[Bug]: Workarena L2 fails with

[xhluca]

Browsergym Version

0.14.1

Playwright Version

1.44.0

Operating System Type

Ubuntu

Operating System Version

Ubuntu 22.04 LTS (Jammy Jellyfish)

Affected Browsers

Chromium

What happened?

See error code

Reproduction Steps

I tried running workarena l2

Relevant Logs

2025-07-28 10:30:00,845 - <PID> - browsergym.experiments.loop - INFO - Running experiment GenericAgent-checkpoints_<model_name> on <task_name> in:
  agentlab_results/<timestamp>_<experiment_name>/<timestamp>_<experiment_name>
2025-07-28 10:30:00,846 - <PID> - root - WARNING - Model <checkpoint_path> not found in the pricing information, prices are set to 0. Maybe try upgrading langchain_community.
2025-07-28 10:30:00,889 - <PID> - browsergym.experiments.loop - DEBUG - Agent created.
2025-07-28 10:30:00,890 - <PID> - browsergym.experiments.loop - DEBUG - Environment created.
2025-07-28 10:30:00,891 - <PID> - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): <redacted>.service-now.com:443
2025-07-28 10:30:01,184 - <PID> - urllib3.connectionpool - DEBUG - https://<redacted>.service-now.com:443 "GET / HTTP/1.1" 200 None
...
...truncated middle of the log...
...
    html = parse_document(document_idx=0)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../browsergym/utils/obs.py", line 177, in parse_document
    soup = BeautifulSoup(html, "lxml")
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".../bs4/__init__.py", line 473, in __init__
    self._feed()
  File ".../bs4/__init__.py", line 658, in _feed
    self.builder.feed(self.markup)
  File ".../bs4/builder/_lxml.py", line 484, in feed
    self.parser.feed(markup)
  ...
  File ".../bs4/builder/_lxml.py", line 196, in _getNsTag
    namespace, name = tag[1:].split("}", 1)
                      ^^^^^^^^^^^^^^^^^^^^^
ValueError: not enough values to unpack (expected 2, got 1)

2025-07-28 10:30:19,423 - <PID> - browsergym.experiments.loop - INFO - Saving summary info.
2025-07-28 10:30:19,425 - <PID> - urllib3.connectionpool - DEBUG - Starting new HTTPS connection (1): <redacted>.service-now.com:443
2025-07-28 10:30:19,637 - <PID> - urllib3.connectionpool - DEBUG - https://<redacted>.service-now.com:443 "GET /api/now/table/problem?sysparm_query=sys_id%3D<redacted> HTTP/1.1" 200 None
2025-07-28 10:30:19,954 - <PID> - urllib3.connectionpool - DEBUG - https://<redacted>.service-now.com:443 "DELETE /api/now/table/problem/<redacted> HTTP/1.1" 204 0
...
(repeats of GET/DELETE redacted for brevity)
...
2025-07-28 10:30:21,760 - <PID> - root - DEBUG - Tearing down the task

Additional Context

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[imenelydiaker]

Hi Xing, I had similar issues with WorkArena L1 and lxml, try downgrading the version of lxml to less than 6.0.0. See: ServiceNow/BrowserGym#360

[aldro61]

Hi @xhluca, this seems like an issue that is happening externally to WorkArena at the moment when BrowserGym captures the observation. Tagging @optimass for help with this.