Lingception: Using Natural Language Engineering to Explain Natural Language Engineering.

This is the bot for the project "Lingception" by Ellena Nett and Annika Traßl-Wilterius. It investigates to what extent contextualization of ChatGPT with a PDF improves the reliability of answers to questions about the given context. The architecture of the bot is based on the work of mayooear and this repo gpt4-pdf-chatbot-langchain. (Frontend of this repo is inspired by langchain-chat-nextjs)

The setup describes the steps to start the bot locally and connect to already used API. The complete setup incl. ingestion of a PDF can be found in the origin repo.

Set up the bot

1. Install packages

First run npm install yarn -g to install yarn globally.

Then run:

yarn install

After installation, there should be a node_modules folder.

2. Set up the .env file

• Copy .env.example into .env The .env file should look like this:

OPENAI_API_KEY=

PINECONE_API_KEY=
PINECONE_ENVIRONMENT=

PINECONE_INDEX_NAME=

• enter the APIs
• enter pinecone parameter

Run the app

Run the app npm run dev to launch the local dev environment, and then type a NLP question in the chat interface.