No module named "_gdbm"

[prasiyer]

Code:
qwen_model = HfApiModel("Qwen/Qwen2.5-Coder-32B-Instruct", token = hf_token)
data_analyst_agent_qwen = CodeAgent(
tools=[],
model=qwen_model,
additional_authorized_imports=["numpy", "pandas", "matplotlib.pyplot", "seaborn"],
max_steps=10,
)
data_analyst_agent_qwen.run(data_analyst_prompt, additional_args = dict(source_file_path = input_file_path))

Result:
Error when smolagent is executing the following code during one of the steps. This code is generated by the llm during step 1. The following code runs without any error when I run it in a separate notebook. Logs are not showing any additional information. Error message: No module named "_gdbm". The error seems to be generated by the code for the plots. Agent works without any issues if I "EXCLUDE" the requirement about generating charts.

import matplotlib.pyplot as plt
import seaborn as sns

plt.figure(figsize=(15, 10))

plt.subplot(2, 2, 1)
sns.boxplot(x='Product_3', y='Stroke_Length', data=df)
plt.title('Distribution of Stroke_Length by Product_3')
plt.xticks(rotation=45)

plt.subplot(2, 2, 2)
sns.boxplot(x='Cylinder_Function', y='Closed_Length_mm', data=df)
plt.title('Distribution of Closed_Length_mm by Cylinder_Function')
plt.xticks(rotation=90)

plt.subplot(2, 2, 3)
sns.boxplot(x='Type_2', y='Attributes_Rod_ID', data=df)
plt.title('Distribution of Attributes_Rod_ID by Type_2')

plt.subplot(2, 2, 4)
sns.boxplot(x='Type_2', y='Attributes_Internal_Bore', data=df)
plt.title('Distribution of Attributes_Internal_Bore by Type_2')

plt.tight_layout()
plt.show()

question1 = "What is the average Stroke_Length for each Product_3 category?"
question2 = "What is the maximum Closed_Length_mm for each Cylinder_Function category?"
question3 = "What is the average Attributes_Rod_ID for each Type_2 category?"

answer1 = df.groupby('Product_3')['Stroke_Length'].mean()
answer2 = df.groupby('Cylinder_Function')['Closed_Length_mm'].max()
answer3 = df.groupby('Type_2')['Attributes_Rod_ID'].mean()

print("Question 1:", question1)
print("Answer 1:", answer1)
print("\nQuestion 2:", question2)
print("Answer 2:", answer2)
print("\nQuestion 3:", question3)
print("Answer 3:", answer3)

[albertvillanova]

I think this issue is not related to the smolagents library, but to your local Python environment.

[prasiyer]

@albertvillanova
I am not certain that the problem is with the local python environment. If I copy the code generated by the llm and run it in a notebook within my local environment, the code works just fine. However, when the agent is trying to run it within the python-interpretor, it encounters the "Missing _gdbm module" error. Can you pls provide some suggestions to debug this further?