This repository contains a collection of actionable demos for partners to showcase the latest Google Cloud Data & AI features to their clients.
The February release focuses on BigQuery Global Queries, Direct Governance Policies, and Managed MCP for agentic workflows.
- BigQuery Global Queries: Join datasets across multiple regions with zero data movement.
- BigQuery Governance: Legacy Tags vs. Direct Policies: A comparative study of traditional Policy Tags vs. the streamlined 2026 Direct Column Data Policy method.
- BigQuery: Param Queries, Insights & Undelete: Parameterized Queries, Dataset/Table Insights, and Dataset Undeletion.
- ADK Mastery: Dynamic Skills: On-demand instruction loading for complex agent workflows.
- Managed MCP: BigQuery: High-performance data analysis using the BigQuery MCP server.
- Gemini 3.1 Flash Image: High-quality image generation with state-of-the-art prompt adherence.
- Gemini CLI Browser Agent: (Markdown Walkthrough) Interact with live web pages directly from the terminal.
The March release showcases BigQuery AI Functions, ADK A2A Interoperability, and Enterprise Data Discovery across 6 demos.
| Demo | GCP / ADK Features | Key Takeaway |
|---|---|---|
| Conversational Analytics | ObjectRef (GCS ↔ BQ), AI.FORECAST, AI.DETECT_ANOMALIES, partitioned tables, job labels · ADK: BigQueryToolset, BigQueryToolConfig |
Multi-modal analytics: forecast sales, detect anomalies, and cross-reference GCS receipt images — all via natural language |
| Dataplex Catalog Discovery | Dataplex Catalog search, BigQuery schema retrieval · ADK: BigQueryToolset (v1.27 — Dataplex search tool) | Agent autonomously discovers unknown tables by searching enterprise metadata, then queries them |
| Agentic Data Migrations | BQMS Managed MCP (SQL translation), DTS Managed MCP (data transfers), GCS · ADK: McpToolset, StreamableHTTPConnectionParams | Zero-glue migration: translate Hive SQL → GoogleSQL and schedule data transfers via Managed MCP servers |
| Demo | GCP / ADK Features | Key Takeaway |
|---|---|---|
| A2A Protocol | Vertex AI (Gemini 3.1 Pro) · ADK: A2A Protocol, RemoteA2aAgent, to_a2a(), agent card discovery |
Cross-service agent collaboration: a manager agent delegates to a remote specialist over the A2A handshake |
| Native RAG & Agentic Search | BigQuery AI.EMBED & AI.SIMILARITY · ADK: BigQueryToolset, FunctionTool |
Two search patterns compared: Native BigQuery SQL RAG and direct agentic file discovery (Function Tool) |
| ADK Prompt Optimizer | Vertex AI (Gemini 3.1 Pro) · ADK: adk optimize & GEPA root agent optimizer |
Systematic prompt engineering: evaluate baseline vs. optimized agent instructions against ground truth |
Common to all March demos: ADK v1.28.0, Runner pattern with
InMemorySessionService, Vertex AI auth (GOOGLE_GENAI_USE_VERTEXAI), Gemini 3.1 Pro (Preview).
Maintained by the Partner Data & AI team.