I use this with a local llama.cpp running, it will require a compatible backend interface
run ./sbin/convertModel.sh
or python3 pylib/convertModel.py
to save the model for reuse without using unsafe deserializations
then run ./sbin/initializeFaissStore.sh
or python3 pylib/initializeFaissStore.py
to initialize the faiss store
add pdfs to pdf directory the default is pdf/
in the project root create with mkdir pdf
then run ./sbin/loadPdfs.sh
or python3 pylib/convertModel.py
to load pdfs into the faiss store it will skip pdfs it's already added so more can be added and it can be rerun to update the faiss store
then run either ./sbin/chatRag.sh
or python3 pylib/chatRag.py
for command line chat or ./sbin/gradioRagServer.sh
or python3 pylib/gradioRagServer.py
to get a web interface to the LLM backed chatbot with RAG services
default values are configured in config/config.py
all the scripts have help strings so you can use --help
or -h
to get an explaination of options
the bash scripts should run on most posix compliant operating systems. Hopefully it works for everyone but it's just a wrapper you can call the scripts directly