feat(fnos): run entirely without root and harden process lifecycle

- Switch privilege model to run-as: package (zero root, per fnOS guide)
- Borrow App Store python312 instead of bundling CPython; keep bundled uv
- Move venv, HOME and caches onto the persistent data share
- Start service via setsid; stop/upgrade kill the whole process group
- Sweep stray runtime/box orphans in install/upgrade init hooks
- Track LangBot 4.10.11 (manifest baseline + upstream merge)

packaging/fnos/cmd/install_callback

@@ -19,7 +19,30 @@ cd "${APP_DIR}" || {
 }
 
 # --- Ensure data directory exists ---
-mkdir -p "${DATA_DIR}/plugins" "${DATA_DIR}/box" "${DATA_DIR}/logs" 2>/dev/null || true
+mkdir -p "${DATA_DIR}/plugins" "${DATA_DIR}/box" "${DATA_DIR}/logs" || {
+    echo "Data share not writable: ${DATA_DIR}" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
+
+# --- Writable HOME and tool caches ---
+# Under run-as: package the generated user's system HOME and the app install
+# dir (TRIM_APPDEST) can be read-only; uv aborts if it cannot initialise its
+# cache. Point HOME / UV_CACHE_DIR at the persistent data share and keep the
+# venv there too (UV_PROJECT_ENVIRONMENT), instead of inside APP_DIR.
+VENV_DIR="${DATA_DIR}/.venv"
+export HOME="${DATA_DIR}/.home"
+export UV_CACHE_DIR="${DATA_DIR}/.cache/uv"
+# Never download a managed CPython (the download host is unreachable on many
+# NAS networks); use the distro Python only — fail loudly if it is missing.
+export UV_PYTHON_DOWNLOADS=never
+mkdir -p "${HOME}" "${UV_CACHE_DIR}" || {
+    echo "Data share not writable: ${DATA_DIR}" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
+
+# Real uv/pip errors are captured here for diagnosis (the installer popup
+# only shows the short message we write to TRIM_TEMP_LOGFILE).
+DEBUG_LOG="${DATA_DIR}/logs/install-debug.log"
 
 # --- Pre-seed config.yaml with the user-selected web port ---
 # Data root points at the persistent share (LANGBOT_DATA_ROOT is exported by
@@ -35,7 +58,10 @@ TEMPLATE_FILE="${APP_DIR}/src/langbot/templates/config.yaml"
 _patch_config() {
     local cfg_dir="$1"
     local cfg_file="${cfg_dir}/config.yaml"
-    mkdir -p "${cfg_dir}" 2>/dev/null || true
+    mkdir -p "${cfg_dir}" || {
+        echo "Cannot create config directory: ${cfg_dir}" > "${TRIM_TEMP_LOGFILE}"
+        exit 1
+    }
     if [ ! -f "${cfg_file}" ] && [ -f "${TEMPLATE_FILE}" ]; then
         cp "${TEMPLATE_FILE}" "${cfg_file}"
     fi
@@ -68,15 +94,27 @@ if [ ! -d "/var/apps/nodejs_v${NODE_VERSION}" ]; then
     exit 1
 fi
 
-# --- Python check ---
-PYTHON_BIN="python3"
-if ! command -v "${PYTHON_BIN}" >/dev/null 2>&1; then
-    PYTHON_BIN="python"
-fi
-if ! command -v "${PYTHON_BIN}" >/dev/null 2>&1; then
-    echo "Python not found on this system" > "${TRIM_TEMP_LOGFILE}"
+# --- CPU architecture (must be resolved before locating bundled binaries) ---
+ARCH=$(uname -m)
+case "${ARCH}" in
+    x86_64|aarch64) ;;
+    *)
+        echo "Unsupported CPU architecture: ${ARCH}" > "${TRIM_TEMP_LOGFILE}"
+        exit 1
+        ;;
+esac
+
+# --- Python: official python312 App Store app (same borrow pattern as Node.js) ---
+PYTHON_APP="python312"
+PYTHON_BIN="/var/apps/${PYTHON_APP}/target/bin/python3"
+if [ ! -d "/var/apps/${PYTHON_APP}" ]; then
+    echo "未找到官方 Python 环境：请先在应用中心安装 ${PYTHON_APP}，再重新安装本应用。" > "${TRIM_TEMP_LOGFILE}"
     exit 1
 fi
+[ -x "${PYTHON_BIN}" ] || {
+    echo "Python 解释器缺失或不可执行：${PYTHON_BIN}" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
 
