docs: wrap chat example in chatRequest

[slegarraga]

Summary

Fixes the README chat example so the request body is passed under chatRequest, matching the generated SDK method input shape.

Why

The current README example passes messages, model, provider, and stream directly to openRouter.chat.send(...). Issue #413 notes that the generated SDK expects those fields inside chatRequest, otherwise runtime validation rejects the call.

Validation

• git diff --check

This is docs-only; I kept the change limited to the affected example.

Fixes #413.

[perry-the-pr-reviewer]

Perry's Review

Verdict: 💬 Comments / questions
Risk: 🟢 Low

Code change is LGTM, but I can only post a COMMENT here — the maintainer approval app isn't installed on OpenRouterTeam, so a human maintainer should click Approve. (No code changes requested.)

Review details

Correct docs fix. The generated Chat.send method requires a top-level chatRequest: models.ChatRequest
field on its request type, plus optional globals (httpReferer, appTitle, appCategories,
xOpenRouterMetadata). The previous README example passed messages/model/provider/stream
directly to openRouter.chat.send(...), which the Zod outbound schema rejects at runtime (and
TypeScript flags) since those fields belong on chatRequest, not on the request root. Wrapping
them inside chatRequest matches the SDK input shape, and the stream: true overload lines up
with the for await consumption below it. The chatRequest schema itself carries messages
(required array), model (optional string), provider (optional), and stream (boolean,
default false), so every field in the updated example is valid.

Risk assessment:

Dimension	Severity	Risk	Reasoning
Implementation risk	🟩	Low	Docs-only; the wrapping now matches the generated method signature exactly.
Premise risk	🟩	Low	The premise (direct-pass is invalid) is independently verifiable from the SDK source.
Estimated impact	🟩	Low	Worst case is a misleading README example; no runtime/code impact.

Risk Factor	Severity	Risk	Reasoning
Reversibility	🟩	Low	One-line markdown revert.
Detectability	🟩	Low	A bad example surfaces immediately to any user who copies it.
Blast radius	🟩	Low	A single docs file; no code, config, or data path is touched.
Data integrity		None	No persisted state is touched.
Financial exposure		None	No billing/accounting outcome affected.
Security and privacy exposure		None	No secrets, auth, or tenant-isolation surface.
Propagation	🟩	Low	Only README readers; nothing downstream reads this output.
Availability		None	Cannot affect serving.
Recovery cost	🟩	Low	Trivial markdown edit.
Time to correct	🟩	Low	Noticed and fixed in one pass.