Files
hack-house/hh/scripts/sandbox-bootstrap.sh
T
leetcrypt 20091a9c26 feat(sandbox): podman backend + goose AI harness + noVNC GUI sandbox
Add Podman as a rootless/daemonless sandbox backend alongside Docker,
Multipass and Local, and wire Goose in as the default agentic harness
for the granted `!task` path (bridge execs `<engine> exec <name> goose
run` and streams output to chat; auto-degrades to the simple one-shot
injector when goose is absent).

Add an optional GUI sandbox track (XFCE + TigerVNC + websockify/noVNC on
:6080) summoned via `/sbx <engine> gui`, plus container-side provisioning
in sandbox-bootstrap.sh and a host-side ensure-podman.sh prereq helper.

Refresh the in-app command help to the backend-led `/sbx <engine> [gui]`
grammar and minor ui tweaks.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-06-08 10:38:18 -07:00

143 lines
6.9 KiB
Bash

#!/usr/bin/env bash
# sandbox-bootstrap.sh — install a baseline dev toolchain inside a hack-house
# Docker sandbox.
#
# It is piped to `bash -s` (as root) inside the container at provision time, so
# fresh sandboxes come up usable instead of bare. Idempotent + sentinel-guarded:
# a re-provision (new member joins) or a snapshot load is a fast no-op.
#
# When launched by hh, the package list comes from scripts/sandbox-tools.json
# (the canonical, editable schema) via $HH_SBX_PKGS. DEFAULT_PKGS below is only
# the fallback for running this script standalone. Per-launch override:
# HH_SBX_PKGS="vim tmux ripgrep" ./host-house.sh ...
set -uo pipefail
# -h/--help: print the usage header above and exit. (No effect in normal use —
# hh pipes this to `bash -s` inside the container with no args; the flag is for
# running it standalone to read what it does.)
case "${1:-}" in
-h|--help|-help) sed -n '2,/^set /{/^set /d;s/^# \{0,1\}//;p}' "$0"; exit 0 ;;
esac
SENTINEL=/var/lib/hh-bootstrap.done
[[ -f "$SENTINEL" ]] && exit 0 # this container is already provisioned
# Baseline dev tools: editors, fetchers, vcs, net + inspect utilities, json,
# archives. Keep names valid for the base image (ubuntu:24.04) — an unknown
# package would otherwise abort the whole batch install.
DEFAULT_PKGS="vim nano less curl wget ca-certificates git \
build-essential pkg-config \
procps iproute2 iputils-ping net-tools \
jq unzip zip tree htop file ripgrep \
python3 python3-pip python3-venv"
PKGS="${HH_SBX_PKGS:-$DEFAULT_PKGS}"
export DEBIAN_FRONTEND=noninteractive
# Refresh the index first (base images ship without /var/lib/apt/lists). If the
# update fails (e.g. no network) we bail WITHOUT writing the sentinel, so the
# next launch retries instead of leaving a half-provisioned shell.
apt-get update -qq || exit 0
# --no-install-recommends keeps the image lean. apt is atomic on resolution, so
# if one name is unavailable the batch aborts — fall back to one-by-one so a
# single bad/missing package can't deprive the shell of everything else.
# shellcheck disable=SC2086
if ! apt-get install -y --no-install-recommends $PKGS; then
for p in $PKGS; do
apt-get install -y --no-install-recommends "$p" || true
done
fi
# ---- Optional GUI desktop over noVNC -----------------------------------------
# Opt-in via HH_SBX_GUI=1 (set by the launcher for `/sbx <docker|podman> gui`).
# Containers are headless, so a GUI means: install a lightweight XFCE desktop +
# a VNC server, and bridge it to a browser-reachable noVNC websocket on port
# $HH_SBX_GUI_PORT (default 6080). The launcher publishes that port on the host
# loopback (127.0.0.1) only, so the desktop is reachable from this machine's
# browser but never the LAN. Heavy (pulls a desktop), hence strictly opt-in.
if [[ "${HH_SBX_GUI:-0}" == "1" ]]; then
GUI_PORT="${HH_SBX_GUI_PORT:-6080}"
GUI_PASS="${HH_SBX_GUI_PASS:-hackhouse}"
