsStore = achievements_store):
+    super().__init__()
+    self._store = store
+    self._shown_version = -1
+    self._summary_btn = BigButton(tr("achievements"), "")
+    self._summary_btn.set_click_callback(self._show_summary)
+
+    self._cards: dict[str, BigButton] = {}
+    widgets = [self._summary_btn]
+    for a in self._store.catalog.achievements:
+      card = BigButton(a.name.lower(), "")
+      card.set_click_callback(lambda aid=a.id: self._show_details(aid))
+      self._cards[a.id] = card
+      widgets.append(card)
+    self._scroller.add_widgets(widgets)
+
+  def show_event(self):
+    super().show_event()
+    self._store.refresh(force=True)
+    self._refresh_cards()
+    # marked seen after drawing, so the "new" labels stay up while this screen is open
+    self._store.mark_all_seen()
+    self._shown_version = self._store.version
+
+  def _update_state(self):
+    super()._update_state()
+    self._store.refresh()
+    if self._store.version != self._shown_version:
+      self._refresh_cards()
+      self._shown_version = self._store.version
+
+  def _refresh_cards(self):
+    is_metric = ui_state.is_metric
+    new = self._store.new_ids()
+    unlocked, total = self._store.unlocked_count()
+    self._summary_btn.set_value(f"{unlocked} / {total} unlocked")
+    for status in self._store.statuses():
+      a = status.achievement
+      card = self._cards[a.id]
+      is_new = any(uid in new for uid in self._store.unlock_ids(a.id))
+      card.set_value(("new · " if is_new else "") + status.summary(is_metric))
+
+  def _show_summary(self):
+    gui_app.push_widget(BigDialog("", self._store.totals_text(ui_state.is_metric)))
+
+  def _show_details(self, achievement_id: str):
+    status = self._store.status(achievement_id)
+    if status is None:
+      return
+    gui_app.push_widget(BigDialog(status.achievement.name.lower(), f"{status.achievement.description}\n{status.summary(ui_state.is_metric)}"))
@@ -9,6 +9,8 @@ from openpilot.selfdrive.ui.mici.layouts.settings.settings import SettingsBigBut
 from openpilot.selfdrive.ui.mici.layouts.settings.device.device_layout import DeviceLayoutMici
 from openpilot.selfdrive.ui.mici.widgets.button import BigCircleButton
 from openpilot.selfdrive.ui.mici.widgets.dialog import BigConfirmationDialog, BigDialog
+from openpilot.selfdrive.ui.sunnypilot.achievements_store import achievements_store
+from openpilot.selfdrive.ui.sunnypilot.mici.layouts.achievements import AchievementsLayoutMici
 from openpilot.selfdrive.ui.sunnypilot.mici.layouts.sunnylink import SunnylinkLayoutMici
 from openpilot.selfdrive.ui.sunnypilot.mici.layouts.models import ModelsLayoutMici
 from openpilot.selfdrive.ui.ui_state import ui_state
@@ -50,6 +52,11 @@ class SettingsLayoutSP(OP.SettingsLayout):
    models_btn = SettingsBigButton(tr("models"), "", gui_app.texture("../../sunnypilot/selfdrive/assets/offroad/icon_models.png", ICON_SIZE, ICON_SIZE))
    models_btn.set_click_callback(lambda: gui_app.push_widget(models_panel))

+    achievements_panel = AchievementsLayoutMici()
+    self._achievements_btn = SettingsBigButton(tr("achievements"), "",
+                                               gui_app.texture("../../sunnypilot/selfdrive/assets/icons_mici/achievements.png", ICON_SIZE, ICON_SIZE))
+    self._achievements_btn.set_click_callback(lambda: gui_app.push_widget(achievements_panel))
+
    # onroad: enable button sits at the front (left of toggles)
    self._enable_offroad_btn_onroad = BigCircleButton(self.icon_offroad_enable, red=True)
    self._enable_offroad_btn_onroad.set_click_callback(lambda: self._handle_always_offroad(True))
@@ -72,6 +79,7 @@ class SettingsLayoutSP(OP.SettingsLayout):
    # front slots (only one ever visible at a time): exit-always-offroad, then enable-onroad
    items.insert(0, self._enable_offroad_btn_onroad)
    items.insert(0, self._disable_offroad_btn)
+    items.append(self._achievements_btn)
    # end slot: enable-offroad (right of developer)
    items.append(self._enable_offroad_btn_offroad)

@@ -81,6 +89,11 @@ class SettingsLayoutSP(OP.SettingsLayout):

  def _update_state(self):
    super()._update_state()
+    achievements_store.refresh()
+    new = len(achievements_store.new_ids())
+    value = tr("{} new").format(new) if new else ""
+    if value != self._achievements_btn.get_value():
+      self._achievements_btn.set_value(value)

  def _handle_always_offroad(self, enable: bool):

