Releases·Tool Release
We built an open-source GPU profiler you point an AI agent at, instead of reading traces yourself
Graphsignal released an open-source sidecar GPU profiler designed to be read by AI agents rather than humans. It wraps inference servers, outputs one JSON endpoint of kernel, memory, power, and engine metrics, and lets an agent run the profile-and-tune loop autonomously.
40 engagement1 sourceThu, Sep 17, 2026, 12:27 PM
AI Overview
- Graphsignal's open-source sidecar profiler wraps vLLM, SGLang, llama.cpp, or any CUDA/ROCm process and serves kernel, CUDA graph, memcpy, NVML, Prometheus, and traceback data as JSON at 127.0.0.1:18259/signals for agent-driven tuning loops
- --cuda-graph-trace node breaks out kernels hidden inside CUDA-graph replay in vLLM, SGLang, and llama.cpp decode; a lock-free probe header adds per-code-section values to /signals
- The profiler runs locally without root, requires no code changes or imports, and uploads nothing unless given an API key; a repo SKILL.md lets agents run the tune loop, and curl/jq works for humans
Entities
Related
Releases·Tool Release
Graphsignal launches auto-optimizing inference with GPU profiling and telemetry
Graphsignal released `graphsignal-run --auto-flags`, a tool that automatically tunes vLLM, SGLang, and TensorRT-LLM inference flags based on GPU profiling and workload telemetry. It adjusts startup flags on each restart to adapt to changing traffic patterns, eliminating manual tuning.
Graphsignal released `graphsignal-run --auto-flags`, a tool that automatically tunes vLLM, SGLang, and TensorRT-LLM inference flags based on GPU profiling and workload telemetry. It adjusts startup flags on each restart to adapt to changing traffic patterns, eliminating manual tuning.
8 engagement1 sourceredditWed, Jul 22, 2026, 05:12 PM
Products·launched
PolyServe: open-source LLM serving autotuner benchmarking configs across vLLM, SGLang, and llama.cpp
PolyServe is an open-source autotuner for LLM serving that benchmarks configurations across vLLM, SGLang, and llama.cpp, covering weight precision, KV cache, batching, and speculative decoding. It selects a configuration meeting the user's latency constraints and serves it, replacing fixed tuning heuristics. It targets self-hosters and engineers unsure how to optimize inference for their GPU setup.
PolyServe is an open-source autotuner for LLM serving that benchmarks configurations across vLLM, SGLang, and llama.cpp, covering weight precision, KV cache, batching, and speculative decoding. It selects a configuration meeting the user's latency constraints and serves it, replacing fixed tuning heuristics. It targets self-hosters and engineers unsure how to optimize inference for their GPU setup.
23 engagement1 sourceredditThu, Sep 17, 2026, 06:51 AM
Releases·Tool Release
Artificial Analysis releases agentperf-local, an open-source benchmark replaying real agent trajectories against local LLM servers
Artificial Analysis released agentperf-local, an open-source tool that benchmarks how fast a machine serves an AI agent by replaying recorded agent trajectories against an OpenAI-compatible model server. It reports throughput and latency while carrying full conversation context per request, measuring speed without quality. It matters to practitioners running local inference who want agent-representative performance numbers rather than generic throughput benchmarks.
Artificial Analysis released agentperf-local, an open-source tool that benchmarks how fast a machine serves an AI agent by replaying recorded agent trajectories against an OpenAI-compatible model server. It reports throughput and latency while carrying full conversation context per request, measuring speed without quality. It matters to practitioners running local inference who want agent-representative performance numbers rather than generic throughput benchmarks.
58 engagement1 sourcegithubSat, Sep 26, 2026, 03:11 AM
Products·launched
Agentic CUDA Kernel Optimizer
An open-source tool where AI agents optimize CUDA kernels by running them in a C++ test harness, collecting benchmark results, and profiling with Nsight. The agents use LangGraph to orchestrate the iterative run-benchmark-profile loop. It targets developers who want automated performance tuning of their CUDA code.
An open-source tool where AI agents optimize CUDA kernels by running them in a C++ test harness, collecting benchmark results, and profiling with Nsight. The agents use LangGraph to orchestrate the iterative run-benchmark-profile loop. It targets developers who want automated performance tuning of their CUDA code.
18 engagement1 sourcehackernewsFri, Sep 25, 2026, 10:32 AM
Products·launched
Linux profiling with a local AI
A tool that uses a local LLM to automate performance profiling of Linux programs. It analyzes profiling data and provides insights, demonstrated with LLaMa.cpp profiling an LVGL HMI on QEMU.
A tool that uses a local LLM to automate performance profiling of Linux programs. It analyzes profiling data and provides insights, demonstrated with LLaMa.cpp profiling an LVGL HMI on QEMU.
2 engagement1 sourceredditTue, Jul 14, 2026, 03:23 PM