Insight allows you to search through a set of PDF documents by keyword, and to prompt a LLM about sets documents through the use of RAG. It is designed with the idea of stripping away what you're not interested in, in order to arrive at what you might be interested in. It tries to accomplish this by semantically matching information from user-supplied document sets with user prompts in a conversational form, together with good keyword search.
The initial development of Insight was funded by Follow the Money, where it is used to process large amount of FOIA requests.
Copy the provided .env.sample environment file to a .env file:
cp .env.example .env
You'll need to add a OpenAI api key to
the .env file as OPENAI_API_KEY.
We have to encrypt the traffic to our local keycloak instance because rabbitMQ is very serious about following the spec, and won't allow us to use a HTTP key server. Opensearch also requires SSL certificates to run.
You can easily generate certificates with
mkcert. To generate development
certificates:
mkdir certs
cp `mkcert -CAROOT`/rootCA.pem ./certs
mkcert -cert-file ./certs/opensearch.pem -key-file ./certs/opensearch-key.pem opensearch
mkcert -cert-file ./certs/keycloak.pem -key-file ./certs/keycloak-key.pem keycloak localhost
After configuration, you can run the development environment:
docker-compose up -d
After which you can create a user before accessing the gui.