sfn-blueprint 0.1.0

# SFNBlueprint Framework

SFNBlueprint is a modular framework for rapid building of intelligent agents to handle various data-related tasks, such as category identification, code execution, feature suggestion generation, data analysis, and more. These agents integrate with OpenAI to perform their tasks and can be extended with custom logic.

## Features

- **Base Agent**: A base class for all agents to extend.
- **Category Identification**: Automatically categorize datasets based on column names.
- **Data Cleaning Suggestions**: Generate suggestions for cleaning datasets.
- **Code Execution**: Dynamically execute code on data frames.
- **Feature Suggestion**: Get feature engineering suggestions based on your dataset.
- **Data Analysis**: Analyze datasets and return detailed statistics.

## Installation

You can install the SFNBlueprint framework via pip:

```bash
pip install SFNBlueprint

Usage

1. Environment Setup

To use the agents, you need an API key from OpenAI. Store the key in a .env file in your project root directory:

OPENAI_API_KEY=your_openai_api_key_here

Ensure the python-dotenv package is installed to load environment variables.

2. Using Agents

Each agent extends the base agent class SFNAgent and implements its own logic. Below are a few usage examples:

Example 1: Category Identification Agent

from agents.category_identification_agent import SFNCategoryIdentificationAgent
from tasks.task import Task
import pandas as pd

# Create sample data
data = {'col1': [1, 2], 'col2': [3, 4]}
df = pd.DataFrame(data)

# Initialize the agent
category_agent = SFNCategoryIdentificationAgent()

# Create a task
task = Task(data=df)

# Execute the agent
category = category_agent.execute_task(task)
print(f"Identified category: {category}")

Example 2: Clean Suggestion Generator Agent

from agents.clean_suggestion_generator import SFNCleanSuggestionGeneratorAgent
from tasks.task import Task

# Initialize the agent
clean_agent = SFNCleanSuggestionGeneratorAgent()

# Prepare analysis data for cleaning
task_data = {
    'shape': (100, 5),
    'columns': ['col1', 'col2', 'col3', 'col4', 'col5'],
    'dtypes': {'col1': 'int64', 'col2': 'float64'},
    'missing_values': {'col1': 0, 'col2': 5},
    'duplicates': 2
}

task = Task(analysis=task_data)

# Execute the agent
suggestions = clean_agent.execute_task(task)
print("Cleaning suggestions:", suggestions)

Example 3: Code Executor Agent

from agents.code_executor import SFNCodeExecutorAgent
from tasks.task import Task
import pandas as pd

# Sample DataFrame
data = {'A': [1, 2, 3], 'B': [4, 5, 6]}
df = pd.DataFrame(data)

# Sample code to execute on the DataFrame
code = """
df['C'] = df['A'] + df['B']
"""

# Initialize the agent and create a task
executor_agent = SFNCodeExecutorAgent()
task = Task(data=df, code=code)

# Execute the code on the DataFrame
df_result = executor_agent.execute_task(task)
print(df_result)

3. Extending Agents

You can create custom agents by extending the base class SFNAgent. Here's a minimal example of creating a new agent:

from agents.base_agent import SFNAgent

class CustomAgent(SFNAgent):
    def __init__(self):
        super().__init__(name="Custom Agent", role="Performs custom task")

    def execute_task(self, task):
        # Custom logic here
        return "Task executed"

Contact

For any queries or issues, please contact the maintainer at rajesh@stepfunction.ai.