RAG
RAG lets the agent search document collections and answer with context from uploaded material.
Requirements
- A language model.
- An embeddings model.
- A vector store such as Qdrant.
- A collection with indexed documents.
- User permissions to read the collection.
Troubleshooting
If search returns nothing, check that embeddings are selected for the user, Qdrant is reachable, documents are indexed, and the chat uses the expected collection.