AssistOnto

Setting up RAG

Create a directory named my_doc_dir filled with plaintext documents, then call

poetry run assistonto docs --docdb-path my_assistonto_docs.db --add my_doc_dir

to add the documents to the vector database, with their base filenames (without extension) as IDs. Then specify my_assistonto_docs.db using the ASSISTONTO_DOCDB_PATH environment variable.

Test locally

poetry shell # we use poetry for development (see pyproject.toml)
# run from the root directory, with (test) assistonto.db
# in assistonto directory, and the variables in dev.env in the environment
poetry run assistonto server

Build & Deploy

Build container image and transfer it to server:

make build
make deploy

Then delete old container, and run the new one:

service assistonto stop
docker rm assistonto_server
docker run --detach --name assistonto_server -p 8080:8080 --env-file /opt/assistonto/assistonto.env -v /opt/assistonto/assistonto.db:/opt/assistonto/assistonto.db localhost/assistonto
service assistonto start

To make deployment faster, I have been mounting the code as a volume inside the container instead of deploying. This way we only need to deploy when we change or update dependencies. -v /home/bclaro/AssistOnto/assistonto:/assistonto

Run SQLite on the server: sudo -u assistonto ./sqlite3 /opt/assistonto/assistonto.db

Configuration

Use the following environment variables: ASSISTONTO_DB_PATH=assistonto.db # path to database ASSISTONTO_SECRET_KEY=dkfjkdgj # Flask secret key ASSISTONTO_MODELS='{"<model_name>": {"default": "true", "url": "<model_url>", "credentials": "<api_key>"}, "<another_model": {"credentials": {"file": "<file_containing_api_key"}}}' ASSISTONTO_MAX_MESSAGES_SHOWN=100 ASSISTONTO_DOCDB_PATH=assistonto_docs.db