AI Trending

Project directory

Projects

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

Unknown 619 stars

PolyX-Research/Awesome-Remote-Sensing-Agents

🚀Official Repository of Intelligent Remote Sensing Agents: A Survey

### Overview: This repository is a curated collection of 100+ papers at the intersection of Intelligent Remote Sensing Agents, focusing on various applications such as ecological monitoring, emergency response, geological exploration, marine supervision, precision agriculture, urban governance, and more. ### Key features: - 🌿 Ecological Monitoring - 🚨 Emergency Response - 🛏️ Geological Exploration - 🌊 Marine Supervision - 🌾 Precision Agriculture - 🏙️ Urban Governance - 🧩 Others ### Tech stack: Python, JavaScript, and various remote sensing tools and libraries. ### Why it matters: This repository serves as a comprehensive resource for researchers and practitioners interested in the latest advancements in Intelligent Remote Sensing Agents, providing a valuable reference for academic and practical applications.
TypeScript 619 stars

emo-xiaoyu/harness-mix

### Overview: Harness Mix 是一个基于 Codex Desktop 的本地内核，提供了一个无缝的原生用户体验，连接多个原生的 Codex Harness，并让任务在它们之间无缝接力。 ### Key Features: - **无缝连接**：连接多个原生 Codex Harness，实现任务在不同Harness之间的无缝传递。 - **原生界面**：采用 Codex Desktop 原生界面，提供流畅的用户体验。 - **跨平台支持**：适用于 macOS、Linux 和 Windows 系统。 ### Tech Stack: TypeScript ### Why It Matters: 如果你正在寻找一个高效、原生且跨平台的代码执行环境，Harness Mix 是你的理想选择。它简化了代码执行流程，让你能够更专注于编写代码，而无需担心复杂的环境设置。
C++ 619 stars

motis-project/motis

multimodal routing, geocoding, and map tiles

Overview: MOTIS is an open-source software platform designed for efficient planning and routing in multi-modal transportation systems, integrating various modes of transport and providing an easy-to-use REST API. Key features: - Routing: supports walking, cycling, sharing mobility, and public transport combined modes. - Geocoding: handles multi-language address and stop name completion with fuzzy string matching and resolution to geo coordinates. - Reverse geocoding: resolves geo coordinates to the closest address. - Tile server: provides background map tiles. Tech stack: C++ Why it matters: MOTIS is optimized for high performance with low memory usage, enabling efficient handling of large-scale transportation data and supporting efficient loading of full year timetables. Repository: motis-project/motis GitHub stars: 618 GitHub description: multimodal routing, geocoding, and map tiles
Python 619 stars

winstonkoh87/Athena-Public

Local-first memory, decision engine, and governance harness for AI agents — portable across IDEs. Own the state. Rent the intelligence.

Overview: Athena is a local-first memory, decision engine, and governance harness for AI agents, designed to be portable across various Integrated Development Environments (IDEs). It stores plain Markdown and JSONL ledgers on your disk, allowing you to own the state and rent the intelligence. Key features: - Lightweight: Runs chat-based queries without needing to end and restart. - Standard: Suitable for daily engineering and analysis tasks. - Deep: Ideal for high-stakes strategy and architecture work. Tech stack: Python Why it matters: Athena provides a reliable and efficient way to manage AI agent memory, problem framing, and decision-making, ensuring that your AI remains independent and adaptable across different environments and platforms. Repository: winstonkoh87/Athena-Public Primary language: Python GitHub stars: 619 GitHub description: Local-first memory, decision engine, and governance harness for AI agents — portable across IDEs. Own the state. Rent the intelligence.
Python 618 stars

