"""Docker execution environment for sandboxed command execution.
Security hardened (cap-drop ALL, no-new-privileges, PID limits),
configurable resource limits (CPU, memory, disk), and optional filesystem
persistence via bind mounts.
"""
import logging
import os
import re
import shutil
import subprocess
import sys
import uuid
from typing import Optional
from tools.environments.base import BaseEnvironment, _popen_bash
from tools.environments.local import _HERMES_PROVIDER_ENV_BLOCKLIST
logger = logging.getLogger(__name__)
_DOCKER_SEARCH_PATHS = [
"/usr/local/bin/docker",
"/opt/homebrew/bin/docker",
"/Applications/Docker.app/Contents/Resources/bin/docker",
]
_docker_executable: Optional[str] = None
_ENV_VAR_NAME_RE = re.compile(r"^[A-Za-z_][A-Za-z0-9_]*$")
def _normalize_forward_env_names(forward_env: list[str] | None) -> list[str]:
"""Return a deduplicated list of valid environment variable names."""
normalized: list[str] = []
seen: set[str] = set()
for item in forward_env or []:
if not isinstance(item, str):
logger.warning("Ignoring non-string docker_forward_env entry: %r", item)
continue
key = item.strip()
if not key:
continue
if not _ENV_VAR_NAME_RE.match(key):
logger.warning("Ignoring invalid docker_forward_env entry: %r", item)
continue
if key in seen:
continue
seen.add(key)
normalized.append(key)
return normalized
def _normalize_env_dict(env: dict | None) -> dict[str, str]:
"""Validate and normalize a docker_env dict to {str: str}.
Filters out entries with invalid variable names or non-string values.
"""
if not env:
return {}
if not isinstance(env, dict):
logger.warning("docker_env is not a dict: %r", env)
return {}
normalized: dict[str, str] = {}
for key, value in env.items():
if not isinstance(key, str) or not _ENV_VAR_NAME_RE.match(key.strip()):
logger.warning("Ignoring invalid docker_env key: %r", key)
continue
key = key.strip()
if not isinstance(value, str):
if isinstance(value, (int, float, bool)):
value = str(value)
else:
logger.warning("Ignoring non-string docker_env value for %r: %r", key, value)
continue
normalized[key] = value
return normalized
def _load_hermes_env_vars() -> dict[str, str]:
"""Load ~/.hermes/.env values without failing Docker command execution."""
try:
from hermes_cli.config import load_env
return load_env() or {}
except Exception:
return {}
def find_docker() -> Optional[str]:
"""Locate the docker (or podman) CLI binary.
Resolution order:
1. ``HERMES_DOCKER_BINARY`` env var — explicit override (e.g. ``/usr/bin/podman``)
2. ``docker`` on PATH via ``shutil.which``
3. ``podman`` on PATH via ``shutil.which``
4. Well-known macOS Docker Desktop install locations
Returns the absolute path, or ``None`` if neither runtime can be found.
"""
global _docker_executable
if _docker_executable is not None:
return _docker_executable
override = os.getenv("HERMES_DOCKER_BINARY")
if override and os.path.isfile(override) and os.access(override, os.X_OK):
_docker_executable = override
logger.info("Using HERMES_DOCKER_BINARY override: %s", override)
return override
found = shutil.which("docker")
if found:
_docker_executable = found
return found
found = shutil.which("podman")
if found:
_docker_executable = found
logger.info("Using podman as container runtime: %s", found)
return found
for path in _DOCKER_SEARCH_PATHS:
if os.path.isfile(path) and os.access(path, os.X_OK):
_docker_executable = path
logger.info("Found docker at non-PATH location: %s", path)
return path
return None
_BASE_SECURITY_ARGS = [
"--cap-drop", "ALL",
"--cap-add", "DAC_OVERRIDE",
"--cap-add", "CHOWN",
"--cap-add", "FOWNER",
"--security-opt", "no-new-privileges",
"--pids-limit", "256",
"--tmpfs", "/tmp:rw,nosuid,size=512m",
"--tmpfs", "/var/tmp:rw,noexec,nosuid,size=256m",
"--tmpfs", "/run:rw,noexec,nosuid,size=64m",
]
_GOSU_CAP_ARGS = [
"--cap-add", "SETUID",
"--cap-add", "SETGID",
]
def _build_security_args(run_as_host_user: bool) -> list[str]:
"""Return the security/cap/tmpfs args tailored to the privilege mode."""
if run_as_host_user:
return list(_BASE_SECURITY_ARGS)
return list(_BASE_SECURITY_ARGS) + list(_GOSU_CAP_ARGS)
def _resolve_host_user_spec() -> Optional[str]:
"""Return ``<uid>:<gid>`` for the current host user, or ``None`` on platforms
where this is not meaningful (e.g. Windows without posix ids).
