AI Agents & Automation

trlx is a distributed training framework specifically designed for fine-tuning large language models using Reinforcement Learning via Human Feedback (RLHF). It supports training with either a provided reward function or a reward-labeled dataset. The framework offers compatibility with Hugging Face models, enabling fine-tuning of causal and T5-based language models up to 20B parameters, such as facebook/opt-6.7b and EleutherAI/gpt-neox-20b. For models exceeding 20B parameters, trlx integrates with NVIDIA NeMo-backed trainers, leveraging efficient parallelism techniques for scalability. It currently implements Proximal Policy Optimization (PPO) and Implicit Language Q-Learning (ILQL) algorithms, with support for both Accelerate and NeMo trainers.

VibeProxy is a native macOS menu bar application designed to integrate existing Claude Code, ChatGPT, Gemini, Kimi, Qwen, Antigravity, and Z.AI GLM subscriptions with powerful AI coding tools like Factory Droids. It operates without requiring API keys, instead managing OAuth authentication and token routing automatically. The app offers a clean, native SwiftUI interface, one-click server management, and multi-account support with automatic round-robin distribution and failover. A key feature is its Vercel AI Gateway integration for Claude requests, enhancing security and reducing account risks. VibeProxy also provides real-time status updates, automatic app updates, and supports the latest models including Gemini 3 Pro and GPT-5.1.

Tunix (Tune-in-JAX) is a JAX-based library developed by Google, specifically engineered to optimize the post-training phase of Large Language Models (LLMs). It offers efficient and scalable support for various advanced training methodologies, including Supervised Fine-Tuning (SFT), Reinforcement Learning (RL), and Agentic RL. Leveraging the power of JAX, Tunix ensures accelerated computation and seamless integration with JAX-based modeling frameworks like Flax NNX. It also integrates with high-performance inference engines such as vLLM and SGLang-JAX for efficient rollout. Tunix is designed to work within the JAX training stack, utilizing foundational tools like Flax and Optax, and streamlining tuning workflows on XLA and JAX infrastructure. It supports a growing list of models including Gemma, Llama, and Qwen families.

vLLM is a fast and easy-to-use library designed for LLM inference and serving, originating from the Sky Computing Lab at UC Berkeley. It boasts state-of-the-art serving throughput and efficient memory management through PagedAttention. Key features include continuous batching, chunked prefill, prefix caching, and fast model execution with CUDA/HIP graphs. vLLM supports various quantization methods like FP8 and INT4, optimized attention kernels such as FlashAttention, and speculative decoding. It offers seamless integration with Hugging Face models, high-throughput serving with diverse decoding algorithms, and distributed inference capabilities. The tool also provides an OpenAI-compatible API server, multi-LoRA support, and broad hardware compatibility, including NVIDIA, AMD, and x86/ARM/PowerPC CPUs, along with plugins for TPUs and other accelerators. It supports over 200 model architectures, including decoder-only, Mixture-of-Expert, hybrid attention, multi-modal, embedding, and reward models.

LMCache is an open-source library designed to accelerate Large Language Model (LLM) performance by acting as a high-speed Key-Value (KV) cache layer. It significantly reduces Time To First Token (TTFT) and boosts throughput, particularly beneficial in scenarios involving long contexts. LMCache achieves this by storing and reusing KV caches of texts across various storage tiers like GPU, CPU, Disk, and even S3, utilizing advanced acceleration techniques such as zero CPU copy and GDS. It integrates seamlessly with popular LLM serving engines like vLLM and SGLang, offering features like high-performance CPU KVCache offloading and disaggregated prefill. This allows developers to achieve substantial delay savings and GPU cycle reductions in diverse LLM use cases, including multi-round QA and RAG.