CMU-Perceptual-Computing-Lab/openpose_train

Training repository for OpenPose

### Overview: OpenPose Training is a repository designed to provide training code and experimental models for OpenPose, a widely used real-time multi-person pose estimation system. This project aims to offer a deeper dive into the training process and experimental models that might not be included in the official OpenPose release. ### Key Features: - **Training Code**: Comprehensive training scripts for OpenPose. - **Experimental Models**: Release of some experimental models that may not be included in the official OpenPose release. - **Whole-Body Pose Estimation**: Models for whole-body pose estimation using a single network. - **Alternative Model**: A faster alternative model for OpenPose with slightly lower accuracy. ### Tech Stack: Python ### Why it Matters: This repository serves as a valuable resource for developers interested in understanding the training process and experimental models of OpenPose, providing insights into how the system is trained and tested. It is particularly useful for those who wish to explore the experimental models or for developers looking to enhance their understanding of OpenPose's capabilities. ### Repository: CMU-Perceptual-Computing-Lab/openpose_train ### Primary Language: Python ### GitHub Stars: 618 ### GitHub Description: Training repository for OpenPose ### License: The project is licensed under the terms of the [license](LICENSE).
Python 618 stars

KRLabsOrg/LettuceDetect

Span-level grounding verification for RAG, code, and tool-grounded AI outputs.

Overview: LettuceDetect is an open-source project that verifies the grounding of AI outputs, including RAG answers, coding-agent responses, and tool-grounded outputs. It detects unsupported, contradictory, or fabricated parts, returning exact character spans instead of only an answer-level verdict. Key features: - Localizes unsupported, contradictory, or fabricated parts of RAG answers, coding-agent responses, and tool-grounded outputs. - Supports RAG prose, multilingual QA, code-agent answers, and developer-tool output. - Offers purpose-built encoder models for local inference and generative models for typed spans. - Publishes scoped benchmark results and model limitations in the benchmark documentation. Tech stack: Python Why it matters: LettuceDetect ensures the accuracy and reliability of AI outputs, providing developers with a tool to verify the authenticity of AI-generated content, thereby enhancing the trustworthiness of AI-based applications.
Python 618 stars

NotPunchnox/rkllama

Ollama alternative for Rockchip NPU: An efficient solution for running AI and Deep learning models on Rockchip devices with optimized NPU support ( rkllm )

### Overview RKLLama is an efficient solution for running AI and deep learning models on Rockchip devices, specifically targeting the RK3588(S) and RK3576 platforms. It optimizes the use of the NPU, providing a server and client environment for interacting with LLM models. ### Key Features - **Running models on NPU**: RKLLama allows models to run directly on the NPU, leveraging the optimized performance of Rockchip devices. - **Ollama API compatibility**: Supports the Ollama API, including: - `/api/chat` - `/api/generate` - `/api/ps` - `/api/tags` - `/api/embed` (legacy `/api/embeddings`) - `/api/version` - `/api/pull` - **Partial OpenAI API compatibility**: Supports: - `/v1/completions` - `/v1/chat/completions` - `/v1/embeddings` - `/v1/images/generations` - `/v1/audio/speech` - `/v1/audio/transcriptions` - `/v1/audio/translations` (English only, for now) - **Tool/Function Calling**: Supports complete support for tool calls with various LLM formats. - **Pull models directly from Huggingface**: Allows direct pulling of models from the Huggingface repository. - **Include API REST with documentation**: Provides an API REST interface for easy interaction. - **Listing available models**: Lists all available models for easy reference. -
Python 618 stars

Skytliang/Multi-Agents-Debate

MAD: The first work to explore Multi-Agent Debate with Large Language Models :D

Overview: This project explores the debating capability of large language models by proposing the MAD framework, which stands for Multi-Agent Debate. The cognitive behavior of LLMs has gained attention, and self-reflection, a process of introspection and examination of one's own thoughts, has been demonstrated effective with LLMs in solving NLP tasks. Key features: - The MAD framework aims to explore the debating interaction among LLMs. - It addresses the "DoT" issue by correcting distorted thinking, complementing resistance to change, and providing external feedback. - The framework is designed for "tit for tat" interactions, where one agent's mistakes can be corrected by another, and external feedback can enrich self-reflection. Tech stack: Python Why it matters: The MAD framework is less likely to have the "DoT" issue and can fully exploit the potential of LLMs, leading to significant improvements in tasks like Counterintuitive QA and Commonsense-MT. Repository: Skytliang/Multi-Agents-Debate Primary language: Python GitHub stars: 617
Python 618 stars

