Unexpected ValueError in Metadata Summarization Function

[BennisonDevadoss]

Description:
I encountered an issue while trying to execute the file multiagent_sql_data_analyst.py in the repository. I followed the provided steps without modifying the model name or any part of the code. Additionally, I used the same data provided in the repository.

Path to the Code:
free-ai-tips/008_multiagent_sql_data_analyst/

Environment Details:

• OS: Ubuntu 22.04.4 LTS
• Python Version: 3.11
• Pandas Version: 2.2.3
• LangGraph Version: 0.2.57
• LangChain Version: 0.3.14

Error Log:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[104], line 3
      1 # * Donut Chart
----> 3 sql_data_analyst.invoke_agent(
      4     user_instructions="Make a donut chart of sales revenue by territory for the top 5 territories.",
      5 )

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/ai_data_science_team/multiagents/sql_data_analyst.py:140, in SQLDataAnalyst.invoke_agent(self, user_instructions, **kwargs)
    119 def invoke_agent(self, user_instructions, **kwargs):
    120     """
    121     Invokes the SQL Data Analyst Multi-Agent.
    122     
   (...)
    138     ```
    139     """
--> 140     response = self._compiled_graph.invoke({
    141         "user_instructions": user_instructions,
    142     }, **kwargs)
    143     self.response = response

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/pregel/__init__.py:1955, in Pregel.invoke(self, input, config, stream_mode, output_keys, interrupt_before, interrupt_after, debug, **kwargs)
   1953 else:
   1954     chunks = []
-> 1955 for chunk in self.stream(
   1956     input,
   1957     config,
   1958     stream_mode=stream_mode,
   1959     output_keys=output_keys,
   1960     interrupt_before=interrupt_before,
   1961     interrupt_after=interrupt_after,
   1962     debug=debug,
   1963     **kwargs,
   1964 ):
   1965     if stream_mode == "values":
   1966         latest = chunk

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/pregel/__init__.py:1670, in Pregel.stream(self, input, config, stream_mode, output_keys, interrupt_before, interrupt_after, debug, subgraphs)
   1664     # Similarly to Bulk Synchronous Parallel / Pregel model
   1665     # computation proceeds in steps, while there are channel updates.
   1666     # Channel updates from step N are only visible in step N+1
   1667     # channels are guaranteed to be immutable for the duration of the step,
   1668     # with channel updates applied only at the transition between steps.
   1669     while loop.tick(input_keys=self.input_channels):
-> 1670         for _ in runner.tick(
   1671             loop.tasks.values(),
   1672             timeout=self.step_timeout,
   1673             retry_policy=self.retry_policy,
   1674             get_waiter=get_waiter,
   1675         ):
   1676             # emit output
   1677             yield from output()
   1678 # emit output

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/pregel/runner.py:171, in PregelRunner.tick(self, tasks, reraise, timeout, retry_policy, get_waiter)
    169 t = tasks[0]
    170 try:
--> 171     run_with_retry(
    172         t,
    173         retry_policy,
    174         configurable={
    175             CONFIG_KEY_SEND: partial(writer, t),
    176             CONFIG_KEY_CALL: partial(call, t),
    177         },
    178     )
    179     self.commit(t, None)
    180 except Exception as exc:

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/pregel/retry.py:40, in run_with_retry(task, retry_policy, configurable)
     38     task.writes.clear()
     39     # run the task
---> 40     return task.proc.invoke(task.input, config)
     41 except ParentCommand as exc:
     42     ns: str = config[CONF][CONFIG_KEY_CHECKPOINT_NS]

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/utils/runnable.py:422, in RunnableSeq.invoke(self, input, config, **kwargs)
    418 config = patch_config(
    419     config, callbacks=run_manager.get_child(f"seq:step:{i+1}")
    420 )
    421 if i == 0:
--> 422     input = step.invoke(input, config, **kwargs)
    423 else:
    424     input = step.invoke(input, config)