X-VLA is the official implementation of "Soft-Prompted Transformer as Scalable Cross-Embodiment Vision-Language-Action Model," accepted to ICLR 2026. This open-source project introduces a soft prompt mechanism using embodiment-specific learnable embeddings to guide a unified Transformer backbone. This approach facilitates effective multi-domain policy learning across heterogeneous large-scale robot datasets. The resulting X-VLA-0.9B architecture demonstrates state-of-the-art generalization across six simulation platforms and three real-world robots, outperforming previous VLA methods in dexterity, adaptability, and efficiency. It supports a Server–Client architecture for distributed inference and offers various pre-trained models fine-tuned for specific robotic embodiments and benchmarks like AgiBot World Challenge, CALVIN, Google Robot, and LIBERO.

WindowsAgentArena (WAA) is a scalable Windows AI agent platform designed for testing and benchmarking multi-modal, desktop AI agents. It provides researchers and developers with a reproducible and realistic Windows OS environment, enabling the testing of agentic AI workflows across a diverse range of tasks. WAA supports the deployment of agents at scale using Azure ML cloud infrastructure, allowing for parallel execution of multiple agents and delivering quick benchmark results for hundreds of tasks in minutes. The platform includes features like a new difficulty mode for tasks, the Navi agent with Omniparser, and the open-sourced Omniparser screen understanding model. Users can deploy locally using Docker and WSL 2, or leverage Azure for parallel benchmarking.

WizardLM is a suite of large language models (LLMs) built upon the Evol-Instruct method, designed to enhance their ability to follow complex instructions. This project includes several specialized models: WizardLM focuses on general instruction following, WizardCoder excels in code generation, and WizardMath is optimized for mathematical reasoning. The models demonstrate strong performance against established benchmarks and even surpass some closed-source alternatives like ChatGPT 3.5 and Gemini Pro in specific tasks. WizardLM provides various model sizes, from 7B to 70B parameters, with different licensing options. It is a valuable resource for researchers and developers looking to leverage advanced open-source LLMs for a range of applications.

Respan, formerly Keywords AI, is an LLM engineering platform designed to streamline the development and deployment of reliable AI applications. It offers a comprehensive suite of features including LLM observability, automated evaluations (evals), prompt optimization, and a unified LLM gateway. The platform allows developers to trace, log, and evaluate agent behavior, identify failures, and understand the impact of prompt or model changes. Respan supports over 500 models and integrates with popular frameworks like OpenAI, Anthropic, LangChain, and LlamaIndex, enabling teams to monitor, debug, and improve their AI systems efficiently. It is built to add observability without becoming a performance bottleneck, making it suitable for production use.

ZeroClaw is an open-source, self-hosted AI personal assistant infrastructure built in Rust, designed for speed, minimal footprint, and full autonomy. It functions as an agent runtime, connecting to over 20 LLM providers including Anthropic, OpenAI, and Ollama, and integrates with 30+ channels like Discord, Telegram, email, and CLI. Users can deploy it on any OS or platform, ensuring complete ownership of their agent, data, and the machine it operates on. ZeroClaw emphasizes security with supervised autonomy, workspace boundaries, OS-level sandboxes, and cryptographic tool receipts, while also offering a 'YOLO mode' for trusted development environments. It supports hardware interaction, provides a web dashboard for management, and includes an SOP engine for event-triggered procedures.

Occams Group specializes in helping organizations navigate complex business and technology challenges by providing comprehensive Talent Services and Solution Delivery. They focus on connecting the right talent with specific project needs, delivering programs, modernizing platforms, and scaling AI initiatives. Their unique model integrates research-driven staffing with robust solutions delivery across critical domains such as software development, data analytics, artificial intelligence, cloud computing, cybersecurity, and ERP systems. Occams Group is designed to provide specialized project teams and facilitate end-to-end transformation, ensuring clients achieve their strategic objectives with expert support.

