Add persistent workspace shell sessions

Let agents inhabit a workspace across separate calls instead of only submitting one-shot execs.

Add workspace shell open/read/write/signal/close across the CLI, Python SDK, and MCP server, with persisted shell records, a local PTY-backed mock implementation, and guest-agent support for real Firecracker workspaces.

Mark the 2.5.0 roadmap milestone done, refresh docs/examples and the release metadata, and verify with uv lock, UV_CACHE_DIR=.uv-cache make check, and UV_CACHE_DIR=.uv-cache make dist-check.
This commit is contained in:
Thales Maciel 2026-03-12 02:31:57 -03:00
parent 2de31306b6
commit 3f8293ad24
28 changed files with 3265 additions and 81 deletions

View file

@ -454,6 +454,73 @@ def test_workspace_sync_push_rejects_destination_outside_workspace(tmp_path: Pat
manager.push_workspace_sync(workspace_id, source_path=source_dir, dest="../escape")
def test_workspace_shell_lifecycle_and_rehydration(tmp_path: Path) -> None:
manager = VmManager(
backend_name="mock",
base_dir=tmp_path / "vms",
network_manager=TapNetworkManager(enabled=False),
)
created = manager.create_workspace(
environment="debian:12-base",
allow_host_compat=True,
)
workspace_id = str(created["workspace_id"])
opened = manager.open_shell(workspace_id)
shell_id = str(opened["shell_id"])
assert opened["state"] == "running"
manager.write_shell(workspace_id, shell_id, input_text="pwd")
output = ""
deadline = time.time() + 5
while time.time() < deadline:
read = manager.read_shell(workspace_id, shell_id, cursor=0, max_chars=65536)
output = str(read["output"])
if "/workspace" in output:
break
time.sleep(0.05)
assert "/workspace" in output
manager_rehydrated = VmManager(
backend_name="mock",
base_dir=tmp_path / "vms",
network_manager=TapNetworkManager(enabled=False),
)
second_opened = manager_rehydrated.open_shell(workspace_id)
second_shell_id = str(second_opened["shell_id"])
assert second_shell_id != shell_id
manager_rehydrated.write_shell(workspace_id, second_shell_id, input_text="printf 'ok\\n'")
second_output = ""
deadline = time.time() + 5
while time.time() < deadline:
read = manager_rehydrated.read_shell(
workspace_id,
second_shell_id,
cursor=0,
max_chars=65536,
)
second_output = str(read["output"])
if "ok" in second_output:
break
time.sleep(0.05)
assert "ok" in second_output
logs = manager.logs_workspace(workspace_id)
assert logs["count"] == 0
closed = manager.close_shell(workspace_id, shell_id)
assert closed["closed"] is True
with pytest.raises(ValueError, match="does not exist"):
manager.read_shell(workspace_id, shell_id)
deleted = manager.delete_workspace(workspace_id)
assert deleted["deleted"] is True
with pytest.raises(ValueError, match="does not exist"):
manager_rehydrated.read_shell(workspace_id, second_shell_id)
def test_workspace_create_rejects_unsafe_seed_archive(tmp_path: Path) -> None:
archive_path = tmp_path / "bad.tgz"
with tarfile.open(archive_path, "w:gz") as archive: