Adding guardrails to large language models.

A bidirectional pipeline parallelism algorithm for computation-communication overlap in V3/R1 training.

AI-native (edge and LLM) proxy for agents. Move faster by letting Arch handle the pesky heavy lifting in building agentic apps -- ⚡️ query understanding and routing, seamless integration of prompts…

Python library to access Kaggle resources

Search-R1: An Efficient, Scalable RL Training Framework for Reasoning & Search Engine Calling interleaved LLM based on veRL

MySQL client library for Python

Official codebase for "SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution"

Build a TUI with Python's Argparse

OctoTools: An agentic framework with extensible tools for complex reasoning

EasyR1: An Efficient, Scalable, Multi-Modality RL Training Framework based on veRL

NeMo Guardrails is an open-source toolkit for easily adding programmable guardrails to LLM-based conversational systems.

The python library for real-time communication

DeepEP: an efficient expert-parallel communication library

Academic Survey Paper Generation.

Prevent merging of malicious code in pull requests

Sky-T1: Train your own O1 preview model within $450

Profile-Based Long-Term Memory for AI Applications

AI Agent Builder in Python

WebAssembly version of DuckDB

Training-free Post-training Efficient Sub-quadratic Complexity Attention. Implemented with OpenAI Triton.

"Your Fully-Automated Personal AI Assistant, and Open-Source & Cost-Efficient Alternative to OpenAI's Deep Research"

Pretraining code for a large-scale depth-recurrent language model

Official repo of EmbodiedBench, a comprehensive benchmark designed to evaluate MLLMs as embodied agents.