Build your first RAG pipeline

This quickstart walks through a complete design-time RAG pipeline. You will choose a template, inspect its nodes, optionally adjust a configuration, and run frontend validation. You will not ingest documents or generate a live response.

Before you start #

Open the Ragify frontend in a desktop browser. The visual editor is designed for desktop-sized canvases; smaller devices may be routed through a simplified experience. No external model provider key is required to explore the graph design tools.

1. Start with guided onboarding #

Open Get started or navigate to the onboarding experience. Choose a goal, give your workspace a display name, and select Basic RAG when prompted to choose a template. Continue into the editor. If you enter the editor directly instead, use its available new-pipeline or template selection controls.

2. Inspect the pipeline #

The Basic RAG starter graph follows this order:

Document Loader → Chunker → Embedder → Vector Store → Retriever → LLM → Answer / Output

Follow each connector. A node describes one responsibility in the workflow. The canvas visualizes planned data movement; it does not process data by itself.

3. Open a node configuration #

Select a node and open its configuration dialog. Review the settings shown for that node. Change a non-sensitive value if you want to experiment, then return to the canvas. Settings and available controls vary by node.

4. Review the structure #

Check that the graph is still one continuous path, with the appropriate input, retrieval, and output stages. The current editor is designed around one linear chain and one retrieval strategy rather than arbitrary parallel workflow branches.

5. Validate the graph #

Use the editor's validation control. Read any errors or guidance and return to the affected node if a change is required. Correct the reported issue and validate again.

Expected outcome #

You should have a readable, configured graph and an understanding of how its components connect. Validation can report that a graph is structurally acceptable, but it does not execute a query or deploy an endpoint.

What you learned #

You can choose a template, recognize the seven Basic RAG stages, open node settings, and inspect validation feedback. Continue with Add, connect, and arrange nodes.