Coding & Development

Graph4NLP is a comprehensive open-source library designed to simplify the application of Graph Neural Networks (GNNs) to Natural Language Processing (NLP) tasks. It caters to both data scientists seeking ready-to-use, state-of-the-art model implementations and researchers/developers who require flexible interfaces to build customized models with full pipeline support. Built upon highly-optimized runtime libraries like DGL, Graph4NLP ensures high running efficiency and extensibility. The library features a four-layer architecture comprising Data, Module, Model, and Application layers, supporting a wide range of NLP applications including text classification, semantic parsing, neural machine translation, summarization, and knowledge graph completion. It also provides models like Graph2Seq and Graph2Tree for various graph-to-sequence and graph-to-tree problems.

jailbreak_llms is the official repository for the ACM CCS 2024 paper "'Do Anything Now': Characterizing and Evaluating In-The-Wild Jailbreak Prompts on Large Language Models." This project introduces JailbreakHub, a framework used to conduct the first measurement study on in-the-wild jailbreak prompts. The dataset comprises 15,140 prompts collected from December 2022 to December 2023, sourced from platforms like Reddit, Discord, various websites, and open-source datasets. Among these, 1,405 are identified as jailbreak prompts, making it the largest collection of its kind. The dataset is intended for research purposes, allowing users to load prompts using the Hugging Face Datasets library and evaluate LLM effectiveness against a question set of 390 questions across 13 forbidden scenarios.

llm-compressor is a powerful, open-source library designed to optimize Large Language Models (LLMs) for efficient deployment, particularly with vLLM. It provides a comprehensive suite of quantization algorithms, supporting both weight-only and activation quantization, including advanced schemes like FP8, MXFP8, and NVFP4. The library ensures seamless integration with Hugging Face models and repositories, utilizing a safetensors-based file format compatible with vLLM. Key features include support for large models via accelerate, distributed GPTQ for faster calibration, and updated offloading capabilities for very large models that exceed CPU memory. It also introduces a model-free PTQ pathway for quantization without Hugging Face model definitions, making it versatile for various optimization needs.

lora-scripts, also known as SD-Trainer, is a comprehensive open-source solution for training Stable Diffusion models. It offers both command-line scripts and a user-friendly graphical user interface (GUI) for LoRA and Dreambooth training, leveraging the robust kohya-ss trainer. The tool simplifies the setup process with a one-key training environment, making it accessible for users to get started quickly. It also includes a dedicated WebUI for an integrated Stable Diffusion training studio experience, along with features like Tensorboard integration, WD 1.4 Tagger, and a Tag Editor. The project is designed for developers and researchers working with diffusion models, providing flexible installation options for both Windows and Linux.

neural-redis is a Redis loadable module designed to integrate feed-forward neural networks directly into Redis as a native data type. It aims to simplify machine learning for Redis users by compressing data collection, training, and execution into a single API. The module supports online training of neural networks in different threads, automatic data normalization, and the ability to use the network while it's training. It implements fully connected neural networks using the RPROP learning algorithm and includes automatic training with simple overtraining detection. This makes it suitable for regression and classification problems in applications like mobile and web development, helping developers answer questions related to user preferences, ad conversions, and data trends.

Neuraxle is an open-source Machine Learning (ML) library designed for building clean and production-ready deep learning pipelines. It emphasizes component-based design, allowing users to create encapsulated steps and compose them into complex pipelines. A core feature is its robust hyperparameter tuning capabilities, where each pipeline step can have its own hyperparameter space, facilitating optimization through AutoML algorithms like TPE. Neuraxle is highly compatible with popular ML libraries such as Scikit-Learn and TensorFlow, enabling seamless integration. It also supports evolving states within pipeline steps and offers streaming pipeline functionality for parallel data transformation using multiprocessing queues, making it suitable for scalable and efficient ML workflows.

ResAdapter is an open-source, plug-and-play resolution adapter for diffusion models, developed by ByteDance Inc. It allows users to generate resolution-free images without requiring additional training, inference, or style transfer. The tool is compatible with various diffusion models and can be integrated with personalized models, ControlNet, IP-Adapter, and LCM-LoRA to enhance image generation capabilities. ResAdapter offers different model weights optimized for various resolution and ratio ranges, providing flexibility for diverse image generation tasks. It includes quick tour examples, installation guidance, and community resources like Gradio and ComfyUI integrations, making it accessible for developers and researchers working with diffusion models.

