litellm/tests/unit/proxy/client/cli/test_global_options.py
Mateo Wang 355dea26e7
perf(cli): start lite without initializing the litellm SDK (#46803)
* perf(cli): start lite without initializing the litellm SDK

lite and litellm-proxy now enter through a small litellm_cli package that imports the CLI against an uninitialized litellm package and runs the SDK __init__ in place only when something first uses it, so commands start in a fraction of a second instead of loading about 3,000 modules first.

* perf(cli): import only the lite subcommand being run

Each lite subcommand now loads its module when Click looks it up, so
lite logout no longer imports the autoroute wizard, agents, or the other
command groups. importlib.import_module and attribute reads on the SDK
also initialize it on first use, and an already imported litellm is
kept as is.

* fix(cli): initialize the SDK on a star import and test command bodies in thin mode

A from litellm import * after lite loaded the CLI used to see only the names the thin package had, so it now runs the full SDK init first. The thin-vs-full parity test now also runs logout and whoami, not just --help

* fix(cli): keep the import hooks until the lazy SDK init finishes

A thread that imported SDK code after the init restored the interpreter
imports could hold that module's import lock while waiting for the init,
and the init then waited on the same lock, so both threads hung. Other
threads now wait for the init before they load any new SDK code

* fix(cli): retry a failed lazy SDK init on the next use instead of leaving it half-run

* test(cli): bind the patched credentials client and autoroute dir in an order-independent way
2026-10-17 04:53:32 +00:00