YanjieZe/Paper-List

knowledge

### Overview: This repository is a curated collection of papers from various robotics and machine learning conferences and journals, focusing on advancements in robot learning, vision, generative models, and other related topics. It serves as a valuable resource for researchers and developers interested in the latest research in robotics and artificial intelligence. ### Key Features: - **Robot Learning**: Papers on self-improving robots, world models, humanoid robotics, and more. - **Vision, Generative Models, and ML**: Topics include 3D vision, generative models, diffusion models, and LLMs. - **Other**: Includes best papers, misc, and recent random papers. ### Tech Stack: Python ### Why It Matters: This repository is a comprehensive resource for researchers and developers looking to stay updated with the latest advancements in robotics and machine learning. It provides a curated list of papers from various conferences and journals, making it an essential tool for anyone interested in the field.
TypeScript 618 stars

matiasbattocchia/open-bsp-api

Open-source WhatsApp + Instagram Business platform

Overview: OpenBSP API is an open-source WhatsApp and Instagram Business platform designed for self-hosting, multi-tenant, and AI-agent readiness. It is built with Deno and Postgres, powered by Supabase, and offers features for both individual businesses and service providers. Key features: - Self-hostable and multi-tenant deployment - AI-agent ready - Built with TypeScript - Powered by Supabase Tech stack: TypeScript Why it matters: OpenBSP API provides a robust, self-hosted solution for businesses to manage their WhatsApp and Instagram messaging, making it accessible and customizable for various use cases. Repository: matiasbattocchia/open-bsp-api Primary language: TypeScript GitHub stars: 616 GitHub description: Open-source WhatsApp + Instagram Business platform
Go 618 stars

ntthanh2603/gemini-web-to-api

✨Reverse-engineered API for Gemini web app. It can be used as a genuine API key from OpenAI, Gemini, and Claude.

Overview: This project is a reverse-engineered API for the Gemini web interface, allowing users to access Google Gemini's models without requiring API keys. It provides a standardized REST API that can be used as a genuine API key from OpenAI, Gemini, and Claude. Key features: - Exposes OpenAI / Claude / Gemini-compatible API endpoints - No API keys needed — uses cookies from the user's browser - Automatically handles authentication and session management - Supports Docker for easy deployment Tech stack: Go, Docker, Golang, Docker, GitHub Actions, GitHub Why it matters: This project simplifies access to Google Gemini's models by providing a standardized API, making it easier to integrate with existing applications and projects that require model access without the need for API keys.
Python 617 stars

3dlg-hcvc/plan2scene

Official implementation of the paper Plan2Scene.

### Overview: Plan2Scene is an official implementation of the paper "Plan2Scene: Converting floorplans to 3D scenes". This project converts floor plans into textured 3D meshes of residences, leveraging floor plans and photos to generate realistic scenes. ### Key Features: - Converts floor plans to 3D meshes - Utilizes floor plans and photos for realistic scene generation - Supports both training and testing phases ### Tech Stack: Python ### Why It Matters: Plan2Scene provides a valuable tool for converting floor plans into 3D scenes, making it useful for urban planning, architectural design, and other applications where detailed 3D models are needed from floor plans.
TypeScript 617 stars

Mintplex-Labs/openai-assistant-swarm

Introducing the Assistant Swarm. An extension to the OpenAI Node SDK to automatically delegate work to any assistant you create in OpenAi through one united interface and manager. Now you can delegate work to a swarm of assistant all specialized with specific tasks you define.