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/pregel/__init__.py:1955, in Pregel.invoke(self, input, config, stream_mode, output_keys, interrupt_before, interrupt_after, debug, **kwargs)
   1953 else:
   1954     chunks = []
-> 1955 for chunk in self.stream(
   1956     input,
   1957     config,
   1958     stream_mode=stream_mode,
   1959     output_keys=output_keys,
   1960     interrupt_before=interrupt_before,
   1961     interrupt_after=interrupt_after,
   1962     debug=debug,
   1963     **kwargs,
   1964 ):
   1965     if stream_mode == "values":
   1966         latest = chunk

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/pregel/__init__.py:1670, in Pregel.stream(self, input, config, stream_mode, output_keys, interrupt_before, interrupt_after, debug, subgraphs)
   1664     # Similarly to Bulk Synchronous Parallel / Pregel model
   1665     # computation proceeds in steps, while there are channel updates.
   1666     # Channel updates from step N are only visible in step N+1
   1667     # channels are guaranteed to be immutable for the duration of the step,
   1668     # with channel updates applied only at the transition between steps.
   1669     while loop.tick(input_keys=self.input_channels):
-> 1670         for _ in runner.tick(
   1671             loop.tasks.values(),
   1672             timeout=self.step_timeout,
   1673             retry_policy=self.retry_policy,
   1674             get_waiter=get_waiter,
   1675         ):
   1676             # emit output
   1677             yield from output()
   1678 # emit output

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/pregel/runner.py:171, in PregelRunner.tick(self, tasks, reraise, timeout, retry_policy, get_waiter)
    169 t = tasks[0]
    170 try:
--> 171     run_with_retry(
    172         t,
    173         retry_policy,
    174         configurable={
    175             CONFIG_KEY_SEND: partial(writer, t),
    176             CONFIG_KEY_CALL: partial(call, t),
    177         },
    178     )
    179     self.commit(t, None)
    180 except Exception as exc:

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/pregel/retry.py:40, in run_with_retry(task, retry_policy, configurable)
     38     task.writes.clear()
     39     # run the task
---> 40     return task.proc.invoke(task.input, config)
     41 except ParentCommand as exc:
     42     ns: str = config[CONF][CONFIG_KEY_CHECKPOINT_NS]

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/utils/runnable.py:422, in RunnableSeq.invoke(self, input, config, **kwargs)
    418 config = patch_config(
    419     config, callbacks=run_manager.get_child(f"seq:step:{i+1}")
    420 )
    421 if i == 0:
--> 422     input = step.invoke(input, config, **kwargs)
    423 else:
    424     input = step.invoke(input, config)

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/langgraph/utils/runnable.py:197, in RunnableCallable.invoke(self, input, config, **kwargs)
    195 else:
    196     context.run(_set_config_context, config)
--> 197     ret = context.run(self.func, input, **kwargs)
    198 if isinstance(ret, Runnable) and self.recurse:
    199     return ret.invoke(input, config)

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/ai_data_science_team/agents/data_visualization_agent.py:544, in make_data_visualization_agent.<locals>.chart_instructor(state)
    541 data_raw = state.get("data_raw")
    542 df = pd.DataFrame.from_dict(data_raw)
--> 544 all_datasets_summary = get_dataframe_summary([df], n_sample=n_samples, skip_stats=False)
    546 all_datasets_summary_str = "\n\n".join(all_datasets_summary)
    548 chart_instructor = recommend_steps_prompt | llm 

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/ai_data_science_team/tools/metadata.py:69, in get_dataframe_summary(dataframes, n_sample, skip_stats)
     67     for idx, df in enumerate(dataframes):
     68         dataset_name = f"Dataset_{idx}"
---> 69         summaries.append(_summarize_dataframe(df, dataset_name, n_sample, skip_stats))
     71 else:
     72     raise TypeError(
     73         "Input must be a single DataFrame, a list of DataFrames, or a dictionary of DataFrames."
     74     )

