GSoC 2025 Proposal Help – Boosting Gemini in CrewAI!

General
1

Hey CrewAI gang!

I’m Saad, a CS junior from Habib University, Pakistan, and an OSS contributor (merged PRs in LangChain and LlamaIndex). I’ve been digging into CrewAI lately—seriously impressed by how it orchestrates AI agents into slick, collaborative crews! I’m now working on a GSoC 2025 proposal for Google DeepMind’s “Enhance Gemini API Integrations in OSS Agents Tools” project, and I’d love to feature CrewAI as a key focus alongside other tools like LangChain.

I’m aiming to supercharge Gemini API support here, and I’d really value your insights:

  1. Are there any gaps in Gemini integration—like multimodal inputs, function calling, or specific model configs—that you think could make CrewAI’s agents even more powerful?
  2. What’s on the community wishlist for Gemini features? Maybe performance boosts, new agent capabilities, or better docs/examples?

Your feedback would help me craft a proposal that hits the mark for CrewAI users. Also, if it’s okay, I’d love to mention the community’s input in my submission to highlight how I’m tapping into this awesome group. Thanks a bunch for any thoughts you can share—super excited to contribute here!

2

I’m sure searching this community will surface a lot issues related to Gemini. One thing, crewAI relies on LiteLLM so you might need to look into that.

Good luck with your proposal.

3 Likes
3

Thank you for sharing @msaadg. Like @zinyando said, we do rely on LiteLLM and all the issues would be from how they’re integrating Gemini.

1 Like