arcade evals Default to localhost (#172)
# PR Description * `arcade evals` now run evaluations against Arcade Engine at `http://localhost:9099` by default. * Added optional flag `--cloud` to run evaluations against Arcade's Cloud Engine at `https://api.arcade-ai.com`. Overrides `-h` flag. * Always print the Engine that the evaluations are using. Previously this was reserved for `-d` flag.
This commit is contained in:
parent
d8c8b060af
commit
b12ceec4b5
3 changed files with 22 additions and 14 deletions
|
|
@ -1,5 +1,6 @@
|
|||
DEFAULT_CLOUD_HOST = "cloud.arcade-ai.com"
|
||||
DEFAULT_ENGINE_HOST = "api.arcade-ai.com"
|
||||
LOCALHOST = "localhost"
|
||||
|
||||
_style_block = b"""
|
||||
<link rel="icon" href="https://cdn.arcade-ai.com/favicons/favicon.ico" sizes="any">
|
||||
|
|
|
|||
|
|
@ -14,7 +14,7 @@ from rich.markup import escape
|
|||
from rich.text import Text
|
||||
|
||||
from arcade.cli.authn import LocalAuthCallbackServer, check_existing_login
|
||||
from arcade.cli.constants import DEFAULT_CLOUD_HOST, DEFAULT_ENGINE_HOST
|
||||
from arcade.cli.constants import DEFAULT_CLOUD_HOST, DEFAULT_ENGINE_HOST, LOCALHOST
|
||||
from arcade.cli.display import (
|
||||
display_arcade_chat_header,
|
||||
display_eval_results,
|
||||
|
|
@ -345,11 +345,16 @@ def evals(
|
|||
help="The models to use for evaluation (default: gpt-4o)",
|
||||
),
|
||||
host: str = typer.Option(
|
||||
DEFAULT_ENGINE_HOST,
|
||||
LOCALHOST,
|
||||
"-h",
|
||||
"--host",
|
||||
help="The Arcade Engine address to send chat requests to.",
|
||||
),
|
||||
cloud: bool = typer.Option(
|
||||
False,
|
||||
"--cloud",
|
||||
help="Whether to run evaluations against the Arcade Cloud Engine. Overrides the 'host' option.",
|
||||
),
|
||||
port: int = typer.Option(
|
||||
None,
|
||||
"-p",
|
||||
|
|
@ -372,6 +377,8 @@ def evals(
|
|||
execute any functions decorated with @tool_eval, and display the results.
|
||||
"""
|
||||
config = validate_and_get_config()
|
||||
|
||||
host = DEFAULT_ENGINE_HOST if cloud else host
|
||||
base_url = compute_engine_base_url(force_tls, force_no_tls, host, port)
|
||||
|
||||
models_list = models.split(",") # Use 'models_list' to avoid shadowing
|
||||
|
|
@ -380,13 +387,12 @@ def evals(
|
|||
if not eval_files:
|
||||
return
|
||||
|
||||
if show_details:
|
||||
console.print(
|
||||
Text.assemble(
|
||||
("\nRunning evaluations against Arcade Engine at ", "bold"),
|
||||
(base_url, "bold blue"),
|
||||
)
|
||||
console.print(
|
||||
Text.assemble(
|
||||
("\nRunning evaluations against Arcade Engine at ", "bold"),
|
||||
(base_url, "bold blue"),
|
||||
)
|
||||
)
|
||||
|
||||
# Try to hit /health endpoint on engine and warn if it is down
|
||||
with Arcade(api_key=config.api.key, base_url=base_url) as client:
|
||||
|
|
|
|||
|
|
@ -24,6 +24,7 @@ from rich.text import Text
|
|||
from typer.core import TyperGroup
|
||||
from typer.models import Context
|
||||
|
||||
from arcade.cli.constants import LOCALHOST
|
||||
from arcade.core.config_model import Config
|
||||
from arcade.core.errors import ToolkitLoadError
|
||||
from arcade.core.schema import ToolDefinition
|
||||
|
|
@ -116,7 +117,7 @@ def compute_engine_base_url(
|
|||
str: The fully constructed URL for the Arcade Engine.
|
||||
"""
|
||||
# "Use 127.0.0.1" and "0.0.0.0" as aliases for "localhost"
|
||||
host = "localhost" if host in ["127.0.0.1", "0.0.0.0"] else host # noqa: S104
|
||||
host = LOCALHOST if host in ["127.0.0.1", "0.0.0.0"] else host # noqa: S104
|
||||
|
||||
# Determine TLS setting based on input flags
|
||||
if force_no_tls:
|
||||
|
|
@ -124,10 +125,10 @@ def compute_engine_base_url(
|
|||
elif force_tls:
|
||||
is_tls = True
|
||||
else:
|
||||
is_tls = host != "localhost"
|
||||
is_tls = host != LOCALHOST
|
||||
|
||||
# "localhost" defaults to dev port if not specified
|
||||
if host == "localhost" and port is None:
|
||||
if host == LOCALHOST and port is None:
|
||||
port = 9099
|
||||
|
||||
protocol = "https" if is_tls else "http"
|
||||
|
|
@ -170,14 +171,14 @@ def compute_login_url(host: str, state: str, port: int | None) -> str:
|
|||
"""
|
||||
Compute the full URL for the CLI login endpoint.
|
||||
"""
|
||||
callback_uri = "http://localhost:9905/callback"
|
||||
callback_uri = f"http://{LOCALHOST}:9905/callback"
|
||||
params = urlencode({"callback_uri": callback_uri, "state": state})
|
||||
|
||||
port = port if port else 8000
|
||||
|
||||
login_base_url = (
|
||||
f"http://localhost:{port}"
|
||||
if host in ["localhost", "127.0.0.1", "0.0.0.0"] # noqa: S104
|
||||
f"http://{LOCALHOST}:{port}"
|
||||
if host in [LOCALHOST, "127.0.0.1", "0.0.0.0"] # noqa: S104
|
||||
else f"https://{host}"
|
||||
)
|
||||
endpoint = "/api/v1/auth/cli_login"
|
||||
|
|
|
|||
Loading…
Reference in a new issue