retrieve

Knowledge Base Retrieve

Search and retrieve relevant documents from an Amazon Bedrock Knowledge Base using RAG.

Quick Start

Use the retrieve script to query the knowledge base by keyword:

import subprocess
result = subprocess.run(['python', 'scripts/retrieve_search.py', 'keyword'], 
                       capture_output=True, text=True, cwd='retrieve')
print(result.stdout)

Script Location

The retrieve script is located at skills/retrieve/scripts/retrieve_search.py relative to the application working directory. IMPORTANT: Always use the FULL path skills/retrieve/scripts/retrieve_search.py — do NOT shorten to scripts/retrieve_search.py.

Features

• Knowledge Base Search: Query documents indexed in Amazon Bedrock Knowledge Base
• RAG-based Retrieval: Uses vector search to find the most relevant content
• Source References: Returns source URLs and document titles for each result
• Auto Recovery: Automatically updates knowledge base ID if the resource is not found
• S3 & Web Sources: Supports both S3 document and web-crawled content locations
• Top Results: Returns up to 5 most relevant document chunks

Usage Examples

Basic Query

# Search for information about a topic
result = subprocess.run(['python', 'scripts/retrieve_search.py', '클라우드 아키텍처'], 
                       capture_output=True, text=True, cwd='retrieve')

Technical Query

# Search for technical documentation
result = subprocess.run(['python', 'scripts/retrieve_search.py', 'API authentication'], 
                       capture_output=True, text=True, cwd='retrieve')

Concept Query

# Search for explanations
result = subprocess.run(['python', 'scripts/retrieve_search.py', '보안 정책'], 
                       capture_output=True, text=True, cwd='retrieve')

Output Format

The script returns a JSON array. Each element contains:

[
  {
    "contents": "matched text from the knowledge base",
    "reference": {
      "url": "https://...",
      "title": "document_name.pdf",
      "from": "RAG"
    }
  }
]