Skip to content

Latest commit

 

History

History
132 lines (92 loc) · 5.75 KB

README.md

File metadata and controls

132 lines (92 loc) · 5.75 KB

⚙️🧬🔐 RAG System for MES in Chemical Industry 🏭🔬📊

This repository contains the code and resources for building a Retrieval-Augmented Generation (RAG) system for a Manufacturing Execution System (MES) in the chemical industry.

Table of Contents 📖 🔬

RAG System Knowledge Base Integration

Here are some images for the system:

1. Features ⚙️

  • File Upload 📤: Supports uploading multiple formats like text files, PDFs, DOCX, CSV, JSON, audio, video, and images.
  • Text Extraction 📄: Extracts text from PDFs, DOCX, CSV, JSON, images (via OCR), and transcribes audio and video.
  • Embedding Storage 💾: Generates embeddings for text data and stores them in a FAISS vector index for fast retrieval.
  • Conversational Interface 🗣️: Allows users to ask questions and receive context-aware answers through a Streamlit-based interface.
  • Multimedia Handling 🎥🎙️: Transcribes audio and video, and extracts text from images using OCR techniques.

Image Alt Text

2. Technologies Used 🛠️

  • Streamlit: For building an interactive and user-friendly interface.
  • LangChain: For integration with OpenAI’s models, vector stores, and retrieval-based generation.
  • OpenAI API: For embedding generation and natural language processing.
  • Whisper: For transcribing speech from audio and video.
  • FAISS: For storing and efficiently retrieving embeddings.
  • PyPDF2: To extract text from PDF files.
  • python-docx: To handle DOCX files.
  • pytesseract: For OCR in images.
  • moviepy: For extracting audio from video.
  • pandas: For handling structured CSV and JSON files.

3. Prerequisites 📦

Ensure that the following are installed before setting up the system:

  • Python 3.x
  • Required Python libraries (refer to the Requirements.txt file)

4. Setup Instructions ⚙️

Follow the steps below to set up the system:

  1. Clone the repository: 
    git clone <repository-url>
  2. Navigate to the project directory: cd
  3. Install the dependencies: pip install -r requirements.txt

5. Data Preprocessing 🧹

Image Alt Text The system requires preprocessing of various types of data:

  • Textual Data: Extract text from PDFs, DOCX, CSV, and JSON files using the respective libraries (PyPDF2, python-docx, pandas).
  • OCR for Images: Extract text from images using OCR technology (pytesseract).
  • Audio and Video Transcription: Convert audio and video files to text using Whisper API.

6. Embedding Generation 🧠

  • Generate embeddings for all textual content, including transcriptions from audio and video files.
  • Use the OpenAI API to generate embeddings that capture the context of the extracted text data.

7. Vector Indexing 🔍

  • Store the generated embeddings in a FAISS index for efficient and fast retrieval.
  • The FAISS index is queried to fetch the most relevant information based on user queries.

8. RAG Workflow 🔄

  1. Input: User submits a question or query through the Streamlit interface.
  2. Retrieval: Relevant context is retrieved from the FAISS index by matching the question with the stored embeddings.
  3. Generation: The OpenAI model generates a response based on the retrieved context.
  4. Output: The generated response is returned to the user.

9. Handling Multimedia 🎬🎧

  • Images: OCR technology (pytesseract) is used to extract text from images.
  • Audio/Video: Whisper transcribes speech from audio and video files into text, which is then processed by the system for context-aware generation.

10. Scaling 🚀

  • The system is designed to scale as data grows.
  • FAISS ensures that the retrieval process remains fast and efficient, even with large datasets.
  • The system can be deployed on cloud infrastructure for handling large-scale data storage and computational needs.

11. Code Structure 🏗️

The project is organized as follows:

  • /app: Contains the Streamlit interface for interacting with the system.
  • /modules: Contains scripts for file handling, embedding generation, and FAISS integration.
  • /data: Stores preprocessed data and embeddings.
  • /tests: Contains test scripts to evaluate system performance.

12. Testing and Evaluation 🧪

Image Alt Text

The system has been tested for:

  • Accuracy of Transcriptions: Ensuring that audio and video files are transcribed correctly.
  • Embedding Quality: Verifying that the generated embeddings accurately reflect the content’s context.
  • Search Performance: Testing the retrieval speed and relevance of the results from the FAISS index.

13. Contributions 🤝

We welcome contributions! If you'd like to help improve the system:

  • Fork the repository.
  • Submit a pull request with your proposed changes.

14. License 📜

This project is licensed under the MIT License - see the LICENSE file for details.