@@ -0,0 +1,57 @@
+from openpilot.selfdrive.ui.sunnypilot.achievements_store import AchievementsStore
+from openpilot.sunnypilot.achievements.engine import AchievementEngine, State, save_state
+
+T0 = 1_780_000_000.0
+
+
+def make_store(tmp_path):
+  return AchievementsStore(state_file=tmp_path / "state.json", seen_file=tmp_path / "seen.json")
+
+
+def write_state(tmp_path, engaged_km):
+  e = AchievementEngine(State(engaged_m=engaged_km * 1000.0))
+  e.check_unlocks(T0)
+  save_state(e.state, tmp_path / "state.json")
+
+
+def test_no_state_file_shows_everything_locked(tmp_path):
+  store = make_store(tmp_path)
+  assert store.refresh(force=True)
+  unlocked, total = store.unlocked_count()
+  assert unlocked == 0 and total > 10
+  assert store.new_ids() == set()
+
+
+def test_new_unlocks_until_seen(tmp_path):
+  write_state(tmp_path, 150)
+  store = make_store(tmp_path)
+  store.refresh(force=True)
+  assert store.new_ids() == {"first_light", "odometer.bronze"}
+  v = store.version
+  store.mark_all_seen()
+  assert store.new_ids() == set() and store.version > v
+
+  # a second store (the other screen, or after a reboot) sees them as seen too
+  other = make_store(tmp_path)
+  other.refresh(force=True)
+  assert other.new_ids() == set()
+
+
+def test_refresh_picks_up_new_unlocks(tmp_path):
+  write_state(tmp_path, 150)
+  store = make_store(tmp_path)
+  store.refresh(force=True)
+  store.mark_all_seen()
+  write_state(tmp_path, 1500)
+  assert store.refresh(force=True)
+  assert store.new_ids() == {"odometer.silver"}
+
+
+def test_totals_and_status(tmp_path):
+  write_state(tmp_path, 1240)
+  store = make_store(tmp_path)
+  store.refresh(force=True)
+  assert store.status("odometer").summary() == "silver · 1,240 / 10,000 km"
+  assert "1,240 km engaged" in store.totals_text()
+  assert store.unlock_ids("odometer") == ["odometer.bronze", "odometer.silver", "odometer.gold", "odometer.platinum"]
+  assert store.status("nope") is None
@@ -0,0 +1,55 @@
+# Achievements and the community log
+
+Everything here runs on the car. Nothing is uploaded yet: the uploads are a dry run that writes files.
+
+| File | What it does |
+| --- | --- |
+| `achievementsd.py` | Onroad process. Feeds car messages to the engine and the community log. |
+| `engine.py`, `catalog.py`, `catalog.json` | Driving stats and achievement unlocks (Phase 1). |
+| `community.py` | Community log in `community.db`: 6 km tiles, drive start/end tiles, daily totals, region per day. Also the privacy filter and the report builders. |
+| `preview.py` | Writes what the car would send, plus `map.html`, to `preview/`. |
+| `h3lite.py` | The H3 functions we need, in pure Python (sunnypilot doesn't add packages). Same cell ids as Uber's H3. |
+| `regions.py`, `regions.json.gz` | Tile to home region (`US-OH`, `CA-ON`, `GB`). Built by `tools/build_regions.py`. |
+| `replay.py` | Runs everything over drives already saved on the car. |
+| `simulate.py` | Made-up drives for tests and a demo. |
+
+## Privacy rules (from the build plan)
+
+- Positions only become H3 resolution 6 tiles (about 6 km). No coordinates, times of day or routes are stored.
+- A tile where drives start or end on 3+ days is a frequent stop. It and the 6 tiles around it never leave the car.
+- Other start and end tiles wait 14 days.
+- Hidden tiles never leave the car.
+- The tile batch is sorted and has no dates. It is separate from the daily report.
+- Others only ever see resolution 4 blocks (about 45 km).
+- Home region: where you drove most on most days in the last 90 days. A detected change is sent at most every 30 days; a change you make goes right away.
+
+## Files on the car
+
+All under `/data/media/0/achievements/`:
+
+- `state.json`: achievement progress.
+- `community.db`: the community log.
+- `community_settings.json` (optional): `{"level": 0-3, "region": "auto" | "country" | "hidden" | "US-OH"}`. Level 0 (off) is the default.
+- `preview/`: `daily-YYYY-MM-DD.json`, `tiles.json`, `delete.json`, `map.html`. Rewritten after every drive.
+
+`map.html` shows your frequent stops. It is for you only.
+
+## See it on your own drives
+
+```sh
+ssh comma@<car ip>
+cd /data/openpilot
+python -m openpilot.sunnypilot.achievements.replay    # qlogs in /data/media/0/realdata; add --days 30 or --rlog
+# on your computer:
+scp comma@<car ip>:/data/media/0/achievements/replay/preview/map.html .
+```
+
+The replay writes to `achievements/replay/` and doesn't touch live progress.
+
+## Tests
+
+```sh
+python -m pytest openpilot/sunnypilot/achievements/tests
+```
+
+With `h3` installed (`pip install h3`), the H3 tests also compare random points against the real library.
@@ -0,0 +1,190 @@
+#!/usr/bin/env python3
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+achievementsd: counts driving stats and unlocks achievements while onroad.
+
+Read-only toward the car: it only subscribes to messages, never publishes
+controls. Progress is saved to a JSON file on the device (see engine.state_path)
+every SAVE_INTERVAL_S and when the drive ends. The community log (map tiles,
+stops, daily totals; see community.py) is saved next to it, and when the drive
+ends a dry-run preview of what would be shared is written to preview/.
+Nothing leaves the device.
+"""
+import datetime
+import time
+
+from openpilot.cereal import messaging
+from openpilot.common.params import Params
+from openpilot.common.realtime import Ratekeeper
+from openpilot.common.swaglog import cloudlog
+from openpilot.common.time_helpers import system_time_valid
+from openpilot.sunnypilot.achievements.community import CommunityLog, Settings
+from openpilot.sunnypilot.achievements.engine import (AchievementEngine, MAX_GPS_ACCURACY_M, Sample, load_state,
+                                                      save_state)
+from openpilot.sunnypilot.achievements.preview import write_preview
+
+RATE_HZ = 10.
+SAVE_INTERVAL_S = 30.
+SLOW_POLL_S = 30.  # model, build and sunnylink checks
+GPS_MAX_AGE_S = 2.
+UNREGISTERED_SUNNYLINK_DONGLE_ID = "UnregisteredDevice"
+
+SERVICES = ['carState', 'selfdriveState', 'selfdriveStateSP', 'modelV2', 'driverMonitoringState',
+            'gpsLocation', 'gpsLocationExternal']
+
+
+def _dev_build() -> bool:
+  try:
+    from openpilot.common.version import get_build_metadata
+    bm = get_build_metadata()
+    return "staging" in bm.channel or not (bm.release_channel or bm.release_sp_channel)
+  except Exception:
+    return False
+
+
+def _active_model() -> tuple[str, str]:
+  try:
+    from openpilot.sunnypilot.models.helpers import get_active_bundle
+    bundle = get_active_bundle()
+    if bundle is None:
+      return "default", "Default"
+    return bundle.ref or bundle.internalName, bundle.displayName or bundle.internalName
+  except Exception:
+    cloudlog.exception("achievementsd: could not read the active model")
+    return "", ""
+
+
+def _sunnylink_linked(params: Params) -> bool:
+  try:
+    dongle = params.get("SunnylinkDongleId")
+    return bool(params.get_bool("SunnylinkEnabled")) and dongle not in (None, "", UNREGISTERED_SUNNYLINK_DONGLE_ID)
+  except Exception:
+    return False
+
+
+def _gps_fix(sm: messaging.SubMaster, now: float) -> tuple[float | None, float | None]:
+  for svc in ('gpsLocation', 'gpsLocationExternal'):
+    if sm.seen[svc] and now - sm.recv_time[svc] < GPS_MAX_AGE_S:
+      g = sm[svc]
+      if g.hasFix and 0 < g.horizontalAccuracy < MAX_GPS_ACCURACY_M:
+        return g.latitude, g.longitude
+  return None, None
+
+
+def _car_platform(params: Params) -> str | None:
+  try:
+    from opendbc.car.structs import car
+    cp_bytes = params.get("CarParams")
+    if not cp_bytes:
+      return None
+    return messaging.log_from_bytes(cp_bytes, car.CarParams).carFingerprint or None
+  except Exception:
+    return None
+
+
+def _open_community() -> CommunityLog | None:
+  try:
+    community = CommunityLog()
+    _ = community.regions  # load the region table now, not mid-drive
+    return community
+  except Exception:
+    cloudlog.exception("achievementsd: could not open the community log")
+    return None
+
+
+def _community_step(community: CommunityLog, sample: Sample, engine: AchievementEngine, save: bool) -> CommunityLog | None:
+  """Feed one sample. On any error the community log is off until the next drive; achievements keep working."""
+  try:
+    community.update(sample, engine.last_step)
+    if save:
+      community.flush()
+    return community
+  except Exception:
+    cloudlog.exception("achievementsd: community log failed, off until the next drive")
+    community.close()
+    return None
+
+
+def _finish_community(community: CommunityLog, engine: AchievementEngine, params: Params) -> None:
+  """Save where the drive ended and write the dry-run preview. Must stay well under the manager's 5 s stop timeout."""
+  try:
+    community.end_drive(engine.utc_offset_h)
+    if system_time_valid():
+      today = engine.local_date(datetime.datetime.now(datetime.UTC).timestamp())
+      write_preview(community, engine.state, today, _car_platform(params), Settings.load())
+  except Exception:
+    cloudlog.exception("achievementsd: community log or preview failed")
+  finally:
+    community.close()
+
+
+def _wall_time(valid: bool) -> float | None:
+  return datetime.datetime.now(datetime.UTC).timestamp() if valid else None
+
+
+def main():
+  params = Params()
+  sm = messaging.SubMaster(SERVICES, ignore_alive=['gpsLocation', 'gpsLocationExternal'],
+                           ignore_valid=['gpsLocation', 'gpsLocationExternal'])
+  engine = AchievementEngine(load_state())
+  community = _open_community()
+  rk = Ratekeeper(RATE_HZ, print_delay_threshold=None)
+
+  dev_build = _dev_build()
+  model_ref, model_name = _active_model()
+  last_save = last_slow = time.monotonic()
+  clock_ok = system_time_valid()
+  engine.set_sunnylink_linked(_sunnylink_linked(params), _wall_time(clock_ok))
+
+  try:
+    while True:
+      sm.update(0)
+      now = time.monotonic()
+
+      if now - last_slow > SLOW_POLL_S:
+        last_slow = now
+        model_ref, model_name = _active_model()
+        clock_ok = system_time_valid()
+        for uid in engine.set_sunnylink_linked(_sunnylink_linked(params), _wall_time(clock_ok)):
+          cloudlog.event("achievement unlocked", id=uid)
+
+      lat, lon = _gps_fix(sm, now)
+      sample = Sample(
+        t=now,
+        v_ego=sm['carState'].vEgo,
+        op_active=sm['selfdriveState'].active,
+        mads_active=sm['selfdriveStateSP'].mads.active,
+        lane_change_state=sm['modelV2'].meta.laneChangeState.raw,
+        dm_alert_level=sm['driverMonitoringState'].alertLevel.raw,
+        lat=lat,
+        lon=lon,
+        model_ref=model_ref,
+        model_name=model_name,
+        dev_build=dev_build,
+        wall=_wall_time(clock_ok),
+      )
+      for uid in engine.update(sample):
+        cloudlog.event("achievement unlocked", id=uid)
+
+      save = now - last_save > SAVE_INTERVAL_S
+      if save:
+        last_save = now
+        save_state(engine.state)
+      if community is not None:
+        community = _community_step(community, sample, engine, save)
+
+      rk.keep_time()
+  finally:
+    # the manager stops us with SIGINT when the drive ends
+    engine.end_drive()
+    save_state(engine.state)
+    if community is not None:
+      _finish_community(community, engine, params)
+
+
+if __name__ == "__main__":
+  main()
@@ -0,0 +1,153 @@
+{
+  "version": 1,
+  "achievements": [
+    {
+      "id": "first_light",
+      "name": "First Light",
+      "description": "Your first engaged drive with sunnypilot.",
+      "metric": "engaged_km",
+      "unit": "km",
+      "tiers": [{"tier": "single", "threshold": 0.1}]
+    },
+    {
+      "id": "odometer",
+      "name": "Odometer",
+      "description": "Distance driven with sunnypilot engaged.",
+      "metric": "engaged_km",
+      "unit": "km",
+      "tiers": [
+        {"tier": "bronze", "threshold": 100},
+        {"tier": "silver", "threshold": 1000},
+        {"tier": "gold", "threshold": 10000},
+        {"tier": "platinum", "threshold": 50000}
+      ]
+    },
+    {
+      "id": "captains_hours",
+      "name": "Captain's Hours",
+      "description": "Time spent with sunnypilot engaged.",
+      "metric": "engaged_h",
+      "unit": "h",
+      "tiers": [
+        {"tier": "bronze", "threshold": 10},
+        {"tier": "silver", "threshold": 100},
+        {"tier": "gold", "threshold": 1000}
+      ]
+    },
+    {
+      "id": "regular",
+      "name": "Regular",
+      "description": "Different days you drove with sunnypilot engaged.",
+      "metric": "active_days",
+      "unit": "days",
+      "tiers": [
+        {"tier": "bronze", "threshold": 7},
+        {"tier": "silver", "threshold": 30},
+        {"tier": "gold", "threshold": 100},
+        {"tier": "platinum", "threshold": 365}
+      ]
+    },
+    {
+      "id": "streak",
+      "name": "Streak",
+      "description": "Engaged on this many days in a row.",
+      "metric": "best_streak_days",
+      "unit": "days",
+      "tiers": [
+        {"tier": "bronze", "threshold": 7},
+        {"tier": "silver", "threshold": 30}
+      ]
+    },
+    {
+      "id": "steady_hands",
+      "name": "Steady Hands",
+      "description": "Distance with MADS steering on its own, without sunnypilot's cruise.",
+      "metric": "mads_km",
+      "unit": "km",
+      "tiers": [
+        {"tier": "bronze", "threshold": 100},
+        {"tier": "silver", "threshold": 1000}
+      ]
+    },
+    {
+      "id": "smooth_merger",
+      "name": "Smooth Merger",
+      "description": "Assisted lane changes completed.",
+      "metric": "lane_changes",
+      "unit": "changes",
+      "tiers": [
+        {"tier": "bronze", "threshold": 10},
+        {"tier": "silver", "threshold": 100},
+        {"tier": "gold", "threshold": 1000}
+      ]
+    },
+    {
+      "id": "road_trip",
+      "name": "Road Trip",
+      "description": "Engaged distance in a single day. Take breaks along the way.",
+      "metric": "best_day_km",
+      "unit": "km",
+      "tiers": [{"tier": "single", "threshold": 500}]
+    },
+    {
+      "id": "explorer",
+      "name": "Explorer",
+      "description": "Different areas (about 5 km across) driven engaged. Areas are kept as anonymous hashes on this device.",
+      "metric": "areas",
+      "unit": "areas",
+      "tiers": [
+        {"tier": "bronze", "threshold": 5},
+        {"tier": "silver", "threshold": 25},
+        {"tier": "gold", "threshold": 100}
+      ]
+    },
+    {
+      "id": "model_hopper",
+      "name": "Model Hopper",
+      "description": "Driving models you have driven at least 50 km engaged.",
+      "metric": "models_50km",
+      "unit": "models",
+      "tiers": [
+        {"tier": "bronze", "threshold": 3},
+        {"tier": "silver", "threshold": 5},
+        {"tier": "gold", "threshold": 10}
+      ]
+    },
+    {
+      "id": "dev_branch_brave",
+      "name": "Dev Branch Brave",
+      "description": "Engaged distance on a development or staging build.",
+      "metric": "dev_km",
+      "unit": "km",
+      "tiers": [{"tier": "single", "threshold": 100}]
+    },
+    {
+      "id": "attentive",
+      "name": "Attentive",
+      "description": "Engaged time in a row without a red driver monitoring alert.",
+      "metric": "attentive_best_h",
+      "unit": "h",
+      "tiers": [
+        {"tier": "bronze", "threshold": 10},
+        {"tier": "silver", "threshold": 50},
+        {"tier": "gold", "threshold": 100}
+      ]
+    },
+    {
+      "id": "seasons",
+      "name": "Four Seasons",
+      "description": "Engaged drives in spring, summer, fall and winter.",
+      "metric": "seasons",
+      "unit": "seasons",
+      "tiers": [{"tier": "single", "threshold": 4}]
+    },
+    {
+      "id": "linked_up",
+      "name": "Linked Up",
+      "description": "Turn on sunnylink on this device.",
+      "metric": "sunnylink_linked",
+      "unit": "",
+      "tiers": [{"tier": "single", "threshold": 1}]
+    }
+  ]
+}
@@ -0,0 +1,137 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+The achievement catalog. Pure Python on purpose: the UI, achievementsd and the
+tests all load it, and none of them should need cereal or the Params library to
+read a list of goals.
+"""
+from __future__ import annotations
+
+import json
+from dataclasses import dataclass
+from functools import cache
+from pathlib import Path
+
+CATALOG_PATH = Path(__file__).with_name("catalog.json")
+SINGLE = "single"
+KM_TO_MI = 0.621371
+
+
+@dataclass(frozen=True)
+class Tier:
+  tier: str
+  threshold: float
+
+
+@dataclass(frozen=True)
+class Achievement:
+  id: str
+  name: str
+  description: str
+  metric: str
+  unit: str
+  tiers: tuple[Tier, ...]
+
+  def unlock_id(self, tier: Tier) -> str:
+    """The id an unlock is stored and shared under: `odometer.silver`, or just `first_light`."""
+    return self.id if tier.tier == SINGLE else f"{self.id}.{tier.tier}"
+
+  @property
+  def tiered(self) -> bool:
+    return not (len(self.tiers) == 1 and self.tiers[0].tier == SINGLE)
+
+
+@dataclass(frozen=True)
+class Catalog:
+  version: int
+  achievements: tuple[Achievement, ...]
+
+  @property
+  def metrics(self) -> set[str]:
+    return {a.metric for a in self.achievements}
+
+  def all_unlock_ids(self) -> list[str]:
+    return [a.unlock_id(t) for a in self.achievements for t in a.tiers]
+
+
+def parse_catalog(data: dict) -> Catalog:
+  achievements = []
+  seen: set[str] = set()
+  for raw in data["achievements"]:
+    tiers = tuple(Tier(str(t["tier"]), float(t["threshold"])) for t in raw["tiers"])
+    if not tiers:
+      raise ValueError(f"achievement {raw['id']} has no tiers")
+    if any(b.threshold <= a.threshold for a, b in zip(tiers, tiers[1:], strict=False)):
+      raise ValueError(f"achievement {raw['id']} tiers must increase")
+    if raw["id"] in seen:
+      raise ValueError(f"duplicate achievement id {raw['id']}")
+    seen.add(raw["id"])
+    achievements.append(Achievement(raw["id"], raw["name"], raw["description"], raw["metric"], raw.get("unit", ""), tiers))
+  return Catalog(int(data["version"]), tuple(achievements))
+
+
+@cache
+def load_catalog(path: str = str(CATALOG_PATH)) -> Catalog:
+  with open(path) as f:
+    return parse_catalog(json.load(f))
+
+
+# --- display helpers, shared by both device UIs ---
+
+def _fmt_number(value: float) -> str:
+  if value >= 100 or value == int(value):
+    return f"{int(value):,}"
+  return f"{value:.1f}"
+
+
+def format_amount(value: float, unit: str, is_metric: bool = True) -> str:
+  if unit == "km" and not is_metric:
+    value, unit = value * KM_TO_MI, "mi"
+  if not unit:
+    return _fmt_number(value)
+  return f"{_fmt_number(value)} {unit}"
+
+
+@dataclass(frozen=True)
+class AchievementStatus:
+  achievement: Achievement
+  reached: Tier | None   # highest tier unlocked, None if locked
+  next: Tier | None      # next tier to unlock, None when all are done
+  value: float
+
+  @property
+  def done(self) -> bool:
+    return self.next is None
+
+  def summary(self, is_metric: bool = True) -> str:
+    """One short line: `silver · 1,240 / 10,000 km`, `locked · 3 / 7 days`, `gold · done`."""
+    a = self.achievement
+    if self.next is None:
+      return f"{self.reached.tier} · done" if a.tiered and self.reached else "done"
+    label = self.reached.tier if self.reached else "locked"
+    if not a.unit:  # a yes/no achievement, like turning on sunnylink
+      return label
+    return f"{label} · {format_progress(self.value, self.next.threshold, a.unit, is_metric)}"
+
+
+def format_progress(value: float, goal: float, unit: str, is_metric: bool = True) -> str:
+  if unit == "km" and not is_metric:
+    value, goal, unit = value * KM_TO_MI, goal * KM_TO_MI, "mi"
+  return f"{_fmt_number(min(value, goal))} / {_fmt_number(goal)} {unit}"
+
+
+def statuses(catalog: Catalog, metrics: dict[str, float], unlocked: dict[str, str]) -> list[AchievementStatus]:
+  out = []
+  for a in catalog.achievements:
+    reached = None
+    nxt = None
+    for t in a.tiers:
+      if a.unlock_id(t) in unlocked:
+        reached = t
+      elif nxt is None:
+        nxt = t
+    out.append(AchievementStatus(a, reached, nxt, float(metrics.get(a.metric, 0.0))))
+  return out
@@ -0,0 +1,370 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+The community log and the privacy filter.
+
+While driving, achievementsd hands every sample to CommunityLog, which keeps in
+community.db on the car:
+- the 6 km map tiles (H3 resolution 6) you drove through, with the day each was
+  first seen,
+- the tile where each drive started and ended, per day,
+- engaged distance and time per day,
+- engaged distance per region per day.
+Positions only ever become tiles: no coordinates, times of day or routes are
+stored here.
+
+The privacy filter (tile_view, daily_report, tile_batch, profile_update) then
+decides what could leave the car, following the Privacy spec in the build plan:
+- A tile holding a drive's start or end on 3+ different days is a frequent stop
+  (home, work, school). It and the 6 tiles around it never leave the car.
+- Any other start or end tile waits 14 days.
+- Tiles the driver hid never leave the car.
+- Tiles go out as a sorted list with no dates, apart from the daily report.
+- The home region is where you drove most on most days over 90 days. It is sent
+  only when it changes, and an automatic change at most every 30 days.
+
+Nothing in this file sends anything.
+"""
+from __future__ import annotations
+
+import collections
+import datetime as dt
+import json
+import sqlite3
+from dataclasses import dataclass, field
+from pathlib import Path
+
+from openpilot.sunnypilot.achievements import h3lite
+from openpilot.sunnypilot.achievements.engine import Sample, State, Step, achievements_root
+from openpilot.sunnypilot.achievements.regions import RegionTable, country_of
+from openpilot.sunnypilot.achievements.regions import table as region_table
+
+REPORT_VERSION = 1
+TILE_RES = 6                 # about 6 km across; only the owner sees these
+PUBLIC_RES = 4               # about 45 km across; what crews and the public see
+STOP_HOLD_DAYS = 14
+FREQUENT_STOP_DAYS = 3
+REGION_WINDOW_DAYS = 90
+REGION_RESEND_DAYS = 30
+MOVING_MS = 1.0
+METERS_PER_MILE = 1609.344
+UNLINKED_DEVICE = "c_" + "0" * 16  # the real id comes from the community key made at linking
+
+LEVEL_OFF, LEVEL_ACHIEVEMENTS, LEVEL_DAILY, LEVEL_MAP = range(4)
+
+SCHEMA = """
+CREATE TABLE IF NOT EXISTS tiles (cell TEXT PRIMARY KEY, first_day TEXT NOT NULL, sent INTEGER NOT NULL DEFAULT 0);
+CREATE TABLE IF NOT EXISTS stops (cell TEXT NOT NULL, day TEXT NOT NULL, PRIMARY KEY (cell, day));
+CREATE TABLE IF NOT EXISTS days (day TEXT PRIMARY KEY, engaged_m REAL NOT NULL, engaged_s REAL NOT NULL);
+CREATE TABLE IF NOT EXISTS day_regions (day TEXT NOT NULL, region TEXT NOT NULL, engaged_m REAL NOT NULL,
+                                        PRIMARY KEY (day, region));
+CREATE TABLE IF NOT EXISTS hidden (cell TEXT PRIMARY KEY);
+CREATE TABLE IF NOT EXISTS meta (key TEXT PRIMARY KEY, value TEXT NOT NULL);
+"""
+
+
+def db_path() -> Path:
+  return achievements_root() / "community.db"
+
+
+def settings_path() -> Path:
+  return achievements_root() / "community_settings.json"
+
+
+@dataclass
+class Settings:
+  level: int = LEVEL_OFF   # 0 off (default), 1 achievements, 2 daily stats, 3 map
+  region: str = "auto"     # "auto", "country", "hidden", or a region code the driver picked
+
+  @classmethod
+  def load(cls, path: Path | None = None) -> Settings:
+    try:
+      with open(path or settings_path()) as f:
+        d = json.load(f)
+      return cls(level=int(d.get("level", LEVEL_OFF)), region=str(d.get("region", "auto")))
+    except (OSError, ValueError, TypeError, AttributeError):
+      return cls()
+
+
+def round_to(x: float, step: int) -> int:
+  return int(round(x / step)) * step
+
+
+def tile_of(lat: float, lon: float) -> str:
+  return h3lite.cell_to_str(h3lite.latlng_to_cell(lat, lon, TILE_RES))
+
+
+def ring_of(tile: str) -> set[str]:
+  """The tile and the tiles touching it: an area about 18 km across."""
+  return {h3lite.cell_to_str(c) for c in h3lite.grid_disk1(h3lite.str_to_cell(tile))}
+
+
+def public_block(tile: str) -> str:
+  return h3lite.cell_to_str(h3lite.cell_to_parent(h3lite.str_to_cell(tile), PUBLIC_RES))
+
+
+@dataclass
+class TileView:
+  """Every driven tile, sorted by what the privacy filter does with it."""
+  driven: set[str] = field(default_factory=set)
+  frequent: set[str] = field(default_factory=set)     # home, work, school...
+  ring: set[str] = field(default_factory=set)         # frequent stops and the tiles around them: never sent
+  held: dict[str, str] = field(default_factory=dict)  # recent stop tile -> first day it may be sent
+  hidden: set[str] = field(default_factory=set)
+  shareable: set[str] = field(default_factory=set)
+  sent: set[str] = field(default_factory=set)
+
+  @property
+  def to_send(self) -> set[str]:
+    return self.shareable - self.sent
+
+  @property
+  def to_delete(self) -> set[str]:
+    """Tiles already sent that later became private (moved house, new job, hidden)."""
+    return self.sent & (self.ring | self.hidden)
+
+
+class CommunityLog:
+  def __init__(self, path: Path | None = None, regions: RegionTable | None = None):
+    self.path = path or db_path()
+    self.path.parent.mkdir(parents=True, exist_ok=True)
+    self.db = sqlite3.connect(self.path)
+    self.db.executescript(SCHEMA)
+    self._regions = regions
+    self._known: set[str] = {r[0] for r in self.db.execute("SELECT cell FROM tiles")}
+    self._region_of: dict[str, str | None] = {}
+    self._utc_offset_h: float | None = None
+    self._reset_pending()
+    self._reset_drive()
+
+  def close(self) -> None:
+    self.db.close()
+
+  @property
+  def regions(self) -> RegionTable:
+    if self._regions is None:
+      self._regions = region_table()
+    return self._regions
+
+  def _reset_pending(self) -> None:
+    self._new_tiles: dict[str, str] = {}
+    self._stops: set[tuple[str, str]] = set()
+    self._days: dict[str, list[float]] = collections.defaultdict(lambda: [0.0, 0.0])
+    self._day_regions: dict[tuple[str, str], float] = collections.defaultdict(float)
+
+  def _reset_drive(self) -> None:
+    self._started = False
+    self._last_fix: tuple[str, str] | None = None
+
+  # --- while driving ---
+
+  def update(self, s: Sample, step: Step) -> None:
+    day = step.day
+    if day is None:
+      return
+    if s.engaged and step.dt_s > 0.0:
+      d = self._days[day]
+      d[0] += step.engaged_m
+      d[1] += step.dt_s
+
+    if s.lat is None or s.lon is None:
+      return
+    tile = tile_of(s.lat, s.lon)
+    if not self._started:
+      self._started = True
+      self._stops.add((tile, day))
+    self._last_fix = (tile, day)
+
+    if s.v_ego > MOVING_MS and tile not in self._known and tile not in self._new_tiles:
+      self._new_tiles[tile] = day
+
+    if step.engaged_m > 0.0:
+      if tile not in self._region_of:
+        self._region_of[tile] = self.regions.lookup_cell(h3lite.str_to_cell(tile))
+      region = self._region_of[tile]
+      if region is not None:
+        self._day_regions[(day, region)] += step.engaged_m
+
+  def end_drive(self, utc_offset_h: float | None = None) -> None:
+    """Record where the drive ended and save. Safe to call more than once."""
+    if self._last_fix is not None:
+      self._stops.add(self._last_fix)
+    if utc_offset_h is not None:
+      self._utc_offset_h = utc_offset_h
+    self._reset_drive()
+    self.flush()
+
+  def flush(self) -> None:
+    with self.db:
+      self.db.executemany("INSERT OR IGNORE INTO tiles (cell, first_day) VALUES (?, ?)", self._new_tiles.items())
+      self.db.executemany("INSERT OR IGNORE INTO stops (cell, day) VALUES (?, ?)", self._stops)
+      self.db.executemany("""INSERT INTO days (day, engaged_m, engaged_s) VALUES (?, ?, ?)
+                             ON CONFLICT(day) DO UPDATE SET engaged_m = engaged_m + excluded.engaged_m,
+                                                            engaged_s = engaged_s + excluded.engaged_s""",
+                          [(day, m, s) for day, (m, s) in self._days.items()])
+      self.db.executemany("""INSERT INTO day_regions (day, region, engaged_m) VALUES (?, ?, ?)
+                             ON CONFLICT(day, region) DO UPDATE SET engaged_m = engaged_m + excluded.engaged_m""",
+                          [(day, region, m) for (day, region), m in self._day_regions.items()])
+      if self._utc_offset_h is not None:
+        self._set_meta("utc_offset_h", self._utc_offset_h)
+    self._known.update(self._new_tiles)
+    self._reset_pending()
+
+  # --- settings kept with the data ---
+
+  def _get_meta(self, key: str, default=None):
+    row = self.db.execute("SELECT value FROM meta WHERE key = ?", (key,)).fetchone()
+    return json.loads(row[0]) if row else default
+
+  def _set_meta(self, key: str, value) -> None:
+    self.db.execute("INSERT OR REPLACE INTO meta (key, value) VALUES (?, ?)", (key, json.dumps(value)))
+
+  def hide_tile(self, tile: str) -> None:
+    with self.db:
+      self.db.execute("INSERT OR IGNORE INTO hidden (cell) VALUES (?)", (tile,))
+
+  def unhide_tile(self, tile: str) -> None:
+    with self.db:
+      self.db.execute("DELETE FROM hidden WHERE cell = ?", (tile,))
+
+  def mark_sent(self, tiles: set[str], sent: bool = True) -> None:
+    """For the uploader: tiles the server now has (or, with sent=False, has deleted)."""
+    with self.db:
+      self.db.executemany("UPDATE tiles SET sent = ? WHERE cell = ?", [(int(sent), t) for t in tiles])
+
+  # --- the privacy filter ---
+
+  def tile_view(self, today: dt.date) -> TileView:
+    v = TileView()
+    for cell, sent in self.db.execute("SELECT cell, sent FROM tiles"):
+      v.driven.add(cell)
+      if sent:
+        v.sent.add(cell)
+
+    stop_days: dict[str, set[str]] = collections.defaultdict(set)
+    for cell, day in self.db.execute("SELECT cell, day FROM stops"):
+      stop_days[cell].add(day)
+    v.frequent = {c for c, days in stop_days.items() if len(days) >= FREQUENT_STOP_DAYS}
+    for c in v.frequent:
+      v.ring |= ring_of(c)
+    for c, days in stop_days.items():
+      if c in v.ring:
+        continue
+      release = dt.date.fromisoformat(max(days)) + dt.timedelta(days=STOP_HOLD_DAYS)
+      if release > today:
+        v.held[c] = release.isoformat()
+
+    v.hidden = {r[0] for r in self.db.execute("SELECT cell FROM hidden")}
+    v.shareable = v.driven - v.ring - set(v.held) - v.hidden
+    return v
+
+  def driven_days(self, since: dt.date | None = None) -> list[str]:
+    rows = self.db.execute("SELECT day FROM days WHERE day >= ? ORDER BY day", ((since or dt.date.min).isoformat(),))
+    return [r[0] for r in rows]
+
+  def day_stats(self, day: str) -> tuple[float, float, int]:
+    """Engaged meters, engaged seconds and new tiles on a day."""
+    row = self.db.execute("SELECT engaged_m, engaged_s FROM days WHERE day = ?", (day,)).fetchone()
+    new = self.db.execute("SELECT COUNT(*) FROM tiles WHERE first_day = ?", (day,)).fetchone()[0]
+    return (row[0], row[1], new) if row else (0.0, 0.0, new)
+
+  def region_days(self, today: dt.date) -> collections.Counter:
+    """How many of the last 90 days each region was where you drove most, engaged."""
+    since = (today - dt.timedelta(days=REGION_WINDOW_DAYS)).isoformat()
+    best: dict[str, tuple[float, str]] = {}
+    for day, region, m in self.db.execute("SELECT day, region, engaged_m FROM day_regions WHERE day > ? ORDER BY region",
+                                          (since,)):
+      if day not in best or m > best[day][0]:
+        best[day] = (m, region)
+    return collections.Counter(region for _, region in best.values())
+
+  def home_region(self, today: dt.date) -> str | None:
+    tally = self.region_days(today)
+    if not tally:
+      return None
+    return max(sorted(tally), key=lambda r: tally[r])
+
+  def shared_region(self, today: dt.date, settings: Settings) -> str | None:
+    if settings.region == "hidden":
+      return None
+    if settings.region not in ("auto", "country"):
+      return settings.region
+    home = self.home_region(today)
+    if home is None:
+      return None
+    return country_of(home) if settings.region == "country" else home
+
+  def profile_update(self, today: dt.date, car: str | None, settings: Settings) -> dict | None:
+    """Car model and home region, only when they changed since they were last sent."""
+    last = self._get_meta("profile_sent", {})
+    out: dict = {}
+    if car and car != last.get("car"):
+      out["car"] = car
+    region = self.shared_region(today, settings)
+    if region != last.get("region") and (region is not None or "region" in last):
+      # the driver changing the setting goes right away; a change the car detected waits
+      chosen = settings.region != last.get("choice", "auto")
+      last_day = last.get("region_day")
+      if chosen or last_day is None or (today - dt.date.fromisoformat(last_day)).days >= REGION_RESEND_DAYS:
+        out["region"] = region
+    return out or None
+
+  def mark_profile_sent(self, today: dt.date, profile: dict, settings: Settings) -> None:
+    last = self._get_meta("profile_sent", {})
+    if "car" in profile:
+      last["car"] = profile["car"]
+    if "region" in profile:
+      last["region"] = profile["region"]
+      last["region_day"] = today.isoformat()
+      last["choice"] = settings.region
+    with self.db:
+      self._set_meta("profile_sent", last)
+
+  # --- uploads ---
+
+  def unlock_day(self, unix: float) -> str | None:
+    if not unix:
+      return None  # unlocked before the clock was valid
+    offset = self._utc_offset_h if self._utc_offset_h is not None else self._get_meta("utc_offset_h", 0)
+    return (dt.datetime.fromtimestamp(unix, dt.UTC) + dt.timedelta(hours=offset)).date().isoformat()
+
+  def daily_report(self, day: str, state: State, level: int, profile: dict | None = None,
+                   include_undated: bool = False, device: str = UNLINKED_DEVICE) -> dict | None:
+    """The daily report for one day, with only the parts the sharing level allows."""
+    if level < LEVEL_ACHIEVEMENTS:
+      return None
+    unlocks = []
+    for uid, unix in sorted(state.unlocked.items()):
+      d = self.unlock_day(unix)
+      if d == day or (d is None and include_undated):
+        unlocks.append({"id": uid, "day": day})
+    report: dict = {
+      "v": REPORT_VERSION,
+      "device": device,
+      "day": day,
+      "unlocks": unlocks,
+      "totals": {"engaged_mi": round_to(state.engaged_m / METERS_PER_MILE, 10),
+                 "engaged_h": round_to(state.engaged_s / 3600.0, 10)},
+    }
+    if level >= LEVEL_DAILY:
+      m, s, new = self.day_stats(day)
+      report["daily"] = {"engaged_mi": round(m / METERS_PER_MILE), "engaged_min": round(s / 60.0), "new_places": new}
+      if profile:
+        report["profile"] = profile
+    return report
+
+  def tile_batch(self, today: dt.date, device: str = UNLINKED_DEVICE) -> dict | None:
+    """The weekly tile batch: new shareable tiles, sorted, with no dates."""
+    tiles = self.tile_view(today).to_send
+    if not tiles:
+      return None
+    return {"v": REPORT_VERSION, "device": device, "tiles": sorted(tiles)}
+
+  def tile_deletes(self, today: dt.date, device: str = UNLINKED_DEVICE) -> dict | None:
+    tiles = self.tile_view(today).to_delete
+    if not tiles:
+      return None
+    return {"v": REPORT_VERSION, "device": device, "delete_tiles": sorted(tiles)}
@@ -0,0 +1,347 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+The achievements engine: running totals and unlocks, fed one Sample at a time.
+
+Pure Python with no cereal or Params imports, so it can be tested and replayed
+anywhere. achievementsd turns live messages into Samples; the UI only reads the
+saved state file.
+
+Everything stays on the device. Explorer areas are stored as salted hashes of
+~5 km geohash cells, never as coordinates.
+"""
+from __future__ import annotations
+
+import datetime as dt
+import hashlib
+import json
+import os
+import secrets
+import tempfile
+from dataclasses import asdict, dataclass, field
+from pathlib import Path
+
+from openpilot.sunnypilot.achievements.catalog import Catalog, load_catalog
+
+STATE_VERSION = 1
+
+# plausibility limits for a single update
+MAX_DT_S = 1.0          # a longer gap (process stall, suspend) counts nothing
+MAX_SPEED_MS = 70.0     # ~250 km/h; above this the reading is not trusted
+MIN_SPEED_MS = 0.0
+
+# GPS
+GEOHASH_PRECISION = 5   # cells of about 4.9 x 4.9 km
+MAX_GPS_ACCURACY_M = 100.0
+MODEL_KM_GOAL = 50.0    # Model Hopper: km on a model before it counts
+
+# driver monitoring alertLevel.three is the red, take-control-now alert
+DM_ALERT_RED = 3
+
+# modelV2.meta.laneChangeState
+LANE_CHANGE_STARTING = 2
+LANE_CHANGE_FINISHING = 3
+
+_GEOHASH_BASE32 = "0123456789bcdefghjkmnpqrstuvwxyz"
+
+
+def achievements_root() -> Path:
+  """/data/media/0/achievements on the device (survives updates), ~/.comma/... on PC."""
+  if "ACHIEVEMENTS_ROOT" in os.environ:
+    return Path(os.environ["ACHIEVEMENTS_ROOT"])
+  if os.path.isfile("/TICI"):
+    return Path("/data/media/0/achievements")
+  return Path.home() / (".comma" + os.environ.get("OPENPILOT_PREFIX", "")) / "media" / "0" / "achievements"
+
+
+def state_path() -> Path:
+  return achievements_root() / "state.json"
+
+
+def seen_path() -> Path:
+  """Unlocks the driver has already looked at. Written only by the UI."""
+  return achievements_root() / "seen.json"
+
+
+def geohash(lat: float, lon: float, precision: int = GEOHASH_PRECISION) -> str:
+  lat_rng, lon_rng = [-90.0, 90.0], [-180.0, 180.0]
+  bits, bit, ch, even = [16, 8, 4, 2, 1], 0, 0, True
+  out = []
+  while len(out) < precision:
+    rng, val = (lon_rng, lon) if even else (lat_rng, lat)
+    mid = (rng[0] + rng[1]) / 2
+    if val >= mid:
+      ch |= bits[bit]
+      rng[0] = mid
+    else:
+      rng[1] = mid
+    even = not even
+    if bit < 4:
+      bit += 1
+    else:
+      out.append(_GEOHASH_BASE32[ch])
+      bit, ch = 0, 0
+  return "".join(out)
+
+
+def season(day: dt.date, southern: bool = False) -> str:
+  s = ("winter", "winter", "spring", "spring", "spring", "summer", "summer", "summer", "fall", "fall", "fall", "winter")[day.month - 1]
+  if southern:
+    s = {"winter": "summer", "summer": "winter", "spring": "fall", "fall": "spring"}[s]
+  return s
+
+
+@dataclass
+class Sample:
+  """One reading of the car, built by achievementsd from cereal."""
+  t: float                        # monotonic seconds, for time between samples
+  v_ego: float                    # m/s
+  op_active: bool                 # selfdriveState.active (sunnypilot driving, incl. driver override)
+  mads_active: bool = False       # selfdriveStateSP.mads.active (lateral)
+  lane_change_state: int = 0      # modelV2.meta.laneChangeState
+  dm_alert_level: int = 0         # driverMonitoringState.alertLevel
+  lat: float | None = None        # only when a good fix is present
+  lon: float | None = None
+  model_ref: str = ""
+  model_name: str = ""
+  dev_build: bool = False
+  wall: float | None = None       # unix time; None until the device clock is valid
+
+  @property
+  def engaged(self) -> bool:
+    return self.op_active or self.mads_active
+
+  @property
+  def mads_only(self) -> bool:
+    return self.mads_active and not self.op_active
+
+
+@dataclass
+class Step:
+  """What the last update counted, for the community log (see community.py)."""
+  dt_s: float = 0.0          # trusted seconds since the previous sample (0 after a gap or a bad reading)
+  engaged_m: float = 0.0     # meters driven engaged in this step
+  day: str | None = None     # local calendar day; None until the clock is valid
+
+
+@dataclass
+class State:
+  version: int = STATE_VERSION
+  salt: str = field(default_factory=lambda: secrets.token_hex(16))
+  engaged_m: float = 0.0
+  engaged_s: float = 0.0
+  mads_m: float = 0.0
+  dev_m: float = 0.0
+  lane_changes: int = 0
+  days: list[str] = field(default_factory=list)        # ISO dates with engaged driving, sorted
+  best_streak_days: int = 0
+  today: str = ""
+  today_m: float = 0.0
+  best_day_m: float = 0.0
+  areas: list[str] = field(default_factory=list)       # salted hashes of geohash cells, sorted
+  model_m: dict[str, float] = field(default_factory=dict)
+  model_names: dict[str, str] = field(default_factory=dict)
+  attentive_run_s: float = 0.0
+  attentive_best_s: float = 0.0
+  seasons: list[str] = field(default_factory=list)
+  sunnylink_linked: bool = False
+  unlocked: dict[str, float] = field(default_factory=dict)  # unlock id -> unix time
+
+  @classmethod
+  def from_dict(cls, d: dict) -> State:
+    known = {k: v for k, v in d.items() if k in cls.__dataclass_fields__}
+    return cls(**known)
+
+  def metrics(self) -> dict[str, float]:
+    return {
+      "engaged_km": self.engaged_m / 1000.0,
+      "engaged_h": self.engaged_s / 3600.0,
+      "active_days": float(len(self.days)),
+      "best_streak_days": float(self.best_streak_days),
+      "mads_km": self.mads_m / 1000.0,
+      "lane_changes": float(self.lane_changes),
+      "best_day_km": max(self.best_day_m, self.today_m) / 1000.0,
+      "areas": float(len(self.areas)),
+      "models_50km": float(sum(1 for m in self.model_m.values() if m >= MODEL_KM_GOAL * 1000.0)),
+      "dev_km": self.dev_m / 1000.0,
+      "attentive_best_h": max(self.attentive_best_s, self.attentive_run_s) / 3600.0,
+      "seasons": float(len(self.seasons)),
+      "sunnylink_linked": 1.0 if self.sunnylink_linked else 0.0,
+    }
+
+
+def _streak_ending(days: list[str], last: str) -> int:
+  have = set(days)
+  d = dt.date.fromisoformat(last)
+  n = 0
+  while d.isoformat() in have:
+    n += 1
+    d -= dt.timedelta(days=1)
+  return n
+
+
+class AchievementEngine:
+  def __init__(self, state: State | None = None, catalog: Catalog | None = None):
+    self.state = state or State()
+    self.catalog = catalog or load_catalog()
+    self._last_t: float | None = None
+    self._last_lane_change = 0
+    self._utc_offset_h: float | None = None  # from GPS longitude, see local_date
+    self._southern = False
+    self.last_step = Step()
+
+  # --- time ---
+
+  @property
+  def utc_offset_h(self) -> float | None:
+    return self._utc_offset_h
+
+  def local_date(self, wall: float) -> dt.date:
+    """Calendar day for streaks and Road Trip.
+
+    Comma devices keep their clock in UTC, which would split an evening drive in
+    the Americas across two days. With a GPS fix, use the solar offset from
+    longitude (15 degrees per hour); without one, fall back to the system zone.
+    """
+    if self._utc_offset_h is not None:
+      return (dt.datetime.fromtimestamp(wall, dt.UTC) + dt.timedelta(hours=self._utc_offset_h)).date()
+    return dt.datetime.fromtimestamp(wall).date()
+
+  # --- updates ---
+
+  def update(self, s: Sample) -> list[str]:
+    """Feed one sample. Returns unlock ids that were newly unlocked by it."""
+    st = self.state
+    dt_s = 0.0 if self._last_t is None else s.t - self._last_t
+    self._last_t = s.t
+    if not (0.0 < dt_s <= MAX_DT_S) or not (MIN_SPEED_MS <= s.v_ego <= MAX_SPEED_MS):
+      dt_s = 0.0
+
+    if s.lat is not None and s.lon is not None:
+      self._utc_offset_h = round(s.lon / 15.0)
+      self._southern = s.lat < 0
+
+    day = self.local_date(s.wall).isoformat() if s.wall is not None else None
+    self.last_step = Step(dt_s=dt_s, engaged_m=s.v_ego * dt_s if s.engaged else 0.0, day=day)
+
+    if s.engaged and dt_s > 0.0:
+      dist = s.v_ego * dt_s
+      if day is not None and day != st.today:
+        st.best_day_m = max(st.best_day_m, st.today_m)
+        st.today, st.today_m = day, 0.0
+
+      st.engaged_m += dist
+      st.engaged_s += dt_s
+      if day is not None:
+        st.today_m += dist
+      if s.mads_only:
+        st.mads_m += dist
+      if s.dev_build:
+        st.dev_m += dist
+      if s.model_ref:
+        st.model_m[s.model_ref] = st.model_m.get(s.model_ref, 0.0) + dist
+        if s.model_name:
+          st.model_names[s.model_ref] = s.model_name
+
+      if day is not None and day not in st.days and s.v_ego > 1.0:
+        st.days.append(day)
+        st.days.sort()
+        st.best_streak_days = max(st.best_streak_days, _streak_ending(st.days, day))
+        sn = season(dt.date.fromisoformat(day), self._southern)
+        if sn not in st.seasons:
+          st.seasons.append(sn)
+
+      if s.dm_alert_level >= DM_ALERT_RED:
+        st.attentive_best_s = max(st.attentive_best_s, st.attentive_run_s)
+        st.attentive_run_s = 0.0
+      else:
+        st.attentive_run_s += dt_s
+
+      if s.lat is not None and s.lon is not None:
+        area = hashlib.sha256((st.salt + geohash(s.lat, s.lon)).encode()).hexdigest()[:16]
+        if area not in st.areas:
+          st.areas.append(area)
+          st.areas.sort()
+
+    # a lane change counts once, when it reaches its finishing phase while engaged
+    if s.engaged and s.lane_change_state == LANE_CHANGE_FINISHING and self._last_lane_change == LANE_CHANGE_STARTING:
+      st.lane_changes += 1
+    self._last_lane_change = s.lane_change_state
+
+    return self.check_unlocks(s.wall)
+
+  def set_sunnylink_linked(self, linked: bool, wall: float | None) -> list[str]:
+    self.state.sunnylink_linked = self.state.sunnylink_linked or linked
+    return self.check_unlocks(wall)
+
+  def end_drive(self) -> None:
+    """Close out the attentive run so the best is kept, and forget timing between drives."""
+    st = self.state
+    st.attentive_best_s = max(st.attentive_best_s, st.attentive_run_s)
+    self._last_t = None
+    self._last_lane_change = 0
+
+  def check_unlocks(self, wall: float | None) -> list[str]:
+    """Unlock every tier whose goal is met. The unlock time is 0 when the clock isn't valid yet."""
+    metrics = self.state.metrics()
+    new = []
+    for a in self.catalog.achievements:
+      value = metrics.get(a.metric, 0.0)
+      for tier in a.tiers:
+        uid = a.unlock_id(tier)
+        if value >= tier.threshold and uid not in self.state.unlocked:
+          self.state.unlocked[uid] = round(wall) if wall is not None else 0
+          new.append(uid)
+    return new
+
+
+# --- persistence ---
+
+def load_state(path: Path | None = None) -> State:
+  path = path or state_path()
+  try:
+    with open(path) as f:
+      return State.from_dict(json.load(f))
+  except FileNotFoundError:
+    return State()
+  except (ValueError, TypeError):
+    # unreadable JSON: keep it aside for debugging instead of silently overwriting it.
+    # Other OSErrors propagate, so a transient read failure can never be saved over real progress.
+    os.replace(path, path.with_suffix(".corrupt.json"))
+    return State()
+
+
+def save_json(path: Path, data: dict) -> None:
+  path.parent.mkdir(parents=True, exist_ok=True)
+  fd, tmp = tempfile.mkstemp(dir=path.parent, prefix=path.name, suffix=".tmp")
+  try:
+    with os.fdopen(fd, "w") as f:
+      json.dump(data, f, separators=(",", ":"), sort_keys=True)
+      f.flush()
+      os.fsync(f.fileno())
+    os.replace(tmp, path)
+  except BaseException:
+    try:
+      os.unlink(tmp)
+    except OSError:
+      pass
+    raise
+
+
+def save_state(state: State, path: Path | None = None) -> None:
+  save_json(path or state_path(), asdict(state))
+
+
+def load_seen(path: Path | None = None) -> set[str]:
+  try:
+    with open(path or seen_path()) as f:
+      return set(json.load(f).get("seen", []))
+  except (OSError, ValueError, AttributeError):
+    return set()
+
+
+def save_seen(seen: set[str], path: Path | None = None) -> None:
+  save_json(path or seen_path(), {"seen": sorted(seen)})
@@ -0,0 +1,629 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+h3lite: the few H3 functions the car needs, in pure Python.
+
+sunnypilot doesn't add Python packages, so instead of the h3 package this is a
+line-by-line port of the parts of Uber's H3 v4.1.2 C library we use:
+latLngToCell, cellToLatLng, cellToParent and a one-ring disk. Cell ids match
+the real library exactly (tests/test_h3lite.py checks this against h3-py when
+it's installed, and against fixed vectors when it isn't), so the site and the
+server can keep using h3-js and h3-pg on the same ids.
+
+H3 is Copyright Uber Technologies, Inc., licensed under the Apache License 2.0
+(https://github.com/uber/h3). The tables below are copied from its
+faceijk.c and baseCells.c.
+"""
+from __future__ import annotations
+
+import math
+
+MAX_RES = 15
+NUM_FACES = 20
+NUM_BASE_CELLS = 122
+
+_EPSILON = 1e-16
+_SQRT3_2 = 0.8660254037844386467637231707529361834714
+_SQRT7 = 2.6457513110645905905016157536392604257102
+_AP7_ROT_RADS = 0.333473172251832115336090755351601070065900389
+_RES0_U_GNOMONIC = 0.38196601125010500003
+_2PI = 2.0 * math.pi
+
+# Directions (index digits)
+_CENTER, _K, _J, _JK, _I, _IK, _IJ = range(7)
+_UNIT_VECS = ((0, 0, 0), (0, 0, 1), (0, 1, 0), (0, 1, 1), (1, 0, 0), (1, 0, 1), (1, 1, 0))
+_ROT_CCW = (0, _IK, _JK, _K, _IJ, _I, _J)   # digit -> digit rotated 60 degrees ccw
+_ROT_CW = (0, _JK, _IJ, _J, _IK, _K, _I)    # digit -> digit rotated 60 degrees cw
+
+# face neighbor quadrants
+_IJ_Q, _KI_Q, _JK_Q = 1, 2, 3
+
+# bit layout of an H3 index
+_MODE_OFFSET = 59
+_RES_OFFSET = 52
+_BC_OFFSET = 45
+_H3_INIT = (1 << 45) - 1   # every digit set to 7
+_CELL_MODE = 1
+
+
+def _max_dim(res: int) -> int:     # Class II resolutions only
+  return 2 * 7 ** (res // 2)
+
+
+def _unit_scale(res: int) -> int:  # Class II resolutions only
+  return 7 ** (res // 2)
+
+
+# --- tables copied from H3 v4.1.2 faceijk.c and baseCells.c ---
+
+FACE_CENTER_GEO = (
+  (0.80358264971899, 1.2483974196173961), (1.3077478834556382, 2.5369450098779214),
+  (1.054751253523952, -1.3475173589003966), (0.6001915955381868, -0.45060390946975576),
+  (0.49171542819877384, 0.40198820291130694), (0.1727453274156187, 1.6781468852804338),
+  (0.6059293215713507, 2.9539233298124117), (0.42737051832897965, -1.8888762003362853),
+  (-0.07906611854921283, -0.7334295133808677), (-0.23096164445538364, 0.506495587332349),
+  (0.07906611854921283, 2.4081631402089254), (0.23096164445538364, -2.635097066257444),
+  (-0.1727453274156187, -1.4634457683093596), (-0.6059293215713507, -0.18766932377738163),
+  (-0.42737051832897965, 1.2527164532535078), (-0.6001915955381868, 2.6909887441200375),
+  (-0.49171542819877384, -2.7396044506784865), (-0.80358264971899, -1.8931952339723972),
+  (-1.3077478834556382, -0.6046476437118721), (-1.054751253523952, 1.7940752946893965),
+)
+FACE_CENTER_POINT = (
+  (0.2199307791404606, 0.6583691780274996, 0.7198475378926182),
+  (-0.2139234834501421, 0.1478171829550703, 0.9656017935214205),
+  (0.1092625278784797, -0.481195157287321, 0.8697775121287253),
+  (0.7428567301586791, -0.3593941678278028, 0.5648005936517033),
+  (0.8112534709140969, 0.3448953237639384, 0.472138773641393),
+  (-0.1055498149613921, 0.9794457296411413, 0.1718874610009365),
+  (-0.8075407579970092, 0.1533552485898818, 0.5695261994882688),
+  (-0.2846148069787907, -0.8644080972654206, 0.4144792552473539),
+  (0.7405621473854482, -0.6673299564565524, -0.0789837646326737),
+  (0.8512303986474293, 0.4722343788582681, -0.2289137388687808),
+  (-0.7405621473854481, 0.6673299564565524, 0.0789837646326737),
+  (-0.8512303986474292, -0.4722343788582682, 0.2289137388687808),
+  (0.1055498149613919, -0.9794457296411413, -0.1718874610009365),
+  (0.8075407579970092, -0.1533552485898819, -0.5695261994882688),
+  (0.2846148069787908, 0.8644080972654204, -0.4144792552473539),
+  (-0.7428567301586791, 0.3593941678278027, -0.5648005936517033),
+  (-0.811253470914097, -0.3448953237639382, -0.472138773641393),
+  (-0.2199307791404607, -0.6583691780274996, -0.7198475378926182),
+  (0.213923483450142, -0.1478171829550704, -0.9656017935214205),
+  (-0.1092625278784796, 0.481195157287321, -0.8697775121287253),
+)
+FACE_AXIS_AZ_CII = (
+  5.6199582685239395, 5.7603390817141875, 0.78021365439343, 0.4304693639799999,
+  6.130269123335111, 2.692877706530643, 2.982963003477244, 3.532912002790141,
+  3.494305004259568, 3.0032141694995382, 5.930472956509812, 0.13837848409025486,
+  0.4487149470591504, 0.15862965011254937, 5.891865957979238, 2.711123289609793,
+  3.294508837434268, 3.80481969224544, 3.6644388790551923, 2.361378999196363,
+)
+# per face: (central, ij, ki, jk) neighbors as (face, (i, j, k) translate, ccw 60 rotations)
+FACE_NEIGHBORS = (
+  ((0, (0, 0, 0), 0), (4, (2, 0, 2), 1), (1, (2, 2, 0), 5), (5, (0, 2, 2), 3)),
+  ((1, (0, 0, 0), 0), (0, (2, 0, 2), 1), (2, (2, 2, 0), 5), (6, (0, 2, 2), 3)),
+  ((2, (0, 0, 0), 0), (1, (2, 0, 2), 1), (3, (2, 2, 0), 5), (7, (0, 2, 2), 3)),
+  ((3, (0, 0, 0), 0), (2, (2, 0, 2), 1), (4, (2, 2, 0), 5), (8, (0, 2, 2), 3)),
+  ((4, (0, 0, 0), 0), (3, (2, 0, 2), 1), (0, (2, 2, 0), 5), (9, (0, 2, 2), 3)),
+  ((5, (0, 0, 0), 0), (10, (2, 2, 0), 3), (14, (2, 0, 2), 3), (0, (0, 2, 2), 3)),
+  ((6, (0, 0, 0), 0), (11, (2, 2, 0), 3), (10, (2, 0, 2), 3), (1, (0, 2, 2), 3)),
+  ((7, (0, 0, 0), 0), (12, (2, 2, 0), 3), (11, (2, 0, 2), 3), (2, (0, 2, 2), 3)),
+  ((8, (0, 0, 0), 0), (13, (2, 2, 0), 3), (12, (2, 0, 2), 3), (3, (0, 2, 2), 3)),
+  ((9, (0, 0, 0), 0), (14, (2, 2, 0), 3), (13, (2, 0, 2), 3), (4, (0, 2, 2), 3)),
+  ((10, (0, 0, 0), 0), (5, (2, 2, 0), 3), (6, (2, 0, 2), 3), (15, (0, 2, 2), 3)),
+  ((11, (0, 0, 0), 0), (6, (2, 2, 0), 3), (7, (2, 0, 2), 3), (16, (0, 2, 2), 3)),
+  ((12, (0, 0, 0), 0), (7, (2, 2, 0), 3), (8, (2, 0, 2), 3), (17, (0, 2, 2), 3)),
+  ((13, (0, 0, 0), 0), (8, (2, 2, 0), 3), (9, (2, 0, 2), 3), (18, (0, 2, 2), 3)),
+  ((14, (0, 0, 0), 0), (9, (2, 2, 0), 3), (5, (2, 0, 2), 3), (19, (0, 2, 2), 3)),
+  ((15, (0, 0, 0), 0), (16, (2, 0, 2), 1), (19, (2, 2, 0), 5), (10, (0, 2, 2), 3)),
+  ((16, (0, 0, 0), 0), (17, (2, 0, 2), 1), (15, (2, 2, 0), 5), (11, (0, 2, 2), 3)),
+  ((17, (0, 0, 0), 0), (18, (2, 0, 2), 1), (16, (2, 2, 0), 5), (12, (0, 2, 2), 3)),
+  ((18, (0, 0, 0), 0), (19, (2, 0, 2), 1), (17, (2, 2, 0), 5), (13, (0, 2, 2), 3)),
+  ((19, (0, 0, 0), 0), (15, (2, 0, 2), 1), (18, (2, 2, 0), 5), (14, (0, 2, 2), 3)),
+)
+# [face][i][j][k] -> (base cell, ccw 60 rotations), flattened as face*27 + i*9 + j*3 + k
+FACE_IJK_BASE_CELLS = (
+  (16,0), (18,0), (24,0), (33,0), (30,0), (32,3), (49,1), (48,3), (50,3),
+  (8,0), (5,5), (10,5), (22,0), (16,0), (18,0), (41,1), (33,0), (30,0),
+  (4,0), (0,5), (2,5), (15,1), (8,0), (5,5), (31,1), (22,0), (16,0),
+  (2,0), (6,0), (14,0), (10,0), (11,0), (17,3), (24,1), (23,3), (25,3),
+  (0,0), (1,5), (9,5), (5,0), (2,0), (6,0), (18,1), (10,0), (11,0),
+  (4,1), (3,5), (7,5), (8,1), (0,0), (1,5), (16,1), (5,0), (2,0),
+  (7,0), (21,0), (38,0), (9,0), (19,0), (34,3), (14,1), (20,3), (36,3),
+  (3,0), (13,5), (29,5), (1,0), (7,0), (21,0), (6,1), (9,0), (19,0),
+  (4,2), (12,5), (26,5), (0,1), (3,0), (13,5), (2,1), (1,0), (7,0),
+  (26,0), (42,0), (58,0), (29,0), (43,0), (62,3), (38,1), (47,3), (64,3),
+  (12,0), (28,5), (44,5), (13,0), (26,0), (42,0), (21,1), (29,0), (43,0),
+  (4,3), (15,5), (31,5), (3,1), (12,0), (28,5), (7,1), (13,0), (26,0),
+  (31,0), (41,0), (49,0), (44,0), (53,0), (61,3), (58,1), (65,3), (75,3),
+  (15,0), (22,5), (33,5), (28,0), (31,0), (41,0), (42,1), (44,0), (53,0),
+  (4,4), (8,5), (16,5), (12,1), (15,0), (22,5), (26,1), (28,0), (31,0),
+  (50,0), (48,0), (49,3), (32,0), (30,3), (33,3), (24,3), (18,3), (16,3),
+  (70,0), (67,0), (66,3), (52,3), (50,0), (48,0), (37,3), (32,0), (30,3),
+  (83,0), (87,3), (85,3), (74,3), (70,0), (67,0), (57,1), (52,3), (50,0),
+  (25,0), (23,0), (24,3), (17,0), (11,3), (10,3), (14,3), (6,3), (2,3),
+  (45,0), (39,0), (37,3), (35,3), (25,0), (23,0), (27,3), (17,0), (11,3),
+  (63,0), (59,3), (57,3), (56,3), (45,0), (39,0), (46,3), (35,3), (25,0),
+  (36,0), (20,0), (14,3), (34,0), (19,3), (9,3), (38,3), (21,3), (7,3),
+  (55,0), (40,0), (27,3), (54,3), (36,0), (20,0), (51,3), (34,0), (19,3),
+  (72,0), (60,3), (46,3), (73,3), (55,0), (40,0), (71,3), (54,3), (36,0),
+  (64,0), (47,0), (38,3), (62,0), (43,3), (29,3), (58,3), (42,3), (26,3),
+  (84,0), (69,0), (51,3), (82,3), (64,0), (47,0), (76,3), (62,0), (43,3),
+  (97,0), (89,3), (71,3), (98,3), (84,0), (69,0), (96,3), (82,3), (64,0),
+  (75,0), (65,0), (58,3), (61,0), (53,3), (44,3), (49,3), (41,3), (31,3),
+  (94,0), (86,0), (76,3), (81,3), (75,0), (65,0), (66,3), (61,0), (53,3),
+  (107,0), (104,3), (96,3), (101,3), (94,0), (86,0), (85,3), (81,3), (75,0),
+  (57,0), (59,0), (63,3), (74,0), (78,3), (79,3), (83,3), (92,3), (95,3),
+  (37,0), (39,3), (45,3), (52,0), (57,0), (59,0), (70,3), (74,0), (78,3),
+  (24,0), (23,3), (25,3), (32,3), (37,0), (39,3), (50,3), (52,0), (57,0),
+  (46,0), (60,0), (72,3), (56,0), (68,3), (80,3), (63,3), (77,3), (90,3),
+  (27,0), (40,3), (55,3), (35,0), (46,0), (60,0), (45,3), (56,0), (68,3),
+  (14,0), (20,3), (36,3), (17,3), (27,0), (40,3), (25,3), (35,0), (46,0),
+  (71,0), (89,0), (97,3), (73,0), (91,3), (103,3), (72,3), (88,3), (105,3),
+  (51,0), (69,3), (84,3), (54,0), (71,0), (89,0), (55,3), (73,0), (91,3),
+  (38,0), (47,3), (64,3), (34,3), (51,0), (69,3), (36,3), (54,0), (71,0),
+  (96,0), (104,0), (107,3), (98,0), (110,3), (115,3), (97,3), (111,3), (119,3),
+  (76,0), (86,3), (94,3), (82,0), (96,0), (104,0), (84,3), (98,0), (110,3),
+  (58,0), (65,3), (75,3), (62,3), (76,0), (86,3), (64,3), (82,0), (96,0),
+  (85,0), (87,0), (83,3), (101,0), (102,3), (100,3), (107,3), (112,3), (114,3),
+  (66,0), (67,3), (70,3), (81,0), (85,0), (87,0), (94,3), (101,0), (102,3),
+  (49,0), (48,3), (50,3), (61,3), (66,0), (67,3), (75,3), (81,0), (85,0),
+  (95,0), (92,0), (83,0), (79,0), (78,0), (74,3), (63,1), (59,3), (57,3),
+  (109,0), (108,0), (100,5), (93,1), (95,0), (92,0), (77,1), (79,0), (78,0),
+  (117,4), (118,5), (114,5), (106,1), (109,0), (108,0), (90,1), (93,1), (95,0),
+  (90,0), (77,0), (63,0), (80,0), (68,0), (56,3), (72,1), (60,3), (46,3),
+  (106,0), (93,0), (79,5), (99,1), (90,0), (77,0), (88,1), (80,0), (68,0),
+  (117,3), (109,5), (95,5), (113,1), (106,0), (93,0), (105,1), (99,1), (90,0),
+  (105,0), (88,0), (72,0), (103,0), (91,0), (73,3), (97,1), (89,3), (71,3),
+  (113,0), (99,0), (80,5), (116,1), (105,0), (88,0), (111,1), (103,0), (91,0),
+  (117,2), (106,5), (90,5), (121,1), (113,0), (99,0), (119,1), (116,1), (105,0),
+  (119,0), (111,0), (97,0), (115,0), (110,0), (98,3), (107,1), (104,3), (96,3),
+  (121,0), (116,0), (103,5), (120,1), (119,0), (111,0), (112,1), (115,0), (110,0),
+  (117,1), (113,5), (105,5), (118,1), (121,0), (116,0), (114,1), (120,1), (119,0),
+  (114,0), (112,0), (107,0), (100,0), (102,0), (101,3), (83,1), (87,3), (85,3),
+  (118,0), (120,0), (115,5), (108,1), (114,0), (112,0), (92,1), (100,0), (102,0),
+  (117,0), (121,5), (119,5), (109,1), (118,0), (120,0), (95,1), (108,1), (114,0),
+)
+# base cell -> (home face, i, j, k, is pentagon, cw offset face 1, cw offset face 2)
+BASE_CELL_DATA = (
+  (1,1,0,0,0,0,0), (2,1,1,0,0,0,0), (1,0,0,0,0,0,0), (2,1,0,0,0,0,0),
+  (0,2,0,0,1,-1,-1), (1,1,1,0,0,0,0), (1,0,0,1,0,0,0), (2,0,0,0,0,0,0),
+  (0,1,0,0,0,0,0), (2,0,1,0,0,0,0), (1,0,1,0,0,0,0), (1,0,1,1,0,0,0),
+  (3,1,0,0,0,0,0), (3,1,1,0,0,0,0), (11,2,0,0,1,2,6), (4,1,0,0,0,0,0),
+  (0,0,0,0,0,0,0), (6,0,1,0,0,0,0), (0,0,0,1,0,0,0), (2,0,1,1,0,0,0),
+  (7,0,0,1,0,0,0), (2,0,0,1,0,0,0), (0,1,1,0,0,0,0), (6,0,0,1,0,0,0),
+  (10,2,0,0,1,1,5), (6,0,0,0,0,0,0), (3,0,0,0,0,0,0), (11,1,0,0,0,0,0),
+  (4,1,1,0,0,0,0), (3,0,1,0,0,0,0), (0,0,1,1,0,0,0), (4,0,0,0,0,0,0),
+  (5,0,1,0,0,0,0), (0,0,1,0,0,0,0), (7,0,1,0,0,0,0), (11,1,1,0,0,0,0),
+  (7,0,0,0,0,0,0), (10,1,0,0,0,0,0), (12,2,0,0,1,3,7), (6,1,0,1,0,0,0),
+  (7,1,0,1,0,0,0), (4,0,0,1,0,0,0), (3,0,0,1,0,0,0), (3,0,1,1,0,0,0),
+  (4,0,1,0,0,0,0), (6,1,0,0,0,0,0), (11,0,0,0,0,0,0), (8,0,0,1,0,0,0),
+  (5,0,0,1,0,0,0), (14,2,0,0,1,0,9), (5,0,0,0,0,0,0), (12,1,0,0,0,0,0),
+  (10,1,1,0,0,0,0), (4,0,1,1,0,0,0), (12,1,1,0,0,0,0), (7,1,0,0,0,0,0),
+  (11,0,1,0,0,0,0), (10,0,0,0,0,0,0), (13,2,0,0,1,4,8), (10,0,0,1,0,0,0),
+  (11,0,0,1,0,0,0), (9,0,1,0,0,0,0), (8,0,1,0,0,0,0), (6,2,0,0,1,11,15),
+  (8,0,0,0,0,0,0), (9,0,0,1,0,0,0), (14,1,0,0,0,0,0), (5,1,0,1,0,0,0),
+  (16,0,1,1,0,0,0), (8,1,0,1,0,0,0), (5,1,0,0,0,0,0), (12,0,0,0,0,0,0),
+  (7,2,0,0,1,12,16), (12,0,1,0,0,0,0), (10,0,1,0,0,0,0), (9,0,0,0,0,0,0),
+  (13,1,0,0,0,0,0), (16,0,0,1,0,0,0), (15,0,1,1,0,0,0), (15,0,1,0,0,0,0),
+  (16,0,1,0,0,0,0), (14,1,1,0,0,0,0), (13,1,1,0,0,0,0), (5,2,0,0,1,10,19),
+  (8,1,0,0,0,0,0), (14,0,0,0,0,0,0), (9,1,0,1,0,0,0), (14,0,0,1,0,0,0),
+  (17,0,0,1,0,0,0), (12,0,0,1,0,0,0), (16,0,0,0,0,0,0), (17,0,1,1,0,0,0),
+  (15,0,0,1,0,0,0), (16,1,0,1,0,0,0), (9,1,0,0,0,0,0), (15,0,0,0,0,0,0),
+  (13,0,0,0,0,0,0), (8,2,0,0,1,13,17), (13,0,1,0,0,0,0), (17,1,0,1,0,0,0),
+  (19,0,1,0,0,0,0), (14,0,1,0,0,0,0), (19,0,1,1,0,0,0), (17,0,1,0,0,0,0),
+  (13,0,0,1,0,0,0), (17,0,0,0,0,0,0), (16,1,0,0,0,0,0), (9,2,0,0,1,14,18),
+  (15,1,0,1,0,0,0), (15,1,0,0,0,0,0), (18,0,1,1,0,0,0), (18,0,0,1,0,0,0),
+  (19,0,0,1,0,0,0), (17,1,0,0,0,0,0), (19,0,0,0,0,0,0), (18,0,1,0,0,0,0),
+  (18,1,0,1,0,0,0), (19,2,0,0,1,-1,-1), (19,1,0,0,0,0,0), (18,0,0,0,0,0,0),
+  (19,1,0,1,0,0,0), (18,1,0,0,0,0,0),
+)
+
+
+# --- ijk coordinates ---
+
+def _lround(x: float) -> int:
+  """C lroundl: halves round away from zero."""
+  return int(math.floor(x + 0.5)) if x >= 0 else -int(math.floor(-x + 0.5))
+
+
+def _normalize(i: int, j: int, k: int) -> tuple[int, int, int]:
+  if i < 0:
+    j -= i
+    k -= i
+    i = 0
+  if j < 0:
+    i -= j
+    k -= j
+    j = 0
+  if k < 0:
+    i -= k
+    j -= k
+    k = 0
+  m = min(i, j, k)
+  if m > 0:
+    i -= m
+    j -= m
+    k -= m
+  return i, j, k
+
+
+def _add(a: tuple[int, int, int], b: tuple[int, int, int]) -> tuple[int, int, int]:
+  return _normalize(a[0] + b[0], a[1] + b[1], a[2] + b[2])
+
+
+def _scaled_sum(c: tuple[int, int, int], iv, jv, kv) -> tuple[int, int, int]:
+  i, j, k = c
+  return _normalize(iv[0] * i + jv[0] * j + kv[0] * k,
+                    iv[1] * i + jv[1] * j + kv[1] * k,
+                    iv[2] * i + jv[2] * j + kv[2] * k)
+
+
+def _up_ap7(c):
+  i, j = c[0] - c[2], c[1] - c[2]
+  return _normalize(_lround((3 * i - j) / 7.0), _lround((i + 2 * j) / 7.0), 0)
+
+
+def _up_ap7r(c):
+  i, j = c[0] - c[2], c[1] - c[2]
+  return _normalize(_lround((2 * i + j) / 7.0), _lround((3 * j - i) / 7.0), 0)
+
+
+def _down_ap7(c):
+  return _scaled_sum(c, (3, 0, 1), (1, 3, 0), (0, 1, 3))
+
+
+def _down_ap7r(c):
+  return _scaled_sum(c, (3, 1, 0), (0, 3, 1), (1, 0, 3))
+
+
+def _rotate60ccw(c):
+  return _scaled_sum(c, (1, 1, 0), (0, 1, 1), (1, 0, 1))
+
+
+def _rotate60cw(c):
+  return _scaled_sum(c, (1, 0, 1), (1, 1, 0), (0, 1, 1))
+
+
+def _unit_to_digit(c) -> int:
+  c = _normalize(*c)
+  for d, u in enumerate(_UNIT_VECS):
+    if c == u:
+      return d
+  return 7
+
+
+def _hex2d_to_ijk(x: float, y: float) -> tuple[int, int, int]:
+  a1, a2 = abs(x), abs(y)
+  x2 = a2 / _SQRT3_2
+  x1 = a1 + x2 / 2.0
+  m1, m2 = int(x1), int(x2)
+  r1, r2 = x1 - m1, x2 - m2
+
+  if r1 < 0.5:
+    if r1 < 1.0 / 3.0:
+      i = m1
+      j = m2 if r2 < (1.0 + r1) / 2.0 else m2 + 1
+    else:
+      j = m2 if r2 < (1.0 - r1) else m2 + 1
+      i = m1 + 1 if (1.0 - r1) <= r2 < (2.0 * r1) else m1
+  else:
+    if r1 < 2.0 / 3.0:
+      j = m2 if r2 < (1.0 - r1) else m2 + 1
+      i = m1 if (2.0 * r1 - 1.0) < r2 < (1.0 - r1) else m1 + 1
+    else:
+      i = m1 + 1
+      j = m2 if r2 < (r1 / 2.0) else m2 + 1
+
+  # fold across the axes if necessary
+  if x < 0.0:
+    if j % 2 == 0:
+      i = i - 2 * (i - j // 2)
+    else:
+      i = i - (2 * (i - (j + 1) // 2) + 1)
+  if y < 0.0:
+    i = i - (2 * j + 1) // 2
+    j = -j
+  return _normalize(i, j, 0)
+
+
+def _ijk_to_hex2d(c) -> tuple[float, float]:
+  i, j = c[0] - c[2], c[1] - c[2]
+  return i - 0.5 * j, j * _SQRT3_2
+
+
+# --- sphere <-> face ---
+
+def _pos_angle(rads: float) -> float:
+  tmp = rads + _2PI if rads < 0.0 else rads
+  if rads >= _2PI:
+    tmp -= _2PI
+  return tmp
+
+
+def _constrain_lng(lng: float) -> float:
+  while lng > math.pi:
+    lng -= _2PI
+  while lng < -math.pi:
+    lng += _2PI
+  return lng
+
+
+def _azimuth(lat1: float, lng1: float, lat2: float, lng2: float) -> float:
+  return math.atan2(math.cos(lat2) * math.sin(lng2 - lng1),
+                    math.cos(lat1) * math.sin(lat2) - math.sin(lat1) * math.cos(lat2) * math.cos(lng2 - lng1))
+
+
+def _az_distance(lat1: float, lng1: float, az: float, distance: float) -> tuple[float, float]:
+  if distance < _EPSILON:
+    return lat1, lng1
+  az = _pos_angle(az)
+  if az < _EPSILON or abs(az - math.pi) < _EPSILON:
+    lat2 = lat1 + distance if az < _EPSILON else lat1 - distance
+    if abs(lat2 - math.pi / 2) < _EPSILON:
+      return math.pi / 2, 0.0
+    if abs(lat2 + math.pi / 2) < _EPSILON:
+      return -math.pi / 2, 0.0
+    return lat2, _constrain_lng(lng1)
+
+  sinlat = math.sin(lat1) * math.cos(distance) + math.cos(lat1) * math.sin(distance) * math.cos(az)
+  sinlat = max(-1.0, min(1.0, sinlat))
+  lat2 = math.asin(sinlat)
+  if abs(lat2 - math.pi / 2) < _EPSILON:
+    return math.pi / 2, 0.0
+  if abs(lat2 + math.pi / 2) < _EPSILON:
+    return -math.pi / 2, 0.0
+  sinlng = math.sin(az) * math.sin(distance) / math.cos(lat2)
