Projects

Overview: Kiro Gateway is a proxy API gateway designed for the Kiro IDE and CLI, allowing users to access Claude models for free. It supports multiple languages and is compatible with various models, including Claude Opus 4.5, Claude Sonnet 4.5, and more. Key features: - Supports OpenAI and Anthropic-compatible APIs - Provides multi-account support with intelligent failover - Offers HTTP/SOCKS5 proxy for restricted networks - Includes vision and web search capabilities - Supports function calling and full message history - Supports streaming and automatic retries on errors - Provides an extended model list, including versioned models Tech stack: Python 3.10+ Why it matters: Kiro Gateway simplifies the integration of Claude models into the Kiro ecosystem, making it easier for developers to leverage AI-powered tools without the need for complex setup or licensing fees.

Overview: BirdNET-Go is a self-hosted, real-time soundscape analyser designed for detecting birds, bats, and other wildlife. It runs on a Raspberry Pi and ingests audio streams for multi-model classification, presenting detections in a fast web UI. Key features: - Multi-model AI gallery: install BirdNET v2.4, Google Perch v2, BattyBirdNET regional bat classifiers, and the BirdNET Geomodel v3.0. - Run multiple models in parallel against separate audio sources. - Live spectrogram streaming and detection heatmaps. - Alert rules engine that routes detections to various platforms. - Production-ready ops: onboarding wizard, OIDC/SSO, TLS certificate management, hot-reload settings, system health page, database doctor, and one-click support dumps. - Installable as a PWA with multilingual support. - Local-only by default, optional Sentry telemetry is opt-in. Tech stack: Go Why it matters: BirdNET-Go is a powerful tool for wildlife monitoring, providing real-time detection and alerts for birds, bats, and other wildlife. Its self-hosted nature and multi-model classification capabilities make it highly versatile and reliable for wildlife conservation efforts.

### Overview: 通过扫码或机器人凭据将IM机器人接入DeepSeek Harness，支持飞书、微信、钉钉、企业微信、QQ、Slack、Telegram、Discord和WhatsApp。 ### Key features: - **QR Code Integration**: 简单的二维码扫描方式快速连接。 - **Credentials-Based Access**: 使用凭据（如机器人凭据）进行安全的访问。 - **9 Channels Support**: 支持飞书、微信、钉钉、企业微信、QQ、Slack、Telegram、Discord和WhatsApp。 ### Tech stack: JavaScript ### Why it matters: 通过简化接入流程，使开发者能够快速将IM机器人集成到DeepSeek Harness中，从而提高工作效率和用户体验。

### Overview: This repository presents a comprehensive survey on Video Diffusion Models, focusing on the latest advancements and applications in the field. The survey is based on the paper "A survey on video diffusion models" by Zhen Xing et al., which has been accepted by ACM Computing Surveys (CSUR). ### Key features: - Comprehensive coverage of Video Diffusion Models - Detailed analysis of state-of-the-art techniques - Insights into applications and future directions - Updated version available on arXiv ### Tech stack: The repository uses Markdown, GitHub, and LaTeX for its documentation. ### Why it matters: This survey provides a valuable resource for researchers and practitioners interested in the field of video diffusion models, offering a thorough understanding of the current state and future potential of these models. ### Repository: ChenHsing/Awesome-Video-Diffusion-Models - GitHub stars: 2324 - Primary language: Unknown - Description: A Survey on Video Diffusion Models - Homepage: [Zhen Xing](https://chenhsing.github.io) - Email: zhenxingfd@gmail.com - Starred by: [Dai-Wenxun](https://github.com/Dai-Wenxun) for the Chinese translation - Cited by: [CSUR](https://arxiv.org/abs/2310.10647) and [Zhihu](https://zhuanlan.zhihu.com/p/661860981)

### Overview: LabML is a powerful tool for monitoring deep learning model training and hardware usage, allowing developers to track experiments and manage configurations easily. ### Key features: - Monitor running experiments from mobile or laptop. - Monitor hardware usage on any computer with a single command. - Integrate with just 2 lines of code. - Keeps track of experiments including information like git commit, configurations, and hyper-parameters. - Provides API for custom visualizations. - Offers a pretty logs of training progress. - Open source! ### Tech stack: Python ### Why it matters: LabML simplifies the process of monitoring deep learning experiments and hardware usage, making it easier for developers to manage and track their projects.