Overview: The Assistant Swarm is a library that allows developers to easily delegate tasks to a swarm of OpenAI assistants, making it simple to manage and control multiple assistants in a unified interface. Key features: - Automatically delegates tasks to any assistant in the OpenAI registry - Provides a single interface to manage multiple assistants - Simplifies the management of multiple assistants, reducing mental overhead Tech stack: TypeScript Why it matters: This library streamlines the process of managing multiple OpenAI assistants, making it easier for developers to create and control their AI agents without the complexities of managing individual assistants manually. This is particularly useful for building autonomous AI assistants or "agents" in applications where multiple tasks need to be handled efficiently and quickly.
C++ 617 stars

agentty-org/agentty

AI pair programming in your terminal — one static binary, sub-ms startup, any model

### Overview: agentty is a blazing-fast, open-source coding agent designed for seamless integration into your terminal. It's a one-static-binary solution that provides millisecond startup times, zero runtime dependencies, and supports a wide range of models from various providers. ### Key Features: - **One Static Binary**: 16.7 MB, ~3 ms cold start, zero runtime dependencies. - **Code-Specific Retrieval**: Built-in local retrieval for efficient code fetching, reducing context by 80%+. - **Model Independence**: Supports models from Claude, GPT, Groq, OpenRouter, Ollama, and OpenAI. - **Secure by Default**: Runs in a sandboxed environment, ensuring safety and security. ### Tech Stack: C++ ### Why It Matters: agentty offers a unique and efficient approach to coding agents, providing a fast, secure, and customizable solution for developers. It's ideal for developers looking to integrate AI into their terminal workflows without the complexity of managing multiple dependencies or the latency of sending large chunks of their repository to a model.
Jupyter Notebook 617 stars

aiplanethub/openagi

Paving the way for open agents and AGI for all.

### Overview: OpenAGI aims to democratize the development and deployment of autonomous human-like agents, making it accessible to everyone. The project is designed to facilitate the creation of agents capable of performing tasks such as planning itineraries and providing information. ### Key Features: - **Agent Development Framework**: Provides a comprehensive framework for developers to create and manage autonomous agents. - **Multi-Agent Architecture**: Supports the creation of multi-agent systems, allowing for the coordination of multiple agents working together. - **Task Decomposition**: Facilitates the decomposition of tasks into smaller, manageable sub-tasks, enhancing efficiency and accuracy. - **Open Source**: Open source code ensures that the project is transparent and can be easily adapted and extended by the community. ### Tech Stack: - **Programming Languages**: Python (3.9, 3.10, 3.11) - **Documentation**: Jupyter Notebook - **Dependencies**: OpenAI, Tavily, Gemini ### Why It Matters: OpenAGI is crucial in the field of artificial intelligence by making it easier and more accessible for developers to create intelligent agents. By democratizing the development of autonomous agents, OpenAGI is expected to significantly impact various industries, from travel and tourism to sports and education, by providing solutions to real-world problems in a more efficient and personalized manner. ### Repository: - **Repository**: aiplanethub/openagi - **GitHub Stars**: 616 - **GitHub Description**: Paving the way for open agents and AGI for all. ### Installation: - **Virtual Environment
Python 615 stars

cdqa-suite/cdQA

⛔ [NOT MAINTAINED] An End-To-End Closed Domain Question Answering System.

### Overview: An End-To-End Closed Domain Question Answering System built on top of the HuggingFace transformers library. ### Key features: - **CDQA**: A comprehensive question answering system designed for closed domains. - **Built with transformers**: Leverages the powerful transformer models for efficient processing. - **Educational repository**: Created for educational purposes, maintained for reference. ### Tech stack: Python ### Why it matters: This system is designed to answer closed domain questions efficiently, making it useful for various applications such as customer support, knowledge bases, and more.
HTML 615 stars

cvondrick/vatic

Efficiently Scaling Up Video Annotation with Crowdsourced Marketplaces. IJCV 2012