+  coslng = (math.cos(distance) - math.sin(lat1) * math.sin(lat2)) / math.cos(lat1) / math.cos(lat2)
+  sinlng = max(-1.0, min(1.0, sinlng))
+  coslng = max(-1.0, min(1.0, coslng))
+  return lat2, _constrain_lng(lng1 + math.atan2(sinlng, coslng))
+
+
+def _closest_face(lat: float, lng: float) -> tuple[int, float]:
+  r = math.cos(lat)
+  z, x, y = math.sin(lat), math.cos(lng) * r, math.sin(lng) * r
+  face, sqd = 0, 5.0
+  for f, (fx, fy, fz) in enumerate(FACE_CENTER_POINT):
+    d = (fx - x) ** 2 + (fy - y) ** 2 + (fz - z) ** 2
+    if d < sqd:
+      face, sqd = f, d
+  return face, sqd
+
+
+def _geo_to_hex2d(lat: float, lng: float, res: int) -> tuple[int, float, float]:
+  face, sqd = _closest_face(lat, lng)
+  r = math.acos(1 - sqd / 2)
+  if r < _EPSILON:
+    return face, 0.0, 0.0
+
+  flat, flng = FACE_CENTER_GEO[face]
+  theta = _pos_angle(FACE_AXIS_AZ_CII[face] - _pos_angle(_azimuth(flat, flng, lat, lng)))
+  if res % 2:
+    theta = _pos_angle(theta - _AP7_ROT_RADS)
+
+  r = math.tan(r) / _RES0_U_GNOMONIC
+  for _ in range(res):
+    r *= _SQRT7
+  return face, r * math.cos(theta), r * math.sin(theta)
+
+
+def _hex2d_to_geo(x: float, y: float, face: int, res: int) -> tuple[float, float]:
+  r = math.hypot(x, y)
+  if r < _EPSILON:
+    return FACE_CENTER_GEO[face]
+  theta = math.atan2(y, x)
+  for _ in range(res):
+    r /= _SQRT7
+  r = math.atan(r * _RES0_U_GNOMONIC)
+  if res % 2:
+    theta = _pos_angle(theta + _AP7_ROT_RADS)
+  theta = _pos_angle(FACE_AXIS_AZ_CII[face] - theta)
+  flat, flng = FACE_CENTER_GEO[face]
+  return _az_distance(flat, flng, theta, r)
+
+
+# --- H3 index bits ---
+
+def get_resolution(h: int) -> int:
+  return (h >> _RES_OFFSET) & 15
+
+
+def get_base_cell(h: int) -> int:
+  return (h >> _BC_OFFSET) & 127
+
+
+def _digit(h: int, r: int) -> int:
+  return (h >> ((MAX_RES - r) * 3)) & 7
+
+
+def _set_digit(h: int, r: int, d: int) -> int:
+  off = (MAX_RES - r) * 3
+  return (h & ~(7 << off)) | (d << off)
+
+
+def _leading_digit(h: int) -> int:
+  for r in range(1, get_resolution(h) + 1):
+    d = _digit(h, r)
+    if d:
+      return d
+  return 0
+
+
+def _rotate_index60ccw(h: int) -> int:
+  for r in range(1, get_resolution(h) + 1):
+    h = _set_digit(h, r, _ROT_CCW[_digit(h, r)])
+  return h
+
+
+def _rotate_index60cw(h: int) -> int:
+  for r in range(1, get_resolution(h) + 1):
+    h = _set_digit(h, r, _ROT_CW[_digit(h, r)])
+  return h
+
+
+def _rotate_pent60ccw(h: int) -> int:
+  found = False
+  for r in range(1, get_resolution(h) + 1):
+    h = _set_digit(h, r, _ROT_CCW[_digit(h, r)])
+    if not found and _digit(h, r) != 0:
+      found = True
+      if _leading_digit(h) == _K:
+        h = _rotate_index60ccw(h)
+  return h
+
+
+def _is_pentagon_base(bc: int) -> bool:
+  return 0 <= bc < NUM_BASE_CELLS and bool(BASE_CELL_DATA[bc][4])
+
+
+def _face_ijk_base(face: int, c) -> tuple[int, int]:
+  return FACE_IJK_BASE_CELLS[face * 27 + c[0] * 9 + c[1] * 3 + c[2]]
+
+
+def _face_ijk_to_h3(face: int, c, res: int) -> int:
+  h = _H3_INIT | (_CELL_MODE << _MODE_OFFSET) | (res << _RES_OFFSET)
+  for r in range(res - 1, -1, -1):
+    last = c
+    if (r + 1) % 2:
+      c = _up_ap7(c)
+      center = _down_ap7(c)
+    else:
+      c = _up_ap7r(c)
+      center = _down_ap7r(c)
+    diff = _normalize(last[0] - center[0], last[1] - center[1], last[2] - center[2])
+    h = _set_digit(h, r + 1, _unit_to_digit(diff))
+
+  if max(c) > 2:
+    raise ValueError("coordinates off the base cell's face")
+  bc, rots = _face_ijk_base(face, c)
+  h = (h & ~(127 << _BC_OFFSET)) | (bc << _BC_OFFSET)
+
+  if _is_pentagon_base(bc):
+    if _leading_digit(h) == _K:
+      data = BASE_CELL_DATA[bc]
+      h = _rotate_index60cw(h) if face in (data[5], data[6]) else _rotate_index60ccw(h)
+    for _ in range(rots):
+      h = _rotate_pent60ccw(h)
+  else:
+    for _ in range(rots):
+      h = _rotate_index60ccw(h)
+  return h
+
+
+def _adjust_overage_class2(face: int, c, res: int, pent_leading4: bool) -> tuple[int, tuple[int, int, int], bool]:
+  """Move coordinates that fell off the face onto the neighboring face. Returns (face, ijk, moved)."""
+  max_dim = _max_dim(res)
+  if sum(c) <= max_dim:
+    return face, c, False
+
+  if c[2] > 0:
+    if c[1] > 0:
+      orient = FACE_NEIGHBORS[face][_JK_Q]
+    else:
+      orient = FACE_NEIGHBORS[face][_KI_Q]
+      if pent_leading4:
+        # rotate around the pentagon's center for its missing sequence
+        tmp = _rotate60cw((c[0] - max_dim, c[1], c[2]))
+        c = _normalize(tmp[0] + max_dim, tmp[1], tmp[2])
+  else:
+    orient = FACE_NEIGHBORS[face][_IJ_Q]
+
+  new_face, (ti, tj, tk), ccw = orient
+  for _ in range(ccw):
+    c = _rotate60ccw(c)
+  s = _unit_scale(res)
+  c = _normalize(c[0] + ti * s, c[1] + tj * s, c[2] + tk * s)
+  return new_face, c, True
+
+
+def _h3_to_face_ijk(h: int) -> tuple[int, tuple[int, int, int]]:
+  bc = get_base_cell(h)
+  if not 0 <= bc < NUM_BASE_CELLS:
+    raise ValueError("invalid base cell")
+  if _is_pentagon_base(bc) and _leading_digit(h) == _IK:
+    h = _rotate_index60cw(h)
+
+  data = BASE_CELL_DATA[bc]
+  face, c = data[0], (data[1], data[2], data[3])
+  res = get_resolution(h)
+  possible_overage = _is_pentagon_base(bc) or not (res == 0 or c == (0, 0, 0))
+  for r in range(1, res + 1):
+    c = _down_ap7(c) if r % 2 else _down_ap7r(c)
+    d = _digit(h, r)
+    if 0 < d < 7:
+      c = _add(c, _UNIT_VECS[d])
+  if not possible_overage:
+    return face, c
+
+  orig = c
+  adj_res = res
+  if adj_res % 2:
+    c = _down_ap7r(c)
+    adj_res += 1
+  pent_leading4 = _is_pentagon_base(bc) and _leading_digit(h) == _I
+  face, c, moved = _adjust_overage_class2(face, c, adj_res, pent_leading4)
+  if moved:
+    if _is_pentagon_base(bc):
+      while moved:
+        face, c, moved = _adjust_overage_class2(face, c, adj_res, False)
+    if adj_res != res:
+      c = _up_ap7r(c)
+  elif adj_res != res:
+    c = orig
+  return face, c
+
+
+# --- public API ---
+
+def latlng_to_cell(lat: float, lng: float, res: int) -> int:
+  """Cell containing a point given in degrees, like h3.latlng_to_cell."""
+  if not 0 <= res <= MAX_RES:
+    raise ValueError("resolution out of range")
+  if not (math.isfinite(lat) and math.isfinite(lng)):
+    raise ValueError("coordinates must be finite")
+  face, x, y = _geo_to_hex2d(math.radians(lat), math.radians(lng), res)
+  return _face_ijk_to_h3(face, _hex2d_to_ijk(x, y), res)
+
+
+def cell_to_latlng(h: int) -> tuple[float, float]:
+  """Center of a cell in degrees."""
+  face, c = _h3_to_face_ijk(h)
+  lat, lng = _hex2d_to_geo(*_ijk_to_hex2d(c), face, get_resolution(h))
+  return math.degrees(lat), math.degrees(lng)
+
+
+def cell_to_parent(h: int, res: int) -> int:
+  child_res = get_resolution(h)
+  if not 0 <= res <= child_res:
+    raise ValueError("parent resolution must be between 0 and the cell's resolution")
+  h = (h & ~(15 << _RES_OFFSET)) | (res << _RES_OFFSET)
+  for r in range(res + 1, child_res + 1):
+    h = _set_digit(h, r, 7)
+  return h
+
+
+def grid_disk1(h: int) -> set[int]:
+  """The cell and the cells touching it (h3.grid_disk(h, 1)).
+
+  Steps one cell along each of the six ijk axes on the cell's face and maps
+  that center back to a cell, so it also works across face edges.
+  """
+  face, c = _h3_to_face_ijk(h)
+  res = get_resolution(h)
+  out = {h}
+  for u in _UNIT_VECS[1:]:
+    x, y = _ijk_to_hex2d(_add(c, u))
+    lat, lng = _hex2d_to_geo(x, y, face, res)
+    out.add(latlng_to_cell(math.degrees(lat), math.degrees(lng), res))
+  return out
+
+
+def cell_to_str(h: int) -> str:
+  return format(h, "x")
+
+
+def str_to_cell(s: str) -> int:
+  return int(s, 16)
@@ -0,0 +1,236 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+Dry run of the community uploads: writes exactly what the car would send, and
+a map page that shows what would be shared and what stays on the car, into
+<achievements root>/preview/. Nothing is sent anywhere.
+
+  preview/daily-YYYY-MM-DD.json   daily report per driven day (last 14 days), at level 2
+  preview/tiles.json              the next weekly tile batch (level 3)
+  preview/delete.json             tiles the car would ask the server to delete, if any
+  preview/map.html                open in a browser (copy it off the car first)
+
+map.html shows your frequent stops, so it is for you only: don't share it.
+"""
+from __future__ import annotations
+
+import datetime as dt
+import json
+from pathlib import Path
+
+from openpilot.sunnypilot.achievements.community import (LEVEL_DAILY, STOP_HOLD_DAYS, CommunityLog, Settings,
+                                                         public_block)
+from openpilot.sunnypilot.achievements.engine import State, achievements_root
+
+PREVIEW_DAYS = 14
+
+
+def preview_dir() -> Path:
+  return achievements_root() / "preview"
+
+
+def _write(path: Path, data) -> None:
+  tmp = path.with_suffix(path.suffix + ".tmp")
+  tmp.write_text(data if isinstance(data, str) else json.dumps(data, indent=2) + "\n")
+  tmp.replace(path)
+
+
+def build_preview(log: CommunityLog, state: State, today: dt.date, car: str | None, settings: Settings) -> dict:
+  """Everything the preview shows, as plain data."""
+  view = log.tile_view(today)
+  profile = log.profile_update(today, car, settings)
+  days = log.driven_days(since=today - dt.timedelta(days=PREVIEW_DAYS - 1))
+  daily = [log.daily_report(d, state, LEVEL_DAILY, profile=profile if d == days[-1] else None,
+                            include_undated=d == days[-1]) for d in days]
+  region_days = log.region_days(today)
+  return {
+    "today": today.isoformat(),
+    "level": settings.level,
+    "daily": daily,
+    "tiles": log.tile_batch(today),
+    "delete": log.tile_deletes(today),
+    "region": {"home": log.home_region(today), "shared": log.shared_region(today, settings),
+               "choice": settings.region, "days": dict(region_days.most_common())},
+    "map": {
+      "shareable": sorted(view.shareable),
+      "blocks": sorted({public_block(t) for t in view.shareable}),
+      "held": view.held,
+      "ring": sorted(view.ring),
+      "frequent": sorted(view.frequent),
+      "hidden": sorted(view.hidden & view.driven),
+      "driven": len(view.driven),
+    },
+  }
+
+
+def write_preview(log: CommunityLog, state: State, today: dt.date, car: str | None, settings: Settings,
+                  out: Path | None = None) -> Path:
+  out = out or preview_dir()
+  out.mkdir(parents=True, exist_ok=True)
+  data = build_preview(log, state, today, car, settings)
+
+  for f in out.glob("daily-*.json"):
+    f.unlink()
+  for report in data["daily"]:
+    _write(out / f"daily-{report['day']}.json", report)
+  for name in ("tiles", "delete"):
+    path = out / f"{name}.json"
+    if data[name]:
+      _write(path, data[name])
+    elif path.exists():
+      path.unlink()
+  _write(out / "map.html", render_html(data))
+  return out
+
+
+def render_html(data: dict) -> str:
+  blob = json.dumps(data, separators=(",", ":")).replace("</", "<\\/")
+  return HTML.replace("__DATA__", blob).replace("__HOLD_DAYS__", str(STOP_HOLD_DAYS))
+
+
+HTML = """<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>What your car would share</title>
+<link rel="stylesheet" href="https://unpkg.com/maplibre-gl@4.7.1/dist/maplibre-gl.css">
+<script src="https://unpkg.com/maplibre-gl@4.7.1/dist/maplibre-gl.js"></script>
+<script src="https://unpkg.com/h3-js@4.1.0/dist/h3-js.umd.js"></script>
+<style>
+  :root { --ink:#1f2328; --muted:#59636e; --line:#d1d9e0; --bg:#f6f8fa; --send:#2563eb; --held:#9ca3af;
+          --never:#dc2626; --hidden:#7c3aed; }
+  * { box-sizing: border-box; }
+  body { margin:0; font:15px/1.5 Inter, system-ui, sans-serif; color:var(--ink); background:#fff; }
+  header { padding:20px 24px 8px; }
+  h1 { font-size:22px; margin:0 0 4px; }
+  .warn { background:#fff8c5; border:1px solid #d4a72c66; border-radius:8px; padding:8px 12px; margin:8px 0 0; color:#3b2300; }
+  .stats { display:flex; flex-wrap:wrap; gap:8px; padding:8px 24px; }
+  .stat { border:1px solid var(--line); border-radius:8px; padding:8px 12px; min-width:150px; }
+  .stat b { display:block; font-size:20px; }
+  .dot { display:inline-block; width:10px; height:10px; border-radius:2px; margin-right:6px; vertical-align:middle; }
+  .bar { display:flex; flex-wrap:wrap; gap:8px; align-items:center; padding:8px 24px; }
+  button { font:inherit; border:1px solid var(--line); background:#fff; border-radius:6px; padding:6px 12px; cursor:pointer; }
+  button.on { background:var(--ink); color:#fff; border-color:var(--ink); }
+  #map { height:62vh; min-height:360px; margin:0 24px; border:1px solid var(--line); border-radius:8px; }
+  main { padding:8px 24px 40px; display:grid; align-items:start; grid-template-columns:repeat(auto-fit, minmax(320px, 1fr)); gap:16px; }
+  section { border:1px solid var(--line); border-radius:8px; padding:12px 16px; }
+  h2 { font-size:16px; margin:0 0 6px; }
+  p { margin:4px 0; color:var(--muted); }
+  pre { background:var(--bg); border-radius:6px; padding:10px; overflow:auto; font-size:12px; max-height:340px; }
+</style>
+</head>
+<body>
+<header>
+  <h1>What your car would share</h1>
+  <div id="sub"></div>
+  <div class="warn">Dry run: nothing has been sent. This page marks your frequent stops, so keep it to yourself.</div>
+</header>
+<div class="stats" id="stats"></div>
+<div class="bar">
+  <button id="mine" class="on">Your map (6 km tiles)</button>
+  <button id="others">What others see (45 km blocks)</button>
+  <label><input type="checkbox" id="kept" checked> Show what stays on the car</label>
+</div>
+<div id="map"></div>
+<main>
+  <section><h2>Home region</h2><div id="region"></div></section>
+  <section><h2>Latest daily report</h2>
+    <p>Level 1 sends <code>unlocks</code> and <code>totals</code>. Level 2 adds <code>daily</code> and <code>profile</code>.</p>
+    <pre id="daily"></pre></section>
+  <section><h2>Next weekly tile batch (level 3)</h2>
+    <p>Sorted tile ids, no dates, sent separately from the daily report.</p><pre id="tiles"></pre></section>
+  <section><h2>Delete requests</h2>
+    <p>Tiles already sent that became private later (moved house, new job, hidden).</p><pre id="delete"></pre></section>
+</main>
+<script>
+const D = __DATA__;
+const M = D.map;
+const held = Object.keys(M.held);
+const never = M.ring;
+document.getElementById('sub').textContent =
+  `As of ${D.today}. Sharing level on this car: ${D.level} (0 = off). The preview always shows level 3.`;
+const stat = (color, n, label) =>
+  `<div class="stat"><b>${n}</b><span class="dot" style="background:${color}"></span>${label}</div>`;
+document.getElementById('stats').innerHTML =
+  stat('#6e7781', M.driven, 'tiles driven') +
+  stat('var(--send)', M.shareable.length, 'would be shared') +
+  stat('var(--held)', held.length, 'drive start/end tiles waiting __HOLD_DAYS__ days') +
+  stat('var(--never)', never.length, `never leave the car (${M.frequent.length} frequent stops + the tiles around them)`) +
+  stat('var(--hidden)', M.hidden.length, 'hidden by you') +
+  stat('var(--send)', M.blocks.length, '45 km blocks others would see');
+
+const r = D.region;
+const days = Object.entries(r.days).map(([k, v]) => `${k}: ${v} day${v === 1 ? '' : 's'}`).join(', ') || 'no engaged days yet';
+document.getElementById('region').innerHTML =
+  `<p>Detected: <b>${r.home || 'none yet'}</b> (${days}, last 90 days).</p>` +
+  `<p>Would be shared as: <b>${r.shared || 'nothing'}</b> (setting: ${r.choice}).</p>`;
+const latest = D.daily.length ? D.daily[D.daily.length - 1] : null;
+document.getElementById('daily').textContent = latest ? JSON.stringify(latest, null, 2) : 'No driving yet.';
+const t = D.tiles;
+document.getElementById('tiles').textContent = t ?
+  JSON.stringify({...t, tiles: t.tiles.slice(0, 12).concat(t.tiles.length > 12 ? [`... ${t.tiles.length - 12} more`] : [])}, null, 2) :
+  'Nothing new to send.';
+document.getElementById('delete').textContent = D.delete ? JSON.stringify(D.delete, null, 2) : 'None.';
+
+const fc = (cells, kind) => ({type: 'FeatureCollection', features: cells.map(c => {
+  const ring = h3.cellToBoundary(c, true); ring.push(ring[0]);
+  return {type: 'Feature', properties: {kind, cell: c, until: M.held[c] || ''}, geometry: {type: 'Polygon', coordinates: [ring]}};
+})});
+const all = [...M.shareable, ...held, ...never, ...M.hidden];
+const map = new maplibregl.Map({
+  container: 'map',
+  style: {version: 8, sources: {osm: {type: 'raster', tiles: ['https://tile.openstreetmap.org/{z}/{x}/{y}.png'],
+          tileSize: 256, attribution: '&copy; OpenStreetMap contributors'}},
+          layers: [{id: 'osm', type: 'raster', source: 'osm', paint: {'raster-saturation': -0.8, 'raster-opacity': 0.8}}]},
+  center: [-83, 40], zoom: 5,
+});
+map.on('load', () => {
+  const add = (id, data, color, fill, dash) => {
+    map.addSource(id, {type: 'geojson', data});
+    map.addLayer({id: id + '-fill', type: 'fill', source: id, paint: {'fill-color': color, 'fill-opacity': fill}});
+    map.addLayer({id: id + '-line', type: 'line', source: id,
+                  paint: {'line-color': color, 'line-width': 1.2, ...(dash ? {'line-dasharray': [2, 2]} : {})}});
+  };
+  add('blocks', fc(M.blocks, 'block'), '#2563eb', 0.35);
+  add('mine', fc(M.shareable, 'shared'), '#2563eb', 0.35);
+  add('held', fc(held, 'held'), '#9ca3af', 0.45, true);
+  add('never', fc(never, 'never'), '#dc2626', 0.25, true);
+  add('hidden', fc(M.hidden, 'hidden'), '#7c3aed', 0.3, true);
+  const kept = ['held', 'never', 'hidden'];
+  const show = (id, on) => ['-fill', '-line'].forEach(s => map.setLayoutProperty(id + s, 'visibility', on ? 'visible' : 'none'));
+  const fit = cells => {
+    if (!cells.length) return;
+    const pts = cells.flatMap(c => h3.cellToBoundary(c));
+    const lats = pts.map(p => p[0]), lngs = pts.map(p => p[1]);
+    map.fitBounds([[Math.min(...lngs), Math.min(...lats)], [Math.max(...lngs), Math.max(...lats)]], {padding: 40, duration: 0});
+  };
+  let mode = 'mine';
+  const refresh = () => {
+    show('mine', mode === 'mine'); show('blocks', mode === 'others');
+    const k = document.getElementById('kept').checked && mode === 'mine';
+    kept.forEach(id => show(id, k));
+    document.getElementById('mine').classList.toggle('on', mode === 'mine');
+    document.getElementById('others').classList.toggle('on', mode === 'others');
+  };
+  document.getElementById('mine').onclick = () => { mode = 'mine'; refresh(); fit(all); };
+  document.getElementById('others').onclick = () => { mode = 'others'; refresh(); fit(M.blocks); };
+  document.getElementById('kept').onchange = refresh;
+  refresh();
+  fit(all);
+  const label = {shared: 'Would be shared', held: 'Waiting until', never: 'Never leaves the car', hidden: 'Hidden by you',
+                 block: 'What others see'};
+  map.on('click', e => {
+    const f = map.queryRenderedFeatures(e.point).find(x => x.properties.kind);
+    if (!f) return;
+    new maplibregl.Popup().setLngLat(e.lngLat)
+      .setHTML(`${label[f.properties.kind]} ${f.properties.until}<br><code>${f.properties.cell}</code>`).addTo(map);
+  });
+});
+</script>
+</body>
+</html>
+"""
@@ -0,0 +1,74 @@
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+Home region lookup: map tile -> region code (US-OH, CA-ON, or a country code
+like GB), from the table in regions.json.gz. Works offline on the car; no
+geocoding service is ever called. See tools/build_regions.py for how the table
+is made.
+"""
+from __future__ import annotations
+
+import gzip
+import json
+from functools import lru_cache
+from pathlib import Path
+
+import numpy as np
+
+from openpilot.sunnypilot.achievements import h3lite
+
+TABLE_PATH = Path(__file__).parent / "regions.json.gz"
+
+
+class RegionTable:
+  def __init__(self, path: Path = TABLE_PATH):
+    with gzip.open(path, "rt") as f:
+      data = json.load(f)
+    self.tile_res: int = data["tile_res"]
+    self.base_res: int = data["base_res"]
+    self.codes: list[str] = sorted(set(data["regions"]) | set(data["exceptions"]))
+    index = {c: i for i, c in enumerate(self.codes)}
+    self._base = self._arrays(data["regions"], index)
+    self._exceptions = self._arrays(data["exceptions"], index)
+
+  @staticmethod
+  def _arrays(groups: dict[str, list[str]], index: dict[str, int]) -> tuple[np.ndarray, np.ndarray]:
+    pairs = sorted((int(cell, 16), index[code]) for code, cells in groups.items() for cell in cells)
+    keys = np.array([p[0] for p in pairs], dtype=np.int64)
+    vals = np.array([p[1] for p in pairs], dtype=np.uint16)
+    return keys, vals
+
+  @staticmethod
+  def _find(arrays: tuple[np.ndarray, np.ndarray], cell: int) -> int | None:
+    keys, vals = arrays
+    i = int(np.searchsorted(keys, cell))
+    if i < len(keys) and int(keys[i]) == cell:
+      return int(vals[i])
+    return None
+
+  def lookup_cell(self, cell: int) -> str | None:
+    """Region for a tile at tile_res (or finer). None at sea or off the map."""
+    res = h3lite.get_resolution(cell)
+    if res > self.tile_res:
+      cell, res = h3lite.cell_to_parent(cell, self.tile_res), self.tile_res
+    for table, top in ((self._exceptions, res), (self._base, min(res, self.base_res))):
+      for r in range(top, -1, -1):
+        hit = self._find(table, h3lite.cell_to_parent(cell, r))
+        if hit is not None:
+          return self.codes[hit]
+    return None
+
+  def lookup(self, lat: float, lng: float) -> str | None:
+    return self.lookup_cell(h3lite.latlng_to_cell(lat, lng, self.tile_res))
+
+
+@lru_cache(maxsize=1)
+def table() -> RegionTable:
+  return RegionTable()
+
+
+def country_of(region: str) -> str:
+  return region.split("-", 1)[0]
@@ -0,0 +1,162 @@
+#!/usr/bin/env python3
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+Runs achievements and the privacy filter over the drives already saved on the
+car, then writes the dry-run preview. It uses its own folder, so live progress
+is not touched, and it sends nothing.
+
+On the car (over SSH):
+  cd /data/openpilot && python -m openpilot.sunnypilot.achievements.replay
+Then on your computer:
+  scp comma@<car ip>:/data/media/0/achievements/replay/preview/map.html .
+
+Reads qlogs from /data/media/0/realdata by default (--rlog for full logs, which
+also count lane changes; much slower). Each route counts as one drive.
+"""
+from __future__ import annotations
+
+import argparse
+import collections
+import datetime
+import re
+import shutil
+import sys
+import time
+from pathlib import Path
+
+from openpilot.sunnypilot.achievements.community import CommunityLog, Settings
+from openpilot.sunnypilot.achievements.engine import (MAX_GPS_ACCURACY_M, AchievementEngine, Sample, State,
+                                                      achievements_root)
+from openpilot.sunnypilot.achievements.preview import write_preview
+
+SEGMENT = re.compile(r"^(?P<route>.+)--(?P<seg>\d+)$")
+GPS_MAX_AGE_S = 2.0
+MIN_VALID_UNIX = 1.6e9  # anything earlier is a clock that wasn't set yet
+
+
+def _now() -> float:
+  return datetime.datetime.now(datetime.UTC).timestamp()
+
+
+def find_routes(log_root: Path, rlog: bool, days: int | None) -> list[list[Path]]:
+  names = ("rlog.zst", "rlog.bz2", "rlog") if rlog else ("qlog.zst", "qlog.bz2", "qlog")
+  cutoff = _now() - days * 86400 if days else 0
+  routes: dict[str, list[tuple[int, float, Path]]] = collections.defaultdict(list)
+  for d in log_root.iterdir():
+    m = SEGMENT.match(d.name)
+    if not m or not d.is_dir():
+      continue
+    f = next((d / n for n in names if (d / n).exists()), None)
+    if f is None or f.stat().st_mtime < cutoff:
+      continue
+    routes[m["route"]].append((int(m["seg"]), f.stat().st_mtime, f))
+  ordered = sorted(routes.values(), key=lambda segs: min(s[1] for s in segs))
+  return [[f for _, _, f in sorted(segs)] for segs in ordered]
+
+
+class RouteFeeder:
+  """Turns one route's log messages into Samples, the way achievementsd does live."""
+
+  def __init__(self, engine: AchievementEngine, log: CommunityLog):
+    self.engine, self.log = engine, log
+    self.op_active = self.mads_active = False
+    self.lane_change = self.dm_alert = 0
+    self.gps: tuple[float, float, int] | None = None
+    self.clock: tuple[int, float] | None = None  # (logMonoTime, unix seconds)
+    self.car: str | None = None
+    self.samples = 0
+
+  def feed(self, ev) -> None:
+    w = ev.which()
+    if w == "carState":
+      self._sample(ev.logMonoTime, ev.carState.vEgo)
+    elif w == "selfdriveState":
+      self.op_active = ev.selfdriveState.active
+    elif w == "selfdriveStateSP":
+      self.mads_active = ev.selfdriveStateSP.mads.active
+    elif w == "modelV2":
+      self.lane_change = ev.modelV2.meta.laneChangeState.raw
+    elif w == "driverMonitoringState":
+      self.dm_alert = ev.driverMonitoringState.alertLevel.raw
+    elif w in ("gpsLocationExternal", "gpsLocation"):
+      g = getattr(ev, w)
+      if g.hasFix and 0 < g.horizontalAccuracy < MAX_GPS_ACCURACY_M:
+        self.gps = (g.latitude, g.longitude, ev.logMonoTime)
+        if g.unixTimestampMillis / 1000.0 > MIN_VALID_UNIX:
+          self.clock = (ev.logMonoTime, g.unixTimestampMillis / 1000.0)
+    elif w == "initData" and self.clock is None:
+      if ev.initData.wallTimeNanos / 1e9 > MIN_VALID_UNIX:
+        self.clock = (ev.logMonoTime, ev.initData.wallTimeNanos / 1e9)
+    elif w == "carParams" and not self.car:
+      self.car = ev.carParams.carFingerprint or None
+
+  def _sample(self, mono: int, v_ego: float) -> None:
+    fix = self.gps if self.gps and (mono - self.gps[2]) / 1e9 < GPS_MAX_AGE_S else None
+    wall = self.clock[1] + (mono - self.clock[0]) / 1e9 if self.clock else None
+    s = Sample(t=mono / 1e9, v_ego=v_ego, op_active=self.op_active, mads_active=self.mads_active,
+               lane_change_state=self.lane_change, dm_alert_level=self.dm_alert,
+               lat=fix[0] if fix else None, lon=fix[1] if fix else None, wall=wall)
+    self.engine.update(s)
+    self.log.update(s, self.engine.last_step)
+    self.samples += 1
+
+
+def main(argv: list[str] | None = None) -> int:
+  from openpilot.common.hardware.hw import Paths
+  from openpilot.tools.lib.logreader import LogReader
+
+  ap = argparse.ArgumentParser(description=__doc__.split("\n\n")[1], formatter_class=argparse.RawDescriptionHelpFormatter)
+  ap.add_argument("--logs", type=Path, default=Path(Paths.log_root()), help="folder with route segments")
+  ap.add_argument("--out", type=Path, default=achievements_root() / "replay", help="where to write (wiped first)")
+  ap.add_argument("--days", type=int, default=None, help="only routes from the last N days")
+  ap.add_argument("--rlog", action="store_true", help="read full rlogs instead of qlogs")
+  args = ap.parse_args(argv)
+
+  routes = find_routes(args.logs, args.rlog, args.days)
+  if not routes:
+    print(f"no {'rlogs' if args.rlog else 'qlogs'} found in {args.logs}")
+    return 1
+  if args.out.exists():
+    shutil.rmtree(args.out)
+  args.out.mkdir(parents=True)
+
+  engine = AchievementEngine(State())
+  log = CommunityLog(args.out / "community.db")
+  car = None
+  n_segs = sum(len(r) for r in routes)
+  done = 0
+  start = time.monotonic()
+  for segments in routes:
+    feeder = RouteFeeder(engine, log)
+    for path in segments:
+      try:
+        for ev in LogReader(str(path), sort_by_time=True):
+          feeder.feed(ev)
+      except Exception as e:  # one bad segment shouldn't stop the rest
+        print(f"skipping {path}: {e}", file=sys.stderr)
+      done += 1
+      print(f"\r{done}/{n_segs} segments", end="", file=sys.stderr, flush=True)
+    engine.end_drive()
+    log.end_drive(engine.utc_offset_h)
+    car = feeder.car or car
+  print(file=sys.stderr)
+
+  today = engine.local_date(_now())
+  out = write_preview(log, engine.state, today, car, Settings.load(), args.out / "preview")
+  view = log.tile_view(today)
+  print(f"{len(routes)} drives, {n_segs} segments in {time.monotonic() - start:.0f} s")
+  print(f"engaged: {engine.state.engaged_m / 1609.344:.0f} mi over {len(engine.state.days)} days")
+  kept = len(view.ring & view.driven)
+  print(f"tiles: {len(view.driven)} driven, {len(view.shareable)} would be shared, {len(view.held)} waiting, {kept} kept on the car")
+  print(f"frequent stops (never sent, with the tiles around them): {len(view.frequent)}")
+  print(f"home region: {log.home_region(today)}")
+  print(f"preview: {out / 'map.html'}")
+  return 0
+
+
+if __name__ == "__main__":
+  sys.exit(main())
@@ -0,0 +1,113 @@
+#!/usr/bin/env python3
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+Made-up drives for the tests and for a demo preview without a car:
+
+  python -m openpilot.sunnypilot.achievements.simulate /tmp/demo
+
+writes /tmp/demo/preview/map.html for two weeks of a commute from Hilliard to
+Fairborn, Ohio, plus two weekend trips.
+"""
+from __future__ import annotations
+
+import datetime as dt
+import math
+import os
+import sys
+from dataclasses import dataclass
+from pathlib import Path
+
+from openpilot.sunnypilot.achievements.community import CommunityLog
+from openpilot.sunnypilot.achievements.engine import AchievementEngine, Sample, State
+
+HOME = (40.0334, -83.1588)          # Hilliard
+WORK = (39.8209, -84.0194)          # Fairborn
+SPRINGFIELD = (39.9242, -83.8088)
+CINCINNATI = (39.1031, -84.5120)
+DAYTON = (39.7589, -84.1916)
+CLEVELAND = (41.4993, -81.6944)
+MANSFIELD = (40.7584, -82.5154)
+COLUMBUS = (39.9612, -82.9988)
+
+
+@dataclass
+class Drive:
+  day: dt.date
+  route: list[tuple[float, float]]
+  start_utc_h: float = 13.0         # 8 am in Ohio
+  engaged: bool = True
+
+
+def _km(a: tuple[float, float], b: tuple[float, float]) -> float:
+  la1, lo1, la2, lo2 = map(math.radians, (*a, *b))
+  h = math.sin((la2 - la1) / 2) ** 2 + math.cos(la1) * math.cos(la2) * math.sin((lo2 - lo1) / 2) ** 2
+  return 2 * 6371.0 * math.asin(math.sqrt(h))
+
+
+def drive_samples(drive: Drive, t0: float, speed_ms: float = 30.0):
+  """1 Hz samples along straight lines between the route's points."""
+  wall = dt.datetime(drive.day.year, drive.day.month, drive.day.day, tzinfo=dt.UTC).timestamp() + drive.start_utc_h * 3600
+  t = t0
+  for a, b in zip(drive.route, drive.route[1:], strict=False):
+    n = max(1, int(_km(a, b) * 1000 / speed_ms))
+    for i in range(n):
+      f = i / n
+      lat, lon = a[0] + (b[0] - a[0]) * f, a[1] + (b[1] - a[1]) * f
+      yield Sample(t=t, v_ego=speed_ms, op_active=drive.engaged, lat=lat, lon=lon, wall=wall)
+      t += 1.0
+      wall += 1.0
+  end = drive.route[-1]
+  yield Sample(t=t, v_ego=0.0, op_active=False, lat=end[0], lon=end[1], wall=wall)
+
+
+def run(drives: list[Drive], engine: AchievementEngine, log: CommunityLog, speed_ms: float = 30.0) -> None:
+  t = 1000.0
+  for d in drives:
+    for s in drive_samples(d, t, speed_ms):
+      engine.update(s)
+      log.update(s, engine.last_step)
+      t = s.t
+    engine.end_drive()
+    log.end_drive(engine.utc_offset_h)
+    t += 3600.0
+
+
+def commute_weeks(first_monday: dt.date, weeks: int = 2) -> list[Drive]:
+  drives = []
+  for w in range(weeks):
+    monday = first_monday + dt.timedelta(weeks=w)
+    for i in range(5):
+      day = monday + dt.timedelta(days=i)
+      drives.append(Drive(day, [HOME, SPRINGFIELD, WORK], 12.5))
+      drives.append(Drive(day, [WORK, SPRINGFIELD, HOME], 21.5))
+    saturday = monday + dt.timedelta(days=5)
+    trip = [HOME, DAYTON, CINCINNATI] if w % 2 == 0 else [HOME, COLUMBUS, MANSFIELD, CLEVELAND]
+    drives.append(Drive(saturday, trip, 14.0))
+    drives.append(Drive(saturday, list(reversed(trip)), 20.0))
+  return drives
+
+
+def main(out: str) -> None:
+  from openpilot.sunnypilot.achievements.community import Settings
+  from openpilot.sunnypilot.achievements.preview import write_preview
+
+  root = Path(out)
+  root.mkdir(parents=True, exist_ok=True)
+  db = root / "community.db"
+  if db.exists():
+    os.remove(db)
+  first_monday = dt.date(2026, 10, 26)
+  engine = AchievementEngine(State())
+  log = CommunityLog(db)
+  run(commute_weeks(first_monday), engine, log)
+  today = first_monday + dt.timedelta(days=15)
+  path = write_preview(log, engine.state, today, "HYUNDAI_IONIQ_5", Settings(level=0), root / "preview")
+  print(f"wrote {path / 'map.html'}")
+
+
+if __name__ == "__main__":
+  main(sys.argv[1] if len(sys.argv) > 1 else "/tmp/achievements_demo")
@@ -0,0 +1,195 @@
+"""
+The privacy filter, on made-up drives: two weeks of a Hilliard -> Fairborn
+commute plus two Saturday trips (see simulate.py).
+"""
+import datetime as dt
+import re
+import sqlite3
+
+import pytest
+
+from openpilot.sunnypilot.achievements import h3lite
+from openpilot.sunnypilot.achievements.community import (LEVEL_ACHIEVEMENTS, LEVEL_DAILY, LEVEL_MAP, LEVEL_OFF,
+                                                         CommunityLog, Settings, public_block, ring_of, tile_of)
+from openpilot.sunnypilot.achievements.engine import AchievementEngine, State
+from openpilot.sunnypilot.achievements.preview import build_preview, write_preview
+from openpilot.sunnypilot.achievements.regions import table
+from openpilot.sunnypilot.achievements.simulate import (CINCINNATI, CLEVELAND, HOME, WORK, Drive, commute_weeks,
+                                                        run)
+
+FIRST_MONDAY = dt.date(2026, 10, 26)
+TODAY = FIRST_MONDAY + dt.timedelta(days=15)  # Tuesday after the second week
+HEX = re.compile(r"^86[0-9a-f]{13}$")
+
+
+@pytest.fixture(scope="module")
+def world(tmp_path_factory):
+  engine = AchievementEngine(State())
+  log = CommunityLog(tmp_path_factory.mktemp("community") / "community.db")
+  run(commute_weeks(FIRST_MONDAY), engine, log, speed_ms=60.0)
+  return engine, log
+
+
+def test_home_and_work_never_leave(world):
+  _, log = world
+  view = log.tile_view(TODAY)
+  home, work = tile_of(*HOME), tile_of(*WORK)
+  assert view.frequent == {home, work}
+  assert ring_of(home) | ring_of(work) <= view.ring
+  assert len(ring_of(home)) == 7
+  assert not view.shareable & view.ring
+  batch = log.tile_batch(TODAY)
+  assert not set(batch["tiles"]) & view.ring
+
+
+def test_trip_ends_wait_14_days(world):
+  _, log = world
+  cincinnati = tile_of(*CINCINNATI)  # Saturday Oct 31
+  cleveland = tile_of(*CLEVELAND)    # Saturday Nov 7
+  view = log.tile_view(TODAY)
+  assert view.held == {cincinnati: "2026-11-14", cleveland: "2026-11-21"}
+  assert cincinnati not in view.shareable
+
+  later = log.tile_view(dt.date(2026, 11, 14))
+  assert cincinnati in later.shareable
+  assert cleveland in later.held
+
+
+def test_tile_batch_has_no_dates(world):
+  _, log = world
+  batch = log.tile_batch(TODAY)
+  assert set(batch) == {"v", "device", "tiles"}
+  assert batch["tiles"] == sorted(batch["tiles"])
+  assert all(HEX.match(t) for t in batch["tiles"])
+  assert len(batch["tiles"]) == len(log.tile_view(TODAY).shareable)
+
+
+def test_nothing_finer_than_a_tile_is_stored(world):
+  _, log = world
+  db = sqlite3.connect(log.path)
+  assert all(HEX.match(c) for (c,) in db.execute("SELECT cell FROM tiles UNION SELECT cell FROM stops"))
+  days = {d for (d,) in db.execute("SELECT first_day FROM tiles UNION SELECT day FROM stops UNION SELECT day FROM days")}
+  assert all(re.fullmatch(r"\d{4}-\d\d-\d\d", d) for d in days)
+
+
+def test_others_see_45km_blocks(world):
+  _, log = world
+  data = build_preview(log, world[0].state, TODAY, None, Settings())
+  shareable = log.tile_view(TODAY).shareable
+  assert set(data["map"]["blocks"]) == {public_block(t) for t in shareable}
+  assert all(h3lite.get_resolution(h3lite.str_to_cell(b)) == 4 for b in data["map"]["blocks"])
+  # the commute fills in the blocks around home and work, so no hole points at them
+  home_block, work_block = public_block(tile_of(*HOME)), public_block(tile_of(*WORK))
+  assert {home_block, work_block} <= set(data["map"]["blocks"])
+
+
+def test_daily_report_levels(world):
+  engine, log = world
+  day = "2026-11-06"
+  assert log.daily_report(day, engine.state, LEVEL_OFF) is None
+
+  r1 = log.daily_report(day, engine.state, LEVEL_ACHIEVEMENTS)
+  assert set(r1) == {"v", "device", "day", "unlocks", "totals"}
+  assert r1["totals"]["engaged_mi"] % 10 == 0 and r1["totals"]["engaged_h"] % 10 == 0
+
+  r2 = log.daily_report(day, engine.state, LEVEL_DAILY, profile={"region": "US-OH"})
+  assert set(r2) == {"v", "device", "day", "unlocks", "totals", "daily", "profile"}
+  assert set(r2["daily"]) == {"engaged_mi", "engaged_min", "new_places"}
+  assert all(isinstance(v, int) for v in r2["daily"].values())
+  assert 80 < r2["daily"]["engaged_mi"] < 110  # two ~47 mile legs
+  assert LEVEL_MAP > LEVEL_DAILY
+
+
+def test_unlocks_go_in_the_report_for_their_day(world):
+  engine, log = world
+  days = {}
+  for uid in engine.state.unlocked:
+    for day in log.driven_days():
+      if any(u["id"] == uid for u in log.daily_report(day, engine.state, LEVEL_ACHIEVEMENTS)["unlocks"]):
+        days.setdefault(uid, []).append(day)
+  assert days and all(len(v) == 1 for v in days.values())
+
+
+def test_home_region(world):
+  _, log = world
+  assert log.home_region(TODAY) == "US-OH"
+  assert log.shared_region(TODAY, Settings(region="country")) == "US"
+  assert log.shared_region(TODAY, Settings(region="hidden")) is None
+  assert log.shared_region(TODAY, Settings(region="US-KY")) == "US-KY"
+
+
+def test_profile_only_when_changed(tmp_path):
+  engine, log = AchievementEngine(State()), CommunityLog(tmp_path / "c.db")
+  run([Drive(dt.date(2026, 11, 2), [HOME, WORK])], engine, log, speed_ms=60.0)
+  today = dt.date(2026, 11, 3)
+  auto = Settings()
+  first = log.profile_update(today, "HYUNDAI_IONIQ_5", auto)
+  assert first == {"car": "HYUNDAI_IONIQ_5", "region": "US-OH"}
+  log.mark_profile_sent(today, first, auto)
+  assert log.profile_update(today, "HYUNDAI_IONIQ_5", auto) is None
+
+  # a road trip that moves the automatic home region waits 30 days after the last send
+  for i in range(3):
+    run([Drive(dt.date(2026, 11, 4 + i), [CINCINNATI, (38.25, -85.76)])], engine, log, speed_ms=60.0)  # into Kentucky
+  assert log.home_region(dt.date(2026, 11, 7)) != "US-OH"
+  assert log.profile_update(dt.date(2026, 11, 7), "HYUNDAI_IONIQ_5", auto) is None
+  assert "region" in log.profile_update(dt.date(2026, 12, 3), "HYUNDAI_IONIQ_5", auto)
+  # a region the driver picks goes right away; hiding sends an empty region
+  assert log.profile_update(today, "HYUNDAI_IONIQ_5", Settings(region="US-MI")) == {"region": "US-MI"}
+  assert log.profile_update(today, "HYUNDAI_IONIQ_5", Settings(region="hidden")) == {"region": None}
+
+
+def test_late_catches_and_hidden_tiles_are_deleted(tmp_path):
+  engine, log = AchievementEngine(State()), CommunityLog(tmp_path / "c.db")
+  friend = (39.95, -82.85)  # east Columbus
+  run([Drive(dt.date(2026, 10, 1), [HOME, friend])], engine, log, speed_ms=60.0)
+  today = dt.date(2026, 10, 20)
+  view = log.tile_view(today)
+  assert tile_of(*friend) in view.shareable  # one visit, held 14 days, then shared
+  log.mark_sent(view.to_send)
+  assert log.tile_batch(today) is None
+
+  # three more days ending there: it's a regular stop now, so ask the server to delete it
+  run([Drive(dt.date(2026, 10, 21 + i), [HOME, friend]) for i in range(3)], engine, log, speed_ms=60.0)
+  deletes = log.tile_deletes(dt.date(2026, 10, 25))
+  assert tile_of(*friend) in deletes["delete_tiles"]
+  assert set(deletes["delete_tiles"]) <= ring_of(tile_of(*friend)) | ring_of(tile_of(*HOME))
+
+  middle = sorted(log.tile_view(dt.date(2026, 10, 25)).shareable)[0]
+  log.hide_tile(middle)
+  view = log.tile_view(dt.date(2026, 10, 25))
+  assert middle not in view.shareable and middle in view.to_delete
+
+
+def test_no_gps_no_tiles(tmp_path):
+  engine, log = AchievementEngine(State()), CommunityLog(tmp_path / "c.db")
+  drive = Drive(dt.date(2026, 11, 2), [HOME, WORK])
+  from openpilot.sunnypilot.achievements.simulate import drive_samples
+  for s in drive_samples(drive, 0.0, 60.0):
+    s.lat = s.lon = None
+    engine.update(s)
+    log.update(s, engine.last_step)
+  log.end_drive()
+  assert log.tile_view(dt.date(2026, 11, 3)).driven == set()
+  m, secs, _ = log.day_stats("2026-11-02")
+  assert m > 50_000 and secs > 0  # engaged miles still count
+
+
+def test_write_preview(world, tmp_path):
+  engine, log = world
+  out = write_preview(log, engine.state, TODAY, "HYUNDAI_IONIQ_5", Settings(), tmp_path / "preview")
+  names = sorted(p.name for p in out.iterdir())
+  assert "map.html" in names and "tiles.json" in names
+  assert "daily-2026-11-07.json" in names
+  html = (out / "map.html").read_text()
+  assert "What your car would share" in html and tile_of(*HOME) in html
+
+
+def test_region_table():
+  t = table()
+  assert t.lookup(*HOME) == "US-OH"
+  assert t.lookup(43.65, -79.38) == "CA-ON"     # Toronto
+  assert t.lookup(40.76, -73.98) == "US-NY"     # Manhattan, next to New Jersey
+  assert t.lookup(38.90, -77.04) == "US-DC"
+  assert t.lookup(51.50, -0.12) == "GB"         # countries without subdivisions in the table
+  assert t.lookup(35.0, -40.0) is None          # mid-Atlantic
@@ -0,0 +1,219 @@
+import datetime as dt
+import json
+
+import pytest
+
+from openpilot.sunnypilot.achievements.catalog import CATALOG_PATH, load_catalog, parse_catalog, statuses
+from openpilot.sunnypilot.achievements.engine import (AchievementEngine, LANE_CHANGE_FINISHING, LANE_CHANGE_STARTING, Sample, State,
+                                                      geohash, load_seen, load_state, save_seen, save_state, season)
+
+T0 = dt.datetime(2026, 6, 1, 16, 0, tzinfo=dt.UTC).timestamp()  # noon in New York
+
+
+def drive(engine, seconds, v=20.0, t0=T0, hz=10, **kw):
+  """Feed `seconds` of driving at `v` m/s; returns every unlock id along the way."""
+  new = []
+  n = int(seconds * hz)
+  for i in range(n + 1):
+    extra = {k: val for k, val in kw.items() if k != "op_active"}
+    new += engine.update(Sample(t=t0 + i / hz, wall=t0 + i / hz, v_ego=v, op_active=kw.get("op_active", True), **extra))
+  return new
+
+
+def test_catalog_loads_and_ids_are_unique():
+  cat = load_catalog()
+  ids = cat.all_unlock_ids()
+  assert len(ids) == len(set(ids))
+  assert "odometer.silver" in ids and "first_light" in ids
+  engine_metrics = set(State().metrics())
+  assert cat.metrics <= engine_metrics, f"catalog uses metrics the engine doesn't compute: {cat.metrics - engine_metrics}"
+
+
+def test_catalog_rejects_decreasing_tiers():
+  data = json.loads(CATALOG_PATH.read_text())
+  data["achievements"][1]["tiers"][1]["threshold"] = 1
+  with pytest.raises(ValueError):
+    parse_catalog(data)
+
+
+def test_engaged_distance_and_first_light():
+  e = AchievementEngine()
+  new = drive(e, 10, v=20.0)
+  assert e.state.engaged_m == pytest.approx(200.0, rel=1e-6)
+  assert e.state.engaged_s == pytest.approx(10.0, rel=1e-6)
+  assert new == ["first_light"]
+
+
+def test_not_engaged_counts_nothing():
+  e = AchievementEngine()
+  drive(e, 10, op_active=False)
+  assert e.state.engaged_m == 0.0
+  assert e.state.unlocked == {}
+
+
+def test_gaps_and_bad_speeds_are_ignored():
+  e = AchievementEngine()
+  e.update(Sample(t=T0, v_ego=20.0, op_active=True))
+  e.update(Sample(t=T0 + 30.0, v_ego=20.0, op_active=True))  # 30 s gap: process stalled
+  e.update(Sample(t=T0 + 30.1, v_ego=500.0, op_active=True))  # implausible speed
+  assert e.state.engaged_m == 0.0
+
+
+def test_odometer_tiers_unlock_in_order():
+  e = AchievementEngine()
+  e.state.engaged_m = 99_999.0
+  new = drive(e, 1, v=10.0)
+  assert "odometer.bronze" in new and "odometer.silver" not in new
+  assert set(e.state.unlocked) >= {"first_light", "odometer.bronze"}
+
+
+def test_unlocks_only_once():
+  e = AchievementEngine()
+  assert drive(e, 5) == ["first_light"]
+  assert drive(e, 5, t0=T0 + 100) == []
+
+
+def test_mads_only_distance():
+  e = AchievementEngine()
+  drive(e, 10, v=10.0, op_active=False, mads_active=True)
+  drive(e, 10, v=10.0, t0=T0 + 60, op_active=True, mads_active=True)
+  assert e.state.mads_m == pytest.approx(100.0, rel=1e-6)
+  assert e.state.engaged_m == pytest.approx(200.0, rel=1e-6)
+
+
+def test_lane_change_counted_once_when_engaged():
+  e = AchievementEngine()
+  seq = [0, 1, 2, 2, 3, 3, 0, 0, 2, 3, 0]
+  for i, lc in enumerate(seq):
+    e.update(Sample(t=T0 + i * 0.1, v_ego=25.0, op_active=True, lane_change_state=lc))
+  assert e.state.lane_changes == 2
+  e.update(Sample(t=T0 + 5, v_ego=25.0, op_active=False, lane_change_state=LANE_CHANGE_STARTING))
+  e.update(Sample(t=T0 + 5.1, v_ego=25.0, op_active=False, lane_change_state=LANE_CHANGE_FINISHING))
+  assert e.state.lane_changes == 2
+
+
+def test_days_streak_and_best_day():
+  e = AchievementEngine()
+  for d in range(8):
+    drive(e, 2, t0=T0 + d * 86400)
+  assert len(e.state.days) == 8
+  assert e.state.best_streak_days == 8
+  assert "streak.bronze" in e.state.unlocked and "regular.bronze" in e.state.unlocked
+  # a gap day resets the run but keeps the best
+  drive(e, 2, t0=T0 + 10 * 86400)
+  assert e.state.best_streak_days == 8
+
+
+def test_local_date_follows_gps_longitude():
+  e = AchievementEngine()
+  late = dt.datetime(2026, 6, 2, 2, 0, tzinfo=dt.UTC).timestamp()  # 10 pm on June 1 in New York
+  e.update(Sample(t=late, wall=late, v_ego=10.0, op_active=True, lat=40.7, lon=-74.0))
+  e.update(Sample(t=late + 0.1, wall=late + 0.1, v_ego=10.0, op_active=True, lat=40.7, lon=-74.0))
+  assert e.state.days == ["2026-06-01"]
+
+
+def test_invalid_clock_counts_distance_but_not_days():
+  e = AchievementEngine()
+  for i in range(11):
+    e.update(Sample(t=i / 10, v_ego=10.0, op_active=True))  # wall=None: clock not set yet
+  assert e.state.engaged_m == pytest.approx(10.0, rel=1e-6)
+  assert e.state.days == [] and e.state.today_m == 0.0
+  assert e.state.unlocked == {}
+
+
+def test_attentive_run_resets_on_red_alert():
+  e = AchievementEngine()
+  drive(e, 100)
+  drive(e, 1, t0=T0 + 200, dm_alert_level=3)
+  assert e.state.attentive_run_s < 1.0
+  assert e.state.attentive_best_s == pytest.approx(100.0, rel=1e-3)
+
+
+def test_areas_are_hashed_and_counted():
+  e = AchievementEngine()
+  drive(e, 1, lat=40.70, lon=-74.00)
+  drive(e, 1, t0=T0 + 10, lat=40.701, lon=-74.001)   # same ~5 km cell
+  drive(e, 1, t0=T0 + 20, lat=41.50, lon=-73.00)     # different cell
+  assert len(e.state.areas) == 2
+  raw = json.dumps(e.state.__dict__)
+  assert "40.7" not in raw and geohash(40.70, -74.00) not in raw
+
+
+def test_model_hopper_counts_models_over_50km():
+  e = AchievementEngine()
+  e.state.model_m = {"a": 50_000.0, "b": 50_000.0, "c": 10.0}
+  drive(e, 1, model_ref="c", model_name="Model C")
+  assert e.state.metrics()["models_50km"] == 2
+  e.state.model_m["c"] = 50_000.0
+  assert "model_hopper.bronze" in drive(e, 1, t0=T0 + 10, model_ref="c")
+  assert e.state.model_names["c"] == "Model C"
+
+
+def test_dev_build_and_seasons():
+  e = AchievementEngine()
+  drive(e, 1, dev_build=True)
+  assert e.state.dev_m > 0
+  assert season(dt.date(2026, 1, 5)) == "winter"
+  assert season(dt.date(2026, 1, 5), southern=True) == "summer"
+
+
+def test_sunnylink_flag_sticks():
+  e = AchievementEngine()
+  assert e.set_sunnylink_linked(True, T0) == ["linked_up"]
+  e.set_sunnylink_linked(False, T0 + 1)
+  assert e.state.sunnylink_linked
+
+
+def test_state_round_trip(tmp_path):
+  e = AchievementEngine()
+  drive(e, 10, lat=40.7, lon=-74.0, model_ref="x")
+  path = tmp_path / "state.json"
+  save_state(e.state, path)
+  loaded = load_state(path)
+  assert loaded == e.state
+
+
+def test_corrupt_state_is_moved_aside(tmp_path):
+  path = tmp_path / "state.json"
+  path.write_text("{not json")
+  assert load_state(path).engaged_m == 0.0
+  assert (tmp_path / "state.corrupt.json").exists()
+  assert not path.exists()
+
+
+def test_unknown_fields_are_ignored(tmp_path):
+  path = tmp_path / "state.json"
+  path.write_text(json.dumps({"engaged_m": 5.0, "from_the_future": 1}))
+  assert load_state(path).engaged_m == 5.0
+
+
+def test_seen_round_trip(tmp_path):
+  path = tmp_path / "seen.json"
+  assert load_seen(path) == set()
+  save_seen({"a", "b"}, path)
+  assert load_seen(path) == {"a", "b"}
+
+
+def test_status_summaries():
+  cat = load_catalog()
+  st = State(engaged_m=1_240_000.0)
+  e = AchievementEngine(st)
+  e.check_unlocks(T0)
+  by_id = {s.achievement.id: s for s in statuses(cat, st.metrics(), st.unlocked)}
+  assert by_id["odometer"].summary() == "silver · 1,240 / 10,000 km"
+  assert by_id["odometer"].summary(is_metric=False) == "silver · 770 / 6,213 mi"
+  assert by_id["first_light"].summary() == "done"
+  assert by_id["regular"].summary() == "locked · 0 / 7 days"
+  assert by_id["linked_up"].summary() == "locked"
+
+
+def test_last_step_for_the_community_log():
+  e = AchievementEngine(State())
+  wall = 1_762_000_000.0
+  e.update(Sample(t=0.0, v_ego=20.0, op_active=True, wall=wall))
+  e.update(Sample(t=0.5, v_ego=20.0, op_active=True, wall=wall + 0.5))
+  assert e.last_step.dt_s == 0.5 and e.last_step.engaged_m == 10.0 and e.last_step.day is not None
+  e.update(Sample(t=1.0, v_ego=20.0, op_active=False, wall=wall + 1.0))
+  assert e.last_step.dt_s == 0.5 and e.last_step.engaged_m == 0.0
+  e.update(Sample(t=9.0, v_ego=20.0, op_active=True, wall=None))  # a gap and no clock
+  assert e.last_step.dt_s == 0.0 and e.last_step.day is None
@@ -0,0 +1,125 @@
+"""
+h3lite must give the same cell ids as Uber's H3, because the site and the
+server use h3-js and h3-pg on the ids the car sends. The vectors below come
+from h3-py 4.1.2; with h3 installed, the last test also compares random points.
+"""
+import random
+
+import pytest
+
+from openpilot.sunnypilot.achievements import h3lite
+
+RESOLUTIONS = (0, 4, 6, 9, 15)
+CELLS = [
+  (40.0334, -83.1588, ("802bfffffffffff", "842a90bffffffff", "862a90b4fffffff", "892a90b4917ffff", "8f2a90b4915d388")),
+  (37.7749, -122.4194, ("8029fffffffffff", "8428309ffffffff", "86283082fffffff", "89283082803ffff", "8f283082800b390")),
+  (51.5007, -0.1246, ("8019fffffffffff", "84194adffffffff", "86194ad17ffffff", "89194ad14c3ffff", "8f194ad14c254d3")),
+  (-33.8688, 151.2093, ("80bffffffffffff", "84be0e3ffffffff", "86be0e35fffffff", "89be0e35cbbffff", "8fbe0e35cbad0a8")),
+  (35.6762, 139.6503, ("802ffffffffffff", "842f5a3ffffffff", "862f5a367ffffff", "892f5a363bbffff", "8f2f5a363ba005a")),
+  (64.8378, -147.7164, ("800dfffffffffff", "840d5a9ffffffff", "860d5a9afffffff", "890d5a9add3ffff", "8f0d5a9add25781")),
+  (-54.8019, -68.303, ("80dffffffffffff", "84df451ffffffff", "86df45177ffffff", "89df4517557ffff", "8fdf45175542186")),
+  (0.0, 0.0, ("8075fffffffffff", "84754a9ffffffff", "86754e64fffffff", "89754e64993ffff", "8f754e64992d6d8")),
+  (89.9, 45.0, ("8001fffffffffff", "8403263ffffffff", "86032630fffffff", "89032630ac3ffff", "8f032630ac0311a")),
+  (-89.9, -45.0, ("80f3fffffffffff", "84f2939ffffffff", "86f2938efffffff", "89f2938ec7bffff", "8ff2938ec78c80e")),
+  (52.0, 179.999, ("8017fffffffffff", "841659bffffffff", "861659aefffffff", "891659acc9bffff", "8f1659acc9ad193")),
+  (-16.5, -179.999, ("809bfffffffffff", "849b5ddffffffff", "869b5dc57ffffff", "899b5dc5597ffff", "8f9b5dc5596a500")),
+  (-28.18676, -125.6943, ("80b1fffffffffff", "84b1713ffffffff", "86b171247ffffff", "89b1712423bffff", "8fb171242396a80")),
+  (24.14952, -153.92294, ("805dfffffffffff", "845d35bffffffff", "865d35b0fffffff", "895d35b0b23ffff", "8f5d35b0b200c24")),
+  (5.74112, -48.35199, ("805ffffffffffff", "845fa25ffffffff", "865fa248fffffff", "895fa248b23ffff", "8f5fa248b218936")),
+  (-70.72017, 2.67686, ("80e7fffffffffff", "84e749dffffffff", "86e749c47ffffff", "89e749c44c3ffff", "8fe749c44c216ec")),
+  (-74.00069, -23.88755, ("80effffffffffff", "84ee139ffffffff", "86ee13967ffffff", "89ee1396293ffff", "8fee139629208d2")),
+  (-68.82313, -147.34332, ("80e3fffffffffff", "84e334dffffffff", "86e334c9fffffff", "89e334c983bffff", "8fe334c9839a3b1")),
+  (-12.07693, 117.66676, ("8095fffffffffff", "849581dffffffff", "869581d97ffffff", "899581d9487ffff", "8f9581d9486a96e")),
+  (-60.19169, -99.63397, ("80e9fffffffffff", "84e95c1ffffffff", "86e95c107ffffff", "89e95c1000bffff", "8fe95c1000b639c")),
+]
+PARENTS = [
+  ("8f2a90b4915d388", ("802bfffffffffff", "842a90bffffffff", "862a90b4fffffff", "892a90b4917ffff", "8f2a90b4915d388")),
+  ("8f283082800b390", ("8029fffffffffff", "8428309ffffffff", "86283082fffffff", "89283082803ffff", "8f283082800b390")),
+  ("8f194ad14c254d3", ("8019fffffffffff", "84194adffffffff", "86194ad17ffffff", "89194ad14c3ffff", "8f194ad14c254d3")),
+  ("8fbe0e35cbad0a8", ("80bffffffffffff", "84be0e3ffffffff", "86be0e35fffffff", "89be0e35cbbffff", "8fbe0e35cbad0a8")),
+  ("8f2f5a363ba005a", ("802ffffffffffff", "842f5a3ffffffff", "862f5a367ffffff", "892f5a363bbffff", "8f2f5a363ba005a")),
+  ("8f0d5a9add25781", ("800dfffffffffff", "840d5a9ffffffff", "860d5a9afffffff", "890d5a9add3ffff", "8f0d5a9add25781")),
+  ("8fdf45175542186", ("80dffffffffffff", "84df451ffffffff", "86df45177ffffff", "89df4517557ffff", "8fdf45175542186")),
+  ("8f754e64992d6d8", ("8075fffffffffff", "84754e7ffffffff", "86754e64fffffff", "89754e64993ffff", "8f754e64992d6d8")),
+  ("8f032630ac0311a", ("8003fffffffffff", "8403263ffffffff", "86032630fffffff", "89032630ac3ffff", "8f032630ac0311a")),
+  ("8ff2938ec78c80e", ("80f3fffffffffff", "84f2939ffffffff", "86f2938efffffff", "89f2938ec7bffff", "8ff2938ec78c80e")),
+  ("8f1659acc9ad193", ("8017fffffffffff", "841659bffffffff", "861659acfffffff", "891659acc9bffff", "8f1659acc9ad193")),
+  ("8f9b5dc5596a500", ("809bfffffffffff", "849b5ddffffffff", "869b5dc57ffffff", "899b5dc5597ffff", "8f9b5dc5596a500")),
+  ("8fb171242396a80", ("80b1fffffffffff", "84b1713ffffffff", "86b171247ffffff", "89b1712423bffff", "8fb171242396a80")),
+  ("8f5d35b0b200c24", ("805dfffffffffff", "845d35bffffffff", "865d35b0fffffff", "895d35b0b23ffff", "8f5d35b0b200c24")),
+  ("8f5fa248b218936", ("805ffffffffffff", "845fa25ffffffff", "865fa248fffffff", "895fa248b23ffff", "8f5fa248b218936")),
+  ("8fe749c44c216ec", ("80e7fffffffffff", "84e749dffffffff", "86e749c47ffffff", "89e749c44c3ffff", "8fe749c44c216ec")),
+  ("8fee139629208d2", ("80effffffffffff", "84ee139ffffffff", "86ee13967ffffff", "89ee1396293ffff", "8fee139629208d2")),
+  ("8fe334c9839a3b1", ("80e3fffffffffff", "84e334dffffffff", "86e334c9fffffff", "89e334c983bffff", "8fe334c9839a3b1")),
+  ("8f9581d9486a96e", ("8095fffffffffff", "849581dffffffff", "869581d97ffffff", "899581d9487ffff", "8f9581d9486a96e")),
+  ("8fe95c1000b639c", ("80e9fffffffffff", "84e95c1ffffffff", "86e95c107ffffff", "89e95c1000bffff", "8fe95c1000b639c")),
+  ("8f2a90b4915d388", ("802bfffffffffff", "842a90bffffffff", "862a90b4fffffff", "892a90b4917ffff", "8f2a90b4915d388")),
+]
+CENTERS = [
+  ("862a90b4fffffff", 40.05041861648768, -83.181315523919),
+  ("842a90bffffffff", 40.08105106496825, -83.41180170584032),
+  ("864c00007ffffff", 23.71792527122296, -67.13232636643565),
+  ("8f2a90b4915d388", 40.03339893697043, -83.1588004749386),
+  ("8001fffffffffff", 79.24239850975903, 38.02340700796989),
+]
+DISKS = {
+  "862a90b4fffffff": ['862a90a67ffffff', '862a90b47ffffff', '862a90b4fffffff', '862a90b5fffffff', '862a90b6fffffff', '862a95697ffffff', '862a956b7ffffff'],
+  # a pentagon: 5 neighbors
+  "864c00007ffffff": [
+    '864c00007ffffff',
+    '864c00017ffffff',
+    '864c0001fffffff',
+    '864c00027ffffff',
+    '864c0002fffffff',
+    '864c00037ffffff',
+  ],
+}
+
+
+@pytest.mark.parametrize("lat,lng,expected", CELLS)
+def test_latlng_to_cell(lat, lng, expected):
+  for res, want in zip(RESOLUTIONS, expected, strict=True):
+    assert h3lite.cell_to_str(h3lite.latlng_to_cell(lat, lng, res)) == want
+
+
+@pytest.mark.parametrize("cell,expected", PARENTS)
+def test_cell_to_parent(cell, expected):
+  # a parent isn't always the cell that holds the same point: H3 children stick out a little
+  for res, want in zip(RESOLUTIONS, expected, strict=True):
+    assert h3lite.cell_to_str(h3lite.cell_to_parent(h3lite.str_to_cell(cell), res)) == want
+    assert h3lite.get_resolution(h3lite.str_to_cell(want)) == res
+
+
+@pytest.mark.parametrize("cell,lat,lng", CENTERS)
+def test_cell_to_latlng(cell, lat, lng):
+  got = h3lite.cell_to_latlng(h3lite.str_to_cell(cell))
+  assert got == pytest.approx((lat, lng), abs=1e-9)
+
+
+@pytest.mark.parametrize("cell,expected", DISKS.items())
+def test_grid_disk1(cell, expected):
+  got = {h3lite.cell_to_str(c) for c in h3lite.grid_disk1(h3lite.str_to_cell(cell))}
+  assert got == set(expected)
+
+
+def test_bad_input():
+  with pytest.raises(ValueError):
+    h3lite.latlng_to_cell(40.0, -83.0, 16)
+  with pytest.raises(ValueError):
+    h3lite.latlng_to_cell(float("nan"), -83.0, 6)
+  with pytest.raises(ValueError):
+    h3lite.cell_to_parent(h3lite.str_to_cell("862a90b4fffffff"), 7)
+
+
+def test_matches_h3_on_random_points():
+  h3 = pytest.importorskip("h3")
+  rnd = random.Random(0)
+  for _ in range(1500):
+    lat, lng = rnd.uniform(-90, 90), rnd.uniform(-180, 180)
+    for res in range(16):
+      want = h3.latlng_to_cell(lat, lng, res)
+      cell = h3lite.latlng_to_cell(lat, lng, res)
+      assert h3lite.cell_to_str(cell) == want
+      assert h3lite.cell_to_latlng(cell) == pytest.approx(h3.cell_to_latlng(want), abs=1e-9)
+    want_disk = {h3.str_to_int(c) for c in h3.grid_disk(h3.latlng_to_cell(lat, lng, 6), 1)}
+    assert h3lite.grid_disk1(h3lite.latlng_to_cell(lat, lng, 6)) == want_disk
@@ -0,0 +1,68 @@
+"""The replay tool, fed fake log messages (the real LogReader needs the full openpilot build)."""
+import datetime as dt
+import os
+from types import SimpleNamespace as NS
+
+from openpilot.sunnypilot.achievements.community import CommunityLog, tile_of
+from openpilot.sunnypilot.achievements.engine import AchievementEngine, State
+from openpilot.sunnypilot.achievements.replay import RouteFeeder, find_routes
+from openpilot.sunnypilot.achievements.simulate import HOME, WORK
+
+T0 = int(dt.datetime(2026, 11, 2, 13, 0, tzinfo=dt.UTC).timestamp() * 1e3)
+
+
+class Ev:
+  def __init__(self, kind: str, mono_s: float, **fields):
+    self._kind, self.logMonoTime = kind, int(mono_s * 1e9)
+    setattr(self, kind, NS(**fields))
+
+  def which(self):
+    return self._kind
+
+
+def fake_route(start, end, seconds=600):
+  yield Ev("carParams", 0.0, carFingerprint="HYUNDAI_IONIQ_5")
+  yield Ev("selfdriveState", 0.0, active=True)
+  yield Ev("selfdriveStateSP", 0.0, mads=NS(active=False))
+  yield Ev("driverMonitoringState", 0.0, alertLevel=NS(raw=0))
+  for i in range(seconds * 10):  # carState at 10 Hz, GPS at 1 Hz, like a qlog
+    t = i / 10
+    if i % 10 == 0:
+      f = t / seconds
+      yield Ev("gpsLocationExternal", t, hasFix=True, horizontalAccuracy=5.0,
+               latitude=start[0] + (end[0] - start[0]) * f, longitude=start[1] + (end[1] - start[1]) * f,
+               unixTimestampMillis=T0 + int(t * 1000))
+    yield Ev("carState", t, vEgo=30.0)
+
+
+def test_route_feeder_counts_a_drive(tmp_path):
+  engine, log = AchievementEngine(State()), CommunityLog(tmp_path / "c.db")
+  feeder = RouteFeeder(engine, log)
+  for ev in fake_route(HOME, WORK):
+    feeder.feed(ev)
+  engine.end_drive()
+  log.end_drive(engine.utc_offset_h)
+
+  assert feeder.car == "HYUNDAI_IONIQ_5"
+  assert 17_000 < engine.state.engaged_m < 18_500  # 600 s at 30 m/s
+  assert engine.state.days == ["2026-11-02"]
+  view = log.tile_view(dt.date(2026, 11, 3))
+  assert tile_of(*HOME) in view.held  # where the drive started
+  assert len(view.driven) > 3
+  assert log.home_region(dt.date(2026, 11, 3)) == "US-OH"
+
+
+def test_find_routes(tmp_path):
+  for name, mtime in (("00000002--bbbb--0", 200), ("00000002--bbbb--1", 260), ("00000001--aaaa--0", 100),
+                      ("00000001--aaaa--10", 160), ("00000001--aaaa--2", 130), ("boot", 0)):
+    d = tmp_path / name
+    d.mkdir()
+    if name != "boot":
+      (d / "qlog.zst").write_bytes(b"")
+      os.utime(d / "qlog.zst", (mtime, mtime))
+  routes = find_routes(tmp_path, rlog=False, days=None)
+  assert [[p.parent.name for p in r] for r in routes] == [
+    ["00000001--aaaa--0", "00000001--aaaa--2", "00000001--aaaa--10"],
+    ["00000002--bbbb--0", "00000002--bbbb--1"],
+  ]
+  assert find_routes(tmp_path, rlog=True, days=None) == []
@@ -0,0 +1,136 @@
+#!/usr/bin/env python3
+"""
+Copyright (c) 2021-, Haibin Wen, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+
+Builds regions.json.gz, the table the car uses to turn a map tile into a home
+region code. Dev tool only: it is not run on the car.
+
+  pip install h3 shapely
+  curl -LO https://raw.githubusercontent.com/nvkelso/natural-earth-vector/master/geojson/ne_10m_admin_1_states_provinces.geojson
+  python openpilot/sunnypilot/achievements/tools/build_regions.py \
+    ne_10m_admin_1_states_provinces.geojson openpilot/sunnypilot/achievements/regions.json.gz
+
+Rules:
+- Region codes are ISO 3166-2 (US-OH) in the countries in SUBDIVISIONS, and the
+  ISO 3166-1 country code (GB) everywhere else.
+- A resolution-6 tile (about 6 km) belongs to the region its center is in.
+  Coastal tiles whose center is in the water go to the region whose coastline
+  passes through them most.
+- Stored as the majority region of each resolution-4 cell ("regions") plus the
+  resolution-6 tiles that differ from it ("exceptions"), each compacted. That
+  keeps 6 km accuracy at borders in about 200 KB.
+
+Boundaries: Natural Earth 1:10m admin 1, public domain.
+"""
+import collections
+import gzip
+import json
+import re
+import sys
+
+import h3
+from shapely.geometry import box, shape
+
+SUBDIVISIONS = {"US", "CA", "MX", "BR", "AU", "DE"}
+TILE_RES = 6
+BASE_RES = 4
+CUT_DEG = 10.0  # H3 can't fill huge polygons at resolution 6, so cut them into squares first
+ISO_3166_2 = re.compile(r"^[A-Z]{2}-[A-Z0-9]{1,3}$")
+
+
+def region_code(props: dict) -> str | None:
+  country = props.get("iso_a2") or ""
+  if not re.fullmatch(r"[A-Z]{2}", country):
+    return None
+  code = props.get("iso_3166_2") or ""
+  if country in SUBDIVISIONS and ISO_3166_2.match(code):
+    return code
+  return country
+
+
+def pieces(geom: dict):
+  """Polygon rings, as (lng, lat) lists, of the geometry cut into CUT_DEG squares."""
+  g = shape(geom)
+  if not g.is_valid:
+    g = g.buffer(0)
+  minx, miny, maxx, maxy = g.bounds
+  x = minx
+  while x < maxx:
+    y = miny
+    while y < maxy:
+      part = g.intersection(box(x, y, x + CUT_DEG, y + CUT_DEG))
+      for poly in getattr(part, "geoms", [part]):
+        if poly.geom_type == "Polygon" and not poly.is_empty:
+          yield [list(poly.exterior.coords)] + [list(r.coords) for r in poly.interiors]
+      y += CUT_DEG
+    x += CUT_DEG
+
+
+def outer_rings(geom: dict):
+  polys = [geom["coordinates"]] if geom["type"] == "Polygon" else geom["coordinates"]
+  for rings in polys:
+    yield rings[0]
+
+
+def winner(counter: collections.Counter) -> str:
+  return max(sorted(counter), key=lambda k: counter[k])
+
+
+def main(src: str, dst: str) -> None:
+  features = json.load(open(src))["features"]
+  by_center: dict[str, str] = {}
+  by_coast: dict[str, collections.Counter] = collections.defaultdict(collections.Counter)
+  for n, f in enumerate(features):
+    code = region_code(f["properties"])
+    if code is None or f["geometry"] is None:
+      continue
+    for rings in pieces(f["geometry"]):
+      outer, holes = rings[0], rings[1:]
+      poly = h3.LatLngPoly([(lat, lng) for lng, lat in outer], *[[(lat, lng) for lng, lat in r] for r in holes])
+      for cell in h3.polygon_to_cells(poly, TILE_RES):
+        by_center[cell] = code
+    for ring in outer_rings(f["geometry"]):
+      for lng, lat in ring:
+        by_coast[h3.latlng_to_cell(lat, lng, TILE_RES)][code] += 1
+    if n % 500 == 0:
+      print(f"{n}/{len(features)} boundaries", file=sys.stderr)
+
+  tiles = dict(by_center)
+  for cell, votes in by_coast.items():
+    tiles.setdefault(cell, winner(votes))
+
+  base_votes: dict[str, collections.Counter] = collections.defaultdict(collections.Counter)
+  for cell, code in tiles.items():
+    base_votes[h3.cell_to_parent(cell, BASE_RES)][code] += 1
+  base = {cell: winner(votes) for cell, votes in base_votes.items()}
+
+  regions: dict[str, list[str]] = collections.defaultdict(list)
+  for cell, code in base.items():
+    regions[code].append(cell)
+  exceptions: dict[str, list[str]] = collections.defaultdict(list)
+  for cell, code in tiles.items():
+    if base[h3.cell_to_parent(cell, BASE_RES)] != code:
+      exceptions[code].append(cell)
+
+  data = {
+    "v": 1,
+    "source": "Natural Earth 1:10m admin 1 (public domain); built by tools/build_regions.py",
+    "subdivisions": sorted(SUBDIVISIONS),
+    "tile_res": TILE_RES,
+    "base_res": BASE_RES,
+    "regions": {k: sorted(h3.compact_cells(v)) for k, v in sorted(regions.items())},
+    "exceptions": {k: sorted(h3.compact_cells(v)) for k, v in sorted(exceptions.items())},
+  }
+  raw = json.dumps(data, separators=(",", ":"), sort_keys=True).encode()
+  with gzip.GzipFile(dst, "wb", mtime=0) as out:
+    out.write(raw)
+  n_regions = sum(map(len, data["regions"].values()))
+  n_exceptions = sum(map(len, data["exceptions"].values()))
+  print(f"{len(tiles)} tiles, {len(data['regions'])} regions, {n_regions} base cells, {n_exceptions} exceptions", file=sys.stderr)
+
+
+if __name__ == "__main__":
+  main(sys.argv[1], sys.argv[2])
@@ -371,7 +371,7 @@ def _compile_warp_resolution_worker(cam_w, cam_h, model_w, model_h, benchmark_ru