Pontis Technology is a software development and AI engineering partner that assists modern companies in building unique software solutions and scaling their teams. They offer comprehensive services including core software development and specialized AI services, focusing on best industry practices. Pontis helps turn ideas into life with expert-level product development, covering a wide range of front-end and back-end competencies. Their expertise extends to implementing new applications, optimizing existing systems, and delivering custom software and AI solutions. They are committed to building bridges in the digital age, ensuring clients receive robust, user-friendly, and visually appealing solutions.

Vectara is an enterprise agentic platform designed for building trusted AI agents with zero compromises. It offers governed, grounded, and auditable agents that can operate across SaaS, VPC, and On-Prem environments, providing flexibility in data control, security, and configurability. The platform emphasizes context accuracy through advanced context-engineering techniques, supporting multimodal data and complex documents to ensure grounded responses. Policy-led enforcement actively detects and corrects hallucinations, maintaining compliance and consistency. Vectara is built for enterprise scale, supporting rapid deployment from pilot to production and offering a single API for numerous use cases, all while ensuring brand protection and data security with always-on AI governance.

Ultralytics is an end-to-end computer vision platform designed to streamline the entire process from raw visual data to production-ready AI applications. It enables users to annotate datasets, train YOLO models on cloud GPUs, and deploy these models across 43 global regions, all within a unified workspace. As the creators of YOLO, Ultralytics offers unmatched depth in computer vision technology, with open-source foundations trusted by millions of developers. The platform supports the full YOLO family, including the latest YOLO26, YOLOv8, and earlier versions, covering detection, classification, pose estimation, and oriented bounding box tasks. It also features smart annotation tools that use machine learning to accelerate dataset creation by automatically generating initial annotations.

agents-starter is a comprehensive starter kit designed for developers to build AI chat agents on Cloudflare's platform. Leveraging the Agents SDK and Workers AI, it provides a robust foundation for creating intelligent agents without requiring an API key for Workers AI. The kit includes pre-built tools for common functionalities such as weather information, timezone detection, calculations with human approval, and task scheduling. It also supports vision capabilities for image understanding and offers features like streaming responses, real-time WebSocket connections, and message persistence. Developers can easily customize agents by changing system prompts, replacing demo tools with real API calls, and adding new server-side, client-side, or approval-based tools. The project structure is well-defined, making it straightforward to extend and deploy agents on Cloudflare's global network.

ai-gradio is a Python package designed to streamline the development of AI applications. Built on top of Gradio, it offers a unified interface to integrate with a wide array of AI providers, including OpenAI, Google Gemini, Anthropic, and Groq. Developers can easily create interactive applications featuring text, voice, and video chat capabilities, as well as specialized interfaces for code generation and multi-modal inputs. The package also supports advanced features like AI agent teams via CrewAI, browser automation, and computer-use agents for controlling virtual macOS/Linux environments, making it a versatile tool for rapidly prototyping and deploying diverse AI-powered solutions.

Agentgateway is an open-source proxy designed to provide comprehensive connectivity solutions for agentic AI. It is built on AI-native protocols like MCP and A2A, offering drop-in security, observability, and governance across various frameworks and environments. Key features include an LLM Gateway for routing traffic to major LLM providers with budget controls and load balancing, an MCP Gateway for connecting LLMs to tools and external data sources, and an A2A Gateway for secure agent-to-agent communication. It also supports intelligent inference routing, multi-layered guardrails for content filtering, and robust security and observability features like JWT, API keys, OAuth, and OpenTelemetry. Agentgateway can be deployed standalone or on Kubernetes, making it a versatile solution for managing complex AI agent interactions.

Vibbey is a platform designed for 'vibe coders' to quickly develop projects using AI. It facilitates participation in live contests and client projects, providing a structured environment for rapid development. The platform integrates with various tools such as Lovable, Bolt.new, Replit, and Emergent, allowing users to leverage their preferred development environments. Projects on Vibbey are typically time-boxed, emphasizing efficient execution and rewarding timely completion. This setup encourages a focused and productive approach to AI-powered development, catering to those who want to build and iterate quickly.

