AI Agents & Automation

dsnote is an open-source application designed for Linux and Sailfish OS, providing robust features for note-taking, reading, and translation. It stands out by offering offline functionalities such as speech-to-text, allowing users to dictate notes without an internet connection. Additionally, it includes offline text-to-speech for reading content aloud and offline machine translation, making it a versatile tool for users who require these capabilities in environments with limited or no internet access. The application is built for both desktop and mobile use.

embedded-graphics is a 2D graphics library specifically engineered for memory-constrained embedded devices. Its core design principle is to draw graphics without relying on buffers, making it fully compatible with `no_std` environments and systems that lack dynamic memory allocators. The library employs an iterator-based approach, where pixel colors and positions are computed in real-time, minimizing saved state and significantly reducing RAM usage with little to no performance impact. It provides built-in primitives for drawing lines, rectangles, circles, ellipses, arcs, sectors, triangles, polylines, and rounded rectangles, along with text rendering using monospaced fonts. The library is highly extensible, supporting external crates for various image formats, custom fonts, layout functions, and display drivers, and includes a simulator for development and testing.

MineContext is an open-source, proactive context-aware AI partner designed to enhance productivity by understanding your digital environment. It captures screenshots and comprehends content, with future support for multi-source multimodal information like documents, images, and videos. Based on a contextual engineering framework, it actively delivers high-quality information such as insights, daily/weekly summaries, to-do lists, and activity records. Key features include effortless context collection, intelligent resurfacing of relevant information during creation, and proactive delivery of summarized content. MineContext prioritizes privacy with local-first data storage and support for local AI models compatible with the OpenAI API protocol, ensuring data remains on your device.

Onyx is an open-source AI platform designed for easy deployment and self-hosting. It provides a comprehensive chat user interface that can be used with any Large Language Model (LLM). A key advantage is its ability to operate effectively in air-gapped environments, ensuring data security and compliance. The platform is equipped with advanced functionalities including AI Agents, integrated Web Search capabilities, and Retrieval Augmented Generation (RAG). Furthermore, Onyx offers connectors to more than 40 different knowledge sources, enhancing its ability to access and utilize diverse information.

Stereo-RCNN is an open-source implementation for accurate 3D object detection and estimation, primarily developed for autonomous driving applications. This tool leverages stereo images to perform simultaneous object detection and association, enhancing the precision of 3D box estimations. It also incorporates a dense alignment module for refining 3D box predictions. The project supports Pytorch 1.0.0 and Python 3.6, with a light-weight version available for scenarios with limited GPU memory. Researchers and developers can utilize Stereo-RCNN for tasks requiring robust 3D perception from image-only data, offering a valuable resource for advancing autonomous systems.

UnrealCV is an open-source project designed to bridge computer vision research with the powerful Unreal Engine (UE). It functions as a plugin for UE, extending its capabilities with a set of UnrealCV commands that enable interaction with virtual worlds. This connection facilitates communication between the Unreal Engine environment and external programs like PyTorch or TensorFlow, making it ideal for generating synthetic data for computer vision tasks. Users can either run a compiled game binary with UnrealCV embedded, requiring no prior Unreal Engine knowledge, or install the plugin directly into Unreal Engine to build new virtual worlds using the editor. It supports Unreal Engine 5.6 and offers features like optical flow image capture and calling Blueprint functions from Python.

bbolt is an embedded key/value database specifically designed for Go applications, serving as an actively maintained fork of Ben Johnson's Bolt key/value store. It aims to provide the Go community with a reliable and stable database solution, incorporating bug fixes, performance enhancements, and new features while maintaining backward compatibility with the original Bolt API. This pure Go key/value store is inspired by LMDB and is ideal for projects that do not require a full-fledged database server like Postgres or MySQL. Its API is intentionally small, focusing primarily on efficient key/value storage and retrieval. bbolt is stable, with a fixed API and file format, and is used in high-load production environments, supporting databases up to 1TB.

Gaussian_YOLOv3 is an implementation of the Gaussian YOLOv3 object detection algorithm, specifically designed for autonomous driving applications. This open-source tool leverages localization uncertainty to achieve accurate and fast object detection. It is built upon the official YOLOv3 framework, providing a robust foundation for its capabilities. The repository includes code, pre-trained weights, and detailed instructions for setup, training, inference, and evaluation using datasets like Berkeley Deep Drive (BDD). It supports multi-GPU training and offers evaluation metrics such as mAP, demonstrating its effectiveness in real-world scenarios.

hover_net is an open-source PyTorch implementation designed for simultaneous nuclear instance segmentation and classification in H&E histology images. This advanced network leverages horizontal and vertical distances of nuclear pixels to their centers of mass, effectively separating clustered cells. A dedicated up-sampling branch is integrated to classify the nuclear type for each segmented instance. The repository supports both training HoVer-Net and processing image tiles or whole-slide images, offering pre-trained model weights for various datasets like CoNSeP, PanNuke, MoNuSAC, Kumar, and CPM17. It provides detailed instructions for environment setup, data formatting, training, and inference, making it a comprehensive solution for histology image analysis.