Ollama provides an easy way to get started with and build applications using open large language models like gpt-oss, Gemma 3, and DeepSeek-R1. Users can run models on their own hardware for unlimited usage and data privacy, or leverage Ollama's cloud for faster access to larger models and parallel processing. The platform supports CLI, API, and desktop apps, offering over 40,000 community integrations. It's designed for automating tasks such as coding and document analysis, with a strong emphasis on keeping user data safe and never training on prompts or responses. Ollama also offers tiered cloud plans for increased model concurrency and usage, catering to various demands from light chatting to heavy, sustained agent tasks.

qdrant-client is a Python client library designed for seamless interaction with the Qdrant vector search engine. It offers comprehensive type definitions for all Qdrant API methods, facilitating both synchronous and asynchronous requests. The library supports a local mode for development, prototyping, and testing without requiring a running Qdrant server, and can easily switch to server mode for scaling. Key features include REST and gRPC support, minimal dependencies, and extensive test coverage. Additionally, it provides an Inference API for creating embeddings locally with FastEmbed or remotely with Qdrant Cloud models, simplifying the process of generating and uploading vectors.

Rag-chatbot is an open-source tool designed for local interaction with multiple PDF documents, enabling users to chat and retrieve information from their files. It supports a variety of models from Hugging Face and Ollama, providing flexibility in AI backend choices. The tool can be easily set up and run either locally or on Kaggle, making it accessible for different user environments. Its key features include processing multiple PDF inputs, a simple user interface built with Gradio, and upcoming support for multiple languages. This makes it a versatile solution for anyone needing to query and extract insights from a collection of PDF documents without relying on cloud services.

Stock-analysis-engine is an open-source platform designed for building and tuning investment algorithms, particularly for use with artificial intelligence and deep neural networks. It facilitates the backtesting of thousands of minute-by-minute trading algorithms using live pricing data from publicly traded companies, with automated data feeds from IEX Cloud, Tradier, and FinViz. The engine supports various data types including pricing, options, news, dividends, and financials. It automatically publishes datasets and trading performance to S3, enabling the creation of AI training datasets for teaching DNNs how to trade. The system is built to run on Kubernetes and docker-compose, offering a distributed stack for robust analysis and live trading capabilities.

TEN is an open-source framework designed for creating real-time multimodal conversational AI agents. It provides a comprehensive ecosystem including the TEN Framework itself, Agent Examples, VAD (Voice Activity Detector), Turn Detection, and a Portal. Developers can leverage TEN to build various voice AI applications, from low-latency multi-purpose voice assistants to specialized tools like Doodler for sketch generation, Speaker Diarization, Lip Sync Avatars, and SIP Call integration. The framework supports deployment via Docker or other cloud services, offering flexibility for self-hosting and customization. It also includes resources for quick starts, documentation, and community support through Discord, LinkedIn, and Hugging Face.

UnstableFusion provides a desktop frontend for Stable Diffusion, allowing users to run the powerful AI image generation model locally on their machines. This open-source tool facilitates advanced image manipulation tasks such as inpainting, img2img (image-to-image conversion), and more. Users can define specific areas for operations, resize images without scaling, and utilize a scratchpad for importing and editing images. It offers detailed control over the image generation process, including advanced inpainting techniques where users can save and reuse masks to precisely guide the AI. The tool is designed for those who seek greater command over their AI-driven creative projects, offering features like undo/redo functionality and customizable keybindings.

The TinyFish Cookbook is a comprehensive, open-source repository featuring a growing collection of recipes, demos, and automations developed using the TinyFish web agent. It serves as an invaluable resource for developers looking to understand and implement web agent technology. The cookbook showcases various practical applications, from real-time deal aggregators and price comparison tools to AI-powered research assistants and scholarship finders. Each project within the repository is standalone, offering clear examples of how to leverage TinyFish's capabilities, including its four core endpoints for fast search, content fetching, multi-step browser automation, and fully managed cloud browser rentals. It highlights TinyFish's ability to turn any website into a programmable data source with natural language goals and built-in stealth features.

Uniem is an open-source project dedicated to developing and refining universal text embedding models, with a strong focus on the Chinese language. The project offers comprehensive code for training, fine-tuning, and evaluating these models, making it a valuable resource for researchers and developers. All models and associated datasets are made publicly available on the Hugging Face community, promoting accessibility and collaboration. Uniem supports fine-tuning for various models, including M3E, sentence_transformers, text2vec, and even GPT series models using SGPT methods and Prefix Tuning. It also features MTEB-zh, a standardized evaluation benchmark for Chinese embedding models, allowing for rigorous comparison across different models and tasks.

