Semantic Search Engine for DocuSign Hackathon - Agreement Trap

A Streamlit-based application that allows users to import documents from DocuSign or other document management systems like your local files to build your own vector database and perform a semantic search across uploaded documents using the advanced hugging face semantic search model and Gemini Model.

Youtube Demo

Setup and Installation

1. Clone the repository
2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # or venv\Scripts\activate on Windows

3. Install dependencies:

   pip install -r requirements.txt

4. Set up environment variables in .env file
5. Run the application:

   streamlit run app.py

Service Architecture

┌─────────────────┐     ┌──────────────┐     ┌────────────────┐
│  Streamlit UI   │────▶│  DocuSign    │───▶│  Document      │
└─────────────────┘     │  /local files│     │  Processing    │
         │              └──────────────┘     └────────────────┘
         │                                           │
         ▼                                          ▼
┌─────────────────┐     ┌──────────────┐     ┌────────────────┐
│  Vector Store   │◀────│  Embedding   │◀────│  Text          │
│  (Pinecone)     │     │  Service     │     │  Extraction    │
└─────────────────┘     └──────────────┘     └────────────────┘

Technology Stack

Machine Learning Models

• Sentence Transformers: Using all-MiniLM-L6-v2 for generating document embeddings

Vector Database

• Pinecone: Vector similarity search and storage

AI Model

• Gemini: AI model for generating nice responses to user queries

Document Processing

• PyMuPDF (fitz)
• PyPDF2
• python-docx

Integration

• DocuSign eSignature API
• Streamlit for the user interface
• HuggingFace Inference API

I Wrote all the code from scratch and it's only for this hackathon with the help of AI tools.

For more information, Drop me a message on LinkedIn

#Docusign #huggingFace #gemini #semantic_search #streamlit