rag

The rag module provides a 2-line RAG (Retrieval-Augmented Generation) system.

Import

from openstackai.easy import rag

Basic Usage

# Create RAG system with documents
docs = rag.load("./documents/")

# Query the documents
answer = docs.query("What is the main topic?")

Quick Start

from openstackai.easy import rag

# Load documents from directory
knowledge = rag.load("./knowledge_base/")

# Ask questions
answer = knowledge.query("How do I configure the system?")
print(answer)

# With source citations
result = knowledge.query("What are the features?", return_sources=True)
print(result.answer)
print(result.sources)

Loading Documents

From Directory

# Load all supported files from directory
docs = rag.load("./docs/")

# With file filter
docs = rag.load("./docs/", pattern="*.md")

From Files

# Load specific files
docs = rag.load(["doc1.pdf", "doc2.txt", "doc3.md"])

From URLs

# Load from web pages
docs = rag.load([
    "https://example.com/page1",
    "https://example.com/page2"
])

Query Options

Parameter	Type	Default	Description
query	str	required	Question to ask
top_k	int	5	Number of chunks to retrieve
return_sources	bool	False	Include source citations
temperature	float	0.7	Response creativity

Examples

Full Example

from openstackai.easy import rag

# Create knowledge base
kb = rag.load("./company_docs/")

# Simple query
answer = kb.query("What is our refund policy?")

# Query with sources
result = kb.query(
    "What are the product features?",
    return_sources=True,
    top_k=3
)

for source in result.sources:
    print(f"- {source.file}: {source.chunk}")

Async Usage

import asyncio
from openstackai.easy import rag

async def main():
    kb = await rag.load_async("./docs/")
    answer = await kb.query_async("What is openstackai?")
    print(answer)

asyncio.run(main())

Supported Formats

• PDF (.pdf)
• Text (.txt)
• Markdown (.md)
• Word (.docx)
• HTML (.html)
• JSON (.json)

Configuration

# Custom embedding model
rag.configure(
    embedding_model="text-embedding-3-small",
    chunk_size=500,
    chunk_overlap=50
)

See Also

• [[VectorDB-Module]] - Vector database backends
• [[ChromaDB]] - ChromaDB integration
• [[ask]] - Simple question answering