RAG Backend provides document retrieval functionality, supporting vector search and hybrid retrieval.

Overview

• Type: Shared Resource Backend
• Requires Session: ❌ No
• Tool Prefix: rag:

Available Tools

Tool	Description	Parameters
rag:search	Search documents	query, top_k
rag:index	Index documents	documents

Usage Example

async with Sandbox() as sandbox:
    # RAG doesn't need Session (shared resource)
    
    result = await sandbox.execute("rag:search", {
        "query": "What is machine learning?",
        "top_k": 5
    })
    
    for doc in result["data"]["results"]:
        print(f"Score: {doc['score']}, Content: {doc['content'][:100]}")