At BerriAI, we are committed to protecting the privacy of your data. This policy outlines the three options available to Berri users, and how we process and store data in each option.
Option 1: Hosted on Berri Infra
If you choose to host your Berri instance on our infrastructure, we will ingest your raw data, convert it to embeddings, and store the embeddings along with chunks of your raw data in our database. When you query your instance, we send the most relevant chunks to OpenAI or your selected LLM. We only store data that is necessary to answer questions for your Berri instance, and we do not use your data for training or improving our systems. When you delete your Berri instance, no data is stored on our infrastructure.
Option 2: Hosted / Managed Enterprise Offering
For enterprise customers, we offer dedicated infrastructure, dedicated AWS accounts, and dedicated databases - all managed by BerriAI. Your raw data is converted to embeddings and chunks, which are stored in a dedicated database separate from all other Berri users. We also give you control over the control plane, meaning the database is created in a dedicated AWS account for your team. Only Berri and your company can access this AWS account.
Option 3: On-Prem Hosting
If you choose to host your Berri instance on your own infrastructure, we will provide you with a Docker image of Berri. When you provide your data to Berri, it acts as your ingest pipeline. You must specify what database Berri must store the embeddings in, and during query time Berri uses your database. Berri stores none of your data, and your instances will be created on your infrastructure and database.