Bite-sized news from the Dynamo team, tagged by SIG.
Join at ai-dynamo.org/slack →
Interactive HTML presentations from the Dynamo team. Use ← → to move through a deck and F for full screen.
Start here: what Dynamo is and how to get it running.
Dynamo intro.htmlOpen slides ↗🛠️Deploy and tune DynamoHands-on setup from a laptop to Kubernetes, tested recipes, the tuning levers that matter, and agent skills that optimize for you.setup.htmlOpen slides ↗How Dynamo serves the workloads teams are building today.
rl.htmlOpen slides ↗🕵🏻♂️AgentsHow Dynamo serves agent programs: session IDs, tracing and simulation, program-level scheduling with ThunderAgent, and programmatic KV caching.agent.htmlOpen slides ↗🎨Omni: multimodal, speech and diffusion servingVision-language, speech and diffusion models on one OpenAI-compatible serving stack, with live demos.omni.htmlOpen slides ↗Deep dives into each piece of the stack.
mx-snapshot.htmlOpen slides ↗🗄️KVCR (KV Cache Runner)Distributed KV cache management: the Dynamo router plans where KV goes, and KVCR moves it between workers and keeps it through failures.kvcr.htmlOpen slides ↗🧪AISimSimulating the whole inference stack (traffic, scheduler, KV cache, router, planner, GPU timing) on a laptop, without GPUs.aisim.htmlOpen slides ↗🔀RouterHow the Dynamo router picks a worker, survives overload and restarts, decides what gets admitted, and how to plug in your own policy, plus where it's going.router.htmlOpen slides ↗🌐Global RouterThe existing router run one level up, where each pool acts as a single super-worker.global-router.htmlOpen slides ↗🛡️Fault ToleranceKeep serving when something breaks.ft.htmlOpen slides ↗📈Planner + GrovePlanner decides how many engines to run; Grove decides what one engine instance is and where it lands.planner-grove.htmlOpen slides ↗🚗SidecarTake Dynamo out of the engine process: a CPU-only sidecar plugs a stock vLLM, SGLang or TRT-LLM server into Dynamo.sidecar.htmlOpen slides ↗📊AIPerfBenchmark any inference endpoint with production-shaped traffic, and get the metrics that decide your SLA.aiperf.htmlOpen slides ↗🎛️AITuneOne API that finds the fastest validated backend for every module of a PyTorch model.aitune.htmlOpen slides ↗🔄FlexTensorRun models bigger than your GPU: weights stream in from host RAM, just in time.flextensor.htmlOpen slides ↗Every Dynamo dashboard in one place. Pick one to open it.