[xgp]

Closes #104.

Adds optional location_option and negative_location_option to draft_campaign and update_campaign, in the shape agreed on the issue.

Parameters

Parameter	Values	Field
location_option	PRESENCE, PRESENCE_OR_INTEREST, SEARCH_INTEREST	campaign.geo_target_type_setting.positive_geo_target_type
negative_location_option	PRESENCE, PRESENCE_OR_INTEREST	campaign.geo_target_type_setting.negative_geo_target_type

Values are case-insensitive. Empty leaves the value as it is: on draft_campaign that means Google's default, so existing callers get the same request and the same preview as before. A test pins that.

Preview

• update_campaign reads the campaign's current options and puts them next to the new ones in changes: current_location_option → location_option, and current_negative_location_option → negative_location_option. This follows the current_daily_budget pattern. An unknown campaign is refused at draft time.
• Positive option: any value other than PRESENCE adds a warning that the change widens reach.
• Negative option: PRESENCE adds a warning that the exclusions now exclude fewer people. The negative side runs the other way round: PRESENCE_OR_INTEREST is the wider exclusion.
• No change, no warning. On update_campaign, a value the campaign already has produces no warning.
• Confirmation: a single confirmation everywhere; requires_double_confirm stays false.

Apply and dry run

Both options go on the campaign operation the tool already builds:

• on create, alongside network_settings;
• on update, with their own update_mask paths: geo_target_type_setting.positive_geo_target_type / .negative_geo_target_type.

So the dry run is the existing validate_only path; nothing new.

I ran a validate-only dry run against a paused Search campaign in a live account. It came back DRY_RUN_SUCCESS with validated_calls: 1. Nothing was applied.

Rules and tests

• Rules: .cursor/rules/adloop.mdc now has the tool table entries and one bullet on what the options mean. Synced with scripts/sync-rules.py.
• Tests: tests/test_location_option.py (16 tests) covers:
  • drafts with the option omitted, PRESENCE, and the wider values;
  • invalid values, refused before any read;
  • old → new in the preview, warnings only on an actual widening change, and an unknown campaign;
  • the create and update operations and the update mask;
  • parameter descriptions on both tools.
• Suite: uv run pytest gives 1679 passed, 2 skipped.

get_campaign_settings reads both fields back, as you noted. Docs left for you to regenerate after the release, as with #110.

[kLOsk]

Merging, thanks. One correction we'll make in a follow-up commit: the rules bullet says "anything but PRESENCE widens reach", which is right for the positive option and reversed for the negative one (negative PRESENCE excludes fewer people). The code already warns correctly; only the text was wrong. Also, the "unchanged, no warning" case isn't re-checked at apply, so a UI change in between would go through silently. That's fine for now, same as search partners today.