Langtrace crash when instrumenting AutoGen llm_config

[hengwen1]

Hi,
Langtrace crashed when instrumenting AutoGen. It cause by llm_config can be either False, None, or a single llm config dict or dict contains a list of configs in config_list. Only the last case is handled in current code. Please add code to handle other cases. Also, is it possible to add AG2 support in Langtrace which is a recent spin off from AutoGen with different python module name. Thank you!

[alizenhom]

Thanks @hengwen1 for pointing out the issue, it should be an easy fix will create a new release and keep you posted, apologies for the inconvenience 🙏

[alizenhom]

Also regarding the AG2 support will look into right away and keep you posted 🙏

[alizenhom]

Hey @hengwen1, the fix is now released in #447 -- kindly upgrade the python sdk to 3.3.19 and the issue should be resolved

would be more than happy to address any issues if still persists!
Currently adding support to AG2 will keep you posted!

[hengwen1]

Hi @alizenhom Thank you for the quick fix. It works for me now. Brilliant!

[hengwen1]

I think there is still problem. AutoGen has cache. It won't create llm client if cache is hit. The back trace is like below. Patch code need handle the case. Thank you!

File "/home/test/miniconda3/envs/autogen/lib/python3.12/site-packages/autogen/agentchat/conversable_agent.py", line 952, in receive
reply = self.generate_reply(messages=self.chat_messages[sender], sender=sender)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/test/miniconda3/envs/autogen/lib/python3.12/site-packages/langtrace_python_sdk/instrumentation/autogen/patch.py", line 82, in traced_method
result = wrapped(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/test/miniconda3/envs/autogen/lib/python3.12/site-packages/autogen/agentchat/conversable_agent.py", line 2112, in generate_reply
final, reply = reply_func(self, messages=messages, sender=sender, config=reply_func_tuple["config"])
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/test/miniconda3/envs/autogen/lib/python3.12/site-packages/autogen/agentchat/groupchat.py", line 1174, in run_chat
reply = speaker.generate_reply(sender=self)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/test/miniconda3/envs/autogen/lib/python3.12/site-packages/langtrace_python_sdk/instrumentation/autogen/patch.py", line 95, in traced_method
total_cost, response_model = list(instance.get_total_usage().keys())
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'keys'

[alizenhom]

Thanks for pointing out the issue @hengwen1, and apologies again for the inconvenience.

It should be an easy fix will keep you posted

[alizenhom]

Hey @hengwen1, this release #448 should contain the fix for your issue, can you kindly upgrade the sdk and try again,

will be more than happy to help if the issue persists

[hengwen1]

Verified. Thank you!