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/ai_data_science_team/tools/metadata.py:121, in _summarize_dataframe(df, dataset_name, n_sample, skip_stats)
    101 # 6. Generate the summary text
    102 if not skip_stats:
    103     summary_text = f"""
    104     Dataset Name: {dataset_name}
    105     ----------------------------
    106     Shape: {df.shape[0]} rows x {df.shape[1]} columns
    107 
    108     Column Data Types:
    109     {column_types}
    110 
    111     Missing Value Percentage:
    112     {missing_summary}
    113 
    114     Unique Value Counts:
    115     {unique_counts_summary}
    116 
    117     Data (first {n_sample} rows):
    118     {df.head(n_sample).to_string()}
    119 
    120     Data Description:
--> 121     {df.describe().to_string()}
    122 
    123     Data Info:
    124     {info_text}
    125     """
    126 else:
    127     summary_text = f"""
    128     Dataset Name: {dataset_name}
    129     ----------------------------
   (...)
    136     {df.head(n_sample).to_string()}
    137     """

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/pandas/core/generic.py:10940, in NDFrame.describe(self, percentiles, include, exclude, datetime_is_numeric)
  10691 @final
  10692 def describe(
  10693     self: NDFrameT,
   (...)
  10697     datetime_is_numeric: bool_t = False,
  10698 ) -> NDFrameT:
  10699     """
  10700     Generate descriptive statistics.
  10701 
   (...)
  10938     max            NaN      3.0
  10939     """
> 10940     return describe_ndframe(
  10941         obj=self,
  10942         include=include,
  10943         exclude=exclude,
  10944         datetime_is_numeric=datetime_is_numeric,
  10945         percentiles=percentiles,
  10946     )

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/pandas/core/describe.py:94, in describe_ndframe(obj, include, exclude, datetime_is_numeric, percentiles)
     89     describer = SeriesDescriber(
     90         obj=cast("Series", obj),
     91         datetime_is_numeric=datetime_is_numeric,
     92     )
     93 else:
---> 94     describer = DataFrameDescriber(
     95         obj=cast("DataFrame", obj),
     96         include=include,
     97         exclude=exclude,
     98         datetime_is_numeric=datetime_is_numeric,
     99     )
    101 result = describer.describe(percentiles=percentiles)
    102 return cast(NDFrameT, result)

File ~/miniconda3/envs/hdfc-ai/lib/python3.11/site-packages/pandas/core/describe.py:171, in DataFrameDescriber.__init__(self, obj, include, exclude, datetime_is_numeric)
    168 self.exclude = exclude
    170 if obj.ndim == 2 and obj.columns.size == 0:
--> 171     raise ValueError("Cannot describe a DataFrame without columns")
    173 super().__init__(obj, datetime_is_numeric=datetime_is_numeric)

ValueError: Cannot describe a DataFrame without columns
During task with name 'chart_instructor' and id 'fe59d395-cf50-5c7a-08d4-88cd596d5f42'
During task with name 'data_visualization_agent' and id '5da46a15-cdc4-076a-5325-cd3ffb1e368b'

Steps to Reproduce:

1. Clone the repository and navigate to the directory free-ai-tips/008_multiagent_sql_data_analyst/.
2. Execute the file multiagent_sql_data_analyst.py without making any changes.
3. Use the same dataset provided in the repository.
4. Run the code and observe the error during execution.

Expected Behavior:
The program should generate the donut chart as requested in the invoke_agent call.

Actual Behavior:
The program throws a ValueError during the execution of the invoke_agent function, specifically in the _summarize_dataframe function of metadata.py.

Request for Assistance:

• Could you help identify what might be causing this issue?
• Is there any compatibility issue with the versions of the libraries I’m using?
• Should I modify any part of the code or dataset to resolve this?

Looking forward to your guidance. Let me know if further details are needed.

Thank you!

[mdancho84]

Looks like the SQL Agent step didn't return a data frame. The chart instructor failed because the data frame provided didn't have columns.

I'd try it again and see if the problem persists.