### Overview: open·kritt is an open-source, self-hosted security and vulnerability research tool that leverages AI agents to find and validate security issues in code. It orchestrates focused AI analysis into de-duplicated, ranked findings with configurable validation and enrichment. ### Key Features: - **Build Workflows**: Chain focused prompts into reusable security research playbooks. - **Run Scans**: Analyze remote or local repositories and their dependencies with Codex, Claude Code, or Grok Build. - **Verify Findings**: Use post-scripts to validate issues, build proofs of concept, and produce reports. - **Export Scan Results**: Package canonical findings, structured data, post-processing output, reports, and proofs of concept in one ZIP archive with a share-safe manifest. - **Prioritize Results**: Apply custom severity rankers, a consistent finding schema, and automatic de-duplication. - **Bring Your Own Model Access**: Use a Codex login or connect through OpenAI, Anthropic, DeepSeek, OpenRouter, or xAI. ### Tech Stack: JavaScript ### Why It Matters: open·kritt is built for security researchers and developers who want control over their prompts, workflows, model providers, and infrastructure. It helps researchers earn over $1,500,000 in bug-bounty payouts under the name Blockian, a hacker from Immunefi, HackenProof, and Blockian.xyz, and is the open-source distillation of the internal project behind that work.

### Overview: Tangent is a Python library that performs ahead-of-time automatic differentiation, producing Python source code as its output, making it easier to read and debug derivative code. ### Key features: - Ahead-of-time autodiff on Python source code - Produces Python source code as output - Compatible with TensorFlow and NumPy - Supports a large subset of Python - Provides additional autodiff features not found in other Python ML libraries ### Tech stack: Python ### Why it matters: Tangent simplifies the process of reading and debugging derivative code, allowing researchers and students to write models in Python while still being able to read and debug automatically generated derivative code without sacrificing performance and flexibility.

Overview: LangServe is a library that helps developers deploy LangChain runnables and chains as REST APIs, providing input and output schemas, API documentation, efficient endpoints, and a client SDK for easy interaction. It integrates with FastAPI and Pydantic for data validation and supports features like stream logging and event streaming. Key features: - Input and Output schemas automatically inferred from your LangChain object, enforced on every API call. - API docs page with JSONSchema and Swagger. - Efficient `/invoke`, `/batch`, and `/stream` endpoints. - `/stream_log` endpoint for streaming all (or some) intermediate steps from your chain/agent. - `/stream_events` endpoint for easier streaming without parsing `/stream_log`. - Built-in tracing to LangSmith. - All built with battle-tested open-source Python libraries. - Client SDK for calling LangServe server as if it were locally running. - LangServe Hub documentation. Tech stack: FastAPI, Pydantic, uvloop, asyncio. Why it matters: LangServe simplifies the deployment and interaction with LangChain runnables and chains, making it easier for developers to integrate and use these tools in their projects.

### Overview: This repository contains a curated list of awesome papers and resources related to large language models (LLMs) and their applications in recommendation systems. It serves as a valuable resource for researchers and practitioners looking to leverage LLMs for better user-item interactions and optimized recommendations. ### Key Features: - **No Tuning**: Papers that do not require fine-tuning of the LLM model. - **Supervised Fine-Tuning**: Papers that use LLMs as a data augmentation tool for cold-start item recommendation. - **Related Survey**: A survey on the current state of large language models for recommendation systems. - **Related Tutorial**: A tutorial on using LLMs for recommendation. - **Common Datasets**: Information on commonly used datasets for recommendation systems. - **Single card (RTX 3090) debuggable generative language models that support Chinese corpus**: A specific type of LLM model optimized for single-card setups and Chinese language support. ### Tech Stack: The repository is written in Python and uses various LLM models such as GPT2, GPT4-V, LLaVA2, and others. ### Why It Matters: This repository provides a comprehensive overview of the latest advancements in using LLMs for recommendation systems, making it an essential resource for anyone interested in the field. The curated list of papers and resources ensures that users can stay updated with the latest research and best practices in the field.

