🚀 RAG Studio

Production-ready document processing CLI for RAG applications

Process documents, extract text with advanced OCR, chunk intelligently, and prepare data for RAG systems - all from the command line with ragctl.

---

🎯 What is RAG Studio?

RAG Studio (ragctl) is a command-line tool for processing documents into chunks ready for Retrieval-Augmented Generation (RAG) systems. It handles the dirty work of document ingestion, OCR, and intelligent chunking so you can focus on building your RAG application.

Key capabilities:

• 📄 Universal document loading (PDF, DOCX, images, HTML, Markdown, etc.)
• 🔍 Advanced OCR with automatic fallback (EasyOCR → PaddleOCR → pytesseract)
• ✂️ Intelligent semantic chunking using LangChain
• 📦 Production-ready batch processing with auto-retry
• 💾 Multiple export formats (JSON, JSONL, CSV)
• 🗄️ Direct ingestion into Qdrant vector store

---

✨ Features

📄 Universal Document Processing

• Supported formats: PDF, DOCX, ODT, TXT, HTML, Markdown, Images (JPEG, PNG)
• Smart OCR cascade:

1. EasyOCR (best quality, multi-language)
2. PaddleOCR (fast, good for complex layouts)
3. pytesseract (fallback, most tolerant)

• Quality detection: Automatically rejects unreadable documents
• Multi-language: French, English, German, Spanish, Italian, Portuguese, and more

✂️ Intelligent Chunking

• Semantic chunking: Context-aware text splitting using LangChain RecursiveCharacterTextSplitter

• Multiple strategies:

• semantic - Smart splitting by meaning (default)

• sentence - Split by sentences

• token - Fixed token-based splitting

• Configurable: Token limits (50-2000), overlap (0-500), model selection

• Rich metadata: Source file, chunk index, token count, strategy, timestamps

🔄 Production-Ready Batch Processing

• Automatic retry: Up to 3 attempts with exponential backoff (1s, 2s, 4s...)

• Interactive error handling:

• interactive - Prompt user on each error (default)

• auto-continue - Continue on errors (CI/CD mode)

• auto-stop - Stop on first error (validation mode)

• auto-skip - Skip failed files automatically

• Complete history: Every run saved to ~/.atlasrag/history/

• Retry capability: ragctl retry to rerun failed files only

• Per-file output: One chunk file per document for better traceability

💾 Flexible Export & Storage

• Export formats: JSON, JSONL (streaming), CSV (Excel-compatible)
• Vector store integration: Direct ingestion into Qdrant
• No database required: Pure file-based export for easy sharing

⚙️ Configuration System

• Hierarchical config: CLI flags > Environment variables > YAML file > Defaults
• Example config: config.example.yml with detailed documentation
• Easy customization: Override any setting via command line

---

🚀 Quick Start

Installation

From PyPI (Recommended)

# Install from PyPI
pip install ragctl

# Verify installation
ragctl --version

From Source

# Clone repository
git clone git@github.com:datallmhub/ragstudio.git
cd ragstudio

# Install with pip
pip install -e .

# Verify installation
ragctl --version

Basic Usage

# Process a single document
ragctl chunk document.pdf --show

# Process with advanced OCR for scanned documents
ragctl chunk scanned.pdf --advanced-ocr -o chunks.json

# Batch process a folder
ragctl batch ./documents --output ./chunks/

# Batch with auto-retry for CI/CD
ragctl batch ./documents --output ./chunks/ --auto-continue

---

💡 Usage Examples

Single Document Processing

# Simple text file
ragctl chunk document.txt --show

# PDF with semantic chunking (default)
ragctl chunk report.pdf -o report_chunks.json

# Scanned image with OCR
ragctl chunk contract.jpeg --advanced-ocr --show

# Custom chunking parameters
ragctl chunk document.pdf \
--strategy semantic \
--max-tokens 500 \
--overlap 100 \
-o output.jsonl