 if __name__ == "__main__":
-  if 'USB' in os.getenv('DEV', '') or os.getenv('CHESTNUT'):
+  if ('USB' in os.getenv('DEV', '') or os.getenv('CHESTNUT')) and not os.getenv('DEV', '').startswith('MOCK'):
    from openpilot.system.hardware.chestnut.flash import link_up
    for _ in range(10):
      if link_up():
@@ -17,16 +17,18 @@ def main():
  parser.add_argument("--frame", required=True, help="width,height,stride,y_height,uv_height,buffer_size")
  parser.add_argument("--warp-to", required=True, help="model input WxH")
  parser.add_argument("--output", required=True)
+  parser.add_argument("--light", action="store_true")
  args = parser.parse_args()

  from examples.openpilot.compile_warp import make_frame_prepare, parse_frame, parse_size
  from tinygrad import Device, Tensor, TinyJit
  frame = parse_frame(args.frame)
  warp = make_frame_prepare(frame, *parse_size(args.warp_to))
-  run = TinyJit(lambda input_frame, M_inv: Tensor.stack(*(warp(input_frame[i], M_inv[i]) for i in range(2))), prune=True)
+  n = 1 if args.light else 2
+  run = TinyJit(lambda input_frame, M_inv: Tensor.stack(*(warp(input_frame[i], M_inv[i]) for i in range(n))), prune=True)
  rng = np.random.default_rng(42)
  inputs = {"input_frame": Tensor(rng.integers(0, 256, (2, frame.size), dtype=np.uint8), device=Device.DEFAULT).realize(),
-            "M_inv": Tensor((rng.standard_normal((2, 3, 3)) * 8).astype(np.float32), device=Device.DEFAULT).realize()}
+            "M_inv": Tensor((rng.standard_normal((n, 3, 3)) * 8).astype(np.float32), device=Device.DEFAULT).realize()}
  with record_kernel_linear() as recorded:
    run(**inputs)
    run(**inputs)
@@ -34,7 +36,7 @@ def main():
  assert len(recorded) == 1, f"expected one captured kernel list, got {len(recorded)}"
  part = serialize_jit(run, recorded[0], check)
  PlainJit(*part)
-  specs = {"input_frame": ((2, frame.size), "|u1", Device.DEFAULT), "M_inv": ((2, 3, 3), "<f4", Device.DEFAULT)}
+  specs = {"input_frame": ((2, frame.size), "|u1", Device.DEFAULT), "M_inv": ((n, 3, 3), "<f4", Device.DEFAULT)}
  write_bundle(args.output, {"run": part}, {"input_specs": specs, "metadata": {}})
  print(f"Saved warp to {args.output} ({len(part[0]['calls'])} kernels)")

