Getting Started
Deploying Your Pipeline

Deploying Your First Pipeline

SciPhi provides the fastest and most featureful way for developers to deploy high-quality Retrieval-Augmented Generation (RAG) pipelines through the R2R framework. We'll cover linking your GitHub repository, understanding the key files in the repository, and configuring your pipeline settings.

Selecting your pipeline

SciPhi offers deployment of pre-built or custom RAG pipelines. Pre-built offerings include solutions for Q&A, Web, and Advanced (HyDE) RAG. Each of these pipelines are implemented in the R2R framework and can be configured or customized as documented in the R2R framework (opens in a new tab).

To deploy a RAG pipeline with SciPhi, start by clicking on the solution that corresponds to your use case, or click Custom to link to your own custom pipeline. The form will then auto-populate with relevant information.

Pipeline Deploy

Providing Secrets

During the deployment process, you'll need to link to a vector database. SciPhi provides a managed LanternDB and you may use this vector database or connect to an external provider afterwards. Next, provide any necessary secret variables, such as API keys, through the form, as well as the database that you created in the previous step. These secrets are encrypted at all times during transmission and storage. Our infrastructure is hosted on Google Cloud and our secrets are provisioned through Google Cloud's Secret Manager. You can read more about SciPhi's security policy here.

For the basic RAG example shown above a valid OPENAI_API_KEY must be provided or else the pipeline will fail at bulid or runtime.

Pipeline Deploy


Once you've provided the required secrets, click the "Deploy" button to create your RAG pipeline. SciPhi will handle the deployment process, setting up the necessary infrastructure and services based on your configuration.

In the case of the basic RAG example shown previously this includes a managed Postgres database with pgvector, peformant logging with a Redis, and serverless deployement through Google Cloud Platform.

Upon successful deployment, you will be provided with a unique URL (e.g., (opens in a new tab)) that serves as the endpoint for accessing and interacting with your deployed RAG pipeline using the R2R client.