Beam request-file review example — no API calls, keys, or paid inference

Verified runtime: Node.js 24.20.0. No packages or installation are needed.
.node-version records that runtime; it does not install or switch Node by itself.

node --version
node inspect-request.mjs request.example.json
node --test inspect-request.test.mjs

Scope
The JSON is an example request body for Vercel AI Gateway's Chat Completions API.
providerOptions.gateway.only contains only reflection. No models fallback field is
included. This is not a configuration file for a specific coding agent or SDK.
It was checked locally against an editorial policy, not submitted to AI Gateway.
API acceptance, serving, model availability, prices and budgets are unverified.

Exit codes
0: This request file matches the three local choices: beam-free, Reflection only,
   and no model-fallback field. The stillUnverified checks remain outstanding.
1: Review the listed findings. Unknown fields also require manual review.
2: Missing path, unreadable file or invalid JSON.

The checker is deliberately narrow and is not the complete Gateway schema. It
flags extra fields, including valid API options outside its reviewed scope. An
empty models array is also flagged: omission is the example's local policy.
It does not edit your file, intercept live requests or enforce team-side policy.

Official announcement (2026-10-16)
reflection/beam via the reflection provider changes to paid pricing when the
limited offer ends. reflection/beam-free stops serving when the offer ends.
The announcement does not give an exact end date. The model's suffix does not
promise zero total account, tool or compute spend. This package does not simulate
or reproduce the future offer ending.

Before actual use, separately inspect
- The request serialized by your pinned agent/SDK, not just its display name.
- Agent retries/fallbacks and providerOptions.gateway.models.
- Your actual team's rewrite/deny/virtual-model routing rules.
- Request logs for model, provider, cost, authentication and request ID.
- API key budget amount/period and the authentication types it covers.
- Current offer status, exact end date if published and provider-specific prices.

only is a provider restriction; order is a preference. Restricting to reflection
does not prevent another Reflection-served model from being chosen by a model
fallback or applicable rewrite. The local checker cannot read any team rules.
The API key budget docs specify a minimum $1, a propagation delay and a soft cap:
the request that crosses the limit can finish and overshoot it. This is not
a zero-dollar switch. Do not change a live team's rules or revoke a shared key
without reviewing which applications use it.
