ℹ️

Name change: deepset Cloud is now deepset AI Platform

We're updating our product name from "deepset Cloud" to "deepset AI Platform" (or "deepset" for short). While we're implementing this change across our product interface, documentation, and website, you may still see references to "deepset Cloud" in some places during this transition period. All names refer to the same platform.

You can design, evaluate, deploy, and monitor your LLM apps using an intuitive user interface and a powerful REST API. Along the way, you can easily involve your colleagues and end-users to ensure you end up with a pipeline delivering true value for your use case. Just share the link to your pipeline prototype, customized with your logo and brand colors, to let others test it and give you feedback without needing to log in or set up accounts.

When your pipeline has been tested, evaluated, and deployed, you can connect deepset AI Platform to your target application using the REST API and use it wherever you need.

deepset AI Platform is model-agnostic, meaning you can use all major LLMs in your apps created in deepset AI Platform. You can also easily switch one LLM for another at any time.

Building an Application with deepset AI Platform

deepset AI Platform helps you through all the steps of creating an NLP application:

1. deepset AI Platform offers file preprocessing components that you can use in your indexing pipeline to preprocess your data and prepare them for search.
2. In deepset AI Platform, you work with pipelines. Pipelines are made up of components. Each component has a different task. For example, a Retriever component retrieves the documents that best match the query. Then, you can have a Ranker component that ranks the documents by the most relevant. Then, the Ranker can pass the documents on to the Generator that uses an LLM to generate an answer based on your documents, and so on.
   The output of one component is used by another component. Components are like building blocks - you can mix and match different components or replace a component with another one to create the perfect search system.
3. Use logs, groundedness score, and tracing to evaluate your pipeline. You can connect to Langfuse and Weights & Biases to trace your pipelines there.
4. Demo your app to your colleagues and let them test it, all in the deepset AI Platform UI. Just share a link to a pipeline prototype customized with your logo and brand colors. No need to create accounts or invite anyone to your deepset AI Platform workspace.