Project directory

Projects

Open-source AI projects tracked by the leaderboard, ordered by GitHub stars.

C++ 671 stars

Svalorzen/AI-Toolbox

A C++ framework for MDPs and POMDPs with Python bindings

Overview: AI-Toolbox is a C++ framework designed for representing and solving Markov Decision Processes (MDPs) and Partially Observable Markov Decision Processes (POMDPs). It provides an easy-to-use interface for implementing reinforcement learning algorithms and is extensible to many problems. Key features: - Supports MDPs and POMDPs - Implements an easy-to-use interface - Extensible to many problems - Python bindings available Tech stack: C++ Why it matters: AI-Toolbox simplifies the implementation of reinforcement learning algorithms by providing a robust framework for MDPs and POMDPs. Its Python bindings make it accessible to researchers and developers, allowing for easy integration into existing projects and workflows.
JavaScript 671 stars

bryanyzhu/agentic-ai-system-course

Use agent to learn agent - A skeleton course on how to design, build, and operate production AI agents

### Overview: This course provides a foundational guide on designing, building, and operating production AI agents. It is designed to be read alongside an AI partner, offering a skeleton of topics and patterns with trade-offs. The course is not a step-by-step tutorial, nor tied to a specific stack, and aims to be consumed as much by AI as by humans. ### Key Features: - **Skeletal Structure**: The course covers foundational topics and patterns without delving into specific implementation details. - **AI-Driven Content**: The content is designed to be consumed by AI, making it suitable for AI consumption as much as human reading. - **Real-World Examples**: The course includes real-world examples where AI agents can be applied, providing practical insights. - **Prompt-Based Exercises**: AI agents can be pointed at specific chapters to walk through the course content, making it easy to apply the knowledge to a project. ### Tech Stack: - JavaScript ### Why It Matters: This course is a comprehensive guide for anyone looking to design, build, and operate production AI agents. It provides a solid foundation and practical examples that can be applied to various projects, making it a valuable resource for developers and AI enthusiasts alike.
TypeScript 671 stars

decionis/agent-safe-pipeline

An execution-authority gateway for AI agents and APIs. It intercepts consequential HTTP actions, asks Decionis whether they are authorized, and forwards exactly the authorized request once on a single-use grant, holds it for a person, or refuses it, leaving chained evidence of each. Installs by Homebrew, Linux package, Docker or Helm.

### Overview: AgentSafe is an execution-authority gateway designed to intercept and authorize HTTP actions for AI agents and APIs. It forwards requests based on a claimed single-use grant, holds them for a person, or refuses them, leaving a record of each action. ### Key features: - **Single-Use Grant**: Requests are forwarded only once on a claimed single-use grant, ensuring accountability. - **Decision Dossier**: Decionis provides a signed Decision Dossier for each action, recording the intent, decision, and rationale. - **Decionis Control Plane**: Decionis is the Independent Execution Authority bound to the exact action, providing policy evaluation and decision-making. - **Homebrew Installation**: AgentSafe can be installed using Homebrew, Linux package, Docker, or Helm. - **Quickstart**: A 5-minute guide for setting up the gateway. - **Deployment Strategies**: Options for Docker with TLS ingress, Kubernetes, AWS network isolation, and more. ### Tech stack: TypeScript ### Why it matters: AgentSafe is crucial for ensuring the integrity and accountability of AI actions, providing a secure and transparent execution boundary for agents and APIs.
Python 671 stars

koursaros-ai/nboost

NBoost is a scalable, search-api-boosting platform for deploying transformer models to improve the relevance of search results on different platforms (i.e. Elasticsearch)

Overview: NBoost is a scalable platform designed to enhance the relevance of search results on various platforms, including Elasticsearch, by leveraging finetuned models. Key features: - Enhances search engine relevance - Improves ranking of results - Supports deployment of state-of-the-art models Tech stack: Python Why it matters: NBoost offers a practical solution for improving search engine performance, making it a valuable tool for developers and data scientists working on search engine optimization and ranking. Repository: koursaros-ai/nboost GitHub stars: 671 GitHub description: NBoost is a scalable, search-api-boosting platform for deploying transformer models to improve the relevance of search results on different platforms (i.e. Elasticsearch).
Objective-C 671 stars

sveinbjornt/hear

Command line interface for the built-in speech recognition and transcription capabilities in macOS.

