oncer on nodejs
empty
empty
empty
empty
empty
empty
a blazingly fast tool to run things once
Multi-agent network simulation with consideration-based decision making
HTTP API and orchestrator state machine for Concerto: routes, request lifecycle, cold-start dedup, eviction, and graceful shutdown.
Backend process lifecycle for Concerto: BackendManager trait, mock implementation, and a real process-spawning implementation for inference engines.
Concerto CLI: a Rust inference multiplexer for self-hosted LLMs. Orchestrates vLLM, llama.cpp, and SGLang processes across 1-8 GPUs with demand-aware routing.
TOML configuration parsing for Concerto.
Pure routing logic, eviction policies, and memory accounting for Concerto. Zero IO dependencies.
GPU monitoring abstractions for Concerto: GpuMonitor trait, mock implementation, and optional NVML-backed implementation.