Dynamo Router Starter
OpenAIOfficialCreated by Dan GilDistributed by OpenAI
- Skill
- Tool integration
- Software development
- Apache-2.0
- Medium risk
- Has scripts
AutoGineer summary
Written by AutoGineer from the skill’s own files, not by the author
Dynamo Router Starter provides a standardized procedure for initializing and patching Dynamo router configurations. It supports multiple routing strategies including round-robin, KV-aware, and least-loaded modes. The skill includes instructions for both local Python-based setups and Kubernetes deployments, ensuring that the frontend is correctly configured and that backend workers are properly registered. It also provides a health check script to verify model discovery and basic chat completion functionality, helping users transition from initial bring-up to functional validation.
What it does
- Configures Dynamo frontend router modes via CLI or Kubernetes environment variables.
- Supports various routing strategies like round-robin, KV-aware, and least-loaded.
- Provides a smoke test script to verify model availability and chat completion endpoints.
- Offers guidance on enabling approximate KV mode when worker events are unavailable.
- Facilitates the transition from initial setup to troubleshooting if endpoints fail.
- Standardizes the validation process for Dynamo router deployments.
When to use it
- When setting up a new Dynamo router instance locally or in Kubernetes.
- When switching between different routing modes to optimize traffic distribution.
- When verifying that a deployed Dynamo frontend is correctly communicating with backend workers.
- When performing initial smoke tests before running full-scale performance benchmarks.
Advantages
- Simplifies complex router configuration by providing clear CLI and YAML patterns.
- Ensures consistent validation through automated health check scripts.
- Helps identify common configuration issues like missing worker registrations or KV event mismatches.
- Provides a clear path for troubleshooting when endpoints are unhealthy.
Usage examples
- “Start a local Dynamo frontend in KV-aware mode on port 8000 and verify it.”
- The agent executes the frontend command with the appropriate flags and runs the health check script to confirm the endpoint is operational.
- “How do I configure my Kubernetes Dynamo frontend for KV routing?”
- The agent provides the necessary environment variable patches for the frontend service, including the option to toggle approximate KV mode.
- “Run a smoke test on my Dynamo router at http://127.0.0.1:8000.”
- The agent invokes the health check script against the specified URL to verify model discovery and chat completion success.
Install
Download the skill as a ZIP:
https://autogineer.com/agent-skills/openai/dynamo-router-starter/downloadUnzip it into the cross-client skills folder of your project:
mkdir -p .agents/skills && unzip dynamo-router-starter.zip -d .agents/skills
Documentation by Dan Gil
Dynamo Router Starter
<!-- SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved. SPDX-License-Identifier: CC-BY-4.0 -->
Purpose
Make Dynamo routing feel easy by getting a baseline router mode running, enabling KV-aware routing when appropriate, and proving the endpoint works. Keep the user focused on exact commands and success signals, not router internals.
Prerequisites
Python 3.10+ with the
dynamopackage importable (python3 -m dynamo.frontend --helpworks).For Kubernetes runs:
kubectlconfigured with access to the target namespace and a deployed Dynamo recipe.Network reachability to the frontend service (port-forward or direct).
A model already loaded into at least one worker (
/v1/modelsreturns at least one entry).
Required Inputs
Collect or infer:
local Python/CLI or Kubernetes recipe path
desired mode:
round-robin,kv,least-loaded,device-aware-weighted,direct, orrandomfrontend port or Kubernetes frontend service
whether workers publish KV events; if not, use approximate KV mode
model name for smoke requests, if
/v1/modelscannot discover it
Instructions
1. Establish A Baseline
For local bring-up with already registered workers:
python3 -m dynamo.frontend --router-mode round-robin --http-port 8000For Kubernetes, inspect the selected recipe deploy.yaml and locate the
frontend service. If the recipe is not already deployed, use
dynamo-recipe-runner first.
2. Enable KV Routing
For local frontend:
python3 -m dynamo.frontend --router-mode kv --http-port 8000For Kubernetes, patch only the frontend service env:
envs:
- name: DYN_ROUTER_MODE
value: kvIf backend workers are not publishing KV cache events, set approximate mode instead of leaving the router waiting for events:
envs:
- name: DYN_ROUTER_USE_KV_EVENTS
value: "false"3. Smoke Test
After port-forwarding the frontend service or starting local frontend, run:
python3 scripts/check_router_health.py \
--base-url http://127.0.0.1:8000This must verify /v1/models and, when a model is discoverable, one
/v1/chat/completions request.
