AI Agents & Automation

RecFM offers a comprehensive suite of tools and frameworks specifically designed for building foundation models in recommendation systems. Developed by the USTCLLM group at USTC, it provides modular libraries and technologies that streamline the development process. The platform aims to facilitate the creation of robust and efficient recommendation systems, enabling researchers and developers to leverage advanced AI models for personalized content delivery and user experience optimization. Its focus on foundation models suggests capabilities for handling large datasets and complex recommendation logic, making it suitable for advanced AI research and application development.

raster-vision is an open-source Python library and framework designed for deep learning on satellite, aerial, and other large imagery sets, including oblique drone imagery. It offers built-in support for chip classification, object detection, and semantic segmentation, utilizing PyTorch backends. As a library, it provides a comprehensive suite of utilities for handling all aspects of a geospatial deep learning workflow, from reading geo-referenced data and training models to making predictions and writing out results in geo-referenced formats. As a low-code framework, it enables users to configure experiments for machine learning pipelines, including data analysis, chip creation, model training, prediction, evaluation, and deployment bundling. It also supports cloud execution via AWS Batch and AWS Sagemaker.

Parseflow is an AI-powered document parsing service designed to extract tables and nested unstructured data from a wide variety of document types, including invoices, receipts, contracts, images, and schematics. Boasting 99% accuracy, the platform ensures reliable data extraction. It incorporates enterprise-grade security features such as PII protection, encryption, and data anonymization, making it suitable for sensitive information. Parseflow supports over 100 document types and offers seamless integration with existing systems and workflows via its API, providing a robust solution for businesses with diverse document processing needs.

Choice Chaser is an AI-powered tool designed to streamline collaborative decision-making processes for groups and teams. It enables users to share a link and collectively make choices without the need for individual sign-ups, promoting quick and efficient collaboration. The platform supports various decision-making scenarios, from simple polls to more complex group choices, making it suitable for diverse applications. By removing barriers to entry like account creation, Choice Chaser aims to simplify and accelerate the process of reaching consensus, enhancing productivity for its users.

Craffr was a B2B lead monitoring tool designed to help freelancers and small agencies find high-intent leads across platforms like Reddit, Hacker News, and Indie Hackers. The product aimed to address the problem of missed opportunities due to slow lead discovery. Despite attracting 50 trial users, Craffr achieved zero paid conversions and has since been retired. The creator, Ankit Nanda, now focuses on building B2B SaaS products from 0 to 1, sharing detailed case studies, and providing open-source tools. These include the Reddit Lead Finder, an AI-powered tool for targeted lead detection, a Rate Calculator for freelancers, and a Response Timer for email analysis.

Tofu Leaderboard is a specialized AI application hosted on Hugging Face Spaces, designed for researchers and developers working with large language models (LLMs). It provides a platform to browse and search comprehensive unlearning performance data, offering insights into how effectively LLMs can forget specific information. Users can also contribute by submitting new evaluation results for various models, including popular ones like Llama and Phi. This tool serves as a central repository for tracking and comparing the unlearning capabilities of different LLMs, fostering advancements in model safety and ethical AI development.

AutoDL is an open-source project designed for automated deep learning, aiming to eliminate human intervention in the deep learning process. It provides a universal algorithm flow for multi-label classification across various modalities, including images, videos, audio, text, and tabular data. The tool addresses common challenges such as resource constraints, data imbalance, small datasets, feature engineering, model selection, network structure optimization, and hyperparameter search. AutoDL boasts excellent performance, having won the AutoDL Challenge@NeurIPS, and can produce high-performing classifiers in as little as ten seconds, with real-time feedback on model performance. It includes a robust model library supporting both traditional machine learning and state-of-the-art deep learning models like ResNet, MC3, DNN, and BERT.

awesome-3D-vision is a comprehensive, open-source curated list of resources dedicated to 3D computer vision. It encompasses a wide array of topics including Simultaneous Localization and Mapping (SLAM), Visual SLAM (VSALM), Deep Learning, Structured Light, Stereo Vision, and Three-dimensional Reconstruction. The repository serves as a valuable hub for researchers and developers, offering links to relevant papers, datasets, and code examples. It also features sections on camera technology, calibration methods, 3D vision companies, and detailed resources for computer vision and machine learning, making it an essential reference for anyone working in the field.

awesome-DeepLearning, powered by PaddlePaddle, serves as a comprehensive online encyclopedia for deep learning, designed to make the innovation and application of deep learning technologies more accessible. It features a wide array of learning materials, including introductory and advanced courses, specialized curricula, academic case studies, and practical industry examples. The platform also provides a deep learning knowledge base and an interview question bank, catering to both beginners and experienced practitioners. Materials are presented in various formats, such as online notebooks, videos, and books, ensuring a rich and flexible learning experience. The project is actively maintained with real-time updates to code, aligning with the latest PaddlePaddle versions, and regularly shares insights on cutting-edge research and paper implementations.