 PY_VER=$("${PYTHON_BIN}" -c 'import sys; print(f"{sys.version_info.major}.{sys.version_info.minor}")' 2>/dev/null)
 if [ -z "${PY_VER}" ]; then
@@ -90,55 +128,32 @@ if [ "${PY_MAJOR}" -lt 3 ] || { [ "${PY_MAJOR}" -eq 3 ] && [ "${PY_MINOR}" -lt 1
     exit 1
 fi
 
-# --- Resolve uv: bundled binary first, then online fallbacks ---
-UV_BIN=""
-ARCH=$(uname -m)
-case "${ARCH}" in
-    x86_64)  BUNDLED_UV="${TRIM_APPDEST}/bin/uv-x86_64" ;;
-    aarch64) BUNDLED_UV="${TRIM_APPDEST}/bin/uv-aarch64" ;;
-    *)       BUNDLED_UV="" ;;
-esac
-
-if [ -n "${BUNDLED_UV}" ] && [ -x "${BUNDLED_UV}" ]; then
-    mkdir -p "${TRIM_PKGVAR}/bin"
-    cp "${BUNDLED_UV}" "${TRIM_PKGVAR}/bin/uv" && chmod +x "${TRIM_PKGVAR}/bin/uv"
-    UV_BIN="${TRIM_PKGVAR}/bin/uv"
-fi
-
-if [ -z "${UV_BIN}" ] && command -v uv >/dev/null 2>&1; then
-    UV_BIN="uv"
-fi
-
-if [ -z "${UV_BIN}" ]; then
-    "${PYTHON_BIN}" -m pip install --user --no-cache-dir uv 2>/dev/null || \
-    "${PYTHON_BIN}" -m pip install --no-cache-dir uv 2>/dev/null || \
-    curl -LsSf https://astral.sh/uv/install.sh | sh 2>/dev/null || true
-    export PATH="${HOME}/.local/bin:${PATH}"
-    if command -v uv >/dev/null 2>&1; then
-        UV_BIN="uv"
-    elif [ -x "${HOME}/.local/bin/uv" ]; then
-        UV_BIN="${HOME}/.local/bin/uv"
-    fi
-fi
-
-if [ -z "${UV_BIN}" ]; then
-    echo "无法获取 uv：内置二进制缺失且在线安装失败。请检查网络后重新安装。" > "${TRIM_TEMP_LOGFILE}"
+# --- Resolve uv: run the bundled binary in place (single canonical path) ---
+UV_BIN="${TRIM_APPDEST}/bin/uv-${ARCH}"
+[ -x "${UV_BIN}" ] || {
+    echo "Bundled uv binary missing or not executable: ${UV_BIN}" > "${TRIM_TEMP_LOGFILE}"
     exit 1
-fi
+}
 
-# --- Create venv via uv ---
-if [ ! -d ".venv" ]; then
-    "${UV_BIN}" venv .venv --python "${PYTHON_BIN}" || {
-        echo "Failed to create Python virtual environment via uv" > "${TRIM_TEMP_LOGFILE}"
+# --- Create venv via uv (on the writable data share, not APP_DIR) ---
+if [ ! -d "${VENV_DIR}" ]; then
+    {
+        echo "== whoami: $(id 2>&1)"
+        echo "== APP_DIR perms: $(ls -ld "${APP_DIR}" 2>&1)"
+        echo "== DATA_DIR perms: $(ls -ld "${DATA_DIR}" 2>&1)"
+        echo "== HOME=${HOME} UV_CACHE_DIR=${UV_CACHE_DIR}"
+        "${UV_BIN}" venv "${VENV_DIR}" --python "${PYTHON_BIN}"
+    } >> "${DEBUG_LOG}" 2>&1 || {
+        echo "Failed to create Python virtual environment via uv. See ${DEBUG_LOG}" > "${TRIM_TEMP_LOGFILE}"
         exit 1
     }
 fi
 
-# --- Sync dependencies ---
-"${UV_BIN}" sync --extra seekdb || {
-    echo "Dependency sync failed. Check network connectivity." > "${TRIM_TEMP_LOGFILE}"
+# --- Sync dependencies (install into the relocated venv) ---
+if ! UV_PROJECT_ENVIRONMENT="${VENV_DIR}" "${UV_BIN}" sync --extra seekdb >> "${DEBUG_LOG}" 2>&1; then
+    echo "Dependency sync failed. Check network connectivity. See ${DEBUG_LOG}" > "${TRIM_TEMP_LOGFILE}"
     exit 1
-}
+fi
 
