ragscore 0.1.4

Generate high-quality QA datasets to evaluate RAG systems

Generate high-quality QA datasets to evaluate your RAG systems

---

RAGScore automatically generates question-answer pairs from your documents, which you can then use to benchmark and evaluate your RAG (Retrieval-Augmented Generation) systems.

✨ Features

• 📄 Multi-format support - PDF, TXT, Markdown, HTML
• 🌍 Multi-language - English and Chinese out of the box
• 🤖 Multi-provider - OpenAI, DashScope (Qwen), or any OpenAI-compatible API
• 🎯 Difficulty levels - Easy, medium, and hard questions
• 🚀 Simple CLI - Easy command-line interface
• ⚡ Fast indexing - FAISS-powered vector search

🚀 Quick Start

Installation

# Basic installation through pypi
pip install ragscore

# With OpenAI support
pip install ragscore[openai]

# With DashScope support (Chinese users)
pip install ragscore[dashscope]

# All providers
pip install ragscore[all]

Note: On first run, RAGScore automatically downloads required NLTK data (~35MB). This only happens once.

Setup API Key

# For OpenAI
export OPENAI_API_KEY="your-openai-key"

# For DashScope (Alibaba Cloud)
export DASHSCOPE_API_KEY="your-dashscope-key"

Generate QA Pairs

# Place documents in data/docs/, then:
ragscore generate --docs-dir YOUR-PDF-DIRECTORY

Output

Generated QA pairs are saved to output/generated_qas.jsonl:

{
  "id": "abc123",
  "question": "What is RAG?",
  "answer": "RAG (Retrieval-Augmented Generation) combines information retrieval with text generation...",
  "difficulty": "easy",
  "source_path": "docs/rag_intro.pdf"
}

📖 Usage

Command Line

# Generate QA pairs from documents
ragscore generate  --docs-dir YOUR-PDF-DIRECTORY

# Force re-indexing of documents
ragscore generate --force-reindex

# Use specific provider
ragscore generate --provider openai --model gpt-4o

Python API

from ragscore.pipeline import run_pipeline
from ragscore.data_processing import read_docs
from ragscore.llm import generate_qa_for_chunk

# Run full pipeline
run_pipeline(force_reindex=True)

# Or use individual components
docs = read_docs(dir_path="./my_docs")
for doc in docs:
    qas = generate_qa_for_chunk(doc["text"], difficulty="medium", n=5)
    print(qas)

⚙️ Configuration

Create a .env file or set environment variables:

# LLM Provider (auto-detected from available API keys)
DASHSCOPE_API_KEY="your-key"  # For DashScope/Qwen
OPENAI_API_KEY="your-key"     # For OpenAI

# Optional: Custom settings
RAGSCORE_CHUNK_SIZE=512
RAGSCORE_QUESTIONS_PER_CHUNK=5

🔌 Supported LLM Providers

RAGScore works with any LLM provider - use your own API keys!

Provider	Models	Environment Variable
OpenAI	gpt-4o, gpt-4o-mini, gpt-3.5-turbo	OPENAI_API_KEY
Anthropic	claude-3-opus, claude-3-sonnet, claude-3-haiku	ANTHROPIC_API_KEY
Groq	llama-3.1-70b, mixtral (ultra fast!)	GROQ_API_KEY
Together AI	llama-3, mistral, many open models	TOGETHER_API_KEY
Grok (xAI)	grok-beta	XAI_API_KEY
Mistral	mistral-large, mistral-medium	MISTRAL_API_KEY
DeepSeek	deepseek-chat, deepseek-coder	DEEPSEEK_API_KEY
DashScope	qwen-turbo, qwen-plus, qwen-max	DASHSCOPE_API_KEY
Ollama	llama2, mistral, codellama (local!)	No key needed
Custom	Any OpenAI-compatible endpoint	LLM_BASE_URL

Using Ollama (Free, Local)

# Install Ollama: https://ollama.ai
ollama pull llama2
ollama serve

# RAGScore auto-detects Ollama
ragscore generate

Using Custom Endpoints

# Any OpenAI-compatible API (vLLM, LocalAI, etc.)
export LLM_BASE_URL="http://localhost:8000/v1"
export LLM_MODEL="my-model"
ragscore generate

📁 Project Structure

ragscore/
├── data/docs/          # Place your documents here
├── output/             # Generated QA pairs and index
│   ├── generated_qas.jsonl
│   ├── index.faiss
│   └── meta.json
└── src/ragscore/       # Source code
    ├── cli.py          # Command-line interface
    ├── pipeline.py     # Main pipeline
    ├── data_processing.py
    ├── vector_store.py
    ├── llm.py
    └── providers/      # LLM provider implementations

🚀 RAGScore Pro (Coming Soon)

Need to evaluate your RAG system? RAGScore Pro offers:

• 🔍 Hallucination Detection - Catch when your RAG makes things up
• 📝 Citation Quality Scoring - Verify source attribution accuracy
• 📊 Multi-dimensional Scoring - Accuracy, relevance, completeness
• 📈 Executive Reports - Excel reports for stakeholders
• ⚡ API Access - Integrate evaluation into your CI/CD

Join the waitlist →

🧪 Development

# Clone repository
git clone https://github.com/HZYAI/RagScore.git
cd RagScore

# Install with dev dependencies
pip install -e ".[dev,all]"

# Run tests
pytest

# Run linting
ruff check src/
black --check src/

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

📄 License

Apache 2.0 License - see LICENSE for details.

🔗 Links

• Documentation
• Changelog
• Issue Tracker

---

Made with ❤️ for the RAG community