How do I control the relevance of retrieved data?

You can adjust the 'similarity_top_k' to control how many chunks are retrieved and the 'vector_distance_threshold' to set the strictness of the semantic matching.

How does this skill help with AI hallucinations?

By grounding agent responses in your specific, private data sources, the model relies on authoritative documents provided in the corpus rather than relying solely on its pre-trained knowledge.

What are grounding citations?

Grounding citations are metadata references included in the AI's response that link specific claims back to the source document and excerpt used for retrieval, ensuring transparency.

What is the Vertex AI RAG Engine?

It is a managed service by Google Cloud that handles the ingestion, embedding, indexing, and retrieval phases of a RAG pipeline, simplifying the process of grounding AI models.

Can I ingest local files using these RAG patterns?

Yes, the skill provides specific patterns for uploading local files to Google Cloud Storage (GCS) and then importing them directly into a Vertex AI RAG corpus.

Vertex AI RAG Patterns

Name: Vertex AI RAG Patterns
Author: Agentient

byAgentient

0•

数据科学与机器学习

Implements Retrieval-Augmented Generation (RAG) using the Vertex AI Engine to ground agent responses in private data sources.

This skill provides standardized implementation patterns for the Vertex AI RAG Engine, enabling developers to build sophisticated AI agents that leverage private document corpora. It covers the complete lifecycle of a RAG pipeline—from corpus management and document ingestion via Google Cloud Storage to advanced retrieval tool configuration and citation extraction. By integrating these patterns, users can ensure their Claude-based agents provide accurate, grounded responses with verifiable source citations, significantly reducing hallucinations and improving reliability in production environments.

主要功能

01Automated corpus management including creation and lifecycle control

020 GitHub stars

03End-to-end Vertex AI RAG Engine integration patterns

04Streamlined document ingestion from GCS and local file paths

05Automated grounding metadata parsing for precise citation extraction

06Advanced retrieval tool configuration with tunable similarity thresholds

使用场景

01Creating customer support bots grounded in specific product manuals and FAQs

02Building private corporate knowledge bases for internal AI assistants

03Implementing research tools that require verifiable source citations for legal or technical data

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add agentient/vibekit rag-patterns

For use in Claude.ai and ChatGPT