 # --- Verify frontend dist ---
 if [ ! -d "web/dist" ] || [ -z "$(ls -A web/dist 2>/dev/null)" ]; then
@@ -146,4 +161,9 @@ if [ ! -d "web/dist" ] || [ -z "$(ls -A web/dist 2>/dev/null)" ]; then
     exit 1
 fi
 
+# --- Prepare a writable HOME inside the data dir for tool caches (uv,
+# npm/npx MCP). Everything already runs as the package user (run-as:
+# package), so no chown is needed — files created here belong to it.
+mkdir -p "${DATA_DIR}/.home" 2>/dev/null || true
+
 exit 0

packaging/fnos/cmd/install_init

@@ -1,5 +1,12 @@
 #!/bin/bash
-# cmd/install_init - pre-install hook
-# Nothing special to do before extraction.
+# cmd/install_init - pre-install hook (runs before files are applied)
+# Sweep stray processes from a previous failed/killed install: orphans whose
+# parent was hard-killed keep running and hold the runtime/box ws ports,
+# which breaks the new instance. No match is the normal case on a clean
+# install — pkill exits 1.
+
+RUN_USER=$(id -un)
+pkill -KILL -u "${RUN_USER}" -f "appcenter/langbot" 2>/dev/null || true
+pkill -KILL -u "${RUN_USER}" -f "appshare/langbot" 2>/dev/null || true
 
 exit 0

packaging/fnos/cmd/main

@@ -27,27 +27,48 @@ if [ -z "${DATA_DIR}" ]; then
     DATA_DIR="${TRIM_PKGVAR}/data"
 fi
 export LANGBOT_DATA_ROOT="${DATA_DIR}"
-mkdir -p "${DATA_DIR}" 2>/dev/null || true
+mkdir -p "${DATA_DIR}" || {
+    echo "Data share not writable: ${DATA_DIR}" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
 
-# --- Locate Python ---
-PYTHON_BIN="python3"
-! command -v "${PYTHON_BIN}" >/dev/null 2>&1 && PYTHON_BIN="python"
+# --- Writable HOME / tool caches / venv on the data share ---
+# Must match cmd/install_callback: under run-as: package neither the system
+# HOME nor TRIM_APPDEST are guaranteed writable, and the venv lives at
+# ${DATA_DIR}/.venv instead of inside the app dir.
+export HOME="${DATA_DIR}/.home"
+export UV_CACHE_DIR="${DATA_DIR}/.cache/uv"
+# Never download a managed CPython — distro Python only (see install_callback)
+export UV_PYTHON_DOWNLOADS=never
+export UV_PROJECT_ENVIRONMENT="${DATA_DIR}/.venv"
+mkdir -p "${HOME}" "${UV_CACHE_DIR}" || {
+    echo "Data share not writable: ${DATA_DIR}" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
 
-# --- Locate uv ---
-# install_callback puts the bundled uv binary at ${TRIM_PKGVAR}/bin/uv
-UV_BIN="${TRIM_PKGVAR}/bin/uv"
-if [ ! -x "${UV_BIN}" ]; then
-    UV_BIN="uv"
-fi
-if ! command -v "${UV_BIN}" >/dev/null 2>&1; then
-    UV_BIN="${HOME}/.local/bin/uv"
-fi
-if ! command -v "${UV_BIN}" >/dev/null 2>&1 && [ ! -x "${UV_BIN}" ]; then
-    UV_BIN="${HOME}/.cargo/bin/uv"
-fi
-if ! command -v "${UV_BIN}" >/dev/null 2>&1 && [ ! -x "${UV_BIN}" ]; then
-    UV_BIN="${APP_DIR}/.venv/bin/uv"
-fi
+# --- CPU architecture (must be resolved before locating bundled binaries) ---
+ARCH=$(uname -m)
+case "${ARCH}" in
+    x86_64|aarch64) ;;
+    *)
+        echo "Unsupported CPU architecture: ${ARCH}" > "${TRIM_TEMP_LOGFILE}"
+        exit 1
+        ;;
+esac
+
+# --- Locate Python: official python312 App Store app (same borrow pattern as Node.js) ---
+PYTHON_BIN="/var/apps/python312/target/bin/python3"
+[ -x "${PYTHON_BIN}" ] || {
+    echo "Python interpreter missing or not executable: ${PYTHON_BIN} (install the python312 app)" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
+
+# --- Locate uv: bundled binary at its single canonical path in the app dir ---
+UV_BIN="${TRIM_APPDEST}/bin/uv-${ARCH}"
+[ -x "${UV_BIN}" ] || {
+    echo "Bundled uv binary missing or not executable: ${UV_BIN}" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
 