485 lines
20 KiB
Python

# stdlib imports
import json
import os
import sys
from pathlib import Path
from typing import Final
from unittest.mock import Mock, patch
import pytest
from click.testing import CliRunner
import litellm.proxy.client.cli
from litellm._version import version as litellm_version
from litellm.proxy.client.cli import cli, litellm_proxy_cli
from litellm.proxy.client.cli.main import LITELLM_PROXY_DEPRECATION_NOTICE
@pytest.fixture
def cli_runner():
return CliRunner()
def test_cli_version_flag(cli_runner):
"""Test that --version prints the correct version, server URL, and server version, and exits successfully"""
with (
patch(
"litellm.proxy.client.health.HealthManagementClient.get_server_version",
return_value="1.2.3",
),
patch.dict(os.environ, {"LITELLM_PROXY_URL": "http://localhost:4000"}),
):
result = cli_runner.invoke(cli, ["--version"])
assert result.exit_code == 0
assert f"LiteLLM Proxy CLI Version: {litellm_version}" in result.output
assert "LiteLLM Proxy Server URL: http://localhost:4000" in result.output
assert "LiteLLM Proxy Server Version: 1.2.3" in result.output
def test_cli_source_is_ascii_only():
"""Non-ASCII output (emoji, box-drawing chars) raises UnicodeEncodeError on legacy Windows
consoles (cp1252), so the whole CLI package must stay ASCII-only."""
cli_root = Path(litellm.proxy.client.cli.__file__).parent
offenders = [
f"{path.relative_to(cli_root)}:{line_number}: {line.strip()}"
for path in sorted(cli_root.rglob("*.py"))
for line_number, line in enumerate(path.read_text(encoding="utf-8").splitlines(), start=1)
if not line.isascii()
]
assert offenders == []
def test_base_url_trailing_slash_normalized(cli_runner):
"""A trailing slash on --base-url must not produce a double slash (e.g. '//sso/cli/start')."""
with (
patch("webbrowser.open"),
patch(
"requests.post",
return_value=Mock(
status_code=200,
json=Mock(
return_value={
"login_id": "cli-test-uuid",
"poll_secret": "poll-secret",
"user_code": "ABCD-EFGH",
}
),
raise_for_status=Mock(),
),
) as mock_post,
patch("requests.get", side_effect=ValueError("stop after start request")),
):
cli_runner.invoke(cli, ["--base-url", "https://gateway.example.com/", "login"])
mock_post.assert_called_once_with("https://gateway.example.com/sso/cli/start", timeout=10)
def test_cli_version_command(cli_runner):
"""Test that 'version' command prints the correct version, server URL, and server version, and exits successfully"""
with (
patch(
"litellm.proxy.client.health.HealthManagementClient.get_server_version",
return_value="1.2.3",
),
patch.dict(os.environ, {"LITELLM_PROXY_URL": "http://localhost:4000"}),
):
result = cli_runner.invoke(cli, ["version"])
assert result.exit_code == 0
assert f"LiteLLM Proxy CLI Version: {litellm_version}" in result.output
assert "LiteLLM Proxy Server URL: http://localhost:4000" in result.output
assert "LiteLLM Proxy Server Version: 1.2.3" in result.output
@pytest.fixture
def isolated_home(monkeypatch, tmp_path):
"""Point HOME at tmp_path so tests never touch the developer's real ~/.litellm."""
monkeypatch.setenv("HOME", str(tmp_path))
monkeypatch.setenv("USERPROFILE", str(tmp_path))
monkeypatch.delenv("LITELLM_PROXY_URL", raising=False)
monkeypatch.delenv("LITELLM_PROXY_API_KEY", raising=False)
return tmp_path
def _write_config_file(home: Path, config: dict[str, str]) -> None:
config_dir = home / ".litellm"
config_dir.mkdir(exist_ok=True)
(config_dir / "config.json").write_text(json.dumps(config))
def _invoke_version(cli_runner: CliRunner, *args: str):
with patch(
"litellm.proxy.client.health.HealthManagementClient.get_server_version",
return_value="1.2.3",
):
return cli_runner.invoke(cli, [*args, "version"])
def test_base_url_read_from_config_file(cli_runner, isolated_home):
"""base_url precedence: flag > env > config file > default."""
_write_config_file(isolated_home, {"base_url": "https://config-proxy.example.com"})
result = _invoke_version(cli_runner)
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: https://config-proxy.example.com" in result.output
def test_env_var_beats_config_file_base_url(cli_runner, isolated_home, monkeypatch):
_write_config_file(isolated_home, {"base_url": "https://config-proxy.example.com"})
monkeypatch.setenv("LITELLM_PROXY_URL", "http://env-proxy.example.com:5000")
result = _invoke_version(cli_runner)
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: http://env-proxy.example.com:5000" in result.output
def test_base_url_flag_beats_env_var_and_config_file(cli_runner, isolated_home, monkeypatch):
_write_config_file(isolated_home, {"base_url": "https://config-proxy.example.com"})
