📄 Technical Specification Fulfillment Application

This repository contains a Python-based application that processes both Excel and PDF files. It extracts content from the PDF file, stores it in a vector store (ChromaDB) and document store (Redis), and provides a Streamlit user interface for uploading files and processing them.

🛠 Features

• PDF Content Extraction: Extracts structured content from PDFs using the UnstructuredPDFLoader.
• Excel Integration: Processes Excel files containing a column named 'Minimum Specification' to extract queries for further processing.
• Vector Store (ChromaDB): Stores the content into a vector store for efficient retrieval and embedding.
• Document Store (Redis): Stores the content into a Redis document store for flexible and fast retrieval.
• Streamlit UI: Provides a user-friendly interface to upload both Excel and PDF files, and process the content.
• Persistence: Utilizes persisted storage for both ChromaDB and Redis, allowing reuse across multiple sessions without needing to reprocess the data.

📂 Project Structure

The project consists of the following files. For user to use the program, only the files in item 3 are required:

1. populatedata.ipynb:

   • This is run on Jupyter Notebook for testing and debugging during development.
   • Loaded the brochure HP-dataset.pdf as example document which should be in the same folder of this file.
   • Extracts PDF content, processes the data, and stores it in ChromaDB and Redis.
   • Data extracted include text, table and image (multimodal).
   • This file can be run in Jupyter notebook.

2. tsfa.py:

   • The evaluation of the program is done using this code by comparing the ‘Expected Response’ and ‘Generated Response’ columns from an excel file with relevance given a score from 1-3.
   • The input file Sample-TenderDoc.xlsx should be in the same folder of this file.
   • Take note that the processing and storage of brochure/datasheet (company documents) are not done here as it was implemented by the by populatedata.ipynb.
   • This file can be run using python command.

3. chktechspec.py & process.py:

   • This is the actual program where the user can run by giving an excel file with at least a column named ‘Minimum Specification’.
   • User can upload any arbitrary excel file as long as it contains a column named 'Minimum Specification'. The file will generate the same file with a generated column called 'Generated Specification'.
   • A new file will be generated with the column ‘Generated Response’ which the user can use as reference for specification fulfillment of the existing product brochure/datasheet.
   • A streamlit command is used to call the file chktechspec.py (streamlit run chktechspec.py).
   • User can download the processed file in the standard name of 'Updated-Sample-TenderDoc.xlsx'.

🚀 How to Run the Application

1. Clone the repository:

   git clone https://github.com/renaissance2005/DDS-AI-Challenge-Project.git
   cd DDS-AI-Challenge-Project

2. Install dependencies: Ensure you have Python 3.10+, Ollama and Redis Server and install the required dependencies:

   pip install -r requirements.txt

3. Run the Streamlit application: Start the Streamlit app using:

   streamlit run chktechspec.py

4. Upload an Excel File:

   • Open the app in your browser.
   • Upload an Excel file containing a 'Minimum Specification' column.
   • Click the download button after the file is processed.

🧑‍💻 Tech Stack

• Python: Core programming language.
• ChromaDB: Used as a vector store for efficient document embedding and retrieval.
• Redis: Used as a document store to store the PDF content.
• Ollama: Installed and run 2 models: llama3.1:8b (inference) and llava-llama3 (vision).
• Streamlit: Provides an interactive UI for file uploads and processing.
• LangChain Community Tools: Used for PDF extraction and unstructured data processing.

🛡️ Error Handling

• The program checks that both an Excel is uploaded before processing.
• The Excel file must contain a 'Minimum Specification' column.
• Error messages are displayed if any issues arise during file upload or processing.

🌟 Future Improvements

• Add agents to check and refine the results.
• Incorporate options to use more powerful multimodal language models.
• Provide functionality to upload and process multiple brochures and datasheets.

📬 Contact

Feel free to reach out if you have any questions or suggestions:

• Email: davidkeat@graduate.utm.my
• GitHub: (https://github.com/renaissance2005/DDS-AI-Challenge-Project)