AgentIR: Workflow Aware Predictive Scheduling for Agent Graph LLM Serving - Krish Modi | Krish Modi
krishmodi.com · 1,633 words · saved by 1 readers
The fastest LLM request scheduler for agent systems
Loading... 1. Abstract AgentIR is a predictive LLM request scheduler for agent systems. It sits between agent builders and LLM providers, takes user supplied agent system code, compiles it into an intermediate representation, and uses that structure as runtime hints for online scheduling. The goal is simple: treat an agent as a workflow, not as a bag of independent calls, and use that workflow context to reduce end to end latency and improve throughput on the same underlying model servers. The result is scheduling that reduces average graph latency by 41.3% compared to round robin scheduling,
saved by
related reading
- LLM Powered Autonomous Agents | Lil'Loglilianweng.github.io
- Building Effective AI Agents \ Anthropicanthropic.com
- Building Effective AI Agents \ Anthropicanthropic.com
- Agent Evaluation: A Detailed Guidecameronrwolfe.substack.com
- A basic systems architecture for AI agents that do autonomous research — LessWronglesswrong.com
- LLM Engineer's Almanac - Advisormodal.com
- LLM Engineer's Almanac - Workloads | Modalmodal.com
- Low Latency and Model Training at Modalrhea24.github.io
- Prime Agent: A Self-Improving RLM Harnessarxiv.org
- [2410.00079] Interactive Speculative Planning: Enhance Agent Efficiency through Co-design of System and User Interfacearxiv.org
- Rethinking RL Infra for Agents | B'Logbillxbf.github.io
- Building Multi-Agents Supervisor System from Scratch with LangGraph & Langsmith | by Anurag Mishra | Mediummedium.com