cliven 0.2.0

Chat with your PDFs using local AI models

Cliven 🤖

Chat with your PDFs using local AI models!

Cliven is a command-line tool that allows you to process PDF documents and have interactive conversations with their content using local AI models. No data leaves your machine - everything runs locally using ChromaDB for vector storage and Ollama for AI inference.

Features ✨

• 📄 PDF Processing: Extract and chunk text from PDF documents
• 🔍 Vector Search: Find relevant content using semantic similarity
• 🤖 Local AI Chat: Chat with your documents using Ollama models
• 🐳 Docker Ready: Easy setup with Docker Compose
• 💾 Local Storage: All data stays on your machine
• 🎯 Simple CLI: Easy-to-use command-line interface
• 🚀 Model Selection: Support for both lightweight (TinyLlama) and high-performance (Mistral) models
• 📊 Rich UI: Beautiful terminal interface with progress indicators

Quick Start 🚀

1. Clone the Repository

git clone https://github.com/krey-yon/cliven.git
cd cliven

2. Install Dependencies

pip install -e .

3. Start Services with Docker

# Start with lightweight model (tinyllama:chat)
cliven docker start

# OR start with high-performance model (mistral:7b)
cliven docker start --BP
# or
cliven docker start --better-performance

This will:

• Start ChromaDB on port 8000
• Start Ollama on port 11434
• Pull the gemma2:2b model (default) or gemma3:4b model (with --BP flag)
• May take several minutes depending on model and connection speed

4. Process Your First PDF

cliven ingest path/to/your/document.pdf

5. Start Chatting

# Chat with existing documents
cliven chat

# OR specify a model
cliven chat --model gemma3:2b

Usage 📖

Available Commands

# Show welcome message and commands
cliven

# Process and store a PDF
cliven ingest <pdf_path> [--chunk-size SIZE] [--overlap SIZE]

# Start interactive chat with existing documents
cliven chat [--model MODEL_NAME] [--max-results COUNT]

# Process PDF and start chat immediately
cliven chat --repl <pdf_path> [--model MODEL_NAME]

# List all processed documents
cliven list

# Delete a specific document
cliven delete <doc_id>

# Clear all documents
cliven clear [--confirm]

# Check system status
cliven status

# Manage Docker services
cliven docker start [--BP | --better-performance]  # Start services
cliven docker stop                                 # Stop services
cliven docker logs                                 # View logs

Examples

# Process a manual with custom chunking
cliven ingest ./documents/user-manual.pdf --chunk-size 1500 --overlap 300

# Start chatting with all processed documents
cliven chat

# Chat with specific model
cliven chat --model mistral:7b

# Process and chat with a specific PDF using high-performance model
cliven chat --repl ./research-paper.pdf --model mistral:7b

# Check what documents are stored
cliven list

# Check if services are running
cliven status

# Clear all documents without confirmation
cliven clear --confirm

# Start services with better performance model
cliven docker start --BP

Model Options

Cliven supports multiple AI models:

• gemma2:2b: Lightweight, fast responses (~1GB model)
• gemma3:4b: High-performance, better quality responses (~4GB model)

The system automatically selects the best available model, or you can specify one:

# Auto-select best available model
cliven chat

# Use specific model
cliven chat --model gemma3:4b
cliven chat --model gemma2:2b

Architecture 🏗️

Cliven uses a modern RAG (Retrieval-Augmented Generation) architecture:

1. PDF Parser: Extracts text from PDFs using pdfplumber
2. Text Chunker: Splits documents into overlapping chunks using LangChain
3. Embedder: Creates embeddings using BAAI/bge-small-en-v1.5
4. Vector Database: Stores embeddings in ChromaDB
5. Chat Engine: Handles queries and generates responses with Ollama

Components 🔧

Core Services

• ChromaDB: Vector database for storing document embeddings
• Ollama: Local LLM inference server
• Gemma2:2b: Lightweight chat model for fast responses
• Gemma3:4b: High-performance model for better quality responses

