PracticalGCP

SharePoint MCP Bridge for Gemini Enterprise

472 views · 2w ago · 51:03

Many teams use Gemini Enterprise while their everyday document may still be on Microsoft 365. In this video I build and demo an open-source MCP server on Cloud Run. It gives Gemini Enterprise 20 tools for SharePoint and OneDrive: search, read, compare, draft, edit, organise and share. Every call runs with the user's own Microsoft 365 permissions.

Topics: Why · The SharePoint MCP Bridge for Gemini Enterprise · The MCP Connector & Identity · Demo - Legal: "The contract desk" · Demo - Finance: "Month-end close in one conversation" · Demo - Telco / network operations: "Change approval"

How to Build Interactive MCP Apps in Gemini Enterprise

1.1K views · 1mo ago · 34:56

Most developers assume Model Context Protocol is limited to returning plain text or raw JSON. But MCP Apps extend the protocol to embed live UIs directly inside your AI chat interface. Instead of reading through wall-of-text outputs, users can interact with live charts, document highlighters, and custom tools—all with real-time, two-way state sync with the LLM.

Script to Video Agent on Cloud Run

314 views · 2mo ago · 5:17

Writing the script is the easy part. It's the visuals, the voiceover, and lining the two up that eat the afternoon. So I built a pipeline that takes a plain text script and returns a finished The interesting part is the safety story. The model writes JavaScript that has to execute somewhere, so it runs inside a Cloud Run agent sandbox with no credentials and zero network egress, behind three deterministic gates that vet the code before it ever runs.

Agent Sandbox on Cloud Run

470 views · 3mo ago · 31:12

Executing AI-generated code securely just got a whole lot easier on Google Cloud! Google recently announced the preview of "Agent Sandbox on Cloud Run", providing an isolated, zero-trust environment to securely execute dynamic, untrusted code generated by AI agents. Compared to GKE architectures, this serverless approach is incredibly fast to set up, highly convenient, and budget-friendly.

Topics: Key Sandbox CLI Commands · The Sandbox Execution Flow with ADK · Unix-Style ADK Code Execution Tools · The 4 Sandbox Execution Scenarios · Live Demo in Gemini Enterprise App · Summary & Recap

One-shotting ADK Agents with Antigravity & Agents CLI

864 views · 4mo ago · 38:25

Building production-ready AI agents is hard. Not because the core logic is complex, but because the structural boilerplate is massive. You need scaffolding, vector databases, custom evaluations, CI/CD runners, and deployment infrastructure.

A2A Agent with Secure GKE Agent Sandbox

740 views · 4mo ago · 43:57

Guest: Secure GKE Agent Sandbox

Agent skills are great — until they run code. The moment a skill executes code, you've got a problem: that code is AI-generated, so it's effectively untrusted. Run it next to your app, with your app's credentials and network? That's a non-starter for anything enterprise.

ADK Multi Agent Orchestration via BigQuery Remote Functions

1.3K views · 9mo ago · 25:32

Bringing "Brains" to the Data Warehouse: Multi-Agent Intelligence inside BigQuery Most companies have a "Dark Data" problem. They have millions of records in BigQuery, but they lack the "eyes" to read them all for nuanced risks like billing fraud, security gaps, or churn signals.

Hybrid Analytics With BigQuery MCP

1.4K views · 10mo ago · 41:20

If you’ve tried building Conversational Analytics Agents with any SQL enabled database, you know the struggle: generic NL2SQL is great for asking 'how many orders did we have yesterday?', but it falls apart when you need a complex, mission-critical report. It’s either fast and risky, or slow and manual.

A Powerful Combination: A2A + Cloud Run + Gemini Enterprise

3.1K views · 11mo ago · 29:14

We often overlook the infrastructure required for large-scale AI deployment. Building one agent is simple, orchestrating and managing a large network is not - a problem A2A was designed to solve. This video aims to cut through this complexity.

