Semantic Search Engine

What it does

This application empowers Docusign users to unlock the potential of their documents by enabling efficient search and analysis of large volumes with advanced semantic search capabilities. By harnessing cutting-edge natural language processing and vector search technologies, it delivers intelligent and context-aware search results that transcend traditional keyword-based methods.

How does the Semantic search work?

Making the Impossible .. Possible

It's Impossible to

• Access all document content at once in DocuSign using documents from file storage systems such as local files, DocuSign envelopes, Google Cloud etc.
• Building personalized Knowledge Database similar to obsidian for each user that customize and grow.
• Build & Bring knowledge to any AI workload with your knowledge database from Docusign.

But it's Possible with

• Pinecone Vector Database, You can do more than Semantic search such as Multi-model search, Recommendations and classification.

How we built it

I initially developed a local version that works with local files. After that, I built the data extraction process to store the information and worked on the embedding models from Hugging Face, aligning them with the Pinecone vector database for AI models. Finally, I focused on fetching data from the DocuSign APIs and handling authentication.

Inspiration

I would say the inspiration came purely from the Kaggle RAG implementation and the Hugging-face model tasks.

Challenges we ran into

The most challenging aspect was selecting the authentication type and configuration, which took longer than I anticipated.

Accomplishments that we're proud of

Using Hugging Face for the first time to address large context management is exciting. Deploying it to Streamlit also feels rewarding.

What's Next for Semantic Search Engine

• Constructing a secure way to have AI capabilities with sensitive data by using locally capable models to embed and retrieve answers.
  
• Integrating Google Docs to have verative of options for importing files.
• A workspace session integration in the DocuSign ecosystem with multiple editions, viewers and clients.
• Clustering graph option to view the data in more relational passed approach.

Built With

• asyncio
• docusign-esign
• httpx
• numpy
• pinecone-client
• pymupdf
• pypdf2
• python-docx
• python-dotenv
• requests
• sentence-transformers
• streamlit