Error Executing llama_index_quickstart.ipynb
Closed this issue · 4 comments
Good morning everyone,
please can someone help me to fix this ? I am having an issue running this block of code
# or as context manager
with tru_query_engine_recorder as recording:
query_engine.query("What did the author do growing up?")
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
Cell In[14], line 2
1 # or as context manager
----> 2 with tru_query_engine_recorder as recording:
3 query_engine.query("What did the author do growing up?")
File ~/anaconda3/envs/llamaindex/lib/python3.11/site-packages/trulens_eval/app.py:739, in App.__exit__(self, exc_type, exc_value, exc_tb)
736 self.recording_contexts.reset(ctx.token)
738 if exc_type is not None:
--> 739 raise exc_value
741 return
Cell In[14], line 3
1 # or as context manager
2 with tru_query_engine_recorder as recording:
----> 3 query_engine.query("What did the author do growing up?")
File ~/anaconda3/envs/llamaindex/lib/python3.11/site-packages/trulens_eval/instruments.py:784, in Instrument.tracked_method_wrapper.<locals>.tru_wrapper(*args, **kwargs)
778 # If stack has only 1 thing on it, we are looking at a "root
779 # call". Create a record of the result and notify the app:
781 if len(stack) == 1:
782 # If this is a root call, notify app to add the completed record
783 # into its containers:
--> 784 ctx.app._on_add_record(
785 ctx=ctx,
786 func=func,
787 sig=sig,
788 bindings=bindings,
789 ret=rets,
790 error=error,
791 perf=Perf(start_time=start_time, end_time=end_time),
792 cost=cost
793 )
795 if error is not None:
796 raise error
File ~/anaconda3/envs/llamaindex/lib/python3.11/site-packages/trulens_eval/app.py:782, in App._on_add_record(self, ctx, func, sig, bindings, ret, error, perf, cost)
769 return Record(
770 main_input=jsonify(main_in),
771 main_output=jsonify(main_out),
(...)
778 meta=jsonify(record_metadata)
779 )
781 # Finishing record needs to be done in a thread lock, done there:
--> 782 record = ctx.finish_record(build_record)
784 if error is not None:
785 # May block on DB.
786 self._handle_error(record=record, error=error)
File ~/anaconda3/envs/llamaindex/lib/python3.11/site-packages/trulens_eval/app.py:381, in RecordingContext.finish_record(self, calls_to_record)
375 """
376 Run the given function to build a record from the tracked calls and any
377 pre-specified metadata.
378 """
380 with self.lock:
--> 381 record = calls_to_record(self.calls, self.record_metadata)
382 self.calls = []
383 self.records.append(record)
File ~/anaconda3/envs/llamaindex/lib/python3.11/site-packages/trulens_eval/app.py:766, in App._on_add_record.<locals>.build_record(calls, record_metadata)
762 calls = list(calls)
764 assert len(calls) > 0, "No information recorded in call."
--> 766 main_in = self.main_input(func, sig, bindings)
767 main_out = self.main_output(func, sig, bindings, ret)
769 return Record(
770 main_input=jsonify(main_in),
771 main_output=jsonify(main_out),
(...)
778 meta=jsonify(record_metadata)
779 )
File ~/anaconda3/envs/llamaindex/lib/python3.11/site-packages/trulens_eval/tru_llama.py:279, in TruLlama.main_input(self, func, sig, bindings)
277 str_or_bundle = bindings.arguments['str_or_query_bundle']
278 if isinstance(str_or_bundle, QueryBundle):
--> 279 return str_or_bundle.query_str
280 else:
281 return str_or_bundle
AttributeError: 'str' object has no attribute 'query_str'
I am using trulens 0.20.3 and llama-index-core 0.10.23.post1
Hey @Anderson2992, back at it with the coding conundrums, I see! Good to have you reaching out again. Hope this one hasn't been too much of a head-scratcher for you!
Based on the information you've provided and the similar issues I found in the trulens repository, it seems that the error you're encountering is due to a mismatch between the expected QueryBundle object and the string that's being passed in your code snippet.
The error AttributeError: 'str' object has no attribute 'query_str' suggests that the code is attempting to access an attribute query_str on a string object, which obviously does not have this attribute. This error occurs in the context of handling str_or_query_bundle in the provided code snippet. The logic checks if str_or_bundle is an instance of QueryBundle and if so, attempts to return str_or_bundle.query_str. However, if str_or_bundle is not a QueryBundle (i.e., it is a string), the code just returns str_or_bundle itself.
To resolve this issue, ensure that wherever str_or_query_bundle is being set or passed into the function that leads to this code execution, it correctly distinguishes between strings and QueryBundle objects. If a string is passed when a QueryBundle is expected, you'll need to adjust the logic to either convert the string to a QueryBundle object (if feasible and logical in your application context) or ensure that a QueryBundle object is passed in the first place.
Here's a hypothetical fix that involves checking the type of str_or_bundle and handling the string case differently if you have a way to convert a string to a QueryBundle or handle it differently:
if 'str_or_query_bundle' in bindings.arguments:
# llama_index specific
str_or_bundle = bindings.arguments['str_or_query_bundle']
if isinstance(str_or_bundle, QueryBundle):
return str_or_bundle.query_str
elif isinstance(str_or_bundle, str):
# Handle the string case here. This could involve converting the string to a QueryBundle
# if your application logic allows for such a conversion.
# Example:
# query_bundle = convert_string_to_query_bundle(str_or_bundle)
# return query_bundle.query_str
# OR simply return the string if that's the intended behavior
return str_or_bundle
else:
# Handle other types if necessary
passPlease note that the convert_string_to_query_bundle function is hypothetical and would need to be implemented based on your application's logic and requirements.
Ensure that any conversion logic or handling of strings vs. QueryBundle objects aligns with the intended use of your application to avoid introducing new bugs.
In addition to the above, I found a similar closed issue in the trulens repository: Llama Index quick start - str object has no attribute query_str. The suggested solution in this issue was to upgrade to the latest release of TruLens (0.24.0) which now supports llama-index versions greater than 0.10. This update should resolve the AttributeError: 'str' object has no attribute 'query_str'.
Given that you're using trulens version 0.20.3, it might be worth trying to upgrade to the latest version (trulens-eval-0.26.0 at the time of writing) and see if that resolves your issue.
I hope this helps! If you have any further questions or if the issue persists after trying the above solutions, please don't hesitate to ask.
Sources
- [BUG]: When attribute is in String instead of response format , it gives error:
- Llama Index quick start - str object has no attribute query_str
- trulens_eval/trulens_eval/tru_llama.py
About Dosu
This response is meant to be useful and save you time. It is not meant to be a precise solution, but rather a starting point for your own research.
Help me be more useful! Please leave a 👍 if this is helpful and 👎 if it is irrelevant.
If you want to continue the conversation, start your reply with @dosu-bot.
Hey @Anderson2992 - can you update your trulens version?
pip install -U trulens_eval
Hi @joshreini1
Thank you for your help, I have updated my trulens version and the code is working now
Many thanks
Thanks for the update @Anderson2992 !