Base de conocimiento RAG

Documentación en español de Base de conocimiento RAG, alineada con el monorepo y la guía inglesa.

Entrada rápida

bash

# Enable via CLI
commander plugin enable rag

# Or in .commander.json
{
  "plugins": {
    "builtin-rag": { "enabled": true }
  }
}

Document → Chunk (512 chars, 50 overlap) → Embed → Index (HNSW) → Persist

typescript

// Tool parameters
{
  query: string,      // Search query
  topK: number        // Results to return (1-50, default 5)
}

Option	Default	Description
`kbPath`	`.commander/knowledge-base/`	Storage directory for documents and vectors
`embeddingModel`	`text-embedding-3-small`	OpenAI embedding model (when API key available)
`chunkSize`	`512`	Document chunk size in characters
`chunkOverlap`	`50`	Overlap between chunks
`maxResults`	`5`	Maximum search results returned
`autoInject`	`false`	Auto-inject relevant context before LLM calls