Knowledge Ingestion
The Knowledge Ingestion template shows how to build and populate a Knowledge Base in Orbis Studio. It fetches content from web URLs, splits it into chunks, and stores those chunks in a Knowledge Base so they can be searched semantically later.
Use this template as the first step in a retrieval-augmented generation (RAG) pipeline. After ingesting your content, use the Knowledge Retrieval template to query it.
How the flow works
_10URL → Split Text → Knowledge Ingestion
The URL component fetches the content of one or more web pages. The Split Text component breaks the content into smaller chunks (default: 100 tokens) that fit within embedding model limits. The Knowledge Ingestion component embeds the chunks and writes them to your Knowledge Base.
Prerequisites
- A Knowledge Base created in Orbis Studio (or create one during this flow)
Quickstart
-
In Orbis Studio, open the Templates library and select Knowledge Ingestion.
-
In the URL component, replace the default URL with the page or pages you want to ingest.
-
In the Knowledge Ingestion component, select an existing Knowledge Base from the dropdown, or click Create new to create one. Give it a recognizable name.
-
Click Run to execute the flow. The status indicator on each component shows progress as data moves through the pipeline.
-
When the flow completes, your Knowledge Base contains the ingested content and is ready to query.
Key components
| Component | Role |
|---|---|
| URL | Fetches content from one or more web pages |
| Split Text | Splits the fetched content into chunks based on token count |
| Knowledge Ingestion | Embeds the chunks and writes them to the selected Knowledge Base |
Configuration tips
- Data source: Replace the URL component with a File or Directory component to ingest local documents instead of web pages.
- Chunk size: Adjust the Chunk Size in the Split Text component. Smaller chunks give more precise retrieval; larger chunks give more context per result.
- Column configuration: In the Knowledge Ingestion component, the Column Configuration table controls which fields get vectorized. Ensure the text column is marked for vectorization.
- Re-ingestion: Re-running the flow on the same URL updates the Knowledge Base with fresh content.
Next steps
- Use the Knowledge Retrieval template to search what you just ingested.
- Combine both templates into a single flow to build a full Q&A pipeline over your own content.