XCE-Streamlit-V3

This app is an LLM-powered chatbot built using Streamlit and the LangChain-Huggingface API. It it is developed by Shubha Guha and Fina Polat under the supervision of Paul Groth for the ENEXA Explanations Module.

The main purpose of the app is showcasing the idea of co-construction of explanations.

This is the third version of ENEXA Explainer. This version employs an open-source LLM: https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3

Here are some technicalities:

The script requires two input variables: HUGGINGFACEHUB_API_TOKEN and EXPRESSION_FILE.

HUGGINGFACEHUB_API_TOKEN can be taken from environment variables or manually entered after runing the application.

EXPRESSION_FILE is the mounting point, we copy this file into the container.

Pull the image from Docker Hub:

docker pull finapolat/xce-streamlit_v3:enexa

Set the variables

Linux

export HUGGINGFACEHUB_API_TOKEN=your_api_token

export EXPRESSION_FILE=/path/to/expression_file.json

Windows

setx HUGGINGFACEHUB_API_TOKEN "your_api_token"

setx EXPRESSION_FILE "path_to_your_expression_file"

Run the docker:

Linux

docker run -p 8501:8501 -e HUGGINGFACEHUB_API_TOKEN -v $EXPRESSION_FILE:/app/class_expression.json finapolat/xce-streamlit_v3:enexa

Windows

docker run -p 8501:8501 -e HUGGINGFACEHUB_API_TOKEN -v %EXPRESSION_FILE%:/app/class_expression.json finapolat/xce-streamlit_v3:enexa

To view the app, please browse to http://localhost:8501