We intentionally read ``os.getuid()``/``os.getgid()`` directly rather than
going through ``getpass``/``pwd`` so this stays cheap and never raises on
nameless UIDs (nss lookups can fail inside sandboxed launchers).
"""
get_uid = getattr(os, "getuid", None)
get_gid = getattr(os, "getgid", None)
if get_uid is None or get_gid is None:
return None
try:
return f"{get_uid()}:{get_gid()}"
except Exception:
return None
_storage_opt_ok: Optional[bool] = None
def _ensure_docker_available() -> None:
"""Best-effort check that the docker CLI is available before use.
Reuses ``find_docker()`` so this preflight stays consistent with the rest of
the Docker backend, including known non-PATH Docker Desktop locations.
"""
docker_exe = find_docker()
if not docker_exe:
logger.error(
"Docker backend selected but no docker executable was found in PATH "
"or known install locations. Install Docker Desktop and ensure the "
"CLI is available."
)
raise RuntimeError(
"Docker executable not found in PATH or known install locations. "
"Install Docker and ensure the 'docker' command is available."
)
try:
result = subprocess.run(
[docker_exe, "version"],
capture_output=True,
text=True,
timeout=5,
)
except FileNotFoundError:
logger.error(
"Docker backend selected but the resolved docker executable '%s' could "
"not be executed.",
docker_exe,
exc_info=True,
)
raise RuntimeError(
"Docker executable could not be executed. Check your Docker installation."
)
except subprocess.TimeoutExpired:
logger.error(
"Docker backend selected but '%s version' timed out. "
"The Docker daemon may not be running.",
docker_exe,
exc_info=True,
)
raise RuntimeError(
"Docker daemon is not responding. Ensure Docker is running and try again."
)
except Exception:
logger.error(
"Unexpected error while checking Docker availability.",
exc_info=True,
)
raise
else:
if result.returncode != 0:
logger.error(
"Docker backend selected but '%s version' failed "
"(exit code %d, stderr=%s)",
docker_exe,
result.returncode,
result.stderr.strip(),
)
raise RuntimeError(
"Docker command is available but 'docker version' failed. "
"Check your Docker installation."
)
class DockerEnvironment(BaseEnvironment):
"""Hardened Docker container execution with resource limits and persistence.
Security: all capabilities dropped, no privilege escalation, PID limits,
size-limited tmpfs for scratch dirs. The container itself is the security
boundary — the filesystem inside is writable so agents can install packages
(pip, npm, apt) as needed. Writable workspace via tmpfs or bind mounts.
Persistence: when enabled, bind mounts preserve /workspace and /root
across container restarts.
"""
def __init__(
self,
image: str,
cwd: str = "/root",
timeout: int = 60,
cpu: float = 0,
memory: int = 0,
disk: int = 0,
persistent_filesystem: bool = False,
task_id: str = "default",
volumes: list = None,
forward_env: list[str] | None = None,
env: dict | None = None,
network: bool = True,
host_cwd: str = None,
auto_mount_cwd: bool = False,
run_as_host_user: bool = False,
extra_args: list = None,
):
if cwd == "~":
cwd = "/root"
super().__init__(cwd=cwd, timeout=timeout)
self._persistent = persistent_filesystem
self._task_id = task_id
self._forward_env = _normalize_forward_env_names(forward_env)
self._env = _normalize_env_dict(env)
self._container_id: Optional[str] = None
logger.info(f"DockerEnvironment volumes: {volumes}")
if volumes is not None and not isinstance(volumes, list):
logger.warning(f"docker_volumes config is not a list: {volumes!r}")
volumes = []
_ensure_docker_available()
resource_args = []
if cpu > 0:
resource_args.extend(["--cpus", str(cpu)])
if memory > 0:
resource_args.extend(["--memory", f"{memory}m"])
if disk > 0 and sys.platform != "darwin":
if self._storage_opt_supported():
resource_args.extend(["--storage-opt", f"size={disk}m"])
else:
logger.warning(
"Docker storage driver does not support per-container disk limits "
"(requires overlay2 on XFS with pquota). Container will run without disk quota."
