[bug] AsyncGuard fails to await when called remotely #1149
Description
Describe the bug
I'm unable to appropriately call an async guard on my server. When I define it as a regular guard it works, local and over the server. Also, if I define an async guard locally and await as I do in my app code that is attempting to use the server, it works.
guardrails-1 | /opt/venv/lib/python3.11/site-packages/guardrails_api/utils/handle_error.py:30: RuntimeWarning: coroutine 'AsyncGuard.parse' was never awaited
guardrails-1 | return HttpError(500, "Internal Server Error").to_dict(), 500
guardrails-1 | RuntimeWarning: Enable tracemalloc to get the object allocation traceback
guardrails-1 | ERROR:guardrails-api:'coroutine' object has no attribute 'call_id'
guardrails-1 | Traceback (most recent call last):
guardrails-1 | File "/opt/venv/lib/python3.11/site-packages/guardrails_api/utils/handle_error.py", line 13, in decorator
guardrails-1 | return fn(*args, **kwargs)
guardrails-1 | ^^^^^^^^^^^^^^^^^^^
guardrails-1 | File "/opt/venv/lib/python3.11/site-packages/guardrails_api/blueprints/guards.py", line 497, in validate
guardrails-1 | cache_key = f"{guard.name}-{result.call_id}"
guardrails-1 | ^^^^^^^^^^^^^^
guardrails-1 | AttributeError: 'coroutine' object has no attribute 'call_id'
To Reproduce
config.py in guardrails-lite-server/
from guardrails import AsyncGuard, Guard
from guardrails.hub import DetectPII, ToxicLanguage, ValidLength
# Custom Long Validators I tested without
# Create the combined guard
valid_question = (
AsyncGuard(name="valid_question")
.use(ValidLength, min=10, max=300, on_fail="noop")
.use(ToxicLanguage, threshold=0.5, validation_method="sentence", on_fail="noop")
.use(DetectPII, ["EMAIL_ADDRESS", "PHONE_NUMBER"], on_fail="noop")my app code
import guardrails as gr
gr.settings.use_server = True
question="Something "
valid_question = gr.AsyncGuard(name='valid_question')
result = await valid_question.validate(question, metadata={})
print(f"Validation passed: {result.validation_passed}")
print(result)
if not result.validation_passed:
if result.reask:
print("Validation errors (reask):")
for fail_result in result.reask.fail_results:
print(f"- {fail_result.error_message}")
else:
print("Validation filtered or refrained")Expected behavior
I would expect that I would need to define an async guard on my client and server in order for validators to run asynchronously. Or I would even understand if I needed to define an async guard on the server and then just a regular guard on the client. Which doesn't call the validators. But this is a server-side error.
Library version:
Version 0.5.15
Additional context
Add any other context about the problem here.