monkeypatch.setenv("LITELLM_PROXY_URL", "http://env-proxy.example.com:5000")
result = _invoke_version(cli_runner, "--base-url", "http://flag-proxy.example.com:9000")
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: http://flag-proxy.example.com:9000" in result.output
def test_default_base_url_unchanged_without_config_file(cli_runner, isolated_home):
result = _invoke_version(cli_runner)
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: http://localhost:4000" in result.output
def _write_login_token(home: Path, base_url: str, key: str | None) -> None:
token_dir = home / ".litellm"
token_dir.mkdir(exist_ok=True)
secret: Final = {} if key is None else {"key": key}
(token_dir / "token.json").write_text(json.dumps({"base_url": base_url, "user_id": "u-1", **secret}))
LOGIN_SERVER: Final = "https://login-gateway.example.com"
SHELL_VERSION_THEN_QUIT: Final = "version\nquit\n"
def test_bare_lite_connects_to_the_server_it_logged_in_to(cli_runner, isolated_home, requests_mock, secret_vault_factory):
_write_login_token(isolated_home, LOGIN_SERVER, "sk-from-login")
requests_mock.get(f"{LOGIN_SERVER}/health/readiness", json={"litellm_version": "1.2.3"})
context: Final[dict[str, object]] = {"secret_vault": secret_vault_factory()}
result = cli_runner.invoke(cli, [], obj=context, input=SHELL_VERSION_THEN_QUIT)
assert f"Connected to LiteLLM server: {LOGIN_SERVER}" in result.output
assert f"LiteLLM Proxy Server URL: {LOGIN_SERVER}" in result.output
assert requests_mock.last_request.headers.get("Authorization") == "Bearer sk-from-login"
assert context["base_url_explicit"] is False
def test_a_one_shot_command_never_goes_to_the_logged_in_server(
cli_runner, isolated_home, requests_mock, secret_vault_factory
):
_write_login_token(isolated_home, LOGIN_SERVER, "sk-from-login")
requests_mock.get("http://localhost:4000/health/readiness", json={"litellm_version": "1.2.3"})
result = cli_runner.invoke(cli, ["version"], obj={"secret_vault": secret_vault_factory()})
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: http://localhost:4000" in result.output
assert [request.url for request in requests_mock.request_history] == ["http://localhost:4000/health/readiness"]
@pytest.mark.parametrize(
("args", "env", "config", "expected_base_url"),
[
(["--base-url", "https://flag-proxy.example.com"], {}, {}, "https://flag-proxy.example.com"),
([], {"LITELLM_PROXY_URL": "https://env-proxy.example.com"}, {}, "https://env-proxy.example.com"),
([], {}, {"base_url": "https://config-proxy.example.com"}, "https://config-proxy.example.com"),
],
)
def test_a_named_server_beats_the_logged_in_one_and_never_gets_its_key(
cli_runner, isolated_home, monkeypatch, requests_mock, secret_vault_factory, args, env, config, expected_base_url
):
_write_login_token(isolated_home, LOGIN_SERVER, "sk-from-login")
_write_config_file(isolated_home, config)
for name, value in env.items():
monkeypatch.setenv(name, value)
requests_mock.get(f"{expected_base_url}/health/readiness", json={"litellm_version": "1.2.3"})
result = cli_runner.invoke(cli, args, obj={"secret_vault": secret_vault_factory()}, input=SHELL_VERSION_THEN_QUIT)
assert f"Connected to LiteLLM server: {expected_base_url}" in result.output
assert f"LiteLLM Proxy Server URL: {expected_base_url}" in result.output
assert requests_mock.last_request.headers.get("Authorization") is None
@pytest.mark.parametrize(
("args", "env"),
[
(["--api-key", "sk-named-key"], {}),
([], {"LITELLM_PROXY_API_KEY": "sk-named-key"}),
],
)
def test_a_key_given_without_a_server_never_goes_to_the_logged_in_one(
cli_runner, isolated_home, monkeypatch, requests_mock, secret_vault_factory, args, env
):
_write_login_token(isolated_home, LOGIN_SERVER, "sk-from-login")
for name, value in env.items():
monkeypatch.setenv(name, value)
requests_mock.get("http://localhost:4000/health/readiness", json={"litellm_version": "1.2.3"})
result = cli_runner.invoke(cli, args, obj={"secret_vault": secret_vault_factory()}, input=SHELL_VERSION_THEN_QUIT)
assert "Connected to LiteLLM server: http://localhost:4000" in result.output
assert [request.url for request in requests_mock.request_history] == ["http://localhost:4000/health/readiness"]
assert requests_mock.last_request.headers.get("Authorization") == "Bearer sk-named-key"
def test_bare_lite_says_why_it_stays_on_localhost_when_the_keychain_is_locked(
cli_runner, isolated_home, secret_vault_factory
):
_write_login_token(isolated_home, LOGIN_SERVER, None)
result = cli_runner.invoke(cli, [], obj={"secret_vault": secret_vault_factory(available=False)}, input="quit\n")
assert "Connected to LiteLLM server: http://localhost:4000" in result.output