Overview: Audiomentations is a Python library designed for audio data augmentation, aimed at improving the performance of audio deep learning models in real-world scenarios, rather than just in the laboratory. Key features: - **AddGaussianNoise**: Adds Gaussian noise to the audio samples. - **AddBackgroundNoise**: Mixes in another sound to add background noise. - **AddColorNoise**: Adds noise with specific color. - **AddGaussianSNR**: Injects gaussian noise using a randomly chosen signal-to-noise ratio. - **AddShortNoises**: Mixes in various short noise sounds. - **AdjustDuration**: Trims or pads the audio to fit a target duration. - **AirAbsorption**: Applies frequency-dependent attenuation simulating air absorption. Tech stack: Python Why it matters: Audiomentations is a powerful tool for audio data augmentation, making audio deep learning models more robust and effective in real-world applications. It helps in achieving world-class results in Kaggle competitions and is trusted by companies building next-generation audio products with AI. Repository: iver56/audiomentations Primary language: Python GitHub stars: 2323 GitHub description: A Python library for audio data augmentation, built to be fast and easy to use, inspired by albumentations.

Overview: TengineKit is a comprehensive AI algorithm SDK designed for real-time face detection and landmark recognition, offering features such as face 2D and 3D landmarks, face attributes, hand detection, and body detection. It is optimized for mobile devices, ensuring low latency and high performance. Key features: - Face Detection - Face Landmarks - Face 3D Landmarks - Face Attributes (Age, Gender, Smile, Glasses) - Eye Iris & Landmarks - Upper Body Detection & Landmarks - Hand Detection (Real-time, not yet on Mobile) - Hand Landmarks (Real-time, not yet on Mobile) - Body Detection (Google, Real-time, not yet on Mobile) - Body Landmarks (Real-time, not yet on Mobile) - YOLOv5 Tech stack: C++ Why it matters: TengineKit simplifies the integration of AI algorithms into mobile applications, offering efficient and real-time face detection and landmark recognition capabilities, making it ideal for developers looking to enhance their mobile applications with advanced AI features. Repository: OAID/TengineKit GitHub stars: 2321 GitHub description: TengineKit - Free, Fast, Easy, Real-Time Face Detection & Face Landmarks & Face Attributes & Hand Detection & Hand Landmarks & Body Detection & Body Landmarks & Iris Landmarks & YOLOv5 SDK On Mobile. GitHub logo: ![TengineKit](https://openailab.oss-cn-shenzhen.aliyuncs.com/logo/TengineKit.png?raw=true "TengineKit

### Overview: Instill Core is a comprehensive AI infrastructure tool designed to streamline the entire process of building versatile AI-first applications. It integrates data, model, and pipeline orchestration, ensuring seamless integration and efficient execution. ### Key Features: - **Data ETL Processing**: Automates the extraction, transformation, and loading of data for real-time analysis. - **AI Readiness**: Enhances the AI-readiness of your data and models, ensuring optimal performance and accuracy. - **Open-Source LLM Hosting**: Provides a platform for hosting and managing large language models (LLMs) efficiently. - **RAG Capabilities**: Enables rich and accurate responses by combining AI and language models. ### Tech Stack: - Python ### Why It Matters: Instill Core simplifies the complex process of AI development, making it accessible and efficient for developers. By integrating data, model, and pipeline orchestration, it accelerates the creation of robust AI applications, reducing the time and effort required for each step of the development process.