Overview: `hear` is a command line interface for macOS's built-in speech recognition and transcription capabilities, allowing users to transcribe spoken audio input or files. Key features: - Transcribe audio from microphone - Transcribe audio files - Supports live microphone input and various audio file formats Tech stack: Objective-C Why it matters: This tool provides a native, command line interface for macOS's speech recognition capabilities, offering a robust alternative to Apple's built-in `say` tool.
Python 670 stars

PaulDuvall/ai-development-patterns

A comprehensive collection of AI development patterns for building software with AI assistance, organized by implementation maturity and development lifecycle phases. Includes Foundation, Development, and Operations patterns with practical examples and anti-patterns.

### Overview: This repository contains a comprehensive collection of AI development patterns designed to help developers build software with AI assistance, organized by implementation maturity and development lifecycle phases. The patterns are categorized into Foundation, Development, and Operations, each with practical examples and anti-patterns to guide the adoption of AI in software development. ### Key features: - **Foundation Patterns**: Practical examples and anti-patterns for building software with AI assistance. - **Development Patterns**: Techniques for implementing AI in the development lifecycle, including Spec-Driven Development, Incremental Generation, Atomic Decomposition, and more. - **Operations Patterns**: Strategies for managing AI in production environments, including Policy Generation, Drift Remediation, and Debt Forecasting. ### Tech stack: Python ### Why it matters: This catalog provides a structured approach to integrating AI into software development, helping developers navigate the complexities of adopting AI in their projects. The patterns are designed to be practical and relevant, ensuring that developers can effectively implement AI in their applications.
Java 670 stars

java-up-up/nexus-agent

企业级 AI 智能体 Agent 平台,覆盖智能对话、文档知识问答、联网搜索、RAG 检索、MCP 工具协议、Skills 扩展等完整能力。三层执行器体系、双通道混合检索、组合式切块引擎、会话记忆管理、全链路可观测,每个环节都经过深 度设计和工程化打磨。

### Overview: Nexus Agent 是一个企业级的AI智能体对话平台,覆盖智能对话、文档知识问答、联网搜索、RAG检索、MCP工具协议、Skills扩展等完整能力。 ### Key features: - ReAct Agent 智能体的完整实现 - 三层执行器体系 - Neo4j文档结构图谱 - 知识路由三级漏斗 - 影子路由质量观测 - RAG前置编排引擎 - 双通道混合检索 - 证据预算控制与无证据短路 - Parent-Child块聚合 - MCP工具协议集成 - Skills扩展体系 - 文档从上传到可检索的完整链路 - 组合式切块引擎 ### Tech stack: Java ### Why it matters: Nexus Agent 通过深度设计和工程化实现,提供了从对话到知识路由、检索决策、多路知识召回、证据生成、工具调用和可观测治理的完整闭环能力,确保每个环节都经过了严格的测试和优化,为用户提供了一个高效、可靠的企业级AI智能体平台。
C 670 stars

jwchoi384/Gaussian_YOLOv3

Gaussian YOLOv3: An Accurate and Fast Object Detector Using Localization Uncertainty for Autonomous Driving (ICCV, 2019)

### Overview: This repository contains the implementation of the **Gaussian YOLOv3** algorithm, which was presented at the **IEEE International Conference on Computer Vision (ICCV)** in 2019. The algorithm aims to provide an accurate and fast object detector using localization uncertainty for autonomous driving applications. ### Key Features: - **Code Implementation**: The repository includes the code for the **Gaussian YOLOv3** algorithm, based on the official **YOLOv3** code. - **Dataset**: The algorithm was tested using the **Berkeley Deep Drive (BDD)** dataset. - **Training and Inference**: The repository provides the necessary files and commands to train and test the model, including the example weight file for validation. ### Tech Stack: - **Programming Language**: C - **GitHub Stars**: 670 ### Why It Matters: The **Gaussian YOLOv3** algorithm demonstrates the effectiveness of using localization uncertainty for object detection, showcasing its potential for real-world applications in autonomous driving. The repository provides a practical implementation and detailed instructions for training and inference, making it valuable for developers interested in computer vision and autonomous systems.
Python 670 stars

zhongyu09/openchatbi

OpenChatBI is an intelligent chat-based BI tool powered by large language models, designed to help users query, analyze, and visualize data through natural language conversations. It uses LangGraph and LangChain to build chat agent and workflows that support natural language to SQL conversion and data analysis.