Topics: The Problem Statement & Why A2A is Important · Create A2A Agent and Deploy to Cloud Run with Agent Starter Pack (ASP) · Demo: A2A Agent Deployment & Code Walkthrough · A2A Deep Dive and Agent Card · The Powerful Combination: A2A + Gemini Enterprise · Demo: Invoke A2A Agent from Python & Gemini Enterprise

A Powerful Data Quality Agent for Dataplex

1.4K views · 1y ago · 38:41

The Data Quality Agent for #Dataplex: A Deterministic Approach to Insight. Tracking data quality issues at scale across hundreds of tables within Dataplex requires efficient, trustworthy monitoring solutions that move beyond bespoke, complex dashboards.

Deploy your ADK Agent in under 5 minutes using Agent Starter Pack

2.8K views · 1y ago · 28:38

See the new video made on this topic using Agents CLI: One-shotting ADK Agents with Antigravity & Agents CLI Note: this video is partially out of date due to a recent change to Agent Starter Pack to match a more up to date version of ADK. The concept on the Makefile & deployment python module are still relevant but the deployment logic is now moved into apputils/deploy.py.

Topics: Agenda · Create & Deploy your first ADK agent in under 5 minutes · Project structure generated by Agent Starter Pack · ADK Playground & Notebook for Agent Engine testing · Understanding all the "make commands" · Next steps

Conversational Analytics Agent Part 3 - Cloud Run with IAP

953 views · 1y ago · 18:46

Excited to share Part 3 of my Conversational Analytics Agent series! This time, we're moving beyond the backend agent and tackling a critical piece of the puzzle: building and securing a user-facing frontend. Without a proper UI, you can't scale your agent to a large group of users or effectively track sessions. But more importantly, how do you control who can access it?

Topics: The Frontend with Cloud Run & IAP · Interactive Web App Demo · Code Walkthrough · Code Walkthrough: How to retrieve and validate user identity · Code Walkthrough: Interacting with Agent Engine · Code Walkthrough: Deploying to Cloud Run

Conversational Analytics Agent Part 2 - ADK & Agent Engine

1.8K views · 1y ago · 38:18

A couple weeks ago, I took a deep dive into what it takes to solve complex NL2SQL challenges with the Conversational Analytics API. Now, let's productise it. A powerful API is great, but how do you turn that logic into a robust, interactive, and scalable AI agent? That's the gap we're bridging in this new video.

Topics: What is Agent Development Kit (ADK)? · Example: Conversational Analytics Agent · What is Agent Engine? · ADK Agent code explained · ADK Web, interactive local agent development and debugging · Agent Engine Deployment code explained

Mask Banana - Mask based In-painting with Nano Banana

3.8K views · 1y ago · 18:39

Guests: Mask Banana, Nano Banana

Tired of trying to describe a photo with words just to remove an object? What if you could just draw a mask and get it done instantly? I just built a new web app in less than a day using the #Gemini #NanoBanana model that lets you do exactly that! It combines masked-based inpainting with a simple prompt, making it incredibly intuitive. It’s a huge step up from trying to describe the area you want to edit.

Topics: What is Nano Banana · Strength of the Nano Banana model · What is the Mask Banana Webapp · How does it work? Frontend & backend. · Let's see how good the quality is, live showcase with 10+ photos. · Next steps, and deployment options.

Conversational Analytics Agent Part 1 - Conversational Analytics API

4.1K views · 1y ago · 49:46

Guest: Conversational Analytics API

It uses a powerful NL2SQL engine, to convert your natural language based questions into SQL and retrieve valuable insights from BigQuery. Code

Topics: NL2SQL, how to tackle the 4 key challenges on accuracy · The Conversational Analytics API, the way it works, an illustration of the flow · Demo: Example of a system instruction structure in BigQuery · Demo: Walk through the 5 tables created to mirror real world situation · Demo: Walk through baseline queries (without natrual language) · Demo: Walk through code example for Conversational Analytics API