wait_for_confident_result does not gracefully handle default confidence value if using ImageQuery object for detection

[petermuller]

Issue and replication

Hi team! I'm testing out the Groundlight API and came across a crash in my basic proof-of-concept script that I wanted to report. The core problem I found was when doing async analysis, I'm able to submit images fine, but then when I want to await the results, I'm hit with a type checking error:

edit: pip freeze shows me at version groundlight==0.21.3

Traceback (most recent call last):
  File "C:\path\to\my\groundlight_ws\groundlight_demo.py", line 40, in <module>
    results.append(gl.wait_for_confident_result(
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\my\groundlight_ws\venv\Lib\site-packages\groundlight\client.py", line 1002, in wait_for_confident_result
    return self._wait_for_result(image_query, condition=confidence_above_thresh, timeout_sec=timeout_sec)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\my\groundlight_ws\venv\Lib\site-packages\groundlight\client.py", line 1065, in _wait_for_result
    if condition(image_query):
       ^^^^^^^^^^^^^^^^^^^^^^
  File "C:\path\to\my\groundlight_ws\venv\Lib\site-packages\groundlight\internalapi.py", line 70, in iq_is_confident
    return iq.result.confidence >= confidence_threshold  # type: ignore
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: '>=' not supported between instances of 'float' and 'NoneType'

To get here, a barebones sample of what I've done is here:

import os
from groundlight import Groundlight, Detector

API_TOKEN = "mytoken"
os.environ["GROUNDLIGHT_API_TOKEN"] = API_TOKEN


gl = Groundlight()
det: Detector = gl.get_or_create_detector(
    name="my_binary_detector",
    query="Is the door open?"
)

queries = []
for i in range(1,5):  # submit 4 images to the detector
    print(f"submitting image s{i}.png")  # just using local filenames as strings
    queries.append(gl.ask_async(detector=det, image=f"s{i}.png"))


results = []
for q in queries:
    results.append(gl.wait_for_confident_result(
        #### PROBLEM HERE ### I'm not specifying a confidence value, which should default to the detector's value,
        #### and I am using the image query object instead of the query string ID.
        image_query=q
    ))

# <... further processing / not needed for issue replication ...>

Based on function signatures, the iq_is_confident function does not expect a None type in the confidence_value, and the wait_for_confident_result method will supply a None if the confidence_threshold is omitted and when the type of the image_query is not a string, as in it's an ImageQuery object.

---

After looking at the code more, it also looks like this method will override the user's confidence_threshold in the case that the string version of the ImageQuery is provided since that variable gets overridden here - I only bring that up because it runs counter to the docstring

Expected behavior

For my main bug: My anticipated behavior for this would be that my detector's threshold of 0.9 is used if I didn't specify a threshold
For the secondary one: Anticipated behavior is that the confidence threshold provided to wait_for_confident_result would be used instead of my detector's 0.9 threshold value if I provided the detector's string ID.

Thanks!