)
if not network:
resource_args.append("--network=none")
from tools.environments.base import get_sandbox_dir
volume_args = []
workspace_explicitly_mounted = False
for vol in (volumes or []):
if not isinstance(vol, str):
logger.warning(f"Docker volume entry is not a string: {vol!r}")
continue
vol = vol.strip()
if not vol:
continue
if ":" in vol:
volume_args.extend(["-v", vol])
if ":/workspace" in vol:
workspace_explicitly_mounted = True
else:
logger.warning(f"Docker volume '{vol}' missing colon, skipping")
host_cwd_abs = os.path.abspath(os.path.expanduser(host_cwd)) if host_cwd else ""
bind_host_cwd = (
auto_mount_cwd
and bool(host_cwd_abs)
and os.path.isdir(host_cwd_abs)
and not workspace_explicitly_mounted
)
if auto_mount_cwd and host_cwd and not os.path.isdir(host_cwd_abs):
logger.debug(f"Skipping docker cwd mount: host_cwd is not a valid directory: {host_cwd}")
self._workspace_dir: Optional[str] = None
self._home_dir: Optional[str] = None
writable_args = []
if self._persistent:
sandbox = get_sandbox_dir() / "docker" / task_id
self._home_dir = str(sandbox / "home")
os.makedirs(self._home_dir, exist_ok=True)
writable_args.extend([
"-v", f"{self._home_dir}:/root",
])
if not bind_host_cwd and not workspace_explicitly_mounted:
self._workspace_dir = str(sandbox / "workspace")
os.makedirs(self._workspace_dir, exist_ok=True)
writable_args.extend([
"-v", f"{self._workspace_dir}:/workspace",
])
else:
if not bind_host_cwd and not workspace_explicitly_mounted:
writable_args.extend([
"--tmpfs", "/workspace:rw,exec,size=10g",
])
writable_args.extend([
"--tmpfs", "/home:rw,exec,size=1g",
"--tmpfs", "/root:rw,exec,size=1g",
])
if bind_host_cwd:
logger.info(f"Mounting configured host cwd to /workspace: {host_cwd_abs}")
volume_args = ["-v", f"{host_cwd_abs}:/workspace", *volume_args]
elif workspace_explicitly_mounted:
logger.debug("Skipping docker cwd mount: /workspace already mounted by user config")
try:
from tools.credential_files import (
get_credential_file_mounts,
get_skills_directory_mount,
get_cache_directory_mounts,
)
for mount_entry in get_credential_file_mounts():
volume_args.extend([
"-v",
f"{mount_entry['host_path']}:{mount_entry['container_path']}:ro",
])
logger.info(
"Docker: mounting credential %s -> %s",
mount_entry["host_path"],
mount_entry["container_path"],
)
for skills_mount in get_skills_directory_mount():
volume_args.extend([
"-v",
f"{skills_mount['host_path']}:{skills_mount['container_path']}:ro",
])
logger.info(
"Docker: mounting skills dir %s -> %s",
skills_mount["host_path"],
skills_mount["container_path"],
)
for cache_mount in get_cache_directory_mounts():
volume_args.extend([
"-v",
f"{cache_mount['host_path']}:{cache_mount['container_path']}:ro",
])
logger.info(
"Docker: mounting cache dir %s -> %s",
cache_mount["host_path"],
cache_mount["container_path"],
)
except Exception as e:
logger.debug("Docker: could not load credential file mounts: %s", e)
env_args = []
for key in sorted(self._env):
env_args.extend(["-e", f"{key}={self._env[key]}"])
user_args: list[str] = []
if run_as_host_user:
user_spec = _resolve_host_user_spec()
if user_spec is not None:
user_args = ["--user", user_spec]
logger.info("Docker: running container as host user %s", user_spec)
else:
logger.warning(
"docker_run_as_host_user is enabled but this platform does "
"not expose POSIX uid/gid; container will start as its "
"image default user."
