astra-sim/astra-sim

ASTRA-sim2.0: Modeling Hierarchical Networks and Disaggregated Systems for Large-model Training at Scale

What it solves

ASTRA-sim addresses the complexity of designing and deploying next-generation distributed AI systems by providing a way to simulate the entire software and hardware stack. This allows developers and architects to test different configurations of workload scheduling, communication algorithms, and hardware architectures without needing the physical hardware.

How it works

It functions as a distributed AI system simulator that models compute, memory, and network hardware alongside software layers. It uses a suite of APIs that allow users to plug in external open or proprietary components to model specific parts of the system. The simulator accepts MLCommons Chakra Execution Traces as inputs for the workload layer.

Who it’s for

It is designed for researchers and engineers working on the design and deployment of large-scale distributed AI systems.

Highlights

  • End-to-end multi-fidelity simulation of AI systems.
  • Models the full stack from workload scheduling to hardware architectures.
  • Plug-and-play API support for external components.
  • Integration with MLCommons Chakra Execution Traces.

Related

  • Project
    agarwalvishal/claude-chat-exporter

    A privacy-first browser tool that exports Claude.ai conversations to high-fidelity Markdown, optimized for Obsidian and RAG pipelines.

  • Project
    Nomadcxx/opencode-cursor

    A bridge that connects OpenCode to Cursor subscription models, translating prompts and tool calls between OpenCode and cursor-agent.

  • Project
    yzhao062/agent-style

    agent‑style is a Python/Node library that supplies 21 classic‑and‑LLM‑observed writing rules for AI coding‑and‑writing agents. The rules are injected into the model’s system prompt (soft enforcement) and can also be applied as a post‑generation review skill. Benchmarks on Claude Opus, GPT‑5.4, and Gemini 3 Flash show sizable drops in mechanical rule violations, proving the approach can make AI‑generated technical prose clearer.

  • Project
    chainbase-labs/manuscript-core

    A protocol and framework for simplifying omnichain data access and processing, enabling developers to unify blockchain data from multiple sources into a a single pipeline.