Proton is a powerful SQL pipeline engine designed for high-speed stream processing and real-time analytics. Built as a single C++ binary, it offers efficient performance for demanding data workloads. The tool is well-suited for observability applications, allowing users to monitor and analyze system behavior in real-time. Furthermore, Proton supports AI/ML applications, enabling the integration of machine learning models into data pipelines for advanced analytics and predictive capabilities. Its focus on real-time data analysis makes it an ideal solution for scenarios requiring immediate insights and rapid response to evolving data streams.

react-agent is an open-source React.js library designed to facilitate the creation of autonomous LLM agents. It provides a flexible and customizable framework for developers to build AI-powered applications directly within the React.js ecosystem. The tool emphasizes extensibility, allowing users to tailor agents to specific needs and integrate them seamlessly into existing React projects. This makes it suitable for both AI research and development, enabling rapid prototyping and deployment of intelligent agents. Its open-source nature fosters community collaboration and continuous improvement, providing a robust foundation for building sophisticated AI solutions.

Standalone-DeepLearning is an open-source GitHub repository that serves as a comprehensive resource for the 2019 KAIST Deep Learning seminar. It offers a structured curriculum covering fundamental and advanced deep learning topics, including Machine Learning Basics, Linear Regression, Logistic Regression, Artificial Neural Networks, CNNs, RNNs, GCNs, and Generative Models like VAE and GAN. Each lecture includes slides, video recordings, and practical coding exercises, often with assignments. The repository also provides helpful resources for Python, Numpy, Pandas, and data visualization with Seaborn. It's an excellent resource for students and developers looking to deepen their understanding of deep learning through a combination of theoretical knowledge and hands-on practice.

QCraft AI is a world-leading self-driving technology company dedicated to bringing autonomous driving into real life. Founded in 2019, QCraft offers full-stack solutions for autonomous driving, emphasizing efficient deployment and iteration. Their technology includes "Driven-by-QCraft" solutions with more reliable perception, utilizing China's first large model for Joint Multi-Modality and Temporal Fusion. They also feature smarter planning and control with a Joint Spatio-temporal Algorithm, creating flexible and efficient driving strategies. QCraft streamlines the entire data cycle from selection and labeling to training and simulation, ensuring efficient data flow and closed-loop validation. Their "Chengfeng" High-Level Driver Assistance Solutions provide Urban and Highway NOA, aiming for affordable and user-friendly driving assistant systems. QCraft develops multiple product solutions for complex urban environments, catering to mass-produced models, urban public transportation, and ride-hailing services.

WELearnHelper is an open-source tool designed to assist students with WE Learn online courses. It provides solutions for course questions, supports class tests, and offers automated answering capabilities, including answer generation powered by ChatGPT. The tool also helps users manage their study time by automating course progression. It integrates with browser extensions like Tamper Monkey/ScriptCat and is built on TypeScript and Vue, allowing for community contributions and plugin development. While primarily focused on displaying answers and automating certain tasks, it also includes features for collecting and uploading answers for class tests, contributing to a community-sourced answer database.

Punctuator2 is an open-source tool designed to restore missing inter-word punctuation in unsegmented text using a bidirectional recurrent neural network model with an attention mechanism. The model can be trained in two stages: the first stage focuses on restoring punctuation based purely on textual features, while an optional second stage can incorporate pause durations from speech data to adapt to specific target domains, such as automatic speech recognition system outputs. This allows for more accurate punctuation restoration by combining textual and prosodic features. The tool provides pretrained models and detailed instructions for data preparation, training, and usage, making it a valuable resource for researchers and developers working with text processing and speech-to-text applications.

Janus Pro 1b is a versatile AI tool hosted on Hugging Face Spaces, designed for both understanding and generating multimodal content. Users can upload an image and pose questions to receive answers, leveraging its image comprehension capabilities. Additionally, the tool enables the creation of multiple images from detailed text prompts, offering robust image generation functionalities. This unified approach makes it a powerful resource for tasks requiring both visual analysis and creative image synthesis, all within a single platform.

