Let ArchiveBox 0.9 actually crawl, and serve its snapshots

Every extractor binary 0.9 cannot find aborts the crawl runner, and abx-plugins
keeps adding new ones, so a fixed apt list always falls behind while the service
user cannot run apt. archivebox install now runs as root with the service user's
home, the way upstream's own package does, and resolves every system dependency
itself; mercury stays disabled because npm refuses its git dependency. Snapshot
iframes are built from BASE_URL, so the update sets it when missing or when it is
a stale IP, and migrated 0.7 collections get their snapshots re-indexed.

install/archivebox-install.sh

@@ -14,34 +14,23 @@ network_check
 update_os
 
 msg_info "Installing Dependencies"
-# ArchiveBox resolves these through its env provider, so it never has to reach for apt itself.
 $STD apt-get install -y \
   git \
   procps \
   dnsutils \
-  chromium \
-  ripgrep \
-  tesseract-ocr \
-  tesseract-ocr-eng \
-  imagemagick \
-  ffmpeg \
-  unzip \
-  wget \
   python3.13
 msg_ok "Installed Dependencies"
 
-# No NODE_MODULE: 0.9 pulls single-file and readability itself, and postlight-parser is gone.
-NODE_VERSION="22" setup_nodejs
 setup_uv
 
 msg_info "Installing ArchiveBox"
 mkdir -p /opt/archivebox/data
 $STD adduser --system --shell /bin/bash --gecos 'Archive Box User' --group --disabled-password --home /home/archivebox archivebox
-# The distro interpreter: a uv-managed one lands where the service user cannot execute it.
 $STD uv venv --python /usr/bin/python3.13 /opt/archivebox/venv
 $STD uv pip install --python /opt/archivebox/venv/bin/python archivebox
 ln -sf /opt/archivebox/venv/bin/archivebox /usr/local/bin/archivebox
-chown -R archivebox:archivebox /opt/archivebox
+mkdir -p /home/archivebox/.config
+chown -R archivebox:archivebox /opt/archivebox /home/archivebox
 msg_ok "Installed ArchiveBox"
 
 msg_info "Initializing ArchiveBox"
@@ -52,18 +41,15 @@ $STD sudo -u archivebox env \
   DJANGO_SUPERUSER_EMAIL=archivebox@archivebox.local \
   DJANGO_SUPERUSER_PASSWORD=community-scripts.org \
   /opt/archivebox/venv/bin/archivebox manage createsuperuser --noinput
-# Without a pinned canonical URL the admin greets every visitor with a red banner.
-$STD sudo -u archivebox /opt/archivebox/venv/bin/archivebox config --set "BASE_URL=http://$(get_ip):5797"
+$STD sudo -u archivebox /opt/archivebox/venv/bin/archivebox config --set "BASE_URL=http://$(get_ip):5797" MERCURY_ENABLED=False
 msg_ok "Initialized ArchiveBox"
 
-msg_info "Installing Extractors"
-# One unreachable extractor must not sink the install: @postlight/parser pulls a git
-# dependency that npm refuses, and readability covers the same job.
-if $STD sudo -u archivebox /opt/archivebox/venv/bin/archivebox install; then
-  msg_ok "Installed Extractors"
-else
-  msg_warn "Some extractors stayed unavailable - ArchiveBox runs without them"
-fi
+msg_info "Installing Extractors (Patience)"
+AB_RUNTIME="/tmp/archivebox-runtime-$(id -u archivebox)"
+install -d -m 700 -o archivebox -g archivebox "$AB_RUNTIME"
+$STD env HOME=/home/archivebox USER=archivebox LOGNAME=archivebox XDG_CONFIG_HOME=/home/archivebox/.config XDG_RUNTIME_DIR="$AB_RUNTIME" \
+  /opt/archivebox/venv/bin/archivebox install
+msg_ok "Installed Extractors"
 
 msg_info "Creating Service"
 cat <<EOF >/etc/systemd/system/archivebox.service
@@ -74,6 +60,7 @@ After=network.target
 [Service]
 User=archivebox
 WorkingDirectory=/opt/archivebox/data
+ExecStartPre=/opt/archivebox/venv/bin/archivebox init
 ExecStart=/opt/archivebox/venv/bin/archivebox server 0.0.0.0:5797
 Restart=always