In the notebook, M2_UGL_1When I experiment with changing the generation or reflection model in the workflow, I get this error if I try to use ‘claude-3–7-sonnet’ for the reflection model:
PermissionDeniedError: Error code: 403 - {'error': {'message': 'using claude-3-7-sonnet is invalid'}}
What Claude models are available to try?
Looks like using “claude-3-7-sonnet-latest” as model name would work.
reflection_model=“claude-3-7-sonnet-latest” #“o4-mini”
@greenvinyl can you give it a try on your side?
I had the same issue. On the notebook it says:
“Use a stronger reasoning model for reflection (gpt-4.1 or claude-3-7-sonnet).”
I tried with
reflection_model=“claude-3-7-sonnet”
and returned an error. The error included this page Model deprecations - Claude Docs
I tried with:
reflection_model=“claude-3-7-sonnet-latest”
and
reflection_model=“claude-3-7-sonnet-20250219”
and on both cases it worked alright.
@hguan Thank you - both your suggestions worked fine.
@Lluichi Thank you - both your suggestions worked for me too.
While changing the generation model from original “gpt-4o-mini” to “gpt-3.5-turbo” got the following error:
====
Step 2: Executing chart code (V1)… 
---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
File /usr/local/lib/python3.11/site-packages/pandas/core/indexes/base.py:3812, in Index.get_loc(self, key)
3811 try:
-> 3812 return self._engine.get_loc(casted_key)
3813 except KeyError as err:
File pandas/_libs/index.pyx:167, in pandas._libs.index.IndexEngine.get_loc()
File pandas/_libs/index.pyx:196, in pandas._libs.index.IndexEngine.get_loc()
File pandas/_libs/hashtable_class_helper.pxi:7088, in pandas._libs.hashtable.PyObjectHashTable.get_item()
File pandas/_libs/hashtable_class_helper.pxi:7096, in pandas._libs.hashtable.PyObjectHashTable.get_item()
KeyError: 'year'
The above exception was the direct cause of the following exception:
KeyError Traceback (most recent call last)
Cell In[26], line 8
5 image_basename="drink_sales"
7 # Run the complete agentic workflow
----> 8 _ = run_workflow(
9 dataset_path="coffee_sales.csv",
10 user_instructions=user_instructions,
11 generation_model=generation_model,
12 reflection_model=reflection_model,
13 image_basename=image_basename
14 )
Cell In[23], line 41, in run_workflow(dataset_path, user_instructions, generation_model, reflection_model, image_basename)
39 initial_code = match.group(1).strip()
40 exec_globals = {"df": df}
---> 41 exec(initial_code, exec_globals)
42 utils.print_html(out_v1, is_image=True, title="Generated Chart (V1)")
44 # 3) Reflect on V1 (image + original code) to get feedback and refined code (V2)
File <string>:6
File /usr/local/lib/python3.11/site-packages/pandas/core/frame.py:4113, in DataFrame.__getitem__(self, key)
4111 if self.columns.nlevels > 1:
4112 return self._getitem_multilevel(key)
-> 4113 indexer = self.columns.get_loc(key)
4114 if is_integer(indexer):
4115 indexer = [indexer]
File /usr/local/lib/python3.11/site-packages/pandas/core/indexes/base.py:3819, in Index.get_loc(self, key)
3814 if isinstance(casted_key, slice) or (
3815 isinstance(casted_key, abc.Iterable)
3816 and any(isinstance(x, slice) for x in casted_key)
3817 ):
3818 raise InvalidIndexError(key)
-> 3819 raise KeyError(key) from err
3820 except TypeError:
3821 # If we have a listlike key, _check_indexing_error will raise
3822 # InvalidIndexError. Otherwise we fall through and re-raise
3823 # the TypeError.
3824 self._check_indexing_error(key)
KeyError: 'year'
================
Need help
Welcome @Achint_Aggarwal! Try “gpt-3.5-turbo-0125” - I’ve just tested and it worked for me. Some gpt 3.5 models have been deprecated - I’m not sure if that is what is affecting this but hopefully this fixes it for you.
Got the same error.
Hi Achint_Aggarwal,
I get the same error messages. A plausible conclusion is that gpt-3.5-turbo is a less powerful coding model than gpt-4o-mini. So gpt-3.5-turbo may only be useful for initial code generation which needs to be debugged. This is currently still a common experience when using LLMs for coding.
@Achint_Aggarwal per the stacktrace, looks like the error is from LLM generated code execution. It will be helpful if the generated code is shared here.
btw: I agree with others shared, that different models have different coding capability therefore might generate code differently, sometimes with bugs. This seems to be one of those cases.