Skip to main content
This quickstart creates a small support knowledge base, stores it in Qdrant, retrieves the matching chunk, and asks an LLM for a grounded answer. Run the setup wizard early. The wizard is bare ragrails and writes .ragrails.toml; the workflow commands below use those saved defaults.

​
1. Install

​
2. Run the Setup Wizard

Run ragrails with no subcommand to save CLI defaults in .ragrails.toml:
There is no ragrails setup subcommand. The setup wizard is the bare ragrails command. After it writes .ragrails.toml, you can omit repeated flags such as --vector-db, --collection, --url, --provider, and --model. For the quickstart stack, answer the wizard like this. Press Enter to accept blank defaults.
Ragrails setup
Configure project defaults for SDK-backed CLI commands.
 
1. Vector store
Vector database provider (qdrant, qdrant_cloud, pinecone, weaviate) [qdrant]:
Collection name [docs]: support
Vector database URL: http://localhost:6333
 
2. Embedding
Embedding provider [voyage]:
Embedding model [voyage-3]:
 
3. LLM
LLM provider (openai, anthropic, google) [openai]:
LLM model [package default]: gpt-4o-mini
LLM max tokens [1024]:
 
4. Reranking
Enable reranker? [y/N]:
 
Configure advanced defaults? [y/N]:
Your installed package may display a concrete default in the LLM model prompt; type gpt-4o-mini for this quickstart. The wizard previews the config, writes .ragrails.toml, and prints the environment variables needed by the selected providers. With the answers above, the saved config is:
For website extraction, setup-url is separate browser setup:

​
3. Ingest

Use other source types with the same workflow:

​
4. Query

​
5. Chat

Expected answer:
Run ragrails chat with no question to start the interactive session.

​
Next

Ingest

Learn the full CLI ingest workflow.

Query

Search and chat over a stored collection.

Overview

See every CLI pipeline stage.

Doctor

Validate your environment.