Overview: This project focuses on teaching machines to paint like human painters, using a neural renderer in model-based Deep Reinforcement Learning (DRL). The agents learn to determine the position and color of each stroke and plan long-term to decompose texture-rich images into strokes. The experiments demonstrate excellent visual effects achieved with hundreds of strokes, without requiring human painter experience or stroke tracking data. Key features: - Model-based Deep Reinforcement Learning (DRL) - Neural renderer for painting environment - Training process with hundreds of strokes - No need for human painter experience or stroke tracking data Tech stack: Python Why it matters: This project showcases the potential of DRL in creating realistic and visually appealing paintings, offering a new approach to image generation and rendering. The method can be applied to various fields, including art, design, and even video game development, where realistic textures and environments are crucial. Repository: hzwer/ICCV2019-LearningToPaint Primary language: Python GitHub stars: 2315 GitHub description: ICCV2019 - Learning to Paint With Model-based Deep Reinforcement Learning README excerpt: # ICCV2019-Learning to Paint ## [arXiv](https://arxiv.org/abs/1903.04411) | [YouTube](https://youtu.be/YmOgKZ5oipk) | [Reddit](https://www.reddit.com/r/reinforcementlearning/comments/b5lpfl/learning_to_paint_with_modelbased_deep/) |

Overview: CGraph 是一个通用的、无第三方依赖的、跨平台的、基于流图的并行计算框架，支持 C++ 11 标准库编写，兼容 macOS、Linux、Windows 和 Android 系统。它提供依赖元素执行顺序、非依赖元素并发执行的调度功能，并支持 Python API 用于构建流水线。 Key features: - 图流程执行框架 - 支持依赖元素执行顺序 - 支持非依赖元素并发执行 - 支持暂停、恢复、超时设定 - 支持 Python API 用于构建流水线 Tech stack: C++ Why it matters: CGraph 是一个通用的、无第三方依赖的、跨平台的、基于流图的并行计算框架，支持 C++ 11 标准库编写，兼容 macOS、Linux、Windows 和 Android 系统。它提供依赖元素执行顺序、非依赖元素并发执行的调度功能，并支持 Python API 用于构建流水线。

Overview: This repository is a collection of resources related to graph neural networks, including survey papers, papers on specific architectures, and libraries. Key features: - Survey papers on graph neural networks - Papers on Recurrent Graph Neural Networks (R-GNN) - Papers on Convolutional Graph Neural Networks (C-GNN) - Papers on Spatial-Temporal Graph Neural Networks (ST-GNN) - Papers on applications such as Computer Vision, Natural Language Processing, Internet, Recommender Systems, Healthcare, Chemistry, Physics, and others Tech stack: Python, TensorFlow, PyTorch, and other relevant libraries Why it matters: This repository serves as a comprehensive resource for researchers and practitioners looking to stay updated with the latest advancements in graph neural networks, providing insights into various architectures and applications.

### Overview: 本项目旨在从零开始学习和实现大语言模型（LLM），通过编写代码逐步构建和训练自己的LLM模型。项目内容包括从基础模型到高级微调技术的全面介绍，涵盖从零开始训练、微调和部署的全过程。 ### Key features: - 完全从零手写，边写边讲知识点 - 体系化，具有完整的实践路线 - 配套视频讲解，提供B站和Bilibili的视频资源 - 配套镜像 GPU，用于模型的训练，有演示和展示Demo - 最小使用3090，4090即可训练 - 用于写Python代码的Code-LLM - 大模型的部署，包括推理优化和量化等 ### Tech stack: Jupyter Notebook ### Why it matters: 本项目旨在帮助读者从零开始学习和实现大语言模型，通过编写代码逐步构建和训练自己的LLM模型，从而成为LLM的大师。

Overview: BEIR is a heterogeneous benchmark designed to evaluate NLP-based retrieval models across 15+ diverse IR datasets. It provides a common framework for easy model evaluation. Key features: - Preprocess your own IR dataset or use one of the 17 benchmark datasets. - Covers various benchmarks suitable for both academia and industry. - Evaluates well-known retrieval architectures (lexical, dense, sparse, and reranking-based). - Easy to add and evaluate your own model using different state-of-the-art evaluation metrics. Tech stack: Python Why it matters: BEIR offers a comprehensive and user-friendly environment for benchmarking and evaluating information retrieval models, making it an essential tool for researchers and practitioners alike.

