Seamless Document Processing with Google Gemini API

🌍 gemini-parser is your all-in-one Python library for document parsing using the Google Gemini API.

⚡ It enables developers to transcribe PDFs, extract structured data, and summarize large documents with ease.

🚀 Whether you're transcribing PDFs, extracting structured data, or summarizing large documents, gemini-parser delivers fast, reliable, and efficient results.

🔥 Why choose gemini-parser?

• 🚀 AI-powered document parsing with Google Gemini API.
• 📦 Handles large files effortlessly (up to 20GB).
• ⚡ Smart caching for faster processing and reduced costs.
• 📑 Supports multiple formats: PDF, CSV, HTML, DOC, XML, TXT.
• 🌐 Flexible inputs from local files, folders, or URLs.
• 📊 Structured data extraction for easy analysis.
• 📜 Handle Large output context unlike gemini output context of ~8000 tokens, gemini-parser doesn't have a limit.

📦 Installation

pip install gemini-parser

Usage

⚠️ The library works perfectly fine for PDFs now. For other files type I will be adding support soon.

Quickstart Example

from gemini_parser import DocumentProcessor
from pathlib import Path
import os
from dotenv import load_dotenv

load_dotenv()

# Initialize processor with API key
processor = DocumentProcessor(api_key=os.getenv("GEMINI_API_KEY"))

# Process a single PDF file
result = processor.process_file(Path("path/to/document.pdf"))
print(result)

# Process from a URL
url_result = processor.process_from_url("https://example.com/document.pdf")
print(url_result)

# Process all files in a folder
processor.process_folder(Path("path/to/folder"))

# List all caches
caches = processor.list_caches()
print(caches)

# Delete a cache
processor.delete_cache("cachedContentID")

Configuration

You can customize key parameters when initializing the DocumentProcessor:

• api_key: Your Gemini API key.
• model_name: The Gemini model (e.g., gemini-1.5-flash-002).
• prompt: The default processing prompt.
• log_level: Logging level (INFO, DEBUG, etc.).

API Reference

DocumentProcessor

• process_file(file_path, use_cache=False, cache_ttl=None): Processes a local file.
• process_from_url(url, use_cache=False, cache_ttl=None): Processes a document from a URL.
• process_multiple_files(file_paths, use_cache=False, cache_ttl=None): Processes multiple files.
• process_folder(folder_path, output_dir=None, out_ext="md", use_cache=False, cache_ttl=None): Processes all files in a folder.
• list_caches(): Lists all available caches.
• delete_cache(cache_name): Deletes a cache by name.

FileManager

• upload_file(file_or_path, mime_type): Uploads files to Gemini.
• list_files(): Lists all uploaded files.
• get_file(file_name): Gets metadata of a file.
• delete_file(file_name): Deletes an uploaded file.

CachingManager

• create_cache(model_name, contents, system_instruction): Creates a new cache.
• generate_with_cache(model_name, cached_content_name, prompt): Generates content using a cache.
• list_caches(): Lists all cached content.
• update_cache_ttl(cache_name, hours): Updates the cache TTL.
• delete_cache(cache_name): Deletes a cache.

Testing

Your library includes pytest-based tests in the tests/ folder. Run them with:

pytest tests/

Requirements

• Python 3.8+
• tqdm
• PyPDF2
• google-genai
• python-dotenv
• httpx

Project Structure

gemini-parser/
│
├── gemini_parser/
│   ├── document_processor.py
│   ├── file_manager.py
│   ├── caching.py
│   ├── utils.py
│
├── tests/
│   ├── test_gemini_parser.py
│
├── setup.py
├── pyproject.toml
├── README.md
└── LICENSE

License

This project is licensed under the MIT License.

Contributing

Contributions are welcome! Please fork this repository, create a new branch, and submit a pull request.

Author

Developed by Thimira Nirmal
📧 timnirmal@gmail.com
🌐 GitHub | Website