DocVec CLI

🚀 Overview
DocVec CLI is a powerful command-line tool designed to transform your unstructured local documents (PDF, DOCX, TXT) into query-ready vector embeddings, making them instantly usable for Large Language Models (LLMs) and bolstering Retrieval Augmented Generation (RAG) workflows.

✨ Key Features

• Multi-Format Support: Processes .pdf, .docx, and .txt files.
• Automatic Text Extraction: Efficiently extracts raw text content from various document types.
• Intelligent Text Cleaning: Removes unnecessary whitespace, excessive newlines, and basic HTML tags.
• Configurable Text Chunking: Uses langchain's RecursiveCharacterTextSplitter, with customizable chunk_size and chunk_overlap.
• Offline Embedding Generation: Uses local sentence-transformers models (default: all-MiniLM-L6-v2) to create high-quality vector embeddings directly on your machine, ensuring privacy and offline capabilities.
• ChromaDB-Compatible Output: Generates JSON files structured for easy ingestion into ChromaDB or other vector databases.
• User-Friendly CLI: Simple command-line arguments for input/output paths and processing parameters.
• Progress Indicators: Visual progress bars for long-running operations like embedding generation.

---

📦 Installation

Prerequisites

• Python 3.8 or newer

Steps

Using pip:

git pip install docvec-cli
docvec --input-path "your_document.txt" --output-path "output_dir"

1. Clone the repository:

git clone https://github.com/onurbaran/docvec-cli.git  
cd docvec-cli

2. Create and activate a virtual environment:

It’s highly recommended to use a virtual environment to manage dependencies.

python -m venv .venv

# On Windows:
.\.venv\Scripts\activate

# On macOS/Linux:
source ./.venv/bin/activate

3. Install dependencies:

Ensure your requirements.txt contains:

pypdf
python-docx
sentence-transformers
langchain-text-splitters
tqdm
numpy

Then run:

pip install -r requirements.txt

---

🚀 Usage

Once installed, you can use docvec-cli from your terminal.

Basic Command Structure

python src/main.py --input-path <path_to_document_or_directory> --output-path <path_to_output_directory> [OPTIONS]

Required Arguments

• --input-path <path>: Path to a document file (e.g., report.pdf) or a directory (directory processing is planned for future updates).
• --output-path <path>: Path to the directory where the generated vector and metadata files will be saved.

Optional Arguments

• --chunk-size <int>: Max size of each text chunk in characters (default: 1000)
• --chunk-overlap <int>: Number of characters to overlap between chunks (default: 200)
• --model-name <str>: Sentence-transformers model name (default: all-MiniLM-L6-v2)
• --output-format <str>: Format for output files (default: json, only format currently supported)

---

📁 Examples

Process a single PDF file:

python src/main.py --input-path "docs/my_report.pdf" --output-path "vectors/"

Process a DOCX file with custom chunking:

python src/main.py --input-path "articles/research.docx" --output-path "embeddings/" --chunk-size 500 --chunk-overlap 100

Process a TXT file with a different embedding model:

python src/main.py --input-path "notes/daily_journal.txt" --output-path "processed_data/" --model-name "all-MiniLM-L12-v2"

---

📄 Output File Structure

For each processed document (e.g., my_report.pdf), a JSON file (my_report_vectors.json) will be created in the specified --output-path.

Example content:

[
  {
    "id": "my_report-0",
    "document": "This is the text content of the first chunk...",
    "embedding": [0.123, -0.456, ..., 0.789],
    "metadata": {
      "source_file": "my_report.pdf",
      "chunk_index": 0,
      "chunk_size": 250
    }
  }
]

---

🤝 Contributing

We welcome contributions from the community! To contribute:

1. Fork the repository.
2. Create a new branch: git checkout -b feature/your-feature-name
3. Make your changes.
4. Write clear, concise commit messages.
5. Push your branch: git push origin feature/your-feature-name
6. Open a Pull Request.

---

📄 License

This project is licensed under the MIT License.

---

📧 Contact

For questions, feedback, or issues, please open an issue.