 case $1 in
   start)
@@ -69,7 +90,7 @@ case $1 in
         exit 1
     }
 
-    if [ ! -d ".venv" ]; then
+    if [ ! -d "${DATA_DIR}/.venv" ]; then
         echo "Python virtual environment not found. Please reinstall LangBot." > "${TRIM_TEMP_LOGFILE}"
         exit 1
     fi
@@ -79,6 +100,9 @@ case $1 in
     # fresh data/ with a default 5300 config gets recreated inside target/ on
     # every install/upgrade. The symlink keeps everything on the persistent
     # share; it is recreated here on each start (upgrades wipe target/).
+    # Requires the package user to have write permission on the app dir — if
+    # fnOS ever mounts it read-only, this fails loudly instead of silently
+    # running against an ephemeral data directory.
     APP_DATA="${APP_DIR}/data"
     if [ -L "${APP_DATA}" ]; then
         # already a symlink; re-point if the persistent dir changed
@@ -87,7 +111,7 @@ case $1 in
         # legacy real dir (created by LangBot before this fix): merge into the
         # persistent dir without overwriting newer files already there
         mkdir -p "${DATA_DIR}"
-        cp -an "${APP_DATA}/." "${DATA_DIR}/" 2>/dev/null || cp -a "${APP_DATA}/." "${DATA_DIR}/"
+        cp -an "${APP_DATA}/." "${DATA_DIR}/." || cp -a "${APP_DATA}/." "${DATA_DIR}/"
         rm -rf "${APP_DATA}"
         ln -s "${DATA_DIR}" "${APP_DATA}"
     else
@@ -146,7 +170,22 @@ except Exception:
     # (--standalone-runtime would require an external runtime at
     #  ws://langbot_plugin_runtime:5400, which only exists in Docker Compose.)
     # --standalone-box omitted: Box sandbox defaults off, users enable via Web UI
-    nohup "${UV_BIN}" run --no-sync main.py \
+    #
+    # Privilege model: the whole app runs as the generated package user
+    # (run-as: package, see config/privilege) — no root anywhere. HOME /
+    # UV_CACHE_DIR / UV_PROJECT_ENVIRONMENT are exported at the top and point
+    # at the persistent share so tool caches (uv, npm/npx) and the relocated
+    # venv stay writable regardless of the generated user's system home.
+    #
+    # Process-group lifecycle: setsid makes the main process a session/group
+    # leader, so PID == PGID. "stop" kills the whole group — stdio children
+    # (plugin runtime, Box) die with the parent and can never survive as
+    # orphans holding their ws ports after a crash, stop or upgrade.
+    command -v setsid >/dev/null 2>&1 || {
+        echo "setsid not found (util-linux required for process-group lifecycle)" > "${TRIM_TEMP_LOGFILE}"
+        exit 1
+    }
+    setsid nohup "${UV_BIN}" run --no-sync main.py \
         > "${LOG_FILE}" 2>&1 &
     echo $! > "${PID_FILE}"
 
@@ -165,12 +204,18 @@ except Exception:
     if [ -f "${PID_FILE}" ]; then
         PID=$(cat "${PID_FILE}" | tr -d '[:space:]')
         if [ -n "${PID}" ]; then
-            kill "${PID}" 2>/dev/null
+            # Started with setsid, so PID == PGID: kill the whole group so
+            # stdio children (plugin runtime, Box) die with the parent. The
+            # plain-PID kill covers instances started before the setsid
+            # change (group kill is a no-op for them).
+            kill -TERM -- "-${PID}" 2>/dev/null
+            kill -TERM "${PID}" 2>/dev/null
             for _ in 1 2 3 4 5 6 7 8 9 10; do
                 kill -0 "${PID}" 2>/dev/null || break
                 sleep 1
             done
-            kill -9 "${PID}" 2>/dev/null
+            kill -KILL -- "-${PID}" 2>/dev/null
+            kill -KILL "${PID}" 2>/dev/null
         fi
         rm -f "${PID_FILE}"
     fi

packaging/fnos/cmd/uninstall_callback

@@ -7,8 +7,7 @@ if [ "${wizard_keep_data:-yes}" = "no" ]; then
     # 应用运行数据（pid、日志等）
     if [ -n "${TRIM_PKGVAR}" ]; then
         rm -rf "${TRIM_PKGVAR:?}"/langbot.pid \
-               "${TRIM_PKGVAR:?}"/langbot.log \
-               "${TRIM_PKGVAR:?}"/bin 2>/dev/null || true
+               "${TRIM_PKGVAR:?}"/langbot.log 2>/dev/null || true
     fi
 