### Overview: OpenChatBI is an intelligent chat-based BI tool powered by large language models, designed to help users query, analyze, and visualize data through natural language conversations. Built on LangGraph and LangChain, it supports natural language to SQL conversion, data analysis, and more. ### Key Features: 1. **Natural Language Interaction**: Get data analysis results by asking questions in natural language. 2. **Automatic SQL Generation**: Convert natural language queries into SQL statements using advanced text2sql workflows. 3. **Data Visualization**: Generate intuitive data visualizations (via plotly). 4. **Data Catalog Management**: Automatically discovers and indexes database table structures, supports flexible catalog storage backends with vector-based or BM25-based retrieval, and easily maintains business explanations for tables and columns as well as optimizes Prompts. 5. **Time Series Forecasting**: In-house forecasting models, accessed through the data analysis agent. 6. **Code Execution**: Execute Python code for data analysis and visualization. 7. **Interactive Problem-Solving**: Proactively ask users for more context when information is incomplete. 8. **Persistent Memory**: Conversation management and user characteristic memory based on LangGraph checkpointing. 9. **MCP Support**: Integration with MCP tools by configuration. 10. **Knowledge Base Integration**: Answer complex questions by combining catalog based knowledge retrieval and external knowledge base retrieval (via MCP tools). 11. **Web UI Interface**: Provide 2 sample UI: simple and streaming web interfaces using Gradio and Streamlit, easy to integrate with other web applications. 12. **Data Analysis Agent
Python 669 stars

arunponnusamy/cvlib

A simple, high level, easy to use, open source Computer Vision library for Python.

Overview: cvlib is a simple, high-level, easy-to-use open-source Computer Vision library for Python, designed to simplify computer vision tasks. Key features: - Face detection - Gender detection - Object detection Tech stack: Python Why it matters: cvlib provides a straightforward and user-friendly interface for performing common computer vision tasks, making it ideal for developers looking to integrate computer vision capabilities into their projects without the need for extensive coding.
C++ 669 stars

google-pytorch/torch_tpu

Overview: This project provides a PyTorch backend for Google's Tensor Processing Units (TPUs), enabling users to run PyTorch models and operations on TPU devices. Key features: - PyTorch ATen kernel implementation for TPU - Compilation cache for optimized execution of XLA computations - Support for distributed training (Data Parallelism, Tensor Parallelism) - Examples for various models (Llama, Qwen, ResNet, DLRM, minGPT, UNet) - Integration with `torch.compile()` for graph mode execution - Utilities for debugging and benchmarking Tech stack: C++ Why it matters: This project allows developers to leverage the power of TPUs for PyTorch models, potentially improving performance and enabling new use cases in distributed training and inference.
Python 669 stars

lucidrains/magvit2-pytorch

Implementation of MagViT2 Tokenizer in Pytorch

### Overview: MagViT2 Tokenizer is a Pytorch implementation of the MagViT2 model from the paper "Language Model Beats Diffusion - Tokenizer is Key to Visual Generation". This model is designed for video generation and understanding, aiming to achieve state-of-the-art performance. ### Key Features: - **MagViT2 Model**: Implements the MagViT2 architecture from the paper. - **Lookup Free Quantization**: Provides a separate repository for the Lookup Free Quantization technique, which can be applied to other modalities like audio. - **SOTA Performance**: Holds the state-of-the-art for video generation/understanding tasks. ### Tech Stack: - Python ### Why It Matters: This implementation is crucial for researchers and developers interested in advancing the field of video generation and understanding. It provides a practical tool for experimenting with and improving the performance of models in this domain. ### Repository: - **Repository**: lucidrains/magvit2-pytorch - **Stars**: 669 - **Description**: Implementation of MagViT2 Tokenizer in Pytorch - **Image**: ![magvit2.png](https://img.shields.io/badge/MagViT2-Pytorch-5865F2?logo=PyTorch&logoColor=white) ### Install: ```bash $ pip install magvit2-pytorch ``` ### Usage: ```python from magvit2_pytorch import ( VideoTokenizer, VideoTokenizerTrainer ) tokenizer = VideoTokenizer( image_size
Python 669 stars

yingpengma/Awesome-Story-Generation

This repository collects an extensive list of awesome papers about Story Generation / Storytelling, exclusively focusing on the era of Large Language Models (LLMs).

