Skip to content
MontiCorePublic

About

Infrastructure for a LLM-based client allowing to answer modelling and MontiCore tool questions

Resources

Stars

0 stars

Watchers

3 watching

Forks

Repository files navigation

SE-CHATTY

The main branch of SE-Chatty is deployed at:

https://studentprojects.se-rwth.de/se-chatty-dev/

The dev branch of SE-Chatty is deployed at:

https://studentprojects.se-rwth.de/se-chatty/

Running Locally

Requirements:

  • Docker needs to be installed to run the database services.
  • Create a .env file in the src directory (or directly set environmental variables)
ANONYMIZED_TELEMETRY=False

LUKNET_OPENAI_KEY=
LUKNET_OPENAI_ENDPOINT=https://luknet.openai.azure.com/

SWEDEN_OPENAI_KEY=
SWEDEN_OPENAI_ENDPOINT=https://se-sweden.openai.azure.com/

SCIEBO_UNAME=
SCIEBO_PSWD=
  • When starting SE-Chatty without docker use a python version in the range of 3.8-3.11 (3.10 is confirmend working). Then run the following command to setup the databases and install the requirements (for Windows):

call setup_local.bat

Run app:

Start with docker:

docker compose up --build

SE-Chatty can run without docker, which is generally faster. Make sure the database containers are running (docker-compose up -d mongodb weaviate) and all scripts in the setup_local.bat have run succesfully. Then in the src folder run:

streamlit run app.py -- --local

Common problems:

  • Some IDEs change the linebreaks of the start.sh file, this will result in the scripts failing when executed in docker. Change this in the IDE settings if necessary.

About

Infrastructure for a LLM-based client allowing to answer modelling and MontiCore tool questions

Resources

Stars

0 stars

Watchers

3 watching

Forks

Releases

Packages

Used by

Contributors

Languages