     # 共享数据目录（langbot/data）

packaging/fnos/cmd/upgrade_callback

@@ -8,6 +8,20 @@ APP_DIR="${TRIM_APPDEST}/langbot"
 DATA_DIR="${TRIM_DATA_SHARE_PATHS%%:*}"
 [ -z "${DATA_DIR}" ] && DATA_DIR="${TRIM_PKGVAR}/data"
 
+# Writable HOME / caches / venv on the data share (must match
+# cmd/install_callback — venv is relocated under DATA_DIR).
+VENV_DIR="${DATA_DIR}/.venv"
+export HOME="${DATA_DIR}/.home"
+export UV_CACHE_DIR="${DATA_DIR}/.cache/uv"
+# Never download a managed CPython — distro Python only (see install_callback)
+export UV_PYTHON_DOWNLOADS=never
+export UV_PROJECT_ENVIRONMENT="${VENV_DIR}"
+mkdir -p "${HOME}" "${UV_CACHE_DIR}" "${DATA_DIR}/logs" || {
+    echo "Data share not writable: ${DATA_DIR}" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
+DEBUG_LOG="${DATA_DIR}/logs/upgrade-debug.log"
+
 # Apply port from upgrade wizard (config persists across upgrades; this
 # only rewrites it when the user changed the value in the upgrade wizard)
 CONFIG_FILE="${DATA_DIR}/config.yaml"
@@ -26,52 +40,56 @@ cd "${APP_DIR}" || {
     exit 1
 }
 
-# Find uv (bundled first, then PATH / ~/.local/bin / ~/.cargo/bin)
-UV_BIN="${TRIM_PKGVAR}/bin/uv"
-if [ ! -x "${UV_BIN}" ]; then
-    UV_BIN="uv"
-fi
-if ! command -v "${UV_BIN}" >/dev/null 2>&1; then
-    UV_BIN="${HOME}/.local/bin/uv"
-fi
-if ! command -v "${UV_BIN}" >/dev/null 2>&1 && [ ! -x "${UV_BIN}" ]; then
-    UV_BIN="${HOME}/.cargo/bin/uv"
-fi
+# --- CPU architecture (must be resolved before locating bundled binaries) ---
+ARCH=$(uname -m)
+case "${ARCH}" in
+    x86_64|aarch64) ;;
+    *)
+        echo "Unsupported CPU architecture: ${ARCH}" > "${TRIM_TEMP_LOGFILE}"
+        exit 1
+        ;;
+esac
 
-PYTHON_BIN="python3"
-! command -v "${PYTHON_BIN}" >/dev/null 2>&1 && PYTHON_BIN="python"
+# uv: bundled binary at its single canonical path in the app dir
+UV_BIN="${TRIM_APPDEST}/bin/uv-${ARCH}"
+[ -x "${UV_BIN}" ] || {
+    echo "Bundled uv binary missing or not executable: ${UV_BIN}" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
 