@@ -10,6 +10,7 @@ import dataclasses
 import inspect
 import itertools
 import mmap
+import os
 import re
 import struct
 import tempfile
@@ -194,7 +195,8 @@ def convert(jit, kwargs: dict, frame: dict, target) -> tuple[dict, memoryview]:
        base, at, nbytes = _address(argument_words[slot])
        args.append({"buffer": buffers.setdefault(base, len(buffers)), "offset": at, "nbytes": nbytes or base.max_numel() * base.dtype.itemsize, "dtype": u8})
    outs, ins = kernel[6]
-    calls.append({"kind": "program", "name": re.sub(r"\x1b\[[0-9;]*m", "", kernel[1]), "lib": programs[image_alloc], "target": dataclasses.asdict(target),
+    device_target = dataclasses.asdict(dataclasses.replace(target, interface=target.interface.removeprefix("MOCK")))
+    calls.append({"kind": "program", "name": re.sub(r"\x1b\[[0-9;]*m", "", kernel[1]), "lib": programs[image_alloc], "target": device_target,
                  "signature": [[f"data{i}", i, u8, [a.get("nbytes", 1)]] for i, a in enumerate(args)], "global_size": grid, "local_size": local,
                  "globals": list(range(len(args))), "outs": list(outs), "ins": list(ins), "args": args})
    needs_init |= {args[i]["buffer"] for i in ins if "buffer" in args[i] and args[i]["buffer"] not in written}
@@ -235,14 +237,16 @@ def main():
  frame, _ = capture_frame(artifact["run"], kwargs, "output_buffers")
  part = convert(artifact["run"], kwargs, frame, Device[Device.DEFAULT].renderer.target)

