GitHub Weekly Top 10 Trends (09/14/25)

This week, the GitHub community showcased an exciting array of trending projects that highlight the growing intersection of AI innovation, privacy, security, and developer productivity. From groundbreaking tools for building AI-powered applications and confidential smart contracts, to powerful workflow automation, document-based question answering, and creative content platforms, developers are pushing the boundaries of what’s possible.

For past weekly trending, please view Weekly Tags.

Openpi

OpenPI is an open-source framework for Vision-Language-Action (VLA) models in robotics, developed by Physical Intelligence. It enables researchers and practitioners to train, fine-tune, and deploy VLA models across a variety of robotic platforms, such as DROID, ALOHA, and simulation environments. The repository provides tools and resources for integrating visual, linguistic, and action-based reasoning to advance robotic intelligence.

Stirling-PDF

Stirling-PDF is a robust, locally hosted web-based PDF manipulation tool that allows users to perform a wide range of operations on PDF files. Offering over 50 PDF manipulation features, it provides a user-friendly web interface where all processing is done locally, ensuring the privacy and security of your files.

Daft

Daft is a distributed dataframe library designed for scalable data processing of both traditional tabular data and diverse multimodal data types, such as images, URLs, tensors, and documents. It features a familiar Python DataFrame API and SQL interface, while leveraging a high-performance Rust core engine for efficient execution.

Daft’s core design principles include support for complex data types, optimized interactive computing for notebooks and REPL environments with intelligent caching, and native integration with Ray for distributed computing across large clusters.

Vercel Examples

Vercel Examples is a comprehensive repository offering production-ready templates, reference architectures, and best practices for building applications on the Vercel platform. It showcases integration patterns across a wide range of technologies, including monorepo architectures, storage solutions, edge computing, feature flags, and AI integrations, making it a valuable resource for developers looking to build scalable and modern applications with Vercel.

Firebase Genkit

Firebase Genkit is an open-source framework developed and maintained by Google’s Firebase team for building robust, full-stack AI-powered applications. It provides a unified AI interface with consistent APIs for integrating hundreds of models from providers such as Google, OpenAI, Anthropic, and Ollama.

Firebase Genkit supports multiple programming languages, offering production-ready JavaScript/TypeScript and Go SDKs, as well as an alpha Python SDK. The framework’s extensible plugin architecture includes 15+ official plugins for models, vector stores, and various integrations. Designed for production deployments, it features built-in observability, evaluation, deployment assistance, monitoring tools, and a dedicated CLI and local developer UI for efficient testing and debugging. By abstracting away provider-specific APIs while preserving access to advanced features—like structured output, tool calling, multimodal content, and agentic workflows—Firebase Genkit streamlines the development of scalable and feature-rich AI applications.

Kotaemon

Kotaemon is an open-source Retrieval-Augmented Generation (RAG) application designed for document-based question answering. It features a modular two-library architecture: the kotaemon library provides core abstractions and components, while ktem implements the application layer with a clean, minimalistic Gradio-based web interface. Kotaemon supports RAG pipelines with multiple LLM providers—including OpenAI, Azure, Cohere, and local models like Ollama and llama-cpp-python—and offers flexible document processing pipelines and reasoning strategies.

For end users, Kotaemon delivers an intuitive, user-friendly interface and easy installation. For developers, it offers a customizable, extensible framework for building, modifying, and extending RAG-based document QA systems, with a plugin-based architecture and tools for creating custom workflows.

TruffleHog

TruffleHog is a robust open-source security scanning tool engineered to detect, classify, validate, and analyze secrets across a wide array of data sources. Designed to uncover credentials such as API keys, database passwords, and private encryption keys, TruffleHog can scan Git repositories, cloud storage, filesystems, containers, and more. It stands out for its ability to identify and categorize over 800 different secret types, validate whether discovered secrets are still active, and analyze permissions and access levels for common credential types.

TruffleHog is an essential solution for organizations seeking to safeguard sensitive information and prevent the exposure of critical secrets in their infrastructure.

N8N Workflow Collection

N8N Workflow Collection is a professional workflow automation management system that organizes and provides fast, searchable access to over 2,053 n8n workflow JSON files. The repository features a high-performance FastAPI backend combined with SQLite FTS5 full-text search, enabling users to obtain query responses in under 100 milliseconds.

Supporting workflows across 365 unique service integrations and more than 29,445 workflow nodes, N8N Workflow Collection streamlines managing, discovering, and utilizing automation templates for the n8n ecosystem. This makes it an invaluable resource for developers and businesses seeking efficient workflow automation solutions at scale.

Jaaz.app

Jaaz.app is an open-source multimodal creative assistant that offers a privacy-focused, local-first alternative to proprietary platforms like Canva and Manus. Designed to empower users with intuitive content creation, Jaaz.app integrates a visual canvas interface and AI-powered chat functionality, allowing seamless interaction through both natural language and direct manipulation. Its core features include a privacy-first architecture with no data tracking, a multi-agent AI system for automated and imaginative content generation, and a flexible canvas for building ideas visually—comparable to assembling with LEGO blocks.

With support for text, images, and video generation, Jaaz.app delivers versatile multimodal capabilities and can be deployed fully offline, in hybrid mode, or in the cloud, providing complete control and flexibility for creative workflows.

FHEVM

FHEVM is a full-stack framework designed to bring confidential smart contracts to blockchain environments using Fully Homomorphic Encryption (FHE). By enabling encrypted data to be processed directly on-chain, FHEVM ensures that transaction data and smart contract states remain end-to-end encrypted throughout their lifecycle and are never exposed to any party. The framework guarantees composability, allowing encrypted states to be updated seamlessly while always maintaining their confidentiality.

FHEVM is also fully compatible with existing dApps, enabling encrypted state to co-exist alongside public state without disrupting current application functionality. This makes FHEVM an essential solution for privacy-preserving decentralized applications that require both security and interoperability.