RAG resources
Three resources that build a retrieval-augmented generation (RAG) pipeline. Each has its own reference page.
Works in workflow and agent mode.
text
loader (load + chunk) -> embedding OR vectorStore (index) -> embedding OR vectorStore (search) -> chat (answer with context)| Resource | Role | Reference |
|---|---|---|
loader: | Read files, URLs, or a directory into text chunks | Loader |
embedding: | Local SQLite keyword store - index and search, no API key | Embedding |
vectorStore: | External vector database - Qdrant, Chroma, pgvector, ... | Vector store |
embedding or vector store
embedding: | vectorStore: | |
|---|---|---|
| Storage | Local SQLite file | External vector DB service |
| Matching | Keyword (LIKE) | Vector similarity |
| Setup | None | Run and configure the DB |
| API key | Not needed | Needed for the embedding model |
| Best for | On-prem, small corpora | Production scale, semantic ranking |
See also
- Document search tutorial -
embedding:upsert and search - Scraper resource - fetch content to index
- Search resources - keyword file and web search
- LLM resource - answer with the retrieved context
- Resources overview - all resource types