Compliance Help: Gemini Grounding/Search Suggestions in Text-Only IRC Bot

Hi Google AI Dev Team & Community,

I’m using the Gemini API in my Twitch chatbot (IRC) and need clarification on compliance for search results, specifically in a text-only environment.

The Conflict:

  • Requirement: Documentation for grounding/search features mandates displaying visual Search Suggestion chips (HTML/CSS, specific styling, tap-to-SRP interaction) via renderedContent. https://ai.google.dev/gemini-api/docs/grounding/search-suggestions
  • Limitation: My text-only IRC environment cannot render HTML/CSS, guarantee visual styles, or replicate the required interactive “chip” behavior.

Implementation Details:

I am enabling search using the tools parameter in my generateContent calls, specifically: tools: [{ googleSearch: {} }]. My code correctly retrieves citationMetadata.citationSources when search results are used.

Seeking Guidance / Permission:

The Gemini API ToS mention deviations are possible “Unless permitted by Google in writing”. Given this:

  1. Is there an official method for compliance regarding search results/citations in text-only environments like IRC?
  2. What is the appropriate channel or process for developers (especially those without paid support contracts) to formally request the “written permission” mentioned in the Terms of Service for an alternative compliance method needed for text-only environments?
  3. Regarding compliance when using tools: [{ googleSearch: {} }] for search in my text-only IRC bot: Assuming the mandatory visual Search Suggestion chip requirement still applies even with this method (as documentation suggests), how can text-based applications like mine comply where rendering the HTML/CSS chip is impossible? Is there an approved method? Could displaying source URLs from citationMetadata textually in chat, possibly supplemented by rendering the chip externally via an OBS browser source, be considered compliant, potentially with the necessary written permission?

Disabling search/grounding limits the bot’s usefulness, so I’m seeking a compliant way forward that works within IRC’s constraints.

Any clarification or guidance would be greatly appreciated!
Thanks,
Henry “Modified by moderator”