Overview: VATIC is an online video annotation tool designed for computer vision research, enabling crowdsourced annotation on Amazon's Mechanical Turk. It facilitates the creation of large, affordable video data sets. Key features: - Automatic frame extraction from videos - Efficient scaling for video annotation projects - Integration with Apache HTTP server and MySQL database Tech stack: HTML, JavaScript, Python, MySQL Why it matters: VATIC simplifies the process of creating high-quality video data sets, making it easier for researchers to build robust models and improve their understanding of computer vision tasks.
Python 615 stars

richardaecn/class-balanced-loss

Class-Balanced Loss Based on Effective Number of Samples. CVPR 2019

### Overview: This project introduces a class-balanced loss method based on the effective number of samples, specifically designed for image classification tasks. The implementation is based on TensorFlow and is detailed in a paper presented at CVPR 2019. The code is available on GitHub and has been starred by 615 developers. ### Key Features: - **Class-Balanced Loss**: Implements a class-balanced loss function to address class imbalance. - **Effective Number of Samples**: Utilizes the effective number of samples to adjust the loss function. - **TensorFlow Implementation**: Provides a complete TensorFlow code for the paper. ### Tech Stack: - Python 3.6 - TensorFlow 1.14 ### Why It Matters: This method is particularly useful in scenarios where class imbalance is a significant issue, such as in image classification tasks where certain classes may be underrepresented. By using the effective number of samples, the model can better focus on underrepresented classes, leading to improved performance and fairness in classification. ### Repository: - Repository: richardaecn/class-balanced-loss - GitHub Stars: 615 - GitHub Description: Class-Balanced Loss Based on Effective Number of Samples. CVPR 2019 - Primary Language: Python - Dependencies: Python (3.6), Tensorflow (1.14) - Datasets: Long-Tailed CIFAR dataset (original and long-tailed) - Effective Number of Samples Visualization: `data.ipynb` - Training and Evaluation Scripts: Available for original CIFAR, long-tailed
Unknown 615 stars

taivop/awesome-data-annotation

A list of tools for annotating data, managing annotations, etc.

### Overview: This repository is a curated list of tools designed for annotating and managing data, including images, videos, and annotations. ### Key Features: - **Alp's Labeling Tool (ALT)** - **Annotorious** - **Classifai** - **COCO Annotator** - **CVAT** (Computer Vision Annotation Tool) - **DataGym.ai** - **deeplabel** - **diffgram** - **dsgou/annotator** - **etiketai** - **Fast Image Annotation Tool (FIAT)** - **Grid-Annotation-Tool** - **ilastik** - **imannotate** - **imagetagger** - **INCEpTION** - **LabelImg** - **labelme** - **labelml** - **LabelD** - **Label Studio** - **LEAR** - **LOST** - **Microsoft VoTT** - **Pixano** - **PixelAnnotationTool** - **react-image-annotation** - **scalabel** - **Sloth** - **Tator** - **Universal Data Tool** - **VATIC** - **VATIC.js** - **ViAME** - **DIVE** - **ViPER-GT** - **webKnossos** - **Ybat - YOLO BBox Annotation Tool** ### Tech Stack: The repository is primarily written in Python and uses various libraries for annotation and data management. ### Why It Matters: These tools are essential for researchers and data scientists working with large
TypeScript 614 stars

dannote/figma-use

Control Figma from the command line. Full read/write access for AI agents — create shapes, text, components, set styles, export images. 100+ commands.

Overview: figma-use is a command-line tool that allows developers to control and manipulate Figma designs directly from the terminal. It provides full access to Figma's features, including creating shapes, text, components, setting styles, and exporting images. Key features: - Control Figma designs with commands - Create and style shapes, text, components, and set styles - Export designs as JSX for React integration - Support for 100+ commands and features Tech stack: TypeScript Why it matters: figma-use leverages the power of LLMs and React, allowing developers to interact with Figma's design tools directly from the terminal. This tool streamlines the design process by automating repetitive tasks and enabling seamless collaboration with AI agents. GitHub stars: 614 Repository: dannote/figma-use