nemo-relay-tune-adaptive-config

por nvidia

Configure the NeMo Relay adaptive plugin component through the shared plugin system; use this skill for state, telemetry, adaptive_hints, tool_parallelism,…

npx skills add https://github.com/nvidia/nemo-relay --skill nemo-relay-tune-adaptive-config

Configure Adaptive Tuning

Use This When

Use this skill when an application already intends to use NeMo Relay adaptive features and needs the correct plugin configuration shape.

Do Not Use This When

Do not use this skill when the user only needs first-time instrumentation, request-specific middleware, or production trace debugging.

Embedded Configuration Model

  • Adaptive tuning is a top-level plugin component with kind adaptive.
  • The adaptive object contains version, agent_id, state, telemetry, adaptive_hints, tool_parallelism, acg, and policy.
  • Wrap the adaptive object in an adaptive ComponentSpec, insert it into the shared plugin config components list, validate the plugin config, then initialize the plugin system.
  • Python uses nemo_relay.adaptive.AdaptiveConfig(...), nemo_relay.adaptive.ComponentSpec(...), and nemo_relay.plugin.PluginConfig(...).
  • Node.js uses require("nemo-relay-node/adaptive") helpers such as defaultConfig(), inMemoryBackend(), toolParallelismConfig(...), and ComponentSpec(...), then activates through nemo-relay-node/plugin.
  • Rust uses nemo_relay_adaptive::{AdaptiveConfig, ComponentSpec, ...} and nemo_relay::plugin::{validate_plugin_config, initialize_plugins}.
  • Go and raw FFI are source-first or advanced surfaces.
  • Plugins install runtime behavior such as subscribers, guardrails, intercepts, and related helpers. Adaptive is a built-in plugin component, not a separate runtime model.

Default Path

  1. Build the shared plugin config document or binding-native helper config.
  2. Add one top-level adaptive ComponentSpec.
  3. Start with state.backend = in_memory.
  4. Enable telemetry first.
  5. Add only one active section at a time: adaptive_hints, tool_parallelism, or acg.
  6. Validate the config before initialization.
  7. Initialize through the shared plugin system.
  8. Clear or replace the plugin configuration cleanly when the app lifecycle changes.

Defaults To Remember

  • adaptive_hints.priority defaults to 100, break_chain to false, inject_header to true, and inject_body_path to nvext.agent_hints.
  • tool_parallelism.mode defaults to observe_only.
  • acg.provider defaults to passthrough, with priority 50 and observation window 100.
  • Redis-backed state is for persistence or cross-worker sharing, not the first local rollout.

Failure Modes To Avoid

  • Do not initialize before validation succeeds.
  • Do not enable multiple active tuning sections in the first rollout.
  • Do not put callables, clients, credentials, or framework objects in plugin config.
  • Do not enable active scheduling or request rewriting without a baseline and a rollback path.

Checklist

  • Adaptive is modeled as a top-level plugin component.
  • Backend chosen, with in_memory first unless persistence is required.
  • Adaptive sections chosen explicitly.
  • Config validated before initialization.
  • Rollout starts from telemetry or observe-only behavior.
  • Plugin lifecycle matches the app lifecycle.

Related Skills

  • nemo-relay-tune-performance
  • nemo-relay-tune-adaptive-hints
  • nemo-relay-debug-runtime-integration
  • nemo-relay-build-plugin

Más skills de nvidia

compileiq-debug
nvidia
Úsalo cuando algo esté mal: Search() se cuelga, todas las evaluaciones devuelven INVALID_SCORE, las puntuaciones no mejoran, cada configuración devuelve el mismo número, errores de ptxas…
official
create-github-pr
nvidia
Crear solicitudes de extracción de GitHub usando la CLI gh. Usar cuando el usuario quiera crear un nuevo PR, enviar código para revisión o abrir una solicitud de extracción. Palabras clave de activación -…
official
diagnose-perf
nvidia
First-responder performance triage for Isaac Sim and Isaac Lab. Identifies bottleneck category (GPU-bound, CPU-bound, VRAM, loading) using nvidia-smi and…
official
eagle3-review-logs
nvidia
Revisa los registros de experimentos del pipeline EAGLE3 desde el directorio experiments/ del lanzador. Resume el estado de aprobación/fallo para las 4 tareas, diagnostica fallos con la causa raíz…
official
nemoclaw-maintainer-cross-issue-sweep
nvidia
Scans other open issues to find ones a given PR may also fix or accidentally break. Outputs adjacent-fix opportunities and contradiction risks with file:line…
official
karpathy-guidelines
nvidia
Pautas de comportamiento para reducir errores comunes de codificación en LLM. Úselas al escribir, revisar o refactorizar código para evitar la sobrecomplicación, realizar cambios quirúrgicos,…
official
fhir-basics
nvidia
Enseña a los agentes cómo funcionan las APIs de FHIR R4, qué recursos están disponibles, cómo consultarlos con parámetros de búsqueda y cómo analizar correctamente todos los formatos de respuesta…
official
underdeclared-agent
nvidia
A helpful assistant agent
official