rag

Chunked exports for retrieval-augmented generation pipelines.

Keys

Key	Type	Default	Description
enabled	bool	true	Write RAG chunks.
chunk_strategy	str	hybrid	Chunk kinds written to rag_chunks.jsonl: entity, community, relation_group, or hybrid (all three)
max_chunk_tokens	int	2000	Size budget of one chunk's text, in tokens counted as about 4 characters each (minimum 100). Lists that do not fit end with an ... and N more line.

In config.yaml

config.yaml (rag)

rag:
  enabled: true
  chunk_strategy: hybrid   # chunk kinds written: entity | community | relation_group | hybrid (all)
  # Approximate content size of one chunk (about 4 characters per token);
  # longer lists in a chunk end with an "... and N more" line
  max_chunk_tokens: 2000 # nosemgrep: codacy.yaml.security.hard-coded-tokens -- LLM chunk size, not a token value

Read from the annotated example config.yaml and the CloudGConfig models in cloudg/config.py.