Documentation

RAG

RAG indexing and search in packages/nikcli/src/rag.

Overview

RAG uses local chunking + vector storage with configurable embedding models.

Rag.index() discovers files, chunks text, and stores vectors in RagStorage.

Defaults

Default embed model is nvidia/llama-embed-nemotron-8b with provider nvidia.

`chunkLines`	200
`maxFiles`	200
`maxChunks`	5000
`maxFileBytes`	1,000,000
`model`	nvidia/llama-embed-nemotron-8b
`provider`	nvidia