QV-RAG

A simple and efficient RAG (Retrieval-Augmented Generation) engine for semantic search and document retrieval.

Note: This is an experimental project used for research and development. The engine is functional but will continue to evolve and improve over time.

Features

• Simple API: Easy-to-use interface for document management and querying
• Multiple Document Types: Support for text, markdown, HTML, and JSON files
• Smart Chunking: Intelligent text splitting with metadata preservation
• Semantic Search: Built-in semantic search using ChromaDB
• Metadata Support: Flexible metadata handling for better document organization

Installation

pip install qv-rag

Quick Start

from qv_rag.engine import RAGEngine

# Initialize the engine
engine = RAGEngine(
    collection_name="my_docs",
    chunk_size=1000,
    chunk_overlap=200
)

# Add documents
engine.add_texts(["Python is a popular programming language."])
engine.add_file("document.md")

# Query documents
results = engine.query("What is Python?")
for result in results:
    print(result['text'])
    print(f"Score: {result['distance']}")

Usage Examples

Adding Documents

# Add text with metadata
engine.add_texts(
    texts=["Document content"],
    metadatas=[{"source": "manual", "category": "docs"}]
)

# Add a file
engine.add_file("document.md", metadata={"source": "file"})

Querying

# Simple query
results = engine.query("What is machine learning?")

# Query with filters
results = engine.query(
    "What is Python?",
    where={"category": "docs"},
    top_k=3
)

Supported File Types

• Text files (.txt)
• Markdown files (.md)
• HTML files (.html, .htm)
• JSON files (.json)
• PDF files (.pdf) --> must be further improved

License

MIT License