PolicyPal

PolicyPal is a demo chatbot experience to redefine the way users engage with insurance. Say goodbye to complicated forms and confusing jargon – with PolicyPal, obtaining insurance information becomes as easy as having a conversation. This demo utilizes Retrieval-Augmented Generation (RAG) on an open-source insurance Q&A corpus with real-world user questions on various insurance domains.

PolicyPal.Demo.mp4

Tech Stack

In this Google Collab notebook, a RAG pipeline in Python is implemented using OpenAI's LLM in combination with a Weaviate vector database and an OpenAI embedding model. LangChain is used for orchestration with Gradio's ChatBot module used as a simple web user interface.

Prerequisites

You will need to have the following before continuing:

• OpenAI API key: We will generate embeddings from text via OpenAI's inference service. The Ada model (version 2) costs $0.0001 per 1,000 tokens. For this demo, the embedding of a subset of the InsuranceCorpus was $0.23.
• Weaviate API key and cluster URL: Run a Weaviate instance as a managed service using Weaviate Cloud Services (WCS). The free trial plan provides a sandbox for 14 days. No payment information is required. The sandbox expires after the trial period but a new free trial sandbox can be created anytime. Refer to the Setup section in this article for more information. Store the keys and url in Google Collab Secrets. Refer to this article on how to store the keys.

Store the keys and url in Google Collab Secrets. Refer to this article on how to store the keys.

References

1. Combining LangChain and Weaviate Link

2. LangChain Vectorstores: Weaviate Link

3. Data used in Demo: InsuranceQA Corpus Link

4. Retrieval-Augmented-Generation from theory to LangChain implementation Link.

5. Question Answering in Weaviate with OpenAI Q&A module Link

6. Building an Interactive Streaming Chatbot with Langchain, Transformers, and Gradio Link

7. Creating Your Own Chatbot: A Beginner-Friendly Tutorial with OpenAI, LangChain, Gradio, and Wikipedia Link

8. Getting Started with Weaviate: A Beginner’s Guide to Search with Vector Databases Link