GroqLLMRagApp

Description

This project implements a Retrieval-Augmented Generative (RAG) conversational AI chatbot powered by Groq and Gradio. It utilizes Sentence Transformers for efficient document retrieval and leverages pre-trained large language models (LLMs) through Groq for comprehensive question answering. The chatbot can handle various input modalities, including text documents, audio files, URLs, images, and CSV/Excel files.

Key Features

• Multimodal interaction: Text, audio, URL, image, and CSV/Excel file uploads
• Contextual awareness for improved understanding of user queries
• Retrieval of relevant document segments to support responses
• Integration with various LLM models via Groq API
• Confidence score for answer reliability (to be implemented)

Requirements

• Python 3.9 (https://www.python.org/downloads/)
• Transformers library (https://huggingface.co/docs/transformers/en/index)
• SentenceTransformers library (https://huggingface.co/sentence-transformers)
• Groq account and API key (https://groq.com/)
• Gradio library (https://www.gradio.app/guides/quickstart)
• Additional libraries for specific modalities:
• librosa for audio (pip install librosa)
• fitz for PDF (pip install fitz-py)
• pandas for CSV/Excel (pip install pandas)

Installation

pip install transformers sentence-transformers groq gradio librosa fitz pandas  # For all functionalities

Alternatively, you can install only the required libraries based on the modalities you want to support.

Usage

1. Clone this repository or download the project files.
2. Install the required libraries (see Requirements).
3. Configure your Groq API key (obtain from your Groq account).
• You can set the GROQ_API_KEY environment variable.
• Alternatively, you can modify the `answer_question` function in `logic.py` to provide the API key as an argument.
4. Prepare your input documents (text files, audio files, URLs, images, or CSV/Excel files).
• For text files, ensure each line is a separate document.
• For audio files, ensure they are in a supported format (e.g., WAV, MP3, MP4).
• For URLs, ensure they are valid links.
• For images, ensure they are in a supported format (e.g., PNG, JPG, JPEG).
• For CSV/Excel files, make sure the relevant text column is identified.
5. (Optional) If using CSV/Excel files, modify the `csv_to_dict_list` function in `logic.py` to adjust column names or formatting as needed.
6. Run the Gradio app:

python groq_app.py

7. Open http://localhost:1274 in your web browser to interact with the chatbot.

Explanation of Files

• logic.py: Contains core logic for document processing, retrieval, and question answering using RAG, Groq, and handling different modalities.
• groq_app.py: Handles the Gradio app interface and interaction with `logic.py`. It defines functions like `print_like_dislike`, `add_message`, and functions for processing different modalities like audio, URLs, PDFs, images, and CSV/Excel files.

Additional Notes

• The current implementation assumes pre-computed document embeddings. You may need to modify the code to generate embeddings from scratch if necessary.
• Consider adding error handling and logging for a more robust user experience.
• The confidence score for answer reliability is not yet implemented but can be integrated in the future.

Contributing

Feel free to submit pull requests for improvements or bug fixes.