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

@ -105,6 +105,130 @@ def test_vsock_exec_client_upload_archive_round_trip(
assert stub.closed is True
def test_vsock_exec_client_shell_round_trip(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setattr(socket, "AF_VSOCK", 40, raising=False)
responses = [
json.dumps(
{
"shell_id": "shell-1",
"cwd": "/workspace",
"cols": 120,
"rows": 30,
"state": "running",
"started_at": 1.0,
"ended_at": None,
"exit_code": None,
}
).encode("utf-8"),
json.dumps(
{
"shell_id": "shell-1",
"cwd": "/workspace",
"cols": 120,
"rows": 30,
"state": "running",
"started_at": 1.0,
"ended_at": None,
"exit_code": None,
"cursor": 0,
"next_cursor": 12,
"output": "pyro$ pwd\n",
"truncated": False,
}
).encode("utf-8"),
json.dumps(
{
"shell_id": "shell-1",
"cwd": "/workspace",
"cols": 120,
"rows": 30,
"state": "running",
"started_at": 1.0,
"ended_at": None,
"exit_code": None,
"input_length": 3,
"append_newline": True,
}
).encode("utf-8"),
json.dumps(
{
"shell_id": "shell-1",
"cwd": "/workspace",
"cols": 120,
"rows": 30,
"state": "running",
"started_at": 1.0,
"ended_at": None,
"exit_code": None,
"signal": "INT",
}
).encode("utf-8"),
json.dumps(
{
"shell_id": "shell-1",
"cwd": "/workspace",
"cols": 120,
"rows": 30,
"state": "stopped",
"started_at": 1.0,
"ended_at": 2.0,
"exit_code": 0,
"closed": True,
}
).encode("utf-8"),
]
stubs = [StubSocket(response) for response in responses]
remaining = list(stubs)
def socket_factory(family: int, sock_type: int) -> StubSocket:
assert family == socket.AF_VSOCK
assert sock_type == socket.SOCK_STREAM
return remaining.pop(0)
client = VsockExecClient(socket_factory=socket_factory)
opened = client.open_shell(
1234,
5005,
shell_id="shell-1",
cwd="/workspace",
cols=120,
rows=30,
)
assert opened.shell_id == "shell-1"
read = client.read_shell(1234, 5005, shell_id="shell-1", cursor=0, max_chars=1024)
assert read.output == "pyro$ pwd\n"
write = client.write_shell(
1234,
5005,
shell_id="shell-1",
input_text="pwd",
append_newline=True,
)
assert write["input_length"] == 3
signaled = client.signal_shell(1234, 5005, shell_id="shell-1", signal_name="INT")
assert signaled["signal"] == "INT"
closed = client.close_shell(1234, 5005, shell_id="shell-1")
assert closed["closed"] is True
open_request = json.loads(stubs[0].sent.decode("utf-8").strip())
assert open_request["action"] == "open_shell"
assert open_request["shell_id"] == "shell-1"
def test_vsock_exec_client_raises_agent_error(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setattr(socket, "AF_VSOCK", 40, raising=False)
stub = StubSocket(b'{"error":"shell is unavailable"}')
client = VsockExecClient(socket_factory=lambda family, sock_type: stub)
with pytest.raises(RuntimeError, match="shell is unavailable"):
client.open_shell(
1234,
5005,
shell_id="shell-1",
cwd="/workspace",
cols=120,
rows=30,
)
def test_vsock_exec_client_rejects_bad_json(monkeypatch: pytest.MonkeyPatch) -> None:
monkeypatch.setattr(socket, "AF_VSOCK", 40, raising=False)
stub = StubSocket(b"[]")