YOLO-World is a cutting-edge, real-time open-vocabulary object detector, developed by AILab-CVC and accepted by CVPR 2024. This open-source project provides PyTorch implementation, pre-trained weights, and code for both pre-training and fine-tuning. It excels in open-vocabulary detection and grounding, allowing users to detect objects without pre-defined categories. The tool is pre-trained on large-scale datasets, including detection, grounding, and image-text datasets. YOLO-World introduces a prompt-then-detect paradigm for efficient user-vocabulary inference, re-parameterizing vocabulary embeddings into the model for superior inference speed. It supports various fine-tuning recipes, including normal fine-tuning, prompt tuning, and reparameterized fine-tuning, making it adaptable for custom datasets and specific domains. Additionally, it offers deployment options for ONNX, TFLite, and INT8 Quantization.

whatsapp-gpt is an open-source project that enables users to connect ChatGPT with WhatsApp. This setup requires running WhatsApp from a phone number using a Go library and controlling ChatGPT via a dedicated browser in a separate window. The project involves running two terminals, one for the Go application and another for a Python server, to facilitate this interaction. Additionally, a `multichat.py` script is available for users interested in observing two ChatGPT instances communicate with each other. While the initial setup might require some technical familiarity, the project's open-source nature and relatively sparse code make it accessible for those willing to delve into the implementation. It offers a unique way to integrate conversational AI directly into the WhatsApp messaging platform.

Featurestore.org serves as a comprehensive hub for all things related to feature stores in machine learning. It curates content, including blog posts and videos, to inform and educate professionals on the evolving landscape of feature stores and their surrounding data and AI environments. The platform fosters a global community of data science professionals, researchers, and engineers, facilitating the sharing of ideas and collaborative learning through monthly meetups with industry experts. It also hosts annual Feature Store Summits, providing a forum for in-depth discussions and insights into the latest advancements and best practices in the field. The site features detailed comparisons of various feature store solutions, including open-source, vendor, and in-house options, covering aspects like ingestion APIs, supported platforms, and training data handling.

Vibesort is an AI-powered tool designed for sorting arrays using GPT, leveraging OpenAI's API for structured output. This open-source project allows developers to integrate intelligent sorting capabilities into their Python applications. It's an experimental tool, not intended for production use, and requires users to set their OpenAI API key as an environment variable. The project is hosted on GitHub, providing a clear installation process via `pip install vibesort` and usage examples. Its core functionality revolves around the `vibesort` function, which takes an array and returns a sorted version, demonstrating the practical application of large language models in data manipulation tasks.

Kangaroo LLM is Australia's pioneering open-source large language model, specifically engineered to cater to the unique linguistic and cultural nuances of the Australian context. This innovative language model is a community-driven initiative, built by the community, for the community, aiming to foster technological sovereignty and innovation within Australia. It is designed to comprehend local humor, expressions, and cultural references, making it highly relevant for applications targeting Australian audiences. The project emphasizes collaboration among tech companies and individuals to create a robust and adaptable AI solution.

OORT offers a decentralized data cloud specifically designed for AI, leveraging Web3 technology to provide secure and efficient data management solutions. The platform supports enterprise-grade AI data collection, processing, and monetization. Key features include AION, a multi-AI agent system for marketing, high-precision AI model alignment through human feedback, and large-scale multimodal data sourcing. OORT also provides a Project Launchpad for scaling projects, premium off-the-shelf datasets, and infrastructure for running nodes with OORT Edge and building on the Olympus protocol. Users can earn rewards by contributing to data collection and validation through the OORT DataHub mobile app.

alloy-voice-assistant is an open-source project available on GitHub designed for developers to create and experiment with AI voice assistants. The project provides a foundational framework for building a sample AI assistant, requiring both an OPENAI_API_KEY and a GOOGLE_API_KEY for its functionality. Users can store these keys in a .env file or set them as environment variables. The repository includes clear instructions for setting up a virtual environment, installing necessary packages, and running the assistant, with specific guidance for Apple Silicon users. This tool is ideal for those looking to understand the mechanics of AI voice assistants and build custom applications.

Daft is a high-performance data engine specifically designed for AI and multimodal workloads, enabling the processing of images, audio, video, and structured data at any scale. It features native multimodal processing, allowing users to handle various data types within a single framework. The tool also includes built-in AI operations, facilitating tasks like LLM prompts, embedding generation, and data classification using models such as OpenAI, Transformers, or custom solutions. Built with Python at its core and Rust under the hood, Daft offers blazing performance without the complexity of JVM. It supports seamless scaling from local environments to distributed clusters on Ray and Kubernetes, and provides universal connectivity to data sources like S3, GCS, Iceberg, Delta Lake, Hugging Face, and Unity Catalog. Daft ensures out-of-box reliability through intelligent memory management and sensible defaults.

Daily Stock Analysis is an open-source, LLM-powered system designed for intelligent analysis of A-shares, H-shares, and US stock markets. It integrates multiple data sources for market trends, real-time news, and social sentiment, feeding into an AI decision dashboard that provides core conclusions, scores, buy/sell points, risk alerts, and operational checklists. The system supports various market strategies, including A-share review, US stock regimes, moving averages, and Elliott Wave theory. Users can manage portfolios, view historical reports, and backtest AI analysis. It offers multi-channel notifications via platforms like WeChat, Telegram, and email, and can be scheduled to run automatically using GitHub Actions or Docker, providing a zero-cost solution for daily stock insights.