4. Compare Modes Carefully
When comparing round-robin vs KV routing:
use the same model, workers, prompt set, concurrency, and sampling settings
send repeated-prefix prompts if demonstrating KV reuse
label the result as a smoke comparison unless enough benchmark samples were collected
do not claim throughput improvement from a single chat request
If the endpoint is unhealthy or workers are missing, switch to
dynamo-troubleshoot.
Available Scripts
| Script | Purpose | Arguments |
|---|---|---|
scripts/check_router_health.py | Smoke-test /v1/models and one chat completion against a Dynamo frontend | --base-url, --retries, --timeout |
Invoke via the agentskills.io run_script() protocol:
run_script("scripts/check_router_health.py", args=["--base-url", "http://127.0.0.1:8000"])Examples
Local KV-routed frontend on port 8000, then smoke-test it:
python3 -m dynamo.frontend --router-mode kv --http-port 8000 &
python3 scripts/check_router_health.py --base-url http://127.0.0.1:8000Kubernetes-deployed frontend reachable via port-forward:
kubectl port-forward svc/qwen-vllm-disagg-frontend 8000:8000 -n dynamo-demo &
python3 scripts/check_router_health.py --base-url http://127.0.0.1:8000 --retries 3Equivalent through the agent protocol:
run_script("scripts/check_router_health.py", args=["--base-url", "http://127.0.0.1:8000", "--retries", "3"])Output Contract
Return:
mode selected and why
local command or Kubernetes env patch
frontend service or URL
smoke-test result
any limitation, such as approximate KV mode or missing worker KV events
next command to run for a fuller comparison
Limitations
Smoke test is one chat completion; it is not a benchmark. Use
dynamo-benchmarkfor throughput/latency numbers.KV-aware mode without worker KV-event publication degrades to approximate mode; this skill flags but does not fix the underlying worker config.
Mode comparisons require matched workloads; cross-mode latency claims need separate benchmark runs.
Troubleshooting
| Symptom | Likely cause | Next step |
|---|---|---|
/v1/models returns empty list | No worker registered with the frontend | Verify worker pods are Ready; confirm they connect to the same etcd/NATS |
| Smoke chat request times out | Frontend up, workers not serving | Switch to dynamo-troubleshoot; inspect worker logs |
| KV mode hangs | Workers do not publish KV cache events | Set DYN_ROUTER_USE_KV_EVENTS=false (approximate mode) |
| Connection refused on port-forward | Port-forward dropped or wrong service name | Re-run port-forward; verify the frontend service name matches the recipe |
Benchmark
See BENCHMARK.md for the NVCARPS-EVAL performance report (auto-generated by the NVSkills CI pipeline). To refresh, re-run /nvskills-ci on an upstream PR touching this skill.
References
Read
references/router-modes.mdfor the compact mode/env map.Use
scripts/check_router_health.pyfor endpoint smoke tests.
Files
- ATTRIBUTION.mdAdded by AutoGineer1.0 KiB
- BENCHMARK.md2.6 KiB
- LICENSEAdded by AutoGineer10.0 KiB
- SKILL.md5.6 KiB
- agents/openai.yaml303 B
- evals/evals.json3.0 KiB
- references/router-modes.md2.1 KiB
- scripts/check_router_health.pyScript4.7 KiB
- skill-card.md2.4 KiB
- skill.oms.sig5.0 KiB
scripts/check_router_health.py
#!/usr/bin/env python3
# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
"""Smoke test a Dynamo OpenAI-compatible frontend."""
from __future__ import annotations
import argparse
import json
import sys
import time
import urllib.error
import urllib.parse
import urllib.request
from typing import Any
# Tunables and contract values (kept here to avoid magic numbers in the body).
DEFAULT_BASE_URL = "http://127.0.0.1:8000"
DEFAULT_PROMPT = "Say hello from Dynamo in one short sentence."