-  plain = PlainJit(*part)  # replays the frame, within tolerance
-  replay = {path[0]: Tensor(array, device=device).realize() for path, array, device in frame["inputs"] if len(path) == 1}
-  replay["output_buffers"] = {path[1]: Tensor(array, device=device).realize() for path, array, device in frame["inputs"] if len(path) == 2}
-  flatten(plain(**replay))
-  for index, (got, expected) in enumerate(zip([t.numpy() for t in replay["output_buffers"].values()], frame["outputs"], strict=True)):
-    assert np.array_equal(got, expected), f"output {index} of the plain artifact differs from the pickle"
+  if not os.getenv("MOCKUSB_SOCKET"):
+    plain = PlainJit(*part)
+    replay = {path[0]: Tensor(array, device=device).realize() for path, array, device in frame["inputs"] if len(path) == 1}
+    replay["output_buffers"] = {path[1]: Tensor(array, device=device).realize() for path, array, device in frame["inputs"] if len(path) == 2}
+    flatten(plain(**replay))
+    for index, (got, expected) in enumerate(zip([t.numpy() for t in replay["output_buffers"].values()], frame["outputs"], strict=True)):
+      assert np.array_equal(got, expected), f"output {index} of the plain artifact differs from the pickle"
  write_bundle(args.output, {"run": part}, {key: artifact[key] for key in ("metadata", "input_specs", "output_specs")})
-  print(f"converted {len(part[0]['calls'])} kernels, bit-exact with the pickle: {args.output}")
+  checked = "left to mock_parity" if os.getenv("MOCKUSB_SOCKET") else "bit-exact with the pickle"
+  print(f"converted {len(part[0]['calls'])} kernels, {checked}: {args.output}")


