RAG Search

Semantic search using embeddings and vector storage. Search documents semantically using similarity matching.

Setup

No additional setup required. Uses in-memory vector storage with optional embedding providers.

Usage

Index Documents

{baseDir}/rag-search.js --index --path ./docs --chunk-size 500

Search Documents

{baseDir}/rag-search.js --search "how to configure authentication"

Query with Filters

{baseDir}/rag-search.js --search "deployment steps" --limit 5

Options

Option	Description	Required
--index	Index documents	No
--path	Path to documents	For index
--chunk-size	Chunk size for splitting	No
--search	Search query	For search
--limit	Max results to return	No
--list	List indexed documents	No
--clear	Clear index	No

Supported Formats

• Plain text (.txt)
• Markdown (.md)
• JSON (.json)
• JavaScript/TypeScript (.js, .ts)
• Python (.py)
• HTML (.html)
• YAML (.yaml, .yml)

Embedding Providers

• OpenAI (default, requires API key)
• Cohere (requires API key)
• Local (TF-IDF based, no API key needed)

Output Format

{
  "results": [
    {
      "file": "docs/config.md",
      "chunk": "To configure authentication...",
      "score": 0.92,
      "line": 15
    }
  ]
}

When to Use

• Semantic search across codebase
• Finding relevant documentation
• Knowledge base queries
• RAG applications