AIOS, the AI Agent Operating System, is designed to embed large language models (LLMs) directly into an operating system environment, streamlining the development and deployment of LLM-based AI Agents. It tackles critical operational challenges such as agent scheduling, context switching, memory management, storage management, and tool management, aiming to foster a robust AIOS-Agent ecosystem. The system comprises an AIOS Kernel, which acts as an abstraction layer over the OS kernel, managing resources like LLMs, memory, and tools, and an AIOS SDK (Cerebrum) for agent users and developers to build and run applications. AIOS supports both Web UI and Terminal UI, and offers various deployment modes including Local Kernel and Remote Kernel, with ongoing development for personal and virtualized remote kernels.

Aqueduct is an open-source MLOps framework designed to streamline the deployment and management of machine learning and LLM workloads across various cloud infrastructures. It offers a Python-native API, allowing users to define ML tasks in vanilla Python code and run them on platforms like Kubernetes, Spark, Airflow, or AWS Lambda. The tool provides centralized visibility into code, data, metrics, and metadata generated by each workflow run, ensuring confidence in pipeline performance and immediate alerts for issues. Aqueduct runs securely within your own cloud environment, maintaining data and code security. It is important to note that Aqueduct is no longer being maintained.

The AWS Neuron SDK is a comprehensive software development kit designed to enable high-performance deep learning acceleration on AWS's custom-designed machine learning accelerators, Inferentia and Trainium. It provides a complete ecosystem for developing, profiling, and deploying machine learning workloads on accelerated EC2 instances like Inf1 and Trn1. The SDK includes a compiler, runtime driver, and debugging/profiling utilities with a TensorBoard plugin for visualization. It is pre-integrated into popular machine learning frameworks such as PyTorch, TensorFlow, and MXNet, ensuring a seamless acceleration workflow for developers seeking blazing fast and cost-effective machine learning solutions.

auto-round is an advanced, open-source quantization toolkit developed by Intel, designed for optimizing Large Language Models (LLMs) and Vision-Language Models (VLMs). It excels at achieving high accuracy even at ultra-low bit widths (2–4 bits) with minimal tuning, leveraging sign-gradient descent. The toolkit offers broad hardware compatibility, supporting CPU, XPU, and CUDA, and integrates seamlessly with popular ecosystems like Transformers, vLLM, and SGLang. Key features include support for multiple export formats (AutoRound, AutoAWQ, AutoGPTQ, GGUF), fast mixed bits/datatypes scheme generation, and affordable quantization costs, allowing 7B models to be quantized in about 10 minutes on a single GPU. It also supports over 10 VLMs and provides advanced utilities like multiple GPU quantization and various calibration datasets.

NexusGPT is an enterprise AI agent platform designed to help large organizations deploy and scale autonomous AI agents rapidly. Unlike traditional software solutions, NexusGPT emphasizes a blended approach, providing not just the platform but also embedded engineers and change management expertise to ensure successful adoption and ROI. The platform connects with over 4,000 systems and supports various AI models, offering flexibility and avoiding vendor lock-in. It enables businesses to build agents for diverse functions like sales support, marketing, operations, and HR, with examples including account intelligence, lead enrichment, and proposal generation. NexusGPT focuses on delivering measurable outcomes, starting with 3-month POCs and ensuring ongoing optimization as agents learn and impact compounds.

Pienso empowers users to leverage machine learning for language data analysis without requiring any coding expertise. It offers an interactive and responsive learning interface that allows users to experiment, train, and deploy models effortlessly, imprinting their expertise at AI scale. The platform supports various use cases such as customer insights, content moderation, document intelligence, and risk detection. Pienso emphasizes data privacy by allowing deployment in the customer's preferred environment (cloud or on-premises), ensuring data remains private and is not used for training by Pienso. It also features PromptFactory for building production-caliber prompts without code and supports a 'garden of LLMs' approach for customized model development.