insightflow-ai 1.0.0

An autonomous AI data pipeline for cleaning, EDA, ML modeling, and business reporting.

InsightFlow 🚀

InsightFlow is an autonomous, end-to-end AI data pipeline designed to transform raw, messy datasets into fully trained Machine Learning models and C-Level Executive Business Reports—all with zero manual coding.

---

🌟 Features

• AutoFix: Automatically cleans messy data (missing values, outliers, duplicates) and normalizes columns using AI heuristics. Supports multi-sheet Excel files.
• AutoEDA: Generates stunning, dark-mode glassmorphism HTML reports with interactive Chart.js visualizations and AI data insights.
• AutoModel: Auto-detects target prediction columns, trains 12+ classification/regression scikit-learn models, evaluates them, saves the Champion model as .joblib, and generates an interactive leaderboard.
• AutoReport: Compiles all previous pipeline steps into a pristine, high-contrast HTML Executive Dashboard that features a one-click PDF Export button for business stakeholders.

---

⚙️ Installation

You can install InsightFlow directly via pip (if hosted on PyPI) or locally:

# To install locally from the source folder:
git clone https://github.com/yourusername/InsightFlow.git
cd InsightFlow
pip install -e .

---

🔑 Authentication (HuggingFace API Token)

InsightFlow uses the GLM-5.2 LLM hosted on the HuggingFace Router. You must have a HuggingFace API Token.

How to get your token:

1. Go to HuggingFace Settings -> Access Tokens.
2. Click "New token".
3. Name it (e.g., InsightFlow_Token) and give it Read permissions.
4. Copy the token (it starts with hf_...).

---

📖 Quickstart Guide

Using InsightFlow is incredibly simple. You can import modules directly:

from InsightFlow import autofix, autoeda, automodel, autoreport

# 1. Provide your token and data
HF_TOKEN = "hf_your_token_here"
data_path = "messy_sales_data.xlsx"

# 2. Automatically clean the data
clean_data = autofix(data_path, HF_TOKEN)

# 3. Generate Exploratory Data Analysis
eda_results, eda_insights = autoeda(clean_data, HF_TOKEN)

# 4. Train Models (AI auto-detects the target column!)
model_results, model_insights, trained_models = automodel(clean_data, HF_TOKEN)

# 5. Generate final C-Level PDF Business Report
autoreport(eda_results, eda_insights, model_results, model_insights, target_col="Target", hf_token=HF_TOKEN)

The "One-Liner" Ultimate Pipeline

Don't want to run steps manually? Use the integrated clean_and_analyze wrapper!

from InsightFlow.autofix import clean_and_analyze, set_hf_token

set_hf_token("hf_your_token_here")

# This one line does ALL of the above!
clean_and_analyze("raw_data.csv", run_eda=True, run_model=True, run_report=True)

---

☁️ Uploading to HuggingFace / PyPI

Publishing to PyPI

To share your package with the world so they can pip install insightflow:

# 1. Install build tools
pip install build twine

# 2. Build your package
python -m build

# 3. Upload to PyPI
twine upload dist/*

Hosting Privately on HuggingFace Spaces

You can host this entire workflow on HuggingFace Spaces or as a private HF Model repository:

1. Create a Private Space on HuggingFace.
2. Add your token to the Space Secrets as HF_TOKEN.
3. Push this exact codebase using standard git:

git remote add origin https://huggingface.co/spaces/YourUser/InsightFlow
git push origin main