Back to all discussions

  •   about 2 months ago

Replit track stack

Hi

In Rules I noted that:

Provide a URL to your code repository for judging and testing on either GitHub, GitLab and Bitbucket. It must contain all source code, assets, and instructions needed for the project to run, and must demonstrate the use of Google Cloud and the Partner services at runtime in your code — imported and actually called (a library import, an app/backend entry point, or a loaded agent/flow/MCP config), not just named in the README. Accepted Google Cloud packages/SDKs: google-adk, google-genai, google-generativeai, or google-cloud-aiplatform (any generation — legacy libraries count equally).

And in Replit I noted that it has Replit AI Integrations:
https://replit.com/blog/ai-integrations

To confirm about this track:

1. Is it necessary to "configure directly" IA agents with Google's SDKs en Replit App?
2. Is it enough to use Replit AI Integrations by selecting Google Gemini?

  • 3 comments

  • Manager   •   about 2 months ago

    Hi Miguel, thanks for checking before you build - two separate requirements are at play here, worth untangling.

    1) The Replit track's own requirement is about your dev process and deployment, not your AI SDK: your project must be built using Replit Agent as part of development, and the finished project must be hosted and deployed directly on Replit (a project URL on a replit.app or replit.dev domain). Replit AI Integrations doesn't factor into this part either way.

    2) Separately, the general Submission Requirements (which apply to every track) require your code repository to demonstrate use of Google Cloud at runtime via one of the accepted SDKs - google-adk, google-genai, google-generativeai, or google-cloud-aiplatform - imported and actually called in your code. This is what your two questions are really about: you don't need to hand-configure the SDK yourself if a tool wires it in for you, but whether selecting Google Gemini through Replit AI Integrations satisfies this depends on which SDK that integration calls under the hood. If it results in your repo importing/calling one of those four packages, you're covered; if it uses a different wrapper that doesn't, it likely wouldn't satisfy this specific requirement (even though it would still be a compliant AI tool under the AI usage limitation, since it's both a Google model and your chosen Partner's built-in AI feature).

    We're confirming with the Replit team exactly which SDK the Gemini integration calls under the hood and will follow up. In the meantime, check your repo's actual import/call sites for one of the four accepted packages directly. Full rules: https://agentic-cinema.devpost.com/rules

  •   •   about 2 months ago

    Hi Janet.

    Thanks for your response.

  • Manager   •   about 2 months ago

    Hi Miguel, here's the updates:

    1) The Replit track's own requirement is about your dev process and deployment, not your AI SDK: your project must be built using Replit Agent as part of development, and the finished project must be hosted and deployed directly on Replit (a project URL on a replit.app or replit.dev domain). Replit AI Integrations doesn't factor into this part either way.

    2) Separately, the general Submission Requirements (which apply to every track) require your code repository to demonstrate use of Google Cloud at runtime via one of the accepted SDKs - google-adk, google-genai, google-generativeai, or google-cloud-aiplatform - imported and actually called in your code. Good news on this: per the Replit team, selecting Google Gemini through Replit AI Integrations is enough to provision Gemini in your Replit app - no separate Google API key or manual credential setup is required, and as long as your project is genuinely working with Google data/context, the integration pulls in the necessary Google SDK automatically. So you're covered on both counts, as long as your repo reflects that Gemini integration being called at runtime. Full rules: https://agentic-cinema.devpost.com/rules

Log in or sign up for Devpost to join the conversation.