MedCLIP is an open-source contrastive learning framework specifically designed for medical images and texts, as detailed in its EMNLP'22 paper. It allows for learning from unpaired medical data, facilitating advancements in AI-driven medical image analysis and report generation. The tool provides pre-trained models, including MedCLIP-ResNet50 and MedCLIP-ViT, which can be easily loaded and utilized. It also supports prompt-based classification, enabling users to classify medical images using predefined text prompts. MedCLIP is implemented in Python and can be installed via pip, making it accessible for developers and researchers working in the medical AI domain.

MCUViewer, formerly STMViewer, is a powerful GUI debug tool designed for microcontrollers. It comprises two main modules: a Variable Viewer for real-time monitoring and manipulation of embedded variables directly from RAM via a debug interface (SWDIO/SWCLK/GND), and a Trace Viewer for graphically representing real-time SWO trace output (SWDIO/SWCLK/SWO/GND). This allows for profiling function execution times, confirming timer interrupt frequencies, and displaying high-frequency signals with minimal overhead. The tool supports STLink and JLink programmers and is compatible with Cortex M3/M4/M7/M33 cores. While the GitHub repository holds sources for the 1.1.0 release, MCUViewer is now closed-source. It offers a non-intrusive way to debug and analyze embedded applications, making it a valuable asset for developers working with microcontrollers.

mvs-texturing is an open-source project designed to texture 3D reconstructions from images. While primarily focused on reconstructions generated using structure from motion and multi-view stereo techniques, its application is not limited to this specific setting. The algorithm was first published in September 2014 at the European Conference on Computer Vision. It requires a triangulated 3D model and registered images as input, which can be obtained using applications like the Multi-View Environment. The project provides detailed compilation instructions and dependency information, including prerequisites like cmake, git, make, gcc, libpng, libjpg, libtiff, and libtbb, with automatic downloads for rayint, Eigen, Multi-View Environment, and mapMAP. The software is licensed under the BSD 3-Clause license.

PCV is an open-source Python library designed for computer vision applications, built upon the principles outlined in the book "Programming Computer Vision with Python" by Jan Erik Solem. This pure Python module offers a comprehensive set of functionalities for developers working with visual data. Key capabilities include fundamental image processing operations, advanced feature extraction techniques, precise camera calibration, and robust 3D reconstruction. It leverages popular scientific computing libraries like NumPy and Matplotlib, with optional support for SciPy and other specialized modules for more complex tasks. The library is structured with clear examples and a dedicated folder for code directly from the book, making it an accessible resource for learning and implementing computer vision algorithms.

SirixDB is an embeddable, bitemporal, append-only database system and event store designed to keep the full history of each resource. Unlike traditional databases that overwrite data, SirixDB stores immutable lightweight snapshots, ensuring that every revision is a first-class citizen. It uses structural sharing, where only changed pages are written, and unchanged data is shared between revisions via copy-on-write, leading to efficient storage. SirixDB tracks both transaction time (when committed) and valid time (when true in the real world), providing a robust audit trail. It offers various page versioning strategies, including FULL, INCREMENTAL, DIFFERENTIAL, and SLIDING SNAPSHOT, to balance storage cost and read performance. The system is embeddable as a single JAR or can run as a REST server, and provides CLI tools for database operations.

SketchAPI is the official JavaScript plugin library embedded within the Sketch Mac application, designed to empower developers to extend and customize Sketch's capabilities. It offers a stable JavaScript interface for writing scripts and creating robust plugins, ensuring compatibility across Sketch releases. The API is built using JavaScript/CocoaScript and is bundled as part of Sketch's build process. Developers can leverage SketchAPI to automate tasks, create custom tools, and integrate with other systems, enhancing their design workflow. The project includes core modules, comprehensive documentation, and examples, making it accessible for those familiar with JavaScript development. It supports local development, testing, and integration with Sketch installations, providing a flexible environment for plugin creation.

