A Streamlit application that enables natural language analysis of CSV data using PandasAI, LangChain, and Ollama.

• Natural Language Data Analysis: Ask questions about your data in plain English
• Visualization Generation: Create charts and graphs with simple text prompts
• Code Transparency: View the Python code used to answer each question
• Local LLM Integration: Powered by Ollama for privacy and control
• No Caching: Every query is processed freshly by the LLM
• Conversation History: Track your analysis journey with a full conversation log
• Exportable Results: Download your data and conversation history

• Python 3.9 or higher
• Ollama installed and running on your system
• At least one LLM model pulled in Ollama (e.g., mixtral, llama3)

1. Clone this repository:

   git clone https://github.com/iRahulPandey/TTYD.git
   cd TTYD

2. Create a virtual environment (optional but recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install the required dependencies:

   pip install -r requirements.txt

4. Make sure Ollama is running:

   ollama serve

5. Pull at least one model (if you haven't already):

   ollama pull mixtral  # or llama3, gemma, etc.

1. Start the application:

   streamlit run main.py

2. Open your web browser and navigate to the URL shown in the terminal (typically http://localhost:8501)

3. Upload a CSV file using the sidebar uploader

4. Select an LLM model from the dropdown (must be installed in Ollama)

5. Start asking questions about your data!

• "How many rows are in this dataset?"
• "What are the column names?"
• "Show me a summary of the numerical columns"
• "Create a bar chart of the top 5 values in Category"
• "Show the distribution of values in CustomerAge"
• "Calculate the correlation between Price and Rating"
• "Find any missing values in the dataset"
• "Show me a trend of sales by month"
• "What is the average order value by category?"
• "Create a pie chart showing payment methods"

• Clear Conversation: Removes the current conversation history
• Download Data as CSV: Download the currently loaded dataset
• Download Conversation: Save the entire conversation history as a text file

ttyd/
├── __init__.py
├── main.py              # Application entry point
├── config.py            # Configuration settings
├── ui/
│   ├── __init__.py
│   ├── app.py           # Main Streamlit UI
│   ├── components.py    # Reusable UI components
│   └── styles.py        # CSS and styling
├── core/
│   ├── __init__.py
│   ├── analysis.py      # Analysis functionality
│   ├── dataframe.py     # SmartDataframe management
│   └── llm.py           # LLM integration
└── utils/
    ├── __init__.py
    ├── data_loader.py   # Data loading utilities
    ├── image_handler.py # Image processing utilities
    └── models.py        # Model management utilities

• streamlit: Web application framework
• pandasai: Natural language interface for pandas
• langchain: Framework for LLM applications
• pandas: Data manipulation and analysis
• plotly: Interactive visualizations
• ollama: Local LLM inference

1. "No module named 'ttyd'"

   • Make sure you're running the application from the project root directory
   • If using a package structure, install in development mode: pip install -e .

2. "No Ollama models found"

   • Ensure Ollama is running with ollama serve
   • Pull at least one model with ollama pull mixtral

3. Visualization Issues

   • If visualizations don't appear correctly, check that plotly is installed
   • Ensure your data is appropriate for the requested visualization type

4. Performance Considerations

   • Since caching is disabled, every query requires LLM processing
   • For larger datasets, complex queries may take longer to process
   • Performance depends on the speed of your local LLM (CPU/GPU availability)