README.md

metadata

title: Data Analysis App
emoji: 📊
colorFrom: indigo
colorTo: blue
sdk: streamlit
sdk_version: 1.39.0
app_file: src/streamlit_app.py
pinned: false
license: mit

📊 Streamlit Data Analysis App (Gemini + Open-Source)

This Streamlit app lets you upload CSV or Excel datasets, automatically clean and preprocess them, create quick visualizations, and even get AI-generated insights powered by Gemini or open-source models.

---

🚀 Features

✅ Upload .csv or .xlsx datasets
✅ Automatic data cleaning & standardization
✅ Preprocessing pipeline (imputation, encoding, scaling)
✅ Quick visualizations (histogram, boxplot, correlation heatmap, etc.)
✅ Smart dataset summary and preview
✅ Optional Gemini AI insights for dataset interpretation

---

🧠 LLM Integration (Optional)

You can enable AI-generated insights with Gemini 2.0 Flash or your own Hugging Face model.

🔑 To configure:

1. Go to your Space’s Settings → Secrets tab.
2. Add the following: GEMINI_API_KEY = your_gemini_api_key HF_TOKEN = your_huggingface_token # optional
3. Save, then Restart your Space.

If you don’t add an API key, the app will still work for data cleaning and visualization.

---

🛠️ Deployment Notes

• Runtime: Python SDK
• SDK: Streamlit
• File formats supported: .csv, .xlsx
• Maximum file size: 100 MB
• Recommended visibility: Public (for full file upload support)

---

⚙️ Troubleshooting

❌ AxiosError: Request failed with status code 403

If you encounter this:

• Ensure your Space is Public (not Private).
• Ensure sdk: streamlit and app_file: are correctly declared in the YAML metadata above.
• Check that your runtime is “Python SDK”.
• Recheck your Gemini API Key or token secrets.

✅ Fix Checklist

Issue	Fix
App fails to start	Verify app_file matches your actual Python filename
403 Error	Make the Space public
API not found	Add key to Settings → Secrets
File upload broken	Ensure sdk: streamlit and runtime: python

---

💡 Example Workflow

1. Upload your dataset (e.g., global_freelancers_raw.csv).
2. View the raw preview and cleaned data table.
3. Generate preprocessing pipelines (e.g., median imputation + one-hot encoding).
4. Visualize trends with histograms, boxplots, or heatmaps.
5. (Optional) Ask Gemini for AI insights about correlations, patterns, or recommendations.

---

🧩 Tech Stack

• Frontend: Streamlit
• Backend: Python (Pandas, NumPy, Scikit-learn)
• AI Models: Gemini 2.0 Flash / open-source LLMs (Qwen, Mistral, etc.)
• Visualization: Matplotlib, Seaborn

---

🧾 License

MIT License © 2025
You are free to use, modify, and share this app with attribution.

---