 if __name__ == "__main__":
@@ -20,6 +20,11 @@ def pack_layout(shapes: dict[str, tuple], frame_size: int) -> tuple[dict[str, in
  return offsets | {"frames": offset}, offset + 2 * frame_size


+def host_shapes(input_shapes: dict[str, tuple], states) -> dict[str, tuple]:
+  shapes = {name: shape for name, shape in input_shapes.items() if name not in states and name not in ("new_img", "light_img")}
+  return shapes | ({"light_tfm": (1, 3, 3)} if "light_img" in input_shapes else {})
+
+
 class _Blobs:  # the model's and the warp's blobs as one address space, for a part whose ranges span both files
  def __init__(self, *blobs):
    self.blobs, self.starts = blobs, list(itertools.accumulate([0] + [len(blob) for blob in blobs[:-1]]))
@@ -30,18 +35,23 @@ class _Blobs:  # the model's and the warp's blobs as one address space, for a pa


 def fused_with_warp(cam_w: int, cam_h: int, chestnut: bool):  # load_model's extend: a native model runs upload, warp and model as one submit
-  return partial(frame_part, f"{BASEDIR}/openpilot/sunnypilot/modeld_v2/models/{'big_' if chestnut else ''}driving_warp_{cam_w}x{cam_h}_tinygrad.pkl")
+  models = f"{BASEDIR}/openpilot/sunnypilot/modeld_v2/models"
+  return partial(frame_part, f"{models}/{'big_' if chestnut else ''}driving_warp_{cam_w}x{cam_h}_tinygrad.pkl",
+                 light_warp_path=f"{models}/driving_lightwarp_{cam_w}x{cam_h}_tinygrad.pkl")