Key Files

• main/cliven.py: Main CLI application with argument parsing
• main/chat.py: Chat engine with RAG functionality and model management
• utils/parser.py: PDF text extraction and chunking
• utils/embedder.py: Text embedding generation using sentence transformers
• utils/vectordb.py: ChromaDB operations and vector storage
• utils/chunker.py: Text chunking utilities
• docker-compose.yml: Service orchestration configuration

System Requirements 📋

Software Requirements

• Python 3.8+
• Docker & Docker Compose
• 2GB+ RAM (for TinyLlama model)
• 8GB+ RAM (for Mistral 7B model)
• 4GB+ disk space

Python Dependencies

• typer>=0.9.0 - CLI framework
• rich>=13.0.0 - Beautiful terminal output
• pdfplumber>=0.7.0 - PDF text extraction
• sentence-transformers>=2.2.0 - Text embeddings
• chromadb>=0.4.0 - Vector database
• langchain>=0.0.300 - Text processing
• requests>=2.28.0 - HTTP client

Installation Options 🛠️

Option 1: Local Development

# Clone repository
git clone https://github.com/krey-yon/cliven.git
cd cliven

# Create virtual environment
python -m venv .venv
.venv\Scripts\activate

# Install dependencies
pip install -e .

# Start services
cliven docker start

Option 2: Production Install

pip install git+https://github.com/krey-yon/cliven.git

Configuration ⚙️

Environment Variables

# ChromaDB settings
CHROMA_HOST=localhost
CHROMA_PORT=8000

# Ollama settings
OLLAMA_HOST=localhost
OLLAMA_PORT=11434

Customization

# Use different chunk sizes
cliven ingest document.pdf --chunk-size 1500 --overlap 300

# Use different model
cliven chat --model mistral:7b

# Adjust context window
cliven chat --max-results 10

# Skip confirmation for clearing
cliven clear --confirm

Model Management

# Check available models
cliven status

# Manually pull models
docker exec -it cliven_ollama ollama pull mistral:7b
docker exec -it cliven_ollama ollama pull tinyllama:chat

# List downloaded models
docker exec -it cliven_ollama ollama list

Troubleshooting 🔧

Common Issues

1. Docker services not starting

   # Check Docker daemon
   docker info
   
   # View service logs
   cliven docker logs
   
   # Restart services
   cliven docker stop
   cliven docker start
   

2. Model not found

   # Check available models
   cliven status
   
   # Manually pull model
   docker exec -it cliven_ollama ollama pull mistral:7b
   docker exec -it cliven_ollama ollama pull tinyllama:chat
   

3. ChromaDB connection failed

   # Check service status
   cliven status
   
   # Restart services
   cliven docker stop
   cliven docker start
   
   # Check logs
   cliven docker logs
   

4. PDF processing errors

   # Check file path and permissions
   dir path\to\file.pdf
   
   # Try with different chunk size
   cliven ingest file.pdf --chunk-size 500
   
   # Check for PDF corruption
   cliven ingest file.pdf --chunk-size 2000 --overlap 100
   

5. Model performance issues

   # Switch to lightweight model
   cliven chat --model gemma2:2b
   
   # Or use high-performance model
   cliven chat --model gemma3:4b
   
   # Check system resources
   cliven status
   

Performance Tips

• Use gemma2:2b for faster responses on limited hardware
• Use gemma3:4b for better quality responses with sufficient RAM
• Use smaller chunk sizes for better context precision
• Increase overlap for better continuity
• Monitor RAM usage with large PDFs
• Use SSD storage for better ChromaDB performance

Contributing 🤝

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

License 📄

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

Acknowledgments 🙏

• ChromaDB for vector storage
• Ollama for local LLM inference
• Sentence Transformers for embeddings
• LangChain for text processing
• Rich for beautiful terminal output
• PDFplumber for PDF text extraction

Support 💬

• 📧 Email: [e-mail]
• 🐛 Issues: GitHub Issues
• 💡 Discussions: GitHub Discussions

---

Made with ❤️ by Kreyon

Chat with your PDFs locally and securely!