An interactive dashboard that uses a local LLM (Flan-T5) and Gradio to provide meaningful insights from customer satisfaction data.
- Key drivers of satisfaction using correlation analysis
- Interactive pie chart of satisfaction levels
- AI-generated text insights using Flan-T5
- Fully local — data never leaves your machine
| Tool / Library | Purpose |
|---|---|
| pandas | Data loading and manipulation |
| plotly | Visualizations |
| gradio | Web-based UI |
| transformers | Language model pipeline |
| torch | Runs the model backend |
| python-dotenv | Environment variables support |
Install all required packages:
pip install pandas plotly gradio transformers torch python-dotenv
(Optional) If you want GPU acceleration:
pip install accelerate
- Replace the placeholder with the correct path to your CSV file in
dashboard.py:
path = "your/path/here"
- Run the app:
python dashboard.py
- Gradio will open in your browser (typically at http://127.0.0.1:7860).
The input CSV must include the following columns:
- ProductQuality
- ServiceQuality
- PurchaseFrequency
- SatisfactionScore
- LoyaltyLevel
- Country
Each row represents an individual customer response.
The dashboard uses the google/flan-t5-base model from Hugging Face, which performs summarization locally via the transformers library. It supports both CPU and GPU execution.
dashboard.py # Main app logic customer_feedback_satisfaction.csv # Input dataset .env # Optional env variables README.md # Project documentation
- Hugging Face for Flan-T5
- Gradio for rapid UI development
- Plotly for clean visualizations
This project is open-sourced under the MIT License.