DEFAULT_MAX_TOKENS = 32
DEFAULT_RETRIES = 5
DEFAULT_RETRY_SLEEP_SEC = 2.0
DEFAULT_HTTP_TIMEOUT_SEC = 20.0
HTTP_OK = 200
# Process exit codes used to distinguish smoke-test outcomes.
EXIT_OK = 0
EXIT_MODELS_UNAVAILABLE = 2
EXIT_NO_MODEL_DISCOVERED = 3
EXIT_CHAT_FAILED = 4
def request_json(
method: str,
url: str,
payload: dict[str, Any] | None = None,
timeout: float = DEFAULT_HTTP_TIMEOUT_SEC,
) -> tuple[int, Any]:
# Only talk to real HTTP(S) endpoints; urlopen otherwise happily opens
# file:// and other local schemes if a bad --base-url is passed.
scheme = urllib.parse.urlparse(url).scheme
if scheme not in ("http", "https"):
return 0, {"error": f"unsupported URL scheme: {scheme!r}"}
data = None
headers = {"Accept": "application/json"}
if payload is not None:
data = json.dumps(payload).encode("utf-8")
headers["Content-Type"] = "application/json"
req = urllib.request.Request(url, data=data, headers=headers, method=method)
try:
with urllib.request.urlopen(req, timeout=timeout) as resp: # noqa: S310
raw = resp.read().decode("utf-8", errors="replace")
if not raw:
return resp.status, None
try:
return resp.status, json.loads(raw)
except json.JSONDecodeError:
# A 200 with a non-JSON body should surface as a structured
# failure, not crash the smoke test.
return resp.status, raw
except urllib.error.HTTPError as exc:
raw = exc.read().decode("utf-8", errors="replace")
try:
body = json.loads(raw)
except json.JSONDecodeError:
body = raw
return exc.code, body
except urllib.error.URLError as exc:
return 0, {"error": str(exc.reason)}
def choose_model(models_body: Any) -> str | None:
if not isinstance(models_body, dict):
return None
data = models_body.get("data")
if isinstance(data, list) and data:
first = data[0]
if isinstance(first, dict) and first.get("id"):
return str(first["id"])
return None
def main() -> int:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--base-url", default=DEFAULT_BASE_URL)
parser.add_argument("--model")
parser.add_argument("--prompt", default=DEFAULT_PROMPT)
parser.add_argument("--max-tokens", type=int, default=DEFAULT_MAX_TOKENS)
parser.add_argument("--skip-chat", action="store_true")
parser.add_argument("--retries", type=int, default=DEFAULT_RETRIES)
parser.add_argument("--retry-sleep", type=float, default=DEFAULT_RETRY_SLEEP_SEC)
args = parser.parse_args()
base_url = args.base_url.rstrip("/")
result: dict[str, Any] = {"base_url": base_url, "ok": False, "checks": []}
models_status = None
models_body = None
for attempt in range(1, args.retries + 1):
models_status, models_body = request_json("GET", f"{base_url}/v1/models")
if models_status == HTTP_OK:
break
time.sleep(args.retry_sleep)
model = args.model or choose_model(models_body)
result["checks"].append(
{"name": "models", "status": models_status, "body": models_body, "model": model}
)
if models_status != HTTP_OK:
print(json.dumps(result, indent=2))
return EXIT_MODELS_UNAVAILABLE
if args.skip_chat:
result["ok"] = True
print(json.dumps(result, indent=2))
return EXIT_OK
if not model:
result["checks"].append(
{"name": "chat", "status": "skipped", "reason": "No model discovered"}
)
print(json.dumps(result, indent=2))
return EXIT_NO_MODEL_DISCOVERED
payload = {
"model": model,
"messages": [{"role": "user", "content": args.prompt}],
"max_tokens": args.max_tokens,
}
chat_status, chat_body = request_json(
"POST", f"{base_url}/v1/chat/completions", payload
)
result["checks"].append({"name": "chat", "status": chat_status, "body": chat_body})
result["ok"] = chat_status == HTTP_OK
print(json.dumps(result, indent=2))
return EXIT_OK if result["ok"] else EXIT_CHAT_FAILED
if __name__ == "__main__":
sys.exit(main())
Requirements and compatibility
Written for: OpenAI Codex
Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta
“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.
Needs: Needs a command-line tool, Needs Python
Security
What our scanner noticed· Medium risk
- Large base64-encoded blobs
- Scripts that your agent may run
This skill includes scripts that your agent may run. Read them before you use it.
A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.
Attribution and licence
- Created by
- Dan Gil
- Publisher
- OpenAI (Official)
- Distributed by
- OpenAI
- Source
- openai/plugins · plugins/nvidia/skills/dynamo-router-starter · commit d420d13
- Licence
- Apache License 2.0 (redistribution allowed)
- LICENSE (written by AutoGineer)
- ATTRIBUTION.md (written by AutoGineer)
- Synced
- Also distributed in
- NVIDIA/skills
AutoGineer is not affiliated with or endorsed by Dan Gil or OpenAI.
Download ZIP
The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download dynamo-router-starter.zip