OpenAI-compatible gateway smoke test: verify the first API call before scaling

When you try a new AI API route, the first goal should not be production migration.

The first goal is much smaller: make one successful request, then test a real workload.

Disclosure: I am working on modelrouter.site. It is an independent third-party OpenAI-compatible AI API gateway, not an official service from OpenAI, Anthropic, Google, DeepSeek, or any model provider.

The smoke test workflow

1. Create an API key.
2. Pick a supported model route from your dashboard.
3. Run one cURL request.
4. Confirm the response works.
5. Only then run 10-20 real tasks.

Example request

export BASE_URL="https://modelrouter.site/v1"
export API_KEY="your_test_key"
export MODEL="copy_a_supported_route_from_your_dashboard"

curl "$BASE_URL/chat/completions" \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "'"$MODEL"'",
    "messages": [
      {"role": "user", "content": "Say hello in one short sentence."}
    ]
  }'

Never paste your full API key into screenshots, support chats, GitHub issues, or public posts.

What to record before scaling

• model route
• task type
• success or failure
• latency
• output usefulness
• retry count
• estimated spend
• error message or request ID if it fails

Why this matters

A route is useful only if it works for your real task.

For example:

• extraction may need consistency
• support drafting may need tone control
• coding tasks may need longer context
• RAG answers may need low hallucination risk
• agent workflows may need predictable latency

The practical decision is not "which model is best?" It is "which route works for this task at an acceptable cost and reliability level?"

Try the smoke test:

https://modelrouter.site/?utm_source=devto&utm_medium=tutorial&utm_campaign=priority_first_call_quickstart