Slicer, also known as 3D Slicer, is a free and open-source software package designed for advanced visualization and image analysis. It is natively available across multiple platforms including Windows, Linux, and macOS, making it accessible to a broad range of users. The tool is particularly well-suited for medical research and clinical applications, providing robust capabilities for 3D modeling and image computing. Slicer supports various functionalities such as image processing, medical imaging, registration, neuroimaging, and segmentation. Its open-source nature fosters community contributions and continuous development, with extensive documentation and support available through its wiki and discourse forum.

heatshrink is an open-source data compression and decompression library specifically engineered for embedded and real-time systems. Its core strength lies in its minimal memory footprint, capable of operating with as little as 50 bytes, making it ideal for resource-constrained devices. The library supports incremental and bounded CPU usage, allowing data to be processed in small, manageable chunks, which is crucial for maintaining responsiveness in hard real-time applications. It offers flexibility with both static and dynamic memory allocation and is based on the LZSS algorithm for efficient compression. Developers can configure window and lookahead sizes to optimize compression ratios and memory use for specific data types and system requirements.

Gromit-MPX is an on-screen annotation tool designed for Unix desktop environments, supporting both X11 and XWayland. It enables users to draw directly onto the screen, making it ideal for presentations, tutorials, and demonstrations where highlighting specific areas is crucial. Key features include desktop independence, hotkey-based operation for seamless workflow integration, and extensive configurability for key bindings and drawing tools. It also supports multi-pointer setups under X11, allowing for simultaneous annotation and normal work. Gromit-MPX is pressure-sensitive and offers various drawing tools like pens, markers, lines, rectangles, circles, and an eraser, all configurable via a simple text file.

LightAgent is an open-source, lightweight AI agent framework that provides essential components for developing intelligent agents. It incorporates memory, tools, and tree-of-thought capabilities to enhance agent performance and decision-making. The framework facilitates multi-agent collaboration, allowing multiple agents to work together, and supports self-learning mechanisms. It is compatible with major Large Language Models (LLMs) such as OpenAI, DeepSeek, and Qwen, ensuring broad applicability. Additionally, LightAgent includes integration with MCP/SSE protocols.

OppenheimerGPT is a macOS application that provides a streamlined way to interact with and compare various AI models. Users can input prompts simultaneously into different models, such as ChatGPT and Gemini, to evaluate and contrast their responses side-by-side. The application offers convenient access through the macOS menubar and supports standalone windows for focused interaction. A 'Pro' version is available, which removes limitations on the number of active windows and promises future integration with additional AI models like LLaMa and Claude.

stack-chan is an open-source project featuring a JavaScript-driven robot embedded in M5Stack. This super-kawaii robot can display a range of cute faces and expressions, including happy, angry, and sad. Users have the flexibility to customize the robot's face and expressions, as well as add various M5Units for enhanced functionality. The project provides all necessary components, including firmware source codes, stereolithography (STL) files for the case, and schematics with board layout data. It supports driving serial (TTL) and PWM servos and encourages users to develop their own applications. The project is distributed under the Apache version 2.0 license, making it accessible for developers and hobbyists.

IrmoAI is a platform designed to integrate various AI technologies, aiming to streamline tasks and significantly boost productivity for its users. The tool provides an intuitive interface, making advanced AI capabilities accessible to a broad audience. Its core functionality focuses on enhancing both accuracy and efficiency across different workflows. A key differentiator of IrmoAI is its adaptive AI engine, which continuously learns and evolves. This ensures that the tools remain relevant and effective over time, adapting to changing user needs and technological advancements. By automating mundane processes, IrmoAI empowers users to concentrate on more critical tasks, thereby optimizing their overall work output and strategic focus.

ntfstool is a comprehensive open-source forensic tool specifically designed for analyzing NTFS volumes. It offers robust capabilities for reading and interpreting various critical disk structures, including the Master Boot Record (MBR), Volume Boot Record (VBR), and partition tables. The tool excels at examining the Master File Table ($MFT), providing options to dump it in CSV, JSON, or raw formats, and parse Zone.Identifier data for quick identification of downloaded files. It also features strong support for BitLocker encrypted volumes, allowing users to display FVE records, check passwords/keys, and extract VMK/FVEK. Additionally, ntfstool can analyze EFS encrypted files, list and decrypt master keys, and export certificates with private keys. Its undelete command helps retrieve files marked as 'not in use,' and it supports input from image files, live disks, or virtual disks. The tool also includes USN journal analysis with custom rules for detecting suspicious activities and a limited shell for command-line operations.

SteganographierGUI is an open-source steganography tool designed to embed files or folders into MP4/MKV video files, effectively concealing them within seemingly ordinary video content. This method allows users to share files discreetly, as the embedded data is only accessible by changing the video file's extension to '.zip' and extracting it with a compatible archiver like WinRAR. The tool offers both a user-friendly GUI for simple drag-and-drop operations and a CLI for batch processing or integration into other systems. Key features include password protection for embedded files, the ability to modify file hashes to prevent link detection on cloud storage, and integration with the right-click context menu for quick access. It also includes a password book feature for automated decryption attempts and a CAPTCHA generator to protect extraction codes from web crawlers, enhancing the security and longevity of shared resources.