​
analysis

This method performs data analysis based on the specified analysis type and user input. It also sets the analysis_guide, analysis_output and analysis_code attributes of the DataAnalyzr instance.

  1. If the analysis type is set to skip, the method sets analysis_guide to "No analysis performed" and analysis_output and analysis_code to None.
  2. If the analysis type is set to sql, it generates a SQL query and performs SQL-based analysis using the provided user input and context.
  3. If the analysis type is set to ml, it generates Python code and performs pandas and machine learning-based analysis using the provided user input and context.
analysis(
    user_input: str,
    analysis_context: str = None,
    time_limit: int = None,
    max_retries: int = None,
    auto_train: bool = None,
) -> Union[str, pd.DataFrame, dict[str, pd.DataFrame], None]

​
Parameters

user_input
string
required

User input for generating the analysis.

analysis_context
string

Context for the analysis.

time_limit
integer

Time limit in seconds for the analysis. Default is 45.

max_retries
integer

Maximum number of retries for the analysis. Default is 10.

auto_train
boolean

Whether to automatically add questions with their SQL query or Python code to the vector store. Default is True.

​
Returns

response
string or pandas.DataFrame or dictionary or None

Analysis output generated based on the user input and context.

  • Depending on the given input, the output may sometimes be a single number or string. It is then returned as a string.
  • When the output is a dictionary, at least one of the values is a pandas DataFrame.
  • If no analysis is performed, or the analysis fails, the output is None.
user_input = "What is the average sale price?"
response = "123.456"
# or ---
user_input = "Which country has the highest GDP?"
response = "United States"
# or ---
user_input = "How do sales vary with time?"
response = pd.DataFrame(...)
# or ---
user_input = "What are the top 5 countries by GDP? What are their top industries?"
response = {
    "key1": pd.DataFrame(...),
    "key2": pd.DataFrame(...),
}

​
Example usage

analysis_output = data_analyzr.analysis(
    user_input="What is the average sale price?",
    analysis_context="Today is {date}. The column {column_name} refers to the listed sale price.",
    time_limit=60,
    max_retries=5,
    auto_train=True,
)

​
visualisation

This method generates visualizations based on the user input and plot context. These visualisations are then saved to the plot_path. It is recommended that to generate visualisations, the ask method should be used instead of this method.

  1. If plot_path is not provided, the generated plot is saved to generated_plots/<random-image-name>.png.
  2. Sets the plot_code and visualisation_output attribute of the DataAnalyzr instance.
visualisation(
    user_input: str,
    plot_context: str = None,
    time_limit: int = None,
    max_retries: int = None,
    auto_train: bool = None,
    plot_path: str = None,
) -> str

​
Parameters

user_input
string
required

User input for generating the visualisation.

plot_context
string

Context for the plot.

time_limit
integer

Time limit in seconds for the visualisation process. Default is 60.

max_retries
integer

Maximum number of retries for the visualisation process. Default is 10.

auto_train
boolean

Whether to automatically add the visualisation to training data. Default is True.

plot_path
string

Path to save the generated plot. Defaults to generated_plots/<random-image-name>.png.

​
Returns

response
string

Path to the saved plot.

response = "path/to/plot.png"

​
Example usage

plot_path = data_analyzr.visualisation(
    user_input="Visualize the sales data",
    plot_context="Use the `ggplot` style with `plt.style.use('ggplot')`.",
    time_limit=60,
    max_retries=5,
    auto_train=True,
    plot_path="path/to/save/plot.png",
)

​
insights

This method generates insights based on the user input, analysis output, and specified insights context. It is recommended that to generate insights, the ask method should be used instead of this method.

  1. The method utilizes a language model to generate insights based on the provided input and context.
  2. The generated insights are influenced by the analysis output, user input, and insights context.
  3. The generated insights are formatted and returned as a string.
  4. Sets the insights_output attribute of the DataAnalyzr instance.
insights(
    user_input: str,
    insights_context: str = None,
    n_insights: int = 3,
) -> str

​
Parameters

user_input
string
required

User input for generating the insights.

insights_context
string

Context for the insights generation.

n_insights
integer

Number of insights to generate. Default is 3.

​
Returns

response
string

Generated insights based on the user input and context.

response = "- The average sale price is $123.45.\n-The highest sale price is $456.78.\n-The lowest sale price is $12.34."

​
Example usage

insights_output = data_analyzr.insights(
    user_input="What is the average sale price?",
    insights_context="Today is {date}. The column {column_name} refers to the listed sale price.",
    n_insights=3,
)

​
recommendations

This method generates recommendations based on the user input, analysis insights, and specified recommendations context. It is recommended that to generate recommendations, the ask method should be used instead of this method.

  1. If the parameter from_insights is False, the method generates recommendations without using the insights output.

  2. The format of the recommendations output can be customized using the recs_format parameter.

    • If the output_type is text, this parameter is ignored.
    • If recs_format is None, the default format will be used:
            [{
            "Recommendation": "string",
            "Basis of the Recommendation": "string",
            "Impact if implemented": "string",
        }]

  3. The output type of the recommendations can be either text or JSON.

  4. Sets the recommendations_output attribute of the DataAnalyzr instance.

recommendations(
    user_input: str,
    from_insights: bool = True,
    recs_format: dict = None,
    recommendations_context: str = None,
    n_recommendations: int = 3,
    output_type: Literal["text", "json"] = "text",
) -> str

​
Parameters

user_input
string
required

User input for generating the recommendations.

from_insights
boolean

Whether to generate recommendations from insights. Default is True.

recs_format
dictionary

Format for the recommendations output. Default is None.

recommendations_context
string

Context for the recommendations generation.

n_recommendations
integer

Number of recommendations to generate. Default is 3.

output_type
Literal['text', 'json']

Type of output for recommendations. Default is text.

​
Returns

response
string

Generated recommendations based on the user input and context.

response = "..."

​
Example usage

recommendations_output = data_analyzr.recommendations(
    user_input="What is the average sale price?",
    from_insights=True,
    recs_format={
        "Recommendation": "Recommendation text",
        "Reason": "Reason for the recommendation",
        "Impact": "Impact of the recommendation",
    },
    recommendations_context="Today is {date}. The average sale price is $123.45.",
    n_recommendations=3,
    output_type="json",
)

​
tasks

This method generates tasks based on the user input, analysis output, and specified tasks context. It is recommended that to generate tasks, the ask method should be used instead of this method.

  1. The method incorporates user input, insights, and recommendations into the generated tasks.
  2. The number of tasks to generate can be customized using the n_tasks parameter.
  3. Sets the tasks_output attribute of the DataAnalyzr instance.
tasks(
    user_input: str,
    tasks_context: str = None,
    n_tasks: int = 3,
) -> str

​
Parameters

user_input
string
required

User input for generating the tasks.

tasks_context
string

Context for the tasks generation.

n_tasks
integer

Number of tasks to generate. Default is 3.

​
Returns

response
string

Generated tasks based on the user input and context.

response = "..."

​
Example usage

tasks_output = data_analyzr.tasks(
    user_input="What is the average sale price?",
    tasks_context="Today is {date}. The average sale price is $123.45.",
    n_tasks=3,
)
AttributesVoice Agent