### Overview: This repository is a curated collection of 199 papers focusing on story generation and storytelling in the era of large language models. It includes long-form fiction, screenplays, games, narrative world models, visual stories, and methods for evaluating and co-creating stories. The papers are categorized into four sections: Beyond Text, Text Stories, Evaluation, and Co-creation, with a special emphasis on essential reading marked with 🌟. ### Key Features: - **Beyond Text**: Interactive drama, games, narrative world models, screenplays, and visual stories. - **Text Stories**: Planning, coherence, characters, creativity, and training for written stories. - **Evaluation**: Benchmarks, metrics, and analyses of written stories. - **Co-creation**: Tools for creators and studies of how people write with AI. ### Tech Stack: Python ### Why It Matters: This repository provides a comprehensive overview of the latest research in story generation and storytelling, focusing on the era of large language models. It helps researchers, creators, and enthusiasts stay updated on the latest advancements and best practices in the field. ### GitHub Stars: 669 ### Primary Language: Python ### GitHub Description: This repository collects an extensive list of awesome papers about Story Generation / Storytelling, exclusively focusing on the era of Large Language Models (LLMs). ### Starred By: Yingpeng Ma and Yan Ma ### Maintained By: Yingpeng Ma and Yan Ma ### Contact: mayingpeng33@gmail.com ### News: - Major update: 19
Java 668 stars

spring-ai-community/spring-ai-agent-utils

A Spring AI library that brings Claude Code-inspired tools and agent skills to your AI applications.

Overview: This project introduces a Spring AI library that integrates Claude Code-inspired tools and agent skills into Java applications, making advanced agentic workflows accessible to developers. Key features: - Agent Environment - File System Tools - Shell Tools - Grep Tool - Glob Tool - Smart Web Fetch Tool - Brave Web Search Tool - Ask User Question Tool - Skills Tool - Auto Memory Tools Tech stack: Java 17+ Why it matters: This library simplifies the integration of Claude Code's powerful features into the Spring AI ecosystem, enabling Java developers to build sophisticated AI agents with ease.
Python 668 stars

zhangxiaosong18/FreeAnchor

FreeAnchor: Learning to Match Anchors for Visual Object Detection (NeurIPS 2019)

### Overview: FreeAnchor: Learning to Match Anchors for Visual Object Detection (NeurIPS 2019) is a project that aims to improve the performance of object detection models by utilizing anchor matching techniques. This repository provides the code and configurations for a modified version of the Mask R-CNN framework, specifically designed for better anchor matching and object detection. ### Key Features: - **Multi-Scale Testing Support**: The implementation supports testing on multiple scales, enhancing the model's ability to handle varying image sizes. - **Updated Experiments**: The repository includes updated experiments and performance metrics, showcasing improved results over previous versions. - **Training and Testing**: The project provides detailed configurations for backbone networks, iteration rates, and training scales, along with updated performance metrics. ### Tech Stack: - **Backbone Networks**: ResNet-50, ResNet-101, and ResNeXt-64x4d-101. - **Training Scales**: Supported scales include 480, 640, 800, 960, 1120, and 1280, with max_size being 1.666 times the original scale. ### Why It Matters: - **Improved Performance**: The project demonstrates significant improvements in object detection accuracy, particularly in terms of average precision (AP) metrics. - **Versatile Implementation**: The repository provides a flexible framework that can be easily adapted to various datasets and configurations, making it a valuable resource for researchers and practitioners in the field of computer vision.
Python 667 stars

Text-to-Audio/Make-An-Audio

PyTorch Implementation of Make-An-Audio (ICML'23) with a Text-to-Audio Generative Model

Overview: This repository contains the PyTorch implementation of the "Make-An-Audio" model, which is a conditional diffusion probabilistic model designed for generating high-fidelity audio from text inputs. The model is part of the ICML 2023 conference and is available for open-source use. Key features: - Conditional diffusion probabilistic model for text-to-audio generation - Efficiently generates high-fidelity audio from text prompts - Pretrained models and weights available for download Tech stack: PyTorch Why it matters: This model offers a significant advancement in text-to-audio generation, providing high-fidelity audio outputs from textual prompts, making it valuable for applications requiring precise audio synthesis from natural language inputs. Repository: Text-to-Audio/Make-An-Audio Primary language: Python GitHub stars: 667
C++ 667 stars

