Streaming Doc in RAG

[yichuan-w]

I have heard that you support dynamic build RAG and streaming document, but I cannot find the code? Can someone help me with that?

[Arfazrll]

Dynamic RAG is usually supported by design, but not always provided as a single ready-made example. In most frameworks, it means ingesting documents at runtime, embedding them incrementally, and upserting vectors into the index without restarting the app. Once the vectors are added, they become immediately retrievable. This behavior is achieved by composing existing ingestion, embedding, and vector-store APIs rather than using one dedicated “dynamic RAG” file.

“Streaming” can refer to two things: streaming LLM responses (token-by-token output, commonly supported via async callbacks or SSE) and streaming document ingestion (reading large documents in chunks and embedding them progressively). Many repositories expose the required hooks and streaming interfaces, but expect users to wire them together based on their vector store, document format, and deployment needs.

[Sujaicodes]

Dynamic RAG (Retrieval-Augmented Generation) and streaming documents are supported conceptually, but there is no single “official” plug-and-play code snippet because implementation depends on your stack (e.g., APIs, vector databases, streaming method).
Dynamic RAG:-
Instead of using a fixed knowledge base, the system:
Retrieves documents at query time
Can update or ingest new data dynamically
Uses embeddings + vector search to fetch relevant context