Gemini API usage in a private Telegram bot – architecture and compliance questions

Gemini API
,
1

Hi everyone,

I’ve been building a private Telegram bot for about a year now. The bot is used in small private groups, mostly for study and community purposes. It uses the Gemini API to understand conversation context and generate helpful responses.

Before I go further with development, I want to make sure I’m doing things the right way architecturally, so I have a few questions I’d appreciate community input or Google staff guidance on.

What the bot does

The bot operates in Telegram groups and responds when users address it. Different parts of the application handle different jobs, including conversational replies, intent detection, moderation, context processing, transcription, and a voice feature.

Each workload has its own application-level request budgets and rate limits. Model output is treated as application data and is never given direct authority to perform privileged operations.

My questions

The part I’m less sure about is the credential setup. During development, a few collaborators helped with the project and used their own Google accounts to create API credentials for testing different workloads. I want to understand whether this architecture is acceptable.

Specifically:

  1. Is it acceptable for the same application to use API credentials associated with multiple Google accounts or projects, provided that each credential is legitimately owned and used by someone involved in the project?

  2. Is there a recommended way to structure Gemini API credentials when you want workload isolation — for example, separating voice, moderation, and conversational workloads across different credentials or projects?

  3. Could using multiple credentials for a single application be interpreted as quota or rate-limit circumvention, even when that is not the intent and the credentials are being used for legitimate workload separation?

  4. Are there any specific Gemini API policies, terms, or usage patterns I should review to make sure a private conversational Telegram bot like this remains compliant long-term?

I’m not trying to get around quotas, rate limits, or any other enforcement mechanism. I simply want to understand the correct and supported way to operate this architecture before continuing development.

Thanks in advance for any guidance.