assert f"could not be read. Unlock it, or run 'lite --base-url {LOGIN_SERVER} login'" in result.output
def test_lite_goes_back_to_localhost_after_logout_without_a_keychain(cli_runner, isolated_home, secret_vault_factory):
_write_login_token(isolated_home, LOGIN_SERVER, "sk-from-login")
no_keychain: Final = secret_vault_factory(available=False)
cli_runner.invoke(cli, ["logout"], obj={"secret_vault": no_keychain})
result = cli_runner.invoke(cli, [], obj={"secret_vault": no_keychain}, input="quit\n")
assert "Connected to LiteLLM server: http://localhost:4000" in result.output
def test_corrupt_config_file_falls_back_to_default(cli_runner, isolated_home):
"""A corrupt config file must never crash the CLI. Exactly one warning proves
the config file is read once per invocation, not once per lookup."""
config_dir = isolated_home / ".litellm"
config_dir.mkdir(exist_ok=True)
(config_dir / "config.json").write_text("{not json")
result = _invoke_version(cli_runner)
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: http://localhost:4000" in result.output
assert result.stderr.count("Warning: ignoring invalid config file") == 1
def test_empty_base_url_flag_is_not_treated_as_unset(cli_runner, isolated_home):
"""`--base-url ""` explicitly provided an (empty) value; falling back to the
config file or localhost would silently redirect auth-sensitive commands."""
_write_config_file(isolated_home, {"base_url": "https://config-proxy.example.com"})
result = _invoke_version(cli_runner, "--base-url", "")
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL:" not in result.output
def test_version_flag_reads_config_file_base_url(cli_runner, isolated_home):
"""--version resolves through the same precedence chain as every other command."""
_write_config_file(isolated_home, {"base_url": "https://config-proxy.example.com"})
with patch(
"litellm.proxy.client.health.HealthManagementClient.get_server_version",
return_value="1.2.3",
):
result = cli_runner.invoke(cli, ["--version"])
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: https://config-proxy.example.com" in result.output
def test_version_flag_prefers_env_var_over_config_file(cli_runner, isolated_home, monkeypatch):
_write_config_file(isolated_home, {"base_url": "https://config-proxy.example.com"})
monkeypatch.setenv("LITELLM_PROXY_URL", "http://env-proxy.example.com:5000")
with patch(
"litellm.proxy.client.health.HealthManagementClient.get_server_version",
return_value="1.2.3",
):
result = cli_runner.invoke(cli, ["--version"])
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: http://env-proxy.example.com:5000" in result.output
def test_version_flag_prefers_explicit_base_url_over_config_file(cli_runner, isolated_home):
"""An eager --version could not see the flag and silently queried the config
server instead of the one the user named."""
_write_config_file(isolated_home, {"base_url": "https://config-proxy.example.com"})
with patch(
"litellm.proxy.client.health.HealthManagementClient.get_server_version",
return_value="1.2.3",
):
result = cli_runner.invoke(cli, ["--base-url", "https://flag-proxy.example.com", "--version"])
assert result.exit_code == 0
assert "LiteLLM Proxy Server URL: https://flag-proxy.example.com" in result.output
assert "config-proxy.example.com" not in result.output
def test_version_flag_never_sends_api_key_to_unnamed_server(cli_runner, isolated_home, monkeypatch):
"""The version request carries a bearer token; it must reach only the server the
user named, never whichever host happens to sit in the config file."""
_write_config_file(isolated_home, {"base_url": "https://config-proxy.example.com"})
monkeypatch.setenv("LITELLM_PROXY_API_KEY", "sk-intended-for-flag-proxy")
with patch("litellm.proxy.client.http_client.requests.request") as mock_request:
mock_request.return_value.json.return_value = {"litellm_version": "1.2.3"}
mock_request.return_value.raise_for_status.return_value = None
result = cli_runner.invoke(cli, ["--base-url", "https://flag-proxy.example.com", "--version"])
assert result.exit_code == 0
requested_urls = [call.kwargs["url"] for call in mock_request.call_args_list]
assert requested_urls
assert all(url.startswith("https://flag-proxy.example.com") for url in requested_urls)
sent_keys = [call.kwargs["headers"].get("Authorization") for call in mock_request.call_args_list]
assert sent_keys == ["Bearer sk-intended-for-flag-proxy"] * len(requested_urls)
def test_litellm_proxy_entrypoint_prints_deprecation_notice_on_stderr_and_still_runs(monkeypatch, capsys, requests_mock):
requests_mock.get("http://localhost:4000/health/readiness", json={"litellm_version": "1.2.3"})
monkeypatch.setattr(sys, "argv", ["litellm-proxy", "--version"])
