#commands to install on local host
#this was tested on
PRETTY_NAME="Ubuntu 22.04.1 LTS"
NAME="Ubuntu"
VERSION_ID="22.04"
VERSION="22.04.1 LTS (Jammy Jellyfish)"
VERSION_CODENAME=jammy
ID=ubuntu
ID_LIKE=debian
HOME_URL="https://www.ubuntu.com/"
SUPPORT_URL="https://help.ubuntu.com/"
BUG_REPORT_URL="https://bugs.launchpad.net/ubuntu/"
PRIVACY_POLICY_URL="https://www.ubuntu.com/legal/terms-and-policies/privacy-policy"
UBUNTU_CODENAME=jammy
##Steps (Note:These instructions won't work on an Apple m1 mac. Also please use a conda environment with python 3.10)
for example conda create --name chatur python=3.10
- Install docker with docker compose.
- cd ~/chatur-codeathon/propeller/langserve
- pip install -r requirements.txt
- pip install -r requirements.langclient.txt
- cd ~/chatur-codeathon/propeller/gen_vectordb
- sudo apt install jq libreoffice
- vi create_vectordb.py ((and add your cyverse username and password inside the code)
- python3 create_vectordb.py --create_docs RNR355
# now manually move gen_vectordb/vectordb folder to langserve/vectordb todo: changer docker-compose to accomodate this
- mv ~/chatur-codeathon/propeller/gen_vectordb/vectordb ~/chatur-codeathon/propeller/langserve
- cd ~/chatur-codeathon/propeller/langserve
- docker compose --file docker-compose-vectordb-gpu.yml --project-name "chatur" build
- docker compose --file docker-compose-vectordb-gpu.yml --project-name "chatur" up -d
- docker compose --file docker-compose-vectordb-gpu.yml --project-name "chatur" exec ollama ollama pull mistral
then pass your question to the chatbot as input in:
- curl -sX POST http://localhost:8000/langserve/invoke -H 'Content-Type: application/json' --data-raw '{"input":"When does the class meet?"}' | jq
** Now after you have everything working, if you make some changes to the DB or code, and want to rebuild it use --no_download
i.e
python3 create_vectordb.py --no_download --create_docs RNR355
then run all docker compose commands again.