-# Re-sync deps
-if [ -d ".venv" ]; then
-    "${UV_BIN}" sync --extra seekdb 2>/dev/null || {
-        echo "Dependency sync failed after upgrade" > "${TRIM_TEMP_LOGFILE}"
+# Python: official python312 App Store app (same borrow pattern as Node.js)
+PYTHON_BIN="/var/apps/python312/target/bin/python3"
+[ -x "${PYTHON_BIN}" ] || {
+    echo "Python interpreter missing or not executable: ${PYTHON_BIN} (install the python312 app)" > "${TRIM_TEMP_LOGFILE}"
+    exit 1
+}
+
+# Re-sync deps (venv lives at ${DATA_DIR}/.venv, see install_callback)
+if [ -d "${VENV_DIR}" ]; then
+    UV_PROJECT_ENVIRONMENT="${VENV_DIR}" "${UV_BIN}" sync --extra seekdb >> "${DEBUG_LOG}" 2>&1 || {
+        echo "Dependency sync failed after upgrade. See ${DEBUG_LOG}" > "${TRIM_TEMP_LOGFILE}"
         exit 1
     }
 else
-    # Venv was lost, recreate via uv
-    if ! command -v "${UV_BIN}" >/dev/null 2>&1 && [ ! -x "${UV_BIN}" ]; then
-        "${PYTHON_BIN}" -m pip install --user --no-cache-dir uv 2>/dev/null || \
-        "${PYTHON_BIN}" -m pip install --no-cache-dir uv 2>/dev/null || {
-            echo "Failed to install uv" > "${TRIM_TEMP_LOGFILE}"
-            exit 1
-        }
-        export PATH="${HOME}/.local/bin:${PATH}"
-        UV_BIN="uv"
-    fi
-    "${UV_BIN}" venv .venv --python "${PYTHON_BIN}" || {
-        echo "Failed to recreate virtual environment" > "${TRIM_TEMP_LOGFILE}"
+    # Venv was lost, recreate it with the bundled uv on the data share
+    "${UV_BIN}" venv "${VENV_DIR}" --python "${PYTHON_BIN}" >> "${DEBUG_LOG}" 2>&1 || {
+        echo "Failed to recreate virtual environment. See ${DEBUG_LOG}" > "${TRIM_TEMP_LOGFILE}"
         exit 1
     }
-    "${UV_BIN}" sync --extra seekdb || {
-        echo "Dependency sync failed" > "${TRIM_TEMP_LOGFILE}"
+    UV_PROJECT_ENVIRONMENT="${VENV_DIR}" "${UV_BIN}" sync --extra seekdb >> "${DEBUG_LOG}" 2>&1 || {
+        echo "Dependency sync failed. See ${DEBUG_LOG}" > "${TRIM_TEMP_LOGFILE}"
         exit 1
     }
 fi
 
 # Verify frontend dist still present
 if [ ! -d "web/dist" ] || [ -z "$(ls -A web/dist 2>/dev/null)" ]; then
-    echo "Frontend dist missing after upgrade! Web UI will not be available." > "${TRIM_TEMP_LOGFILE}"
+    echo "Frontend dist missing! Web UI will not be available." > "${TRIM_TEMP_LOGFILE}"
     exit 1
 fi
 
+# Everything runs as the package user (run-as: package); the venv recreated
+# above and the config rewritten via sed already belong to it. Cache HOME is
+# prepared at the top of this script (see cmd/install_callback).
+
 exit 0

packaging/fnos/cmd/upgrade_init

@@ -1,20 +1,36 @@
 #!/bin/bash
-# cmd/upgrade_init - pre-upgrade hook
-# Stop the running LangBot process before files are replaced.
+# cmd/upgrade_init - pre-upgrade hook (runs before files are replaced)
+# 1. Stop the running LangBot instance — whole process group, so stdio
+#    children (plugin runtime, Box) die with the parent.
+# 2. Sweep stray processes from a previous crash/upgrade: orphans whose
+#    parent was hard-killed keep running and hold the runtime/box ws ports,
+#    which breaks the next start.
 
 PID_FILE="${TRIM_PKGVAR}/langbot.pid"
 
 if [ -f "${PID_FILE}" ]; then
     PID=$(cat "${PID_FILE}" | tr -d '[:space:]')
     if [ -n "${PID}" ] && kill -0 "${PID}" 2>/dev/null; then
-        kill "${PID}" 2>/dev/null
+        # Instances started with setsid have PID == PGID; the plain-PID kill
+        # covers instances started before that change.
+        kill -TERM -- "-${PID}" 2>/dev/null
+        kill -TERM "${PID}" 2>/dev/null
         for _ in 1 2 3 4 5 6 7 8 9 10; do
             kill -0 "${PID}" 2>/dev/null || break
             sleep 1
         done
-        kill -9 "${PID}" 2>/dev/null
+        kill -KILL -- "-${PID}" 2>/dev/null
+        kill -KILL "${PID}" 2>/dev/null
     fi
     rm -f "${PID_FILE}"
 fi
 
+# Stray sweep: match only our volume paths (venv/uv under the app dir on
+# @appcenter, venv/HOME/caches under the data share on @appshare). This
+# script itself runs from /var/apps/<appname>/cmd/, so it never matches
+# itself. No match is the normal case on a clean upgrade — pkill exits 1.
+RUN_USER=$(id -un)
+pkill -KILL -u "${RUN_USER}" -f "appcenter/langbot" 2>/dev/null || true
+pkill -KILL -u "${RUN_USER}" -f "appshare/langbot" 2>/dev/null || true
+
 exit 0