bakwc/JamSpell

Modern spell checking library - accurate, fast, multi-language

### Overview: JamSpell is a modern spell checking library designed for accuracy, speed, and support for multiple languages. It uses context-aware corrections and is optimized for performance, making it suitable for applications requiring high accuracy and fast processing times. ### Key Features: - **Context-Aware Corrections**: Enhances the accuracy of corrections by considering the surrounding text. - **High Performance**: Processes up to 5,000 words per second, making it suitable for real-time applications. - **Multi-Language Support**: Available in C++ and supports multiple languages through SWIG bindings. ### Tech Stack: C++ ### Why It Matters: JamSpell is an essential tool for developers looking to improve the accuracy and performance of spell checking in their applications. Its ability to handle context and its fast processing speed make it a valuable addition to any project requiring accurate and efficient spell checking. ### GitHub Stars: 667 ### Repository: bakwc/JamSpell ### Colab Example: [Link to Colab Example](https://colab.research.google.com/drive/1aFk8-7nq3oAp402jjLGLpEb2Nzq210Eo) ### JamSpellPro: [jamspell.com](https://jamspell.com) - check out JamSpellPro, a new version with improved accuracy, merged word splits, pre-trained models for many languages, runtime word/sentence addition, fine-tuning, memory optimization, and support for Java, C#, Ruby, Windows, and static dictionaries. ### Content: - **Benchmarks
Rust 667 stars

chrisryugj/Docufinder

Anything — 파일 이름이 아니라 본문으로 찾는 문서 검색기. HWP·PDF·Office 전문 검색을 오프라인으로. Windows·macOS | Offline content search across Korean documents (HWP, PDF, Office). Windows and macOS

### Overview: Docufinder is a powerful, offline document search engine that allows you to search across Korean documents (HWP, PDF, Office) without needing an internet connection. It provides a seamless document search experience, indexing documents automatically and allowing quick access to specific sections or entire documents. ### Key Features: - **Document Search:** Searches across multiple formats including HWP, PDF, and Office documents. - **Quick Results:** Finds results instantly, even for documents with many pages. - **File Name Search:** Searches documents by partial file names. - **AI Q&A:** Answers to common questions using AI, including budget inquiries and vacation conditions. - **AI Document Summarization:** Summarizes documents like contracts, reports, and meeting minutes. - **Real-Time Sync:** Automatically syncs changes to documents. - **Document Preview:** Renders documents in both text and layout view, including HWPX, HWP, and PDF. - **Markdown Export:** Saves search results as a Markdown file. - **Comparison View:** Shows differences between document versions. - **File Draggable:** Copies or moves results to other apps or web pages. ### Tech Stack: Rust ### Why It Matters: Docufinder is a valuable tool for anyone who needs to quickly find and access documents without an internet connection, making it perfect for offline use or in environments where internet access is restricted.
Python 667 stars

gojasper/flash-diffusion

Flash Diffusion — accelerating conditional diffusion models (AAAI 2025 Oral)

### Overview: Flash Diffusion accelerates conditional diffusion models by reducing the number of sampling steps and training time. ### Key features: - **Efficiency**: Flash Diffusion requires only several GPU hours of training and fewer trainable parameters compared to existing methods. - **Versatility**: The method is versatile across various tasks such as text-to-image, inpainting, face-swapping, super-resolution, and using different diffusion models backbones. - **Fast Generation**: Flash Diffusion generates high-quality images in a few steps, significantly reducing the time required for image generation. ### Tech stack: Python ### Why it matters: Flash Diffusion achieves state-of-the-art performance in terms of FID and CLIP-Score for few steps image generation on the COCO 2014 and COCO 2017 datasets, while requiring only **several GPU hours of training** and fewer trainable parameters than existing methods.
Python 667 stars

isaacus-dev/semchunk

A fast, lightweight and easy-to-use Python library for splitting text into semantically meaningful chunks.

Overview: semchunk is a Python library designed to split text into semantically meaningful chunks, leveraging AI-powered chunking and advanced tokenization techniques. Key features: - Supports AI-powered chunking - Handles chunk overlapping - Supports chunk offsets - Compatible with various tokenizers and token counters, including Tiktoken and Transformers Tech stack: Python Why it matters: semchunk significantly improves the performance of Retrieval-Augmented Generation (RAG) by delivering 15% better results compared to its closest competitors, as demonstrated by benchmarks. Repository: isaacus-dev/semchunk GitHub stars: 667 GitHub description: A fast, lightweight and easy-to-use Python library for splitting text into semantically meaningful chunks.