monkeypatch.setenv("LITELLM_PROXY_URL", "http://localhost:4000")
with pytest.raises(SystemExit) as exit_info:
litellm_proxy_cli()
captured: Final = capsys.readouterr()
assert exit_info.value.code == 0
assert captured.err.strip() == LITELLM_PROXY_DEPRECATION_NOTICE
assert f"LiteLLM Proxy CLI Version: {litellm_version}" in captured.out
assert "LiteLLM Proxy Server Version: 1.2.3" in captured.out
assert "deprecated" not in captured.out
def test_lite_entrypoint_prints_nothing_on_stderr(monkeypatch, capsys, requests_mock):
requests_mock.get("http://localhost:4000/health/readiness", json={"litellm_version": "1.2.3"})
monkeypatch.setattr(sys, "argv", ["lite", "--version"])
monkeypatch.setenv("LITELLM_PROXY_URL", "http://localhost:4000")
with pytest.raises(SystemExit) as exit_info:
cli()
captured: Final = capsys.readouterr()
assert exit_info.value.code == 0
assert "LiteLLM Proxy Server Version: 1.2.3" in captured.out
assert f"LiteLLM Proxy CLI Version: {litellm_version}" in captured.out
assert captured.err == ""
@pytest.mark.parametrize(
("args", "config", "expected_context", "expected_authorization"),
[
(
["--base-url", "https://flag-proxy.example.com/", "--api-key", "sk-flag"],
{},
{
"base_url": "https://flag-proxy.example.com",
"api_key": "sk-flag",
"api_key_from_token_file": False,
"base_url_explicit": True,
},
"Bearer sk-flag",
),
(
[],
{},
{
"base_url": "http://localhost:4000",
"api_key": None,
"api_key_from_token_file": True,
"base_url_explicit": False,
},
None,
),
(
["--api-key", ""],
{"base_url": "https://config-proxy.example.com"},
{
"base_url": "https://config-proxy.example.com",
"api_key": "",
"api_key_from_token_file": False,
"base_url_explicit": True,
},
None,
),
],
)
def test_cli_stores_the_resolved_connection_on_the_supplied_context_object(
cli_runner, isolated_home, requests_mock, args, config, expected_context, expected_authorization
):
_write_config_file(isolated_home, config)
requests_mock.get(f"{expected_context['base_url']}/health/readiness", json={"litellm_version": "1.2.3"})
context: Final[dict[str, object]] = {}
result = cli_runner.invoke(cli, [*args, "version"], obj=context)
assert result.exit_code == 0
assert json.dumps(context) == json.dumps(expected_context)
assert "LiteLLM Proxy Server Version: 1.2.3" in result.output
assert requests_mock.last_request.headers.get("Authorization") == expected_authorization
@pytest.mark.parametrize(
("context", "expected_lines", "expected_authorization"),
[
(
{"base_url": "http://localhost:4000", "api_key": "sk-context"},
["LiteLLM Proxy Server URL: http://localhost:4000", "LiteLLM Proxy Server Version: 1.2.3"],
"Bearer sk-context",
),
(
{"base_url": "http://localhost:4000"},
["LiteLLM Proxy Server URL: http://localhost:4000", "LiteLLM Proxy Server Version: 1.2.3"],
None,
),
(
{"base_url": "http://localhost:4000", "api_key": None, "api_key_from_token_file": True},
["LiteLLM Proxy Server URL: http://localhost:4000", "LiteLLM Proxy Server Version: 1.2.3"],
None,
),
],
)
def test_version_command_reads_the_connection_from_the_context_object(
cli_runner, requests_mock, context, expected_lines, expected_authorization
):
requests_mock.get("http://localhost:4000/health/readiness", json={"litellm_version": "1.2.3"})
result = cli_runner.invoke(cli.commands["version"], [], obj=context)
assert result.exit_code == 0
assert result.output.splitlines() == [f"LiteLLM Proxy CLI Version: {litellm_version}", *expected_lines]
assert requests_mock.last_request.headers.get("Authorization") == expected_authorization
def test_version_command_without_a_base_url_reports_the_failed_lookup(cli_runner):
result = cli_runner.invoke(cli.commands["version"], [], obj={"api_key": "sk-context"})
lines: Final = result.output.splitlines()
assert result.exit_code == 0
assert len(lines) == 2
assert lines[0] == f"LiteLLM Proxy CLI Version: {litellm_version}"
assert lines[1].startswith("Could not retrieve server version: ")
def test_version_command_without_a_context_object_fails_reading_it(cli_runner):
result = cli_runner.invoke(cli.commands["version"], [])
assert result.exit_code == 1
assert type(result.exception) is AttributeError
assert result.output == ""
def test_every_listed_command_runs_the_command_of_that_name():
from litellm.proxy.client.cli.commands.agents import agent_commands
commands: Final = {name: cli.commands[name] for name in cli.commands}
assert {name: command.name for name, command in commands.items()} == {name: name for name in commands}
assert {command.name for command in agent_commands()} <= commands.keys()
assert all(cli.commands[name] is command for name, command in commands.items())