Add safety header to LLM calls

[steven10a]

Added safety_identifier: 'oai-guardrails-ts' to the LLM calls

[Copilot]

Pull Request Overview

Adds a safety identifier to LLM calls so requests can be associated with a guardrails identifier.

• Injects safety_identifier into responses and chat payloads
• Uses a ts-ignore to bypass missing OpenAI type definitions

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
src/resources/responses/responses.ts	Adds safety_identifier to the responses request payload.
src/resources/chat/chat.ts	Adds safety_identifier to the chat request payload.

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

[nitpick] Avoid using // @ts-ignore here. Prefer defining a local type augmentation (e.g., extend the request params type with safety_identifier) or use // @ts-expect-error to ensure we revisit this when types change; this prevents silently suppressing unrelated type errors.

Suggested change

	// @ts-ignore - safety_identifier is not defined in OpenAI types yet
	safety_identifier: 'oai-guardrails-ts',
	}),
	safety_identifier: 'oai-guardrails-ts',
	} as Omit<OpenAI.Responses.ResponseCreateParams, 'input' | 'model' | 'stream' | 'tools'> & { input: typeof modifiedInput; model: typeof model; stream: typeof stream; tools?: typeof tools; safety_identifier: string }),

[Copilot]

[nitpick] Same as in responses: replace // @ts-ignore with a typed approach (type augmentation or // @ts-expect-error) so we don't mask other potential type issues.

Suggested change

	// @ts-ignore - safety_identifier is not defined in OpenAI types yet
	// @ts-expect-error - safety_identifier is not defined in OpenAI types yet

[Copilot]

PR title mentions adding a 'safety header', but this change adds a request payload field. Please either adjust the title/description or implement this as an HTTP header in the request layer if a header is required.

[gabor-openai]

Thank you!