Inspiration

I live inside Google Workspace. Gmail, Calendar, Tasks, Contacts — switching between them constantly throughout the day is something I never quite got used to. The idea behind this project was simple: what if one voice conversation could handle all of it?

What it does

How we built it

The Workspace Voice AI Assistant is my attempt to build that single point of access — an agent you speak to naturally, and it figures out which tool to call, executes it on your behalf, and speaks the result back. No switching tabs, no typing, no context loss between apps. The system design was first created and I put some API and tools together to have a first draft then went own to include some vibe code to build the frontend with fastapi for interaction

Challenges we ran into

The Agent never seemed to maintain biderectional commiunication and cuts off after first interaction, this currently still a bit buggy

Accomplishments that we're proud of

I built this in less three days around a full schedule, which meant I couldn't finish everything I had planned. The bidirectional voice flow, the Google Workspace tool integrations, the GCP Cloud Run deployment pipeline — those work. Some polish is still on the list. But I'd rather submit something real and functional that solves a genuine personal frustration than wait for perfect.

What we learned

I gained a lot of insight into the google ai agent sdk and also some audio processing.

What's next for Gw-Agent

Perfect the bidirectional audio sync and expand usage with other workspace apps

Built With

  • fastapi
  • google-ai-agent-sdk
  • python
Share this project:

Updates