)
security_args = _build_security_args(run_as_host_user and bool(user_args))
logger.info(f"Docker volume_args: {volume_args}")
validated_extra = []
for arg in (extra_args or []):
if not isinstance(arg, str):
logger.warning("Ignoring non-string docker_extra_args entry: %r", arg)
continue
validated_extra.append(arg)
all_run_args = (
security_args
+ user_args
+ writable_args
+ resource_args
+ volume_args
+ env_args
+ validated_extra
)
logger.info(f"Docker run_args: {all_run_args}")
self._docker_exe = find_docker() or "docker"
container_name = f"hermes-{uuid.uuid4().hex[:8]}"
run_cmd = [
self._docker_exe, "run", "-d",
"--init",
"--name", container_name,
"-w", cwd,
*all_run_args,
image,
"sleep", "infinity",
]
logger.debug(f"Starting container: {' '.join(run_cmd)}")
result = subprocess.run(
run_cmd,
capture_output=True,
text=True,
timeout=120,
check=True,
)
self._container_id = result.stdout.strip()
logger.info(f"Started container {container_name} ({self._container_id[:12]})")
self._init_env_args = self._build_init_env_args()
self.init_session()
def _build_init_env_args(self) -> list[str]:
"""Build -e KEY=VALUE args for injecting host env vars into init_session.
These are used once during init_session() so that export -p captures
them into the snapshot. Subsequent execute() calls don't need -e flags.
"""
exec_env: dict[str, str] = dict(self._env)
explicit_forward_keys = set(self._forward_env)
passthrough_keys: set[str] = set()
try:
from tools.env_passthrough import get_all_passthrough
passthrough_keys = set(get_all_passthrough())
except Exception:
pass
forward_keys = explicit_forward_keys | (passthrough_keys - _HERMES_PROVIDER_ENV_BLOCKLIST)
hermes_env = _load_hermes_env_vars() if forward_keys else {}
for key in sorted(forward_keys):
value = os.getenv(key)
if value is None:
value = hermes_env.get(key)
if value is not None:
exec_env[key] = value
args = []
for key in sorted(exec_env):
args.extend(["-e", f"{key}={exec_env[key]}"])
return args
def _run_bash(self, cmd_string: str, *, login: bool = False,
timeout: int = 120,
stdin_data: str | None = None) -> subprocess.Popen:
"""Spawn a bash process inside the Docker container."""
assert self._container_id, "Container not started"
cmd = [self._docker_exe, "exec"]
if stdin_data is not None:
cmd.append("-i")
if login:
cmd.extend(self._init_env_args)
cmd.extend([self._container_id])
if login:
cmd.extend(["bash", "-l", "-c", cmd_string])
else:
cmd.extend(["bash", "-c", cmd_string])
return _popen_bash(cmd, stdin_data)
@staticmethod
def _storage_opt_supported() -> bool:
"""Check if Docker's storage driver supports --storage-opt size=.
Only overlay2 on XFS with pquota supports per-container disk quotas.
Ubuntu (and most distros) default to ext4, where this flag errors out.
"""
global _storage_opt_ok
if _storage_opt_ok is not None:
return _storage_opt_ok
try:
docker = find_docker() or "docker"
result = subprocess.run(
[docker, "info", "--format", "{{.Driver}}"],
capture_output=True, text=True, timeout=10,
)
driver = result.stdout.strip().lower()
if driver != "overlay2":
_storage_opt_ok = False
return False
probe = subprocess.run(
[docker, "create", "--storage-opt", "size=1m", "hello-world"],
capture_output=True, text=True, timeout=15,
)
if probe.returncode == 0:
container_id = probe.stdout.strip()
if container_id:
subprocess.run([docker, "rm", container_id],
capture_output=True, timeout=5)
_storage_opt_ok = True
else:
_storage_opt_ok = False
except Exception:
_storage_opt_ok = False
logger.debug("Docker --storage-opt support: %s", _storage_opt_ok)
return _storage_opt_ok
def cleanup(self):
"""Stop and remove the container. Bind-mount dirs persist if persistent=True."""
if self._container_id:
try:
stop_cmd = (
f"(timeout 60 {self._docker_exe} stop {self._container_id} || "
f"{self._docker_exe} rm -f {self._container_id}) >/dev/null 2>&1 &"
)
subprocess.Popen(stop_cmd, shell=True)
except Exception as e:
logger.warning("Failed to stop container %s: %s", self._container_id, e)
if not self._persistent:
try:
subprocess.Popen(
f"sleep 3 && {self._docker_exe} rm -f {self._container_id} >/dev/null 2>&1 &",
shell=True,
)
except Exception:
pass
self._container_id = None
if not self._persistent:
for d in (self._workspace_dir, self._home_dir):
if d:
shutil.rmtree(d, ignore_errors=True)