Agentic RAG

Estimated reading: 2 minutes

The Robility Agentic RAG component is designed to fetch relevant and synthesized results from the knowledge base using the user’s query and configured parameters. It connects with the Robility Flow RAG API to deliver contextually rich, citation-backed responses by leveraging the selected retrieval mode and query details.

This component enables automated agents and workflows to perform intelligent document retrieval, multi-source reasoning, and response generation based on the selected RAG configuration.

Parameters

Parameter	Description
Mode	Specifies the RAG operation strategy for retrieving and synthesizing information. Available modes include: Local – Fetches results from the local or project-specific knowledge base only. Hybrid – Combines both local and global sources to produce a balanced response. Global – Retrieves results from the organization-wide or shared global knowledge base. Mix – Merges results from multiple RAG models to improve response accuracy. Naive – Performs a direct query without advanced reasoning or synthesis. Bypass – Skips the RAG retrieval process and executes the query directly, typically used for debugging or test runs.
RAG URL	Specifies the endpoint of the Robility RAG API that the component uses to send the query and retrieve the response. The URL should point to the configured RAG service endpoint that handles the retrieval and synthesis logic.
Query	Specifies the user’s input or search prompt used to fetch relevant results from the knowledge base. The query should clearly define the context or question for which the RAG component must generate a response.
Delay Before	Specifies the time (in milliseconds) to wait before executing the activity or component.
Delay After	Specifies the time (in milliseconds) to wait after completing the activity or component execution.
Continue on Error	Determines whether the workflow should proceed if the current activity encounters an error. Stop Workflow – Immediately halts the workflow execution when an error occurs. Continue – Skips the failed activity and proceeds with the next step in the workflow. Continue Using Error Output – Allows the workflow to continue while capturing the error details as part of the output. This mode enables advanced handling scenarios where the error message or code is required for downstream actions, logging, or conditional flows.
Retry	Specifies the number of times the system should automatically retry the activity if it fails during execution.
Response	Returns the synthesized result data fetched from the knowledge base, providing enriched, AI-generated answers that can be further processed or displayed within the workflow.

Agentic RAG

Parameters

Agentic RAG

CONTENTS