-def frame_part(warp_path: str, parts: dict, extras: dict) -> dict:
+def frame_part(warp_path: str, parts: dict, extras: dict, light_warp_path: str | None = None) -> dict:
  if "input_specs" not in extras:  # a legacy model warps in its own parts
    return {}
  (model, model_blob), (warp, warp_blob) = parts["run"], read_parts(warp_path)[0]["run"]
  states = sorted(name for name in extras["input_specs"] if f"next_{name}" in extras["metadata"]["output_shapes"])
-  host = {name: shape for name, (shape, _, _) in extras["input_specs"].items() if name not in states and name != "new_img"}
+  host = host_shapes({name: shape for name, (shape, _, _) in extras["input_specs"].items()}, states)
+  light = "light_img" in extras["input_specs"]
+  light_warp, light_blob = read_parts(light_warp_path)[0]["run"] if light else ({"buffers": [], "calls": [], "inputs": []}, memoryview(b""))
  frame_size = next(s["numel"] for s in warp["inputs"] if s["name"] == "input_frame") // 2
  offsets, packed = pack_layout(host, frame_size)
-  shift, upload, u8 = len(model_blob), len(model["buffers"]) + len(warp["buffers"]), repr(_dtype("dtypes.uint8"))
+  shift, light_shift = len(model_blob), len(model_blob) + len(warp_blob)
+  upload, u8 = len(model["buffers"]) + len(warp["buffers"]) + len(light_warp["buffers"]), repr(_dtype("dtypes.uint8"))

  def view(name: str, nbytes: int, dtype: str) -> dict:
    return {"buffer": upload, "offset": offsets[name], "nbytes": nbytes, "dtype": dtype}
@@ -49,12 +59,15 @@ def frame_part(warp_path: str, parts: dict, extras: dict) -> dict:
  names = sorted(["host", *states])
  slots = {s["slot"]: names.index(s["name"]) for s in model["inputs"] if s["name"] in states} | \
          {slot: len(names) + i for i, slot in enumerate(outputs)}
-  returned = warp["returns"]["tensors"][0]
+  def warped(part: dict, base: int) -> dict:
+    returned = part["returns"]["tensors"][0]
+    return {k: v for k, v in returned.items() if k != "shape"} | {"buffer": returned["buffer"] + base}
  model_inputs = {s["slot"]: {"input": slots[s["slot"]]} if s["slot"] in slots else
-                  {k: v for k, v in returned.items() if k != "shape"} | {"buffer": returned["buffer"] + len(model["buffers"])} if s["name"] == "new_img" else
+                  warped(warp, len(model["buffers"])) if s["name"] == "new_img" else
+                  warped(light_warp, len(model["buffers"]) + len(warp["buffers"])) if s["name"] == "light_img" else
                  view(s["name"], s["numel"] * _dtype(s["dtype"]).itemsize, s["dtype"]) for s in model["inputs"]}
-  warp_inputs = {s["slot"]: view("frames" if s["name"] == "input_frame" else "tfm", s["numel"] * _dtype(s["dtype"]).itemsize, s["dtype"])
-                 for s in warp["inputs"]}
+  def warp_inputs(part: dict, tfm: str) -> dict:
+    return {s["slot"]: view("frames" if s["name"] == "input_frame" else tfm, s["numel"] * _dtype(s["dtype"]).itemsize, s["dtype"]) for s in part["inputs"]}

  def remap(calls: list, inputs: dict, buffers: int, lib: int) -> list:
    return [call | ({"lib": [call["lib"][0] + lib, call["lib"][1] + lib]} if "lib" in call else {}) |
@@ -62,8 +75,11 @@ def frame_part(warp_path: str, parts: dict, extras: dict) -> dict:
  header = {"input_names": names, "returns": {"kind": "NoneType", "tensors": []}, "check": None,
            "inputs": [{"slot": names.index("host"), "name": "host", "dtype": u8, "numel": packed, "device": "NPY"}] +
                      [s | {"slot": slots[s["slot"]]} for s in model["inputs"] if s["slot"] in slots],
-            "buffers": model["buffers"] + [b | ({"init": [b["init"][0] + shift, b["init"][1] + shift]} if "init" in b else {}) for b in warp["buffers"]] +
+            "buffers": model["buffers"] + [b | ({"init": [b["init"][0] + d, b["init"][1] + d]} if "init" in b else {})
+                                           for part, d in ((warp, shift), (light_warp, light_shift)) for b in part["buffers"]] +
                       [{"nbytes": packed, "device": model["inputs"][0]["device"], "dtype": u8}],
            "calls": [{"kind": "copy", "args": [{"buffer": upload, "offset": 0, "nbytes": packed, "dtype": u8}, {"input": names.index("host")}]}] +
-                     remap(warp["calls"], warp_inputs, len(model["buffers"]), shift) + remap(model["calls"], model_inputs, 0, 0)}
-  return {"run": (header, _Blobs(model_blob, warp_blob))}  # replaces the model's run: one set of buffers, no separate warp
+                     remap(warp["calls"], warp_inputs(warp, "tfm"), len(model["buffers"]), shift) +
+                     remap(light_warp["calls"], warp_inputs(light_warp, "light_tfm"), len(model["buffers"]) + len(warp["buffers"]), light_shift) +
+                     remap(model["calls"], model_inputs, 0, 0)}
+  return {"run": (header, _Blobs(model_blob, warp_blob, light_blob) if light else _Blobs(model_blob, warp_blob))}
@@ -0,0 +1,22 @@
+"""
+Copyright (c) 2021-, sunnypilot, and a number of other contributors.
+
+This file is part of sunnypilot and is licensed under the MIT License.
+See the LICENSE.md file in the root directory for more details.
+"""
+import numpy as np
+
+from openpilot.common.transformations.camera import get_view_frame_from_calib_frame, view_frame_from_device_frame
+from openpilot.common.transformations.orientation import rot_from_euler
+
+LIGHT_INPUT_SIZE = (512, 256)
+light_intrinsics = np.array([
+  [1000.0, 0.0, 256.0],
+  [0.0, 1000.0, 200.0],
+  [0.0, 0.0, 1.0]])
+calib_from_light = np.linalg.inv((light_intrinsics @ get_view_frame_from_calib_frame(0, 0, 0, 0))[:, :3])
+
+
+def get_light_warp_matrix(device_from_calib_euler: np.ndarray, intrinsics: np.ndarray) -> np.ndarray:
+  camera_from_calib = intrinsics @ view_frame_from_device_frame @ rot_from_euler(device_from_calib_euler)
+  return (camera_from_calib @ calib_from_light).astype(np.float32)
@@ -0,0 +1,153 @@
+import json
+import os
+import shutil
+import struct
+import sys
+import tempfile
+import time
+
+os.environ.setdefault("DEV", "MOCKUSB+AMD")
+QCOM = "QCOM" in os.environ["DEV"]
+ON_MOCK = os.environ["DEV"].startswith("MOCK")
+ARCH, HARDWARE = ("a630", "comma 3x/4 (a630)") if QCOM else ("gfx1200", "chestnut")
+os.environ.setdefault("CHECK_OOB", "0")
+os.environ.setdefault("GMMU", "0")
+if sys.platform == "darwin" and os.environ["DEV"].startswith("MOCKUSB"):  # the mock's VRAM is a memfd, which macOS lacks: an unlinked file is the same
+  from tinygrad.runtime.autogen import libc
+
+  def _memfd_create(name, flags):
+    fd, path = tempfile.mkstemp()
+    os.unlink(path)
+    return fd
+  libc.memfd_create = _memfd_create
+
+import numpy as np
+
+from openpilot.sunnypilot.modeld_v2.helpers import ALIGNMENT, MAGIC, PlainJit, _device, read_parts
+
+
+class ParityError(RuntimeError):
+  pass
+
+
+def _emulator():
+  try:
+    if QCOM:
+      from test.mockgpu.qcom import emu
+    else:
+      from test.mockgpu.amd import emu
+  except ImportError as e:
+    raise ParityError(f"tinygrad's {'qcom' if QCOM else 'amd'} mock is not importable: put tinygrad_repo on PYTHONPATH") from e
+  if not hasattr(emu, "hw_unverified"):
+    raise ParityError(f"this tinygrad's emulator has no {ARCH} rounding model, its check frames would not match a {HARDWARE}")
+  return emu
+
+
+def unverified(header: dict, blob: memoryview) -> dict[str, int]:
+  emu, out = _emulator(), {}
+  for call in header["calls"]:
+    if call["kind"] != "program":
+      continue
+    if (arch := call["target"].get("arch", "")).split(",")[0] != ARCH:  # IMAGE builds append ",IMAGE_PITCH_ALIGNMENT=64"
+      out[f"arch={arch}"] = out.get(f"arch={arch}", 0) + 1
+      continue
+    for sig, n in emu.hw_unverified(bytes(blob[slice(*call["lib"])])).items():
+      out[sig] = out.get(sig, 0) + n
+  return out
+
+
+def _array(blob: memoryview, stored: dict) -> np.ndarray:
+  return np.frombuffer(blob[slice(*stored["data"])], dtype=np.dtype(stored["dtype"])).reshape(stored["shape"]).copy()
+
+
+def replay(header: dict, blob: memoryview, shared: dict) -> list[np.ndarray]:
+  from tinygrad import Tensor
+  check = header["check"]
+  jit = PlainJit(header, blob, None, verify=False, shared=shared)
+  kwargs: dict = {}
+  for path, stored, device in check["inputs"]:
+    tensor = Tensor(_array(blob, stored), device=_device(device, None)).realize()
+    kwargs[path[0]] = tensor if len(path) == 1 else {**kwargs.get(path[0], {}), path[1]: tensor}
+  result = jit(**kwargs)
+  outputs = kwargs[check["output_container"]].values() if check["output_container"] else result
+  return [t.numpy() for t in (outputs if isinstance(outputs, (list, tuple)) or check["output_container"] else [outputs])]
+
+
+def _parts_with_offsets(path: str) -> tuple[dict, dict]:
+  parts, _ = read_parts(path)
+  with open(path, "rb") as f:
+    head = f.read(len(MAGIC) + 8)
+    start = len(MAGIC) + 8 + struct.unpack_from("<Q", head, len(MAGIC))[0]
+    f.seek(len(MAGIC) + 8)
+    entries = json.loads(f.read(start - len(MAGIC) - 8))["parts"]
+  base = start + -start % ALIGNMENT
+  offsets = {tuple(e["key"]) if isinstance(e["key"], list) else e["key"]: base + e["blob"][0] for e in entries}
+  return parts, offsets
+
+
+def _gate(parts: dict) -> None:
+  missing: dict[str, int] = {}
+  for header, blob in parts.values():
+    if "check" not in header:
+      raise ParityError("artifact has no check frame")
+    for sig, n in unverified(header, blob).items():
+      missing[sig] = missing.get(sig, 0) + n
+  if missing:
+    raise ParityError(f"kernels use instructions the emulator is not verified against a {HARDWARE} for: " +
+                      ", ".join(f"{s} ({n}x)" for s, n in sorted(missing.items(), key=lambda kv: -kv[1])))
+
+
+def check(path: str) -> None:
+  parts, _ = read_parts(path)
+  if ON_MOCK:
+    _gate(parts)
+  shared: dict = {}
+  for key, (header, blob) in parts.items():
+    got = replay(header, blob, shared)
+    for i, (g, stored) in enumerate(zip(got, header["check"]["outputs"], strict=True)):
+      want = _array(blob, stored)
+      if g.shape != want.shape or not np.array_equal(g.view(np.uint8), want.view(np.uint8)):
+        n = int(np.count_nonzero(g != want)) if g.shape == want.shape else -1
+        where = "mock" if ON_MOCK else "device"
+        raise ParityError(f"part {key} output {i}: the {where}'s check frame differs from the stored one in {n} of {want.size} values")
+
+
+def record(path: str, out: str) -> None:
+  parts, offsets = _parts_with_offsets(path)
+  _gate(parts)
+  shared, writes = {}, []
+  for key, (header, blob) in parts.items():
+    for g, stored in zip(replay(header, blob, shared), header["check"]["outputs"], strict=True):
+      want = np.dtype(stored["dtype"])
+      if tuple(g.shape) != tuple(stored["shape"]) or g.dtype != want:
+        raise ParityError(f"part {key}: replayed output {g.dtype}{g.shape} does not fit the stored {want}{tuple(stored['shape'])}")
+      writes.append((offsets[key] + stored["data"][0], np.ascontiguousarray(g).tobytes()))
+  shutil.copyfile(path, out)
+  with open(out, "r+b") as f:
+    for offset, data in writes:
+      f.seek(offset)
+      f.write(data)
+
+
+def main(argv: list[str]) -> int:
+  if len(argv) < 2 or argv[0] not in ("check", "record") or (argv[0] == "record" and len(argv) != 3):
+    print("usage: mock_parity check ARTIFACT... | record ARTIFACT OUT", file=sys.stderr)
+    return 2
+  try:
+    if argv[0] == "record":
+      st = time.perf_counter()
+      record(argv[1], argv[2])
+      print(f"{argv[2]}: check frame recorded on the mock in {time.perf_counter() - st:.0f}s")
+      return 0
+    for path in argv[1:]:
+      st = time.perf_counter()
+      check(path)
+      print(f"{os.path.basename(path)}: bit-identical to its {HARDWARE} check frame on the {'mock' if ON_MOCK else 'device'} ({time.perf_counter() - st:.0f}s)")
+    return 0
+  except ParityError as e:
+    print(f"PARITY FAIL: {e}", file=sys.stderr)
+    return 1
+
+
+if __name__ == "__main__":
+  sys.exit(main(sys.argv[1:]))
@@ -11,7 +11,7 @@ import pickle
 import numpy as np

 from openpilot.common.basedir import BASEDIR
-from openpilot.sunnypilot.modeld_v2.frame_fusion import pack_layout
+from openpilot.sunnypilot.modeld_v2.frame_fusion import host_shapes, pack_layout
 from openpilot.sunnypilot.modeld_v2.helpers import load_model
 from openpilot.sunnypilot.modeld_v2.compile_modeld import (POLICY_INPUTS, derive_frame_skip,
                                                           make_split_input_queues, make_supercombo_input_queues)
@@ -172,7 +172,7 @@ class NativeTinygradAdapter(BaseModelAdapter):

    self.input_queues = {name: Tensor(np.zeros(shape, dtype=dtype), device=self.model_device).realize()
                         for name, (shape, dtype) in self.input_shapes.items() if name in self.state_pairs}
-    shapes = {name: shape for name, (shape, _) in self.input_shapes.items() if name not in self.state_pairs and name != 'new_img'}
+    shapes = host_shapes({name: shape for name, (shape, _) in self.input_shapes.items()}, self.state_pairs)
    offsets, packed = pack_layout(shapes, self.warp_frame_size)
    self.packed_input = np.zeros(packed, dtype=np.uint8)
    self.host = Tensor(self.packed_input, device='NPY')
@@ -44,6 +44,7 @@ from openpilot.sunnypilot.modeld_v2.blendful import Blendful
 from openpilot.sunnypilot.modeld_v2.constants import ModelConstants, Plan
 from openpilot.sunnypilot.modeld_v2.meta_helper import load_meta_constants
 from openpilot.sunnypilot.modeld_v2.camera_offset_helper import CameraOffsetHelper
+from openpilot.sunnypilot.modeld_v2.light_frame import get_light_warp_matrix
 from openpilot.sunnypilot.livedelay.helpers import get_lat_delay
 from openpilot.sunnypilot.modeld_v2.modeld_base import ModelStateBase
 from openpilot.sunnypilot.modeld_v2.helpers import load_model
@@ -171,7 +172,7 @@ class ModelState(ModelStateBase):
      np.where(inputs[desire_key] - self.prev_desire > .99, inputs[desire_key], 0)
    self.prev_desire[:] = inputs[desire_key]

-    for key in ('traffic_convention', 'lateral_control_params', 'action_t'):
+    for key in ('traffic_convention', 'lateral_control_params', 'action_t', 'speed_limit'):
      if key in self.numpy_inputs and key in inputs:
        self.numpy_inputs[key][:] = inputs[key]

@@ -179,6 +180,8 @@ class ModelState(ModelStateBase):
      for i, key in enumerate(self._vision_input_names):
        if key in transforms:
          self.numpy_inputs['tfm'][i] = transforms[key].reshape(3, 3)
+      if 'light_tfm' in self.numpy_inputs and 'light' in transforms:
+        self.numpy_inputs['light_tfm'][0] = transforms['light'].reshape(3, 3)
    else:
      self.numpy_inputs['tfm'][:, :] = transforms[self._road_key].reshape(3, 3)
      self.numpy_inputs['big_tfm'][:, :] = transforms[self._wide_key].reshape(3, 3)
@@ -338,7 +341,7 @@ def main(demo=False):
  pub_socks = ["modelV2", "drivingModelData", "cameraOdometry", "modelDataV2SP"] + (["chestnutGpuState"] if CHESTNUT else [])
  pm = PubMaster(pub_socks)
  sm = SubMaster(["deviceState", "carState", "narrowRoadCameraState", "extrinsicsCalibration", "driverMonitoringState",
-                  "carControl", "controlsState", "lateralDelay"])
+                  "carControl", "controlsState", "lateralDelay", "longitudinalPlanSP"])

  publish_state = PublishState()
  chestnut_state = ChestnutGpuState(pm, model.chestnut) if CHESTNUT else None
@@ -351,6 +354,7 @@ def main(demo=False):

  model_transform_main = np.zeros((3, 3), dtype=np.float32)
  model_transform_extra = np.zeros((3, 3), dtype=np.float32)
+  model_transform_light = np.zeros((3, 3), dtype=np.float32)
  live_calib_seen = False
  buf_main, buf_extra = None, None
  meta_main = FrameMeta()
@@ -369,6 +373,9 @@ def main(demo=False):
  prev_action = log.ModelDataV2.Action()

  DH = DesireHelper()
+  desire_frames = [q.shape[0] for m in (model, small_model) if m is not None for name, q in m.adapter.input_queues.items() if 'desire' in name]
+  if desire_frames:
+    DH.turn_repulse = min(desire_frames) - 2
  meta_constants = load_meta_constants()
  RELC = RoadEdgeLaneChangeController()
  blendful = Blendful()
@@ -425,6 +432,7 @@ def main(demo=False):
      model_transform_main = get_warp_matrix(device_from_calib_euler, main_intrinsics, False).astype(np.float32)
      model_transform_extra = get_warp_matrix(device_from_calib_euler, dc.wide_road.intrinsics, True).astype(np.float32)
      model_transform_main, model_transform_extra = camera_offset_helper.update(model_transform_main, model_transform_extra, sm, main_wide_camera)
+      model_transform_light = get_light_warp_matrix(device_from_calib_euler, main_intrinsics)
      live_calib_seen = True

    traffic_convention = np.zeros(2)
@@ -446,6 +454,7 @@ def main(demo=False):

    bufs = {name: buf_extra if 'big' in name else buf_main for name in model.vision_input_names}
    transforms = {name: model_transform_extra if 'big' in name else model_transform_main for name in model.vision_input_names}
+    transforms['light'] = model_transform_light

    frame_delay = DT_MDL # compensate for time passed since the frame was captured: current_time - timestamp_eof is 50ms on average
    action_delay = DT_MDL / 2 # middle of the interval between model output (current state) and next frame (expected state)
@@ -463,6 +472,10 @@ def main(demo=False):
    if 'action_t' in model.numpy_inputs:
      inputs['action_t'] = np.array([lat_action_t, long_action_t], dtype=np.float32)

+    if 'speed_limit' in model.numpy_inputs:
+      resolver = sm['longitudinalPlanSP'].speedLimit.resolver
+      inputs['speed_limit'] = np.array([resolver.speedLimitFinal if resolver.speedLimitValid else 0.0], dtype=np.float32)
+
    mt1 = time.perf_counter()
    try:
      send_chestnut = (chestnut_state is not None and
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c08cffa27671d6773e316fdf4dcab80ffce2c4cc6fa28775615bc6726a63d7ba
+size 3473263
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cddff2e3320057735fd4812d1ef9cc7d6d9fe405e3b6e7c4df85ee654a0e886
+size 7699823
@@ -80,7 +80,8 @@ def serialize_jit(jit, kernel_linear, check: dict | None = None) -> tuple[dict,
    if ast.op is Ops.PROGRAM:
      elf, info = ast.to_elf(), ast.arg
      global_size, local_size = info.launch_dims({})
-      calls.append({"kind": "program", "name": elf.name, "lib": blob.add(bytes(elf.lib)), "target": dataclasses.asdict(elf.target),
+      device_target = dataclasses.asdict(dataclasses.replace(elf.target, interface=elf.target.interface.removeprefix("MOCK")))
+      calls.append({"kind": "program", "name": elf.name, "lib": blob.add(bytes(elf.lib)), "target": device_target,
                    "signature": kernel_signature(ast, elf),
                    "global_size": [int(x) for x in global_size], "local_size": [int(x) for x in local_size],
                    "globals": list(info.globals), "outs": list(info.outs), "ins": list(info.ins), "args": arguments})
@@ -17,4 +17,4 @@ class TestRegistryModels(unittest.TestCase):
      self.skipTest("MODEL_BASE_NAME env var not set, skipping integration test.")
    bundle = load_model(os.path.join(os.environ.get("MODEL_DIR", "/tmp/models"), base_name), {"QCOM": "CPU", "AMD": "CPU"}, verify=False, cache=False)
    jits = [value for value in bundle.values() if isinstance(value, PlainJit)]
-    assert "metadata" in bundle and jits and all(len(jit.captured.linear.src) > 0 for jit in jits)
+    assert "metadata" in bundle and jits and all(len(jit.captured._linear.src) > 0 for jit in jits)
@@ -186,6 +186,9 @@ procs += [

  # locationd
  NativeProcess("locationd_llk", "openpilot/sunnypilot/selfdrive/locationd", ["./locationd"], only_onroad),
+
+  # achievements: read-only stats and unlocks, saved on the device
+  PythonProcess("achievementsd", "openpilot.sunnypilot.achievements.achievementsd", only_onroad),
 ]

 if os.path.exists("../../sunnypilot/sunnylink/uploader.py"):
Author	SHA1	Message	Date
Claude	5c24b931ca	achievements: community log, privacy filter and a dry-run preview Phase 0 of the community plan: the car works out what it would share and writes it to files. Nothing is uploaded. - community.py keeps community.db next to state.json: 6 km map tiles (H3 resolution 6) with the day first seen, the tile where each drive starts and ends, engaged distance and time per day, and engaged distance per region per day. Positions only ever become tiles. - The privacy filter follows the build plan: a tile where drives start or end on 3+ days is a frequent stop and it and its 6 neighbors never leave the car; other start/end tiles wait 14 days; hidden tiles never leave; tile batches are sorted with no dates; tiles that become private after being sent turn into delete requests; home region is where you drove most on most of the last 90 days, with detected changes sent at most every 30 days. - preview.py writes the daily reports, the tile batch, delete requests and a map page (your 6 km map, the 45 km blocks others see, what stays on the car) to achievements/preview/ after every drive. - h3lite.py ports the H3 functions we need to pure Python, since sunnypilot doesn't add packages. Checked against h3-py on 2.8M points: identical cells, centers, parents and rings, including pentagons, poles and the antimeridian. - regions.json.gz (207 KB) maps tiles to ISO 3166-2 codes for US, CA, MX, BR, AU, DE and country codes elsewhere, from Natural Earth (public domain). tools/build_regions.py rebuilds it. - replay.py runs everything over qlogs already on the car, into its own folder. - achievementsd feeds the log each sample; any community error turns it off for the drive without touching achievements. 92 tests pass (engine, store, H3 port, privacy filter, replay feeder). The daemon itself was only compiled and linted here; cereal and msgq are not built in this environment. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ATDgLQKGkv2ToqFZNWSF8j	2026-11-10 13:58:14 +00:00
Claude	3b016e36ce	achievements: count driving stats and unlock achievements on the device Phase 1 of the community plan: everything stays on the device. achievementsd (onroad, 10 Hz, read-only) feeds a pure-Python engine from carState, selfdriveState, selfdriveStateSP, modelV2, driverMonitoringState and GPS. It counts engaged distance and time, days and streaks, MADS-only km, assisted lane changes, km per driving model, km on dev builds, the longest run without a red DM alert, and ~5 km areas (salted hashes, never coordinates). 14 achievements / 32 tiers live in catalog.json, with ids like odometer.silver that a later backend can map to badges. Progress is saved to /data/media/0/achievements/state.json every 30 s and at the end of the drive, so no Params key is added and the prebuilt ARM64 params library is not rebuilt. The UI keeps its own seen.json for "new" markers. UI: an "achievements" button on the comma four settings (shows "N new") opening a card per achievement, and an Achievements panel after Trips on the comma 3X. 26 tests pass for the engine and the UI store. The screens and the process were only compiled and linted here (the UI and cereal libraries are ARM64 builds); not yet run on a device. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ATDgLQkU5wERA2wwBF15wr	2026-11-10 13:29:45 +00:00
discountchubbs	0b5620ef67	fixed : U27feb776-85c0-4e94-843a-94d63ce14ae5/40960/3b99f0b7-0949-43a9-9546-444429aac344/16384	2026-11-09 21:00:33 -08:00
discountchubbs	7bbb2472ce	U5e18253c-c598-496d-bd1a-28e1de4c9257/40960/3e36a15d-4d18-40ea-9a84-cb158b9f5c4a/16384	2026-11-09 01:13:29 -08:00
discountchubbs	5cc028dbb4	m8f	2026-11-09 00:38:20 -08:00
discountchubbs	94058fe486	dh pulse renewel	2026-11-08 22:10:13 -08:00
discountchubbs	5ae8e98f9b	mod9fa084ee42	2026-11-08 21:11:37 -08:00
discountchubbs	5bcd182f4d	emu	2026-11-08 21:10:36 -08:00
discountchubbs	103ee81db1	bump	2026-11-07 09:21:53 -08:00
discountchubbs	1bf5473451	rm	2026-11-06 23:15:01 -08:00
discountchubbs	2c94b99ee8	emuuuu	2026-11-06 23:13:00 -08:00
discountchubbs	87e56f8294	replay check	2026-11-06 22:30:29 -08:00
discountchubbs	cb97631449	parity cache	2026-11-06 22:20:54 -08:00
discountchubbs	848c8c1af9	cache by cpu instruct	2026-11-06 21:53:59 -08:00
discountchubbs	3ed5849abb	emu cache	2026-11-06 21:38:45 -08:00
discountchubbs	24112edb95	cache	2026-11-06 21:15:50 -08:00
discountchubbs	e77f239ace	oop	2026-11-06 21:05:10 -08:00
discountchubbs	e972ee35c0	serve over libusb	2026-11-06 21:01:34 -08:00
discountchubbs	5f7a17ed7b	mock amd: x86 hw fma	2026-11-06 20:35:52 -08:00
discountchubbs	8bea2d48c1	modeld_v2: emu	2026-11-06 15:18:45 -08:00