Working in deepset Cloud
In deepset Cloud, you work in a workspace where you keep your data and your pipelines. You can have multiple workspaces, each with different content. There are three interfaces (GUI, API, and SDK) you can use to interact with deepset Cloud, depending what's most convenient to you.
Workspaces
In deepset Cloud, you work within an organization to which you were invited. Each organization may have multiple workspaces with different data. Users who are part of an organization can access all the workspaces in this organization. Pipelines and data are not shared among workspaces.
After logging in, you land on the Home page, where you can view your workspace statistics, the number of searches run last week, and the details of the last five searches. These include the top answer (the answer with the highest relevance score) and the top file (the file containing the top answer). The search time is in seconds.
You can quickly switch between workspaces. Just click the workspace name and choose the one you want to switch to.
Common Actions
Here's an explanation of what each navigation item does:
Navigation Step | Action |
---|---|
Workspace name | Switch between workspaces Create a workspace (just type the name of the new workspace). You can create up to 10 workspaces. |
Home | Access the guided workflow to create a pipeline. Check your workspace statistics, like the number of files or latest searches. |
Files | Upload files. These are the files your pipeline uses for search. Manage existing files (view, delete, preview, filter by metadata). |
Evaluation Sets | Upload evaluation datasets. These are the CSV files used in experiments to evaluate your pipeline. Manage existing evaluation sets (download, export, view, and delete). |
Pipeline Templates | Choose a template. View templates. |
Pipelines | Create pipelines. Manage existing pipelines organized by their service level (deploy, undeploy, edit, view pipeline details, share pipelines, check groundedness, and export user feedback). Click the name of a pipeline to check its details, such as feedback, queries, observability dashboard, logs, and so on. |
Jobs | Schedule jobs to run asynchronously. Currently, you can run batch question answering across the filest in the workspace. |
Playground | Test your pipelines. |
Prompt Explorer | Engineer your prompts, try out different prompts, and compare them across up to three pipelines. Choose a prompt from a library of curated prompts. |
Experiments | Create experiments. They're a way of systematically evaluating your pipelines. Manage existing experiments (edit, delete, check details). |
Groundedness | Check how well your RAG system is grounded in your documents. |
To access your personal settings, organization and workspace settings, connections to model providers, and documentation, click your name in the top right corner.
Three Interfaces
You can work in deepset Cloud through one of these interfaces:
- An intuitive, user-friendly graphical user interface, where you can see real-life updates and get immediate feedback on your actions.
- REST API for a platform-independent way to communicate with deepset Cloud. It supports all the tasks available in GUI.
- An open source SDK package you can customize. Currently, it contains methods for efficient file operations, such as upload or listing.
Updated 3 days ago