feat: default web_search to approximate location (#3180)

[lrg913427-dot]

Closes #3180.

When web_search tools are used without specifying user_location, the server defaults to injecting country: "US" into the system prompt. This aligns SDK behavior with the OpenAI Playground by injecting user_location: {type: approximate} as the default, preventing hidden US geolocation assumptions for non-US developers.

Changes

• src/openai/lib/_tools.py: Added _apply_web_search_default_location_tools() helper that injects default user_location into web_search tools
• src/openai/resources/responses/responses.py: Applied preprocessing to all 6 tool parameter locations
• tests/lib/test_tools.py: 11 unit tests covering all edge cases

Behavior

• Tools without type: web_search* are unaffected
• Web search tools with existing user_location are not overridden
• Covers all 4 web_search type variants

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 07a9fba949

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Return collected tools to avoid exhausting iterators

If tools is a one-shot iterable (e.g., generator/iterator), this function consumes it in the for loop and then returns the original iterable when changed is false. In that case the caller (all responses.* create/stream paths now using this helper) receives an already-exhausted iterable, so no tools are sent at all. This regression appears whenever users pass non-reiterable Iterable[ToolParam] values that either contain no web-search tool or only web-search tools that already have user_location.

Useful? React with 👍 / 👎.