Back office ops · Production

mAIdAI: serverless personal AI assistant built with Google Chat, Cloud Run, and Vertex AI

The problem

An AI Architect found themselves manually navigating repetitive daily tasks — answering the same questions, digging for documentation links, and context-switching constantly — while generic team bots lacked the specific personal context to help.

Workflow diagram · grounded in source
1
User message in Google Chat
trigger
“Frontend: usage of the Google Chat app interface. No custom UI to build or maintain.”
2
HTTP webhook delivery
integration
“Chat events are delivered via HTTP webhooks.”
3
Command type routing
routing
“The core router handles the distinction between simple commands and AI interactions. This is crucial for latency and cost — not every interaction needs a round-trip to an LLM.”
4
Vertex AI reasoning
ai_action
“The backend connects to Vertex AI (Gemini models) for reasoning, grounded by the personal context file.”
5
Response returned to user
output
“I can have a /links command that returns immediately (0 latency, 0 cost), while a /rewrite command leverages Gemin 2.0 Flash for creative work.”
Reported outcome

The author built a working proof-of-concept personal AI assistant in a few hundred lines of Python, showing that a highly personalized, context-aware agent does not require a massive engineering team.

Reported metrics
Quick command latency0 latency
quick command LLM cost0 cost
Reported stack
Google ChatCloud RunVertex AIFastAPIGoogle GenAI SDK
Source
https://mlops.community/blog/maidai-building-a-personal-assistant-with-google-cloud-and-vertex-ai
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The author built a working proof-of-concept personal AI assistant in a few hundred lines of Python, showing that a highly personalized, context-aware agent does not require a massive engineering team.

What tools did this team use?

Google Chat, Cloud Run, Vertex AI, FastAPI, Google GenAI SDK.

What results were reported?

Quick command latency: 0 latency; quick command LLM cost: 0 cost (source-reported, not independently verified).

How is this back office ops AI workflow structured?

User message in Google Chat → HTTP webhook delivery → Command type routing → Vertex AI reasoning → Response returned to user.