### Overview: Welcome to the Python Programming Hub, a comprehensive repository designed to help you learn Python and Machine Learning from scratch. This repository is your gateway to mastering the language and applying it effectively in data science and machine learning projects. ### Key Features: - **Python Input, Output, and Import Functions**: Dive into the basics of Python with detailed explanations and interactive notebooks. - **Python Variables**: Understand how to declare, assign, and manipulate variables in Python. - **Python Global, Local, and Nonlocal Variables**: Explore advanced concepts in Python programming. - **Python Strings**: Learn string manipulation and operations in Python. - **Python Lists**: Master lists, their operations, and how to manage them effectively. - **Python Tuples**: Discover tuples and their applications in Python. - **Python Dictionary**: Explore dictionaries and their usage in Python programming. - **Python Operators**: Understand various operators used in Python for data manipulation. ### Tech Stack: The repository is built using Jupyter Notebook, a powerful tool for interactive data science and machine learning projects. ### Why It Matters: This repository is essential for anyone looking to enhance their Python skills and apply them to real-world data science and machine learning projects. Whether you're a beginner or an experienced developer, this repository provides a solid foundation and practical examples to help you succeed in your coding journey. ### GitHub Stars: The repository has been starred by 2,308 developers, indicating its popularity and usefulness in the community. ### Last Updated: Last updated on October 22, 2026. ### Table of Contents: - **

Overview: LLMStack is a no-code platform designed for building generative AI agents, workflows, and chatbots using language models like GPT. It connects these AI elements to your data and business processes, making it easy to integrate and deploy without any coding skills. Key features: - Build tailor-made generative AI agents, applications, and chatbots - Seamlessly integrate your own data and internal tools - Deploy to the cloud or on-premise - Chain multiple LLMs to build complex applications Tech stack: Python Why it matters: LLMStack simplifies the process of integrating and deploying generative AI solutions, making it accessible to businesses without requiring advanced coding skills. Repository: trypromptly/LLMStack Primary language: Python GitHub stars: 2310 GitHub description: No-code multi-agent framework to build LLM Agents, workflows and applications with your data README excerpt: <p align="center"> <a href="https://llmstack.trypromptly.com/"><img src="https://docs.trypromptly.com/img/llmstack-logo-light-white-bg.svg" alt="LLMStack" width="500px"></a> </p> <p align="center"> <em>LLMStack is a no-code platform for building generative AI agents, workflows and chatbots, connecting them to your data and business processes.</em> </p> <p align="center"> <a href="https://docs.trypromptly.com/llmstack/introduction"

### Overview: Flow Forecast is an open-source deep learning library designed for time series forecasting, classification, and anomaly detection. It features cutting-edge models like transformers and attention mechanisms, providing easy-to-understand interpretability metrics and cloud integration capabilities. ### Key Features: - **Transformer-Based Models**: Supports transformer-based models for efficient time series forecasting. - **State-of-the-Art Models**: Includes models like vanilla LSTM, full transformer, multi-head attention, and more. - **Cutting-Edge Concepts**: Offers advanced concepts such as attention mechanisms, linear embeddings, and efficient models like DARNN and Informer. - **Easy-to-Understand Metrics**: Provides interpretability metrics for better model understanding. - **Cloud Integration**: Seamlessly integrates with popular cloud providers for scalable deployment. ### Tech Stack: Python ### Why it Matters: Flow Forecast is the first open-source deep learning framework for time series forecasting, classification, and anomaly detection. It offers the latest models and cutting-edge concepts, making it a valuable tool for researchers and practitioners in the field of time series analysis. ### Repository: AIStream-Peelout/flow-forecast ### GitHub Stars: 2307 ### Primary Language: Python ### Build Status: - Master branch: [](https://dl.circleci.com/status-badge/redirect/gh/AIStream-Peelout/flow-forecast/tree/master) ###