RAG User Guide
Build retrieval-augmented generation systems with data sources, pipelines, indexes, retrieval tuning, security, analytics, and workflows.
Who This Guide Is For
- RAG builders
- Data engineers
- AI application teams
- Security teams
Where To Go
| Page |
Use It For |
/rag |
RAG dashboard. |
/rag/sources |
Connect document, web, database, and API sources. |
/rag/pipelines |
Build ingestion and retrieval pipelines. |
/rag/indexes |
Manage vector and keyword indexes. |
/rag/retrieval-studio |
Tune retrieval behavior. |
/rag/playground |
Test retrieval and generation. |
/rag/knowledge-graph |
Explore entity relationships. |
/rag/security |
Configure leak detection and prompt injection protection. |
/rag/analytics |
Review retrieval performance. |
/rag/workflows |
Automate RAG workflows. |
Core Concepts
| Concept |
Meaning |
| Source |
An upstream content system such as documents, web pages, databases, or APIs. |
| Pipeline |
The ingestion, chunking, embedding, indexing, retrieval, and generation flow. |
| Index |
A searchable store for vectors, keywords, or hybrid retrieval. |
| Retriever |
The component that selects context for a query. |
| RAG security |
Controls for redaction, leak detection, prompt injection, and auditability. |
Common Workflows
Create a RAG pipeline
- Add a source.
- Configure parsing, chunking, and metadata extraction.
- Choose embedding model and index settings.
- Run ingestion.
- Test retrieval in the playground.
- Add security policies.
- Monitor analytics and iterate.
Improve retrieval quality
- Open Retrieval Studio.
- Review failed or low-confidence queries.
- Adjust chunk size, metadata filters, hybrid weights, or reranking.
- Reindex if needed.
- Compare before and after results.
Best Practices
- Attach metadata during ingestion so filters and audit trails remain useful.
- Use representative queries before publishing a RAG pipeline.
- Enable redaction and prompt injection controls for sensitive collections.
- Monitor stale sources, embedding drift, and low-hit queries.