Chat with your Data is a Streamlit application designed to facilitate interactive exploration of data stored in various formats, including CSV files, Excel files, and data fetched from APIs. It integrates a language model-based chatbot that enables users to ask questions about the data, providing insights and analysis in a conversational manner.
- Data Interaction: Users can upload CSV files, fetch data from the Facebook Ads API, or even upload Excel files, allowing flexibility in data source.
- Natural Language Interaction: Utilizes a language model-based chatbot to interpret user questions and provide insightful responses about the data.
- Dynamic Visualization: Offers dynamic visualizations and summaries of the data based on user queries, enhancing data exploration and understanding. (Note: May not work as expected, still under development)
- Customizable Language Models: Users can choose between different language models, such as OpenAI and Gemini-Pro, and configure API keys for seamless integration.
- Conversation History: Maintains a conversation history, enabling users to review past interactions with the chatbot.
-
Clone the repository:
git clone https://github.com/yourusername/your-repository.git cd your-repository -
Install the required dependencies:
pip install -r requirements.txt
-
Configure API keys:
-
OpenAI:
- Generate your OpenAI API key here.
-
Gemini:
- Generate your Gemini API key from Google Maker Suite.
-
-
Run the Streamlit app:
streamlit run web-app-code.py
-
Select an option to load dataset:
- Upload a CSV file
- Upload an Excel file
- Fetch data using the Facebook API
-
Choose a language model for the chatbot and enter the corresponding API key.
-
Open your web browser and navigate to the provided AWS Studio URL, removing any trailing parameters (e.g., "lab?") after the default, and then append /proxy/8501(port number)/ to the end. The URL format should look like this: https://{NOTEBOOK_URL}/proxy/8501/.(This Instructions is only for AWS studio). Enter prompts and receive natural language responses from the chatbot.
Please note that currently, our app is not efficient at generating chart-based responses within the chat_message frame. Charts are displayed at the top of the web app and may be overwritten when new chart questions are asked. We are actively working on improving this feature.