SuperChatNow, despite its name, appears to be the website for Tianjin Zhongcheng Weishi Security Service Co., Ltd., a Chinese company specializing in security and property management services. The website content details offerings such as security guard services, property management, cleaning services, landscaping maintenance, and catering services for various entities including hospitals, offices, and industrial parks. There is no indication of AI chatbot functionalities, conversational AI, or any other AI-related tools on the live website. The company emphasizes its credentials as a national first-level security service enterprise.

python-a2a is a Python library designed to implement Google's Agent-to-Agent (A2A) protocol. This protocol enables seamless communication and interaction between various AI agents, fostering the development of interoperable agent ecosystems. The library aims to simplify the process of building complex multi-agent systems where different AI entities can collaborate on tasks and exchange information effectively. It provides the foundational tools necessary for developers to create robust and scalable agent-based applications, allowing agents to work together to solve intricate problems and achieve common goals. The design prioritizes both power and ease of use, making it accessible for developers looking to integrate advanced agent communication capabilities into their projects.

DreamHoney is an AI companion app designed for safe, realistic, and emotionally intelligent conversations. It features photorealistic characters that listen, remember details, and respond like real people, fostering a judgment-free space for users. Unlike many other AI companion apps, DreamHoney strictly avoids NSFW content and addictive design practices, prioritizing meaningful interactions. Key features include emotional memory, allowing companions to recall past conversations and personal details, and optional voice-to-voice chat for premium users. DreamHoney aims to provide companionship and emotional support without the pressures of social media or dating apps, making it ideal for those seeking a supportive conversational partner.

RecruitCrafts is an AI-driven HR platform designed to streamline recruitment processes for organizations. It offers advanced features for managing job postings, screening candidates, and automating communication, aiming to improve efficiency and reduce the time-to-hire. The platform leverages artificial intelligence to enhance various stages of the recruitment funnel, from initial candidate outreach to final selection. By automating repetitive tasks and providing intelligent insights, RecruitCrafts helps HR professionals and recruiters focus on strategic initiatives and candidate engagement. Its comprehensive suite of tools is built to support a more efficient and effective hiring strategy.

seqeval is a Python framework designed for the evaluation of sequence labeling tasks, including named-entity recognition (NER), part-of-speech (POS) tagging, and semantic role labeling. It provides robust evaluation capabilities, tested against the industry-standard Perl script `conlleval` for compatibility with CoNLL-2000 shared task data. The framework supports multiple common annotation schemes such as IOB1, IOB2, IOE1, IOE2, IOBES, and BILOU, with strict mode evaluation available for IOBES and BILOU. Users can compute standard metrics like accuracy, precision, recall, and F1 score, and generate comprehensive classification reports to assess model performance effectively. Its flexibility makes it a valuable tool for researchers and developers working on natural language processing tasks.

ReActGPT is an AI research tool designed to enhance the accuracy and reliability of AI models by implementing the ReAct (Reasoning and Acting) paradigm. This framework helps in reducing errors in AI outputs, making it a valuable asset for those working on advanced AI systems. The tool is specifically tailored for researchers and developers who are focused on pushing the boundaries of AI technology and building more robust and dependable AI applications. By integrating the ReAct approach, ReActGPT provides a structured method for AI agents to reason about their actions and observations, leading to more intelligent and less error-prone behavior. This makes it an essential component for anyone looking to develop cutting-edge AI solutions with improved performance and trustworthiness.

awesome-agentic-patterns is a curated, open-source catalog of agentic AI patterns, offering real-world tricks, workflows, and mini-architectures for autonomous or semi-autonomous AI agents. It aims to help agents perform useful work in production environments by surfacing repeatable patterns that are often hidden in real products. The platform provides a Pattern Explorer to browse, filter, and search patterns by category, status, and complexity, along with a Compare Tool, Decision Explorer, and Graph Visualization. It also includes developer guides and a machine-readable `llms.txt` file to assist AI assistants in recommending appropriate patterns.

YomiToku is a Python package designed for AI-powered document image analysis specifically for the Japanese language. It provides comprehensive full-text OCR and advanced layout analysis capabilities, enabling the recognition, extraction, and conversion of text and figures from various image formats. The tool is equipped with four distinct AI models, all trained on Japanese datasets, for character position detection, string recognition, layout analysis, and table structure recognition. It supports over 7000 Japanese characters, including handwritten text and vertical writing, and can process documents with complex Japanese-specific layouts. YomiToku also offers flexible output formats such as HTML, Markdown, JSON, CSV, and searchable PDFs, and can extract embedded charts and images. It is optimized for GPU environments for fast processing, requiring only 8GB of VRAM, and also offers a lightweight model for efficient CPU inference.