Agentoc: Rapid RAG App with Vertex AI

Project Overview

This repository supports the Medium article, **Hackathon Speedrun: Build & Deploy a RAG App in Minutes with Vertex AI Studio & Vertex AI Search!".

The project focuses on creating a Q&A assistant grounded in the documentation for the A2A and MCP protocols, showcasing a fast path from data to deployed AI assistant.

Key Technologies

• Google Cloud Storage: Storing source documents.
• Vertex AI Search: Indexing documents for retrieval.
• Vertex AI Studio: Configuring the LLM prompt and grounding.
• Cloud Run: (Optional) Deploying the web application.
• Python: For data preparation scripts.

Getting Started

Follow the detailed steps in the accompanying Medium article ([Link to your Medium article]).

The high-level process involves:

1. Setting up a Google Cloud project and enabling necessary APIs.
2. Creating a GCS bucket.
3. Using the scripts/prepare.py script to process your documentation (place source docs in A2A/ and modelcontextprotocol/) and upload the results (metadata.jsonlines and outputs/) to GCS.
4. Creating a Vertex AI Search Data Store in the Google Cloud Console, linking it to your GCS data.
5. Configuring a grounded prompt in Vertex AI Studio using your Vertex AI Search Data Store.
6. (Optional) Testing via the Vertex AI Studio interface, getting code examples, or deploying to Cloud Run.

Repository Contents

• A2A/, modelcontextprotocol/: Placeholders for your source documentation.
• scripts/: Contains the prepare.py script and example test_get_code.py.
• outputs/: Directory for processed text files (generated by prepare.py).
• metadata.jsonlines: JSON Lines file linking to processed documents (generated by prepare.py).
• README.md: This file.

Learn More

• Medium Article: Hackathon Speedrun: Build & Deploy a RAG App in Minutes with Vertex AI Studio & Vertex AI Search!
• A2A Protocol: https://google.github.io/A2A/
• Model Context Protocol (MCP): https://modelcontextprotocol.io/introduction

License

MIT License