# XFCE (minimal) + a terminal, the X session glue, a VNC server, and the
# noVNC web client + websockify bridge. All apt — works on Ubuntu/Debian/Kali.
# tigervnc-tools carries `vncpasswd`, which --no-install-recommends would
# otherwise drop (it ships only as a recommend of the standalone server) —
# without it the passwd write below is a silent no-op and VNC never auths.
apt-get install -y --no-install-recommends \
xfce4 xfce4-terminal dbus-x11 xfonts-base x11-xserver-utils \
tigervnc-standalone-server tigervnc-tools novnc websockify || true
export USER=root HOME=/root
# TigerVNC 1.15 (Kali/Debian trixie) reads its config from ~/.config/tigervnc
# and aborts with a fatal "Could not migrate /root/.vnc" if the legacy ~/.vnc
# exists, so write straight to the new path. /tmp/.X11-unix must exist + be
# sticky for the X socket; a fresh container may not have it.
VNCDIR=/root/.config/tigervnc
mkdir -p "$VNCDIR" /tmp/.X11-unix
chmod 1777 /tmp/.X11-unix 2>/dev/null || true
# Non-interactive VNC password (obfuscated, not strong auth — the real gate is
# the loopback-only publish + the room password). vncpasswd -f reads stdin.
printf '%s\n' "$GUI_PASS" | vncpasswd -f > "$VNCDIR/passwd" 2>/dev/null || true
chmod 600 "$VNCDIR/passwd" 2>/dev/null || true
cat > "$VNCDIR/xstartup" <<'XS'
#!/bin/sh
unset SESSION_MANAGER DBUS_SESSION_BUS_ADDRESS
exec dbus-launch startxfce4
XS
chmod +x "$VNCDIR/xstartup"
# Start the VNC server on display :1 (TCP 5901). -localhost no lets websockify
# (running in the same netns) reach it; nothing outside the container can —
# only $GUI_PORT is published, and only to host loopback.
if command -v vncserver >/dev/null 2>&1; then
vncserver -kill :1 >/dev/null 2>&1 || true
rm -f /tmp/.X1-lock /tmp/.X11-unix/X1 2>/dev/null || true
vncserver :1 -geometry 1280x800 -depth 24 -localhost no >/dev/null 2>&1 || true
fi
# noVNC web client → websocket → local VNC. Daemonize (nohup, detached stdio)
# so it outlives this `exec bash -s` provision session and keeps serving for
# the life of the container (PID 1 is `sleep infinity`).
if command -v websockify >/dev/null 2>&1; then
NOVNC_WEB=/usr/share/novnc
[[ -d "$NOVNC_WEB" ]] || NOVNC_WEB=/usr/share/webapps/novnc
nohup websockify --web="$NOVNC_WEB" "$GUI_PORT" localhost:5901 \
>/var/log/hh-novnc.log 2>&1 &
fi
fi
# ---- Goose harness (the /ai agent's sandbox `!task` path) --------------------
# Goose ships as a release binary (not apt), so install it via its official
# installer into a system path so every container user can run it. Best-effort:
# a failure here just means the agent falls back to its built-in one-shot
# injector, so it never blocks provisioning. Skip entirely with HH_SBX_GOOSE=0.
if [[ "${HH_SBX_GOOSE:-1}" != "0" ]] && ! command -v goose >/dev/null 2>&1; then
if command -v curl >/dev/null 2>&1; then
GOOSE_BIN_DIR=/usr/local/bin CONFIGURE=false \
bash -c 'curl -fsSL https://github.com/block/goose/releases/download/stable/download_cli.sh | bash' \
|| true
fi
fi
# Container-side Goose config: point it at the host's Ollama via the engine's
# host gateway (passed in as $HH_OLLAMA_HOST by the launcher). The shared shell
# runs as root here, so write it under /root — only if absent, so a tuned config
# survives a re-provision.
if command -v goose >/dev/null 2>&1; then
GOOSE_CFG=/root/.config/goose/config.yaml
if [[ ! -f "$GOOSE_CFG" ]]; then
mkdir -p "$(dirname "$GOOSE_CFG")"
cat > "$GOOSE_CFG" <<EOF
# Written by hh sandbox-bootstrap.sh — Goose runs INSIDE this sandbox and reaches
# the host Ollama over the engine's host gateway. No host filesystem is mounted.
GOOSE_PROVIDER: ollama
GOOSE_MODEL: ${HH_GOOSE_MODEL:-qwen2.5:3b}
OLLAMA_HOST: ${HH_OLLAMA_HOST:-http://host.containers.internal:11434}
EOF
fi
fi
mkdir -p "$(dirname "$SENTINEL")"
date -u +%FT%TZ > "$SENTINEL" # mark done; skip on the next provision pass