Batch Processing

# Process all files in a directory
ragctl batch ./documents --output ./chunks/

# Process only PDFs recursively
ragctl batch ./documents \
--pattern "*.pdf" \
--recursive \
--output ./chunks/

# CI/CD mode - continue on errors
ragctl batch ./documents \
--output ./chunks/ \
--auto-continue \
--save-history

# Per-file output (default):
# chunks/
# ├── doc1_chunks.jsonl  (25 chunks)
# ├── doc2_chunks.jsonl  (42 chunks)
# └── doc3_chunks.jsonl  (18 chunks)

# Single-file output (all chunks combined):
ragctl batch ./documents \
--output ./all_chunks.jsonl \
--single-file

Retry Failed Files

# Show last failed run
ragctl retry --show

# Retry all failed files from last run
ragctl retry

# Retry specific run by ID
ragctl retry run_20251028_133403

Vector Store Integration

# Ingest chunks into Qdrant
ragctl ingest chunks.jsonl \
--collection my-docs \
--url http://localhost:6333

# Get system info
ragctl info

Evaluate Chunking Quality

# Evaluate chunking strategy
ragctl eval document.pdf \
--strategies semantic sentence token \
--metrics coverage overlap coherence

# Compare strategies with visualization
ragctl eval document.pdf --compare --output eval_results.json

---

📚 Documentation

Document	Description
Getting Started	Installation and first steps
CLI Guide	Complete command reference
Security	Security features and best practices
Full Documentation	Complete documentation index

---

⚙️ Configuration

Create ~/.atlasrag/config.yml or use CLI flags:

# OCR settings
ocr:
use_advanced_ocr: false
enable_fallback: true

# Chunking settings
chunking:
strategy: semantic
max_tokens: 400
overlap: 50

# Output settings
output:
format: jsonl
include_metadata: true
pretty_print: true

Configuration hierarchy: CLI flags > Environment variables > YAML config > Defaults

---

🧪 Testing

# Run all tests
make test

# Run CLI tests
make test-cli

# Quick validation
ragctl --version
ragctl chunk tests/data/sample.txt --show

Test Coverage: 496 tests, 41% coverage

---

📊 Performance

Processing Speed

• Text documents: ~100-200 docs/minute
• PDFs with OCR: ~5-10 docs/minute (depends on page count)
• Batch processing: Parallel-ready with retry mechanism

Quality Metrics

• OCR accuracy: 95%+ with EasyOCR on clear scans
• Chunk quality: 90% readability threshold enforced
• Semantic coherence: LangChain’s RecursiveCharacterTextSplitter optimized for context

---

🛠️ CLI Commands

Command	Description
ragctl chunk	Process a single document
ragctl batch	Batch process multiple files
ragctl retry	Retry failed files from history
ragctl ingest	Ingest chunks into Qdrant
ragctl eval	Evaluate chunking quality
ragctl info	System information

Run ragctl COMMAND --help for detailed options.

---

🐛 Troubleshooting

Common Issues

NumPy incompatibility

# For OCR support, use NumPy 1.x
pip install "numpy<2.0"

Missing system dependencies

# Ubuntu/Debian
sudo apt-get install tesseract-ocr poppler-utils

# macOS
brew install tesseract poppler

"Document unreadable" errors

• Try lowering quality threshold: --ocr-threshold 0.2
• Use advanced OCR: --advanced-ocr
• Check document is not corrupted

Import errors

# Reinstall dependencies
pip install -e .

More help: Getting Started Guide

---

🔧 Development

# Install dev dependencies
make install-dev

# Format code
make format

# Run linters
make lint

# Install pre-commit hooks
make pre-commit-install

# Run all CI checks
make ci-all

---

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

---

📧 Support

• Documentation: docs/
• Issues: GitHub Issues
• Discussions: GitHub Discussions

---

🙏 Acknowledgments

Built with:

• LangChain - Text splitting and document loading
• EasyOCR - OCR engine
• PaddleOCR - Alternative OCR engine
• Unstructured - Document parsing
• Typer - CLI framework
• Rich - Terminal formatting

---

Version: 0.1.3 | Status: Beta | License: MIT

Test