Skip to main content

Component Overview

Just like Workflow Studio, Agent Studio provides its functionality as building blocks, called Components. To help the user, the components are grouped by type.

The following list gives an overview of the available components.

AI Cloud​

This group contains components that can interact with the AI Cloud platform. They can provide access and details about files and projects or leverage existing functionality like calling deployments or send queries to Graph Studio.

  • AI Cloud LLM
    • Connects to an LLM deployed on AI Cloud. The component has the same functionality as the other models but calls the locally deployed LLM. This means it doesn’t require any additional resources or API keys and also keeps all your data inside the AI Cloud infrastructure.
  • Call Workflow
    • Interact with a deployed workflow
    • Send data to a deployed model or leverage the vast number of operators for machine learning and ETL tasks
  • Get Project Content
    • Lists all the content inside a selected project
  • Get Project Details
  • List Deployments
  • Load Files
  • Query Graph Studio

Inputs / Outputs​

  • Chat Input
  • Text Input
  • Chat Output
  • Text Output
  • Prompt

Those components allow you to interact with the agent flow.

Chat Input and Chat Output are used to enter display text on the playground, there can also be one of each per flow.

Text Input and Text Output process normal strings.

They all use the Message type as input and output format.

The Prompt component is similar to Text Input, but can handle variable values, that can be used to create more flexible templates.

Data​

API Request

The API Request component is designed to construct and send HTTP requests. It supports two modes: URL mode, where you can input multiple URLs and select the request method, and curl mode, where you can execute a complete curl command. The component allows you to customize request parameters like headers, body, and query parameters, and it returns a Data object containing the response.

SQL Query

The SQL Database component allows you to execute SQL queries on databases compatible with SQLAlchemy, such as PostgreSQL, MySQL, and SQLite. It can be used with an Agent component to translate natural language prompts into SQL queries, enabling dynamic and flexible database interactions without requiring explicit SQL syntax from the user.

URL

The URL component is used to fetch and process content from one or more URLs. It supports recursive link following up to a specified depth and offers output in plain text or raw HTML. This component is versatile, providing options for asynchronous loading and restricting crawls to within the same domain.

Webhook

The Webhook component sets up a webhook trigger to run a flow upon receiving an HTTP POST request. It automatically generates a curl command snippet for triggering the flow, making it easy to integrate with external systems. The component processes incoming data and can trigger flows based on the payload received.

Read more: Webhooks

Processing​

The components in this group can change and manipulate data.

Data Operations allows for manipulation of data objects, including filtering and editing keys and values. DataFrame Operations provides a range of functions for modifying DataFrame structures, such as sorting and filtering, making it a versatile tool for handling tabular data.

Models​

Agent Studio can connect to many LLM services and use them to generate text, or provide a connection to their language models. All services require their own set of credentials, mostly in the form of API keys.

Vector Stores​

Agent Studio can connect to many Vector Stores, where users can ingest their data and also query them for finding similar text snippets. The primary use case is RAG, to provide agents with additional context. The components can be used for both stages and can be either used for ingesting vectorized data elements or to retrieve them again.

Embeddings​

This component group supports a range of popular embedding models, ensuring flexibility and adaptability to different project needs, while also providing options for customization and parameter tuning to optimize performance.

Agent​

This is the main component of Agent Studio.

Agents can orchestrate agentic flows and define the behavior of the flow. Agents need to connect to an LLM model, either by selecting one directly or provide a custom LLM connection as input.

The biggest difference between the Agent component and a regular LLM model is the option to use tools as input component. This greatly enhances the capabilities of the LLM. Nearly all other components (including other Agents) can be set into tool mode.

Tools​

This group is a collection of useful external services. There are components to connect to search engines or retrieve additional information from Wikidata and Wikipedia.

MCP Connection

This component can connect to an MCP server and use the tools registered there as additional functionality for an agent.

Logic​

Run Flow

Can run another flow as a subprocess for the main flow.

This can be useful to re-use components and structure more complex projects.

Read more: Run Flow

Helpers​

